Air Pollution Modeling and its Application XIX (NATO Science for Peace and Security Series C: Environmental Security)

Air Pollution Modeling and Its Application XIX NATO Science for Peace and Security Series This Series presents the re...

Author: Carlos Borrego | Ana Isabel Miranda

13 downloads 1133 Views 19MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Air Pollution Modeling and Its Application XIX

NATO Science for Peace and Security Series This Series presents the results of scientific meetings supported under the NATO Programme: Science for Peace and Security (SPS). The NATO SPS Programme supports meetings in the following Key Priority areas: (1) Defence Against Terrorism; (2) Countering other Threats to Security and (3) NATO, Partner and Mediterranean Dialogue Country Priorities. The types of meeting supported are generally "Advanced Study Institutes" and "Advanced Research Workshops". The NATO SPS Series collects together the results of these meetings. The meetings are coorganized by scientists from NATO countries and scientists from NATO's "Partner" or "Mediterranean Dialogue" countries. The observations and recommendations made at the meetings, as well as the contents of the volumes in the Series, reflect those of participants and contributors only; they should not necessarily be regarded as reflecting NATO views or policy. Advanced Study Institutes (ASI) are high-level tutorial courses intended to convey the latest developments in a subject to an advanced-level audience Advanced Research Workshops (ARW) are expert meetings where an intense but informal exchange of views at the frontiers of a subject aims at identifying directions for future action Following a transformation of the programme in 2006 the Series has been re-named and re-organised. Recent volumes on topics not related to security, which result from meetings supported under the programme earlier, may be found in the NATO Science Series. The Series is published by IOS Press, Amsterdam, and Springer, Dordrecht, in conjunction with the NATO Public Diplomacy Division. Sub-Series A. B. C. D. E.

Chemistry and Biology Physics and Biophysics Environmental Security Information and Communication Security Human and Societal Dynamics

http://www.nato.int/science http://www.springer.com http://www.iospress.nl

Series C: Environmental Security

Springer Springer Springer IOS Press IOS Press

Air Pollution Modeling and Its Application XIX

edited by

Carlos Borrego University of Aveiro, Portugal and

Ana Isabel Miranda University of Aveiro, Portugal

Published in cooperation with NATO Public Diplomacy Division

Proceedings of the 29th NATO/CCMS International Technical Meeting on Air Pollution Modelling and Its Application Aveiro, Portugal 24–28 September 2007

Library of Congress Control Number: 2008928516

ISBN 978-1-4020-8452-2 (PB) ISBN 978-1-4020-8451-5 (HB) ISBN 978-1-4020-8453-9 (e-book)

Published by Springer, P.O. Box 17, 3300 AA Dordrecht, The Netherlands. www.springer.com

Printed on acid-free paper

All Rights Reserved © 2008 Springer Science + Business Media B.V. No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work.

Previous Volumes in this Mini-Series Volumes I-XII were included in the NATO Challenges of Modern Society Series. AIR POLLUTION MODELING AND ITS APPLICATION I Edited by C. De Wispelaere AIR POLLUTION MODELING AND ITS APPLICATION II Edited by C. De Wispelaere AIR POLLUTION MODELING AND ITS APPLICATION III Edited by C. De Wispelaere AIR POLLUTION MODELING AND ITS APPLICATION IV Edited by C. De Wispelaere AIR POLLUTION MODELING AND ITS APPLICATION V Edited by C. De Wispelaere, Francis A. Schiermeier, and Noor V. Gillani AIR POLLUTION MODELING AND ITS APPLICATION VI Edited by Han van Dop AIR POLLUTION MODELING AND ITS APPLICATION VII Edited by Han van Dop AIR POLLUTION MODELING AND ITS APPLICATION VIII Edited by Han van Dop and Douw G. Steyn AIR POLLUTION MODELING AND ITS APPLICATION IX Edited by Han van Dop and George Kallos AIR POLLUTION MODELING AND ITS APPLICATION X Edited by Sven-Erik Gryning and Millán M. Millán AIR POLLUTION MODELING AND ITS APPLICATION XI Edited by Sven-Erik Gryning and Francis A. Schiermeier AIR POLLUTION MODELING AND ITS APPLICATION XII Edited by Sven-Erik Gryning and Nadine Chaumerliac AIR POLLUTION MODELING AND ITS APPLICATION XIII Edited by Sven-Erik Gryning and Ekaterina Batchvarova AIR POLLUTION MODELING AND ITS APPLICATION XIV Edited by Sven-Erik Gryning and Francis A. Schiermeier AIR POLLUTION MODELING AND ITS APPLICATION XV Edited by Carlos Borrego and Guy Schayes AIR POLLUTION MODELING AND ITS APPLICATION XVI Edited by Carlos Borrego and Selahattin Incecik AIR POLLUTION MODELING AND ITS APPLICATION XVII Edited by Carlos Borrego and Ann-Lise Norman AIR POLLUTION MODELING AND ITS APPLICATION XVIII Edited by Carlos Borrego and Eberhard Renner

v

Preface In 1969, the North Atlantic Treaty Organization (NATO) established the Committee on Challenges of Modern Society (CCMS). The subject of air pollution was from the start one of the priority problems under study within the framework of various pilot studies undertaken by this committee. The organization of a periodic conference dealing with air pollution modelling and its application has become one of the main activities within the pilot study relating to air pollution. The first five international conferences were organized by the United States as the pilot country, the second five by the Federal Republic of Germany, the third five by Belgium, the fourth four by The Netherlands, the next five by Denmark and the last five by Portugal. This volume contains the abstracts of papers and posters presented at the 29th NATO/CCMS International Technical Meeting on Air Pollution Modelling and Its Application, held in Aveiro, Portugal, during September 24–28, 2007. This ITM was organized by the University of Aveiro, Portugal (Pilot Country and Host Organization). The key topics distinguished at this ITM included: Local and urban scale modelling; Regional and intercontinental modelling; Data assimilation and air quality forecasting; Model assessment and verification; Aerosols in the atmosphere; Interactions between climate change and air quality; Air quality and human health. The ITM was attended by 156 participants representing 36 countries from Asia, Australia, Europe as well as North and South America. Invited papers were presented by Alexander Baklanov, Denmark (On-line integrated meteorological and chemical transport modelling: advantages and prospectives), Ashok Gadgil, USA (Rapid Data Assimilation in the Indoor Environment: theory and examples from real-time interpretation of indoor plumes of airborne chemicals), Gregory Carmichael, USA (Predicting air quality: current status and future directions) and Michael Brauer, Canada (Models of exposure for use in epidemiological studies of air pollution health impacts). On behalf of the ITM Scientific Committee and as organizers and editors, we should like to thank all the participants who made the meeting so successful. Among the participants, we especially recognize the efforts of the chairpersons and rapporteurs. Finally, special thanks to the sponsoring Institution University of Aveiro, Portugal, and the sponsoring organizations NATO Committee on the Challenges of Modern Society and GRICES (Office for International Relations in Science and Higher Education, Portugal). A special grant was given by EURASAP (European Association for the Sciences of Air Pollution) to award a prize to young researchers for the best paper or poster. The next conference will be held in 2009 in the United States of America. Aveiro, Portugal Aveiro, Portugal

Ana Isabel Miranda (Local Conference Organizer) Carlos Borrego (Scientific Committee Chairperson) vii

The members of the Scientific Committee for the 29th NATO/SPS International Technical Meeting (ITM) on Air Pollution Modeling and Its Application

G. Schayes, Belgium D. Syrakov, Bulgaria D. Steyn, Canada S.-E. Gryning, Denmark N. Chaumerliac, France E. Renner, Germany W. Klug, Germany G. Kallos, Greece D. Anfossi, Italy T. Iversen, Norway

C. Borrego, Portugal (Chairman) A.I. Miranda, Portugal J.M. Baldasano, Spain P. Builtjes, The Netherlands H. Dop, The Netherlands S. Incecik, Turkey B. Fisher, United Kingdom S.T. Rao, United States F. Schiermeier, United States

ix

History of NATO/CCMS air pollution pilot studies

Pilot Study on Air Pollution: International Technical Meetings (ITM) on Air Pollution Modelling and Its Application Dates of Completed Pilot Studies: 1969 1975

- 1974 - 1979

1980

- 1984

Air Pollution Pilot Study (Pilot Country – United States) Air Pollution Assessment Methodology and Modelling (Pilot Country – Germany) Air Pollution Control Strategies and Impact Modelling (Pilot Country – Germany)

Dates and Locations of Pilot Study Follow-Up Meetings: Pilot Country United States (R.A. McCormick, L.E. Niemeyer) February 1971 Eindhoven, First Conference on Low Pollution The Netherlands Power Systems Development July 1971 Paris, Second Meeting of the Expert Panel on France Air Pollution Modelling

From 1972 to 2000 all of the meetings were entitled NATO/CCMS International Technical Meetings (ITM) on Air Pollution Modelling and Its Application October May June

1972 1973 1974

Paris, France Oberursel, Federal Republic of Germany Roskilde, Denmark

3rd ITM 4th ITM 5th ITM

Pilot Country Germany (Erich Weber) September 1975 Frankfurt, Federal Republic of Germany September 1976 Airlie House, Virginia, USA September 1977 Louvain-La-Neuve, Belgium August 1978 Toronto, Ontario, Canada October 1979 Rome, Italy

6th ITM 7th ITM 8th ITM 9th ITM 10th ITM

Pilot Country Belgium (C. De Wispelaere) November 1980 Amsterdam, The Netherlands August 1981 Menlo Park, California, USA September 1982 Ile des Embiez, France September 1983 Copenhagen, Denmark April 1985 St. Louis, Missouri, USA

11th ITM 12th ITM 13th ITM 14th ITM 15th ITM xi

xii

History of NATO/CCMS air Pollution Pilot Studies

Pilot Country The Netherlands (Han van Dop) April 1987 Lindau, Federal Republic of Germany September 1988 Cambridge, United Kingdom May 1990 Vancouver, British Columbia, Canada September 1991 lerapetra, Crete, Greece

16th ITM 17th ITM 18th ITM 19th ITM

Pilot Country Denmark (Sven-Erik Gryning) November 1993 Valencia, Spain November 1995 Baltimore, Maryland, USA June 1997 Clermont-Ferrand, France September 1998 Varna, Bulgaria May 2000 Boulder, Colorado, USA

20th ITM 21st ITM 22nd ITM 23rd ITM 24th ITM

All of the following meetings were entitled NATO/SPS International Technical Meetings (ITM) on Air Pollution Modelling and Its Application. Pilot Country – Portugal (Carlos Borrego) October 2001 Louvain-la-Neuve, Belgium May 2003 Istanbul, Turkey October 2004 Banff, Canada May 2006 Leipzig, Federal Republic of Germany September 2007 Aveiro, Portugal

25th ITM 26th ITM 27th ITM 28th ITM 29th ITM

List of Participants The 29th NATO/CCMS International Technical Meeting on Air Pollution Modeling and Its Application, Aveiro, Portugal, September 24–28, 2007.

Albania Mima, Marieta

Environmental Center for Administration and Technology Rr.A. Frasheri, Pall.16/ Shk.6/ Ap.53, Tirana [email protected]

Australia Hurley, Peter

CSIRO Marine and Atmospheric Research 107-121 Station Street, Private Bag 1, Aspendale, 3195 Melbourne [email protected]

Olaru, Doina

University of Western Australia Business School, 35 Stirling Highway 6009 Crawley [email protected]

Austria Pechinger, Ulrike

ZAMG Environmental Meteorology Hohe Warte 38, A 1191 Vienna [email protected]

Belgium Andy, Delcloo

Royal Meteorological Institute of Belgium (RMI) Observations Ringlaan 3, 1180 Ukkel [email protected]

Clemens, Mensink

VITO NV, Integrated Environmental Studies Boeretang 200, 2400 MOL [email protected]

Schayes, Guy

Université de Louvain, Department de Astronomy e Geophysics Chemin du Cyclotron 2, B-1348 Louvain-laNeuve [email protected] xiii

xiv

Stijn, Janssen

List of Participants

VITO NV, Integrated Environmental Studies Boeretang 200, 2400 MOL [email protected]

Brazil Escada, Marcos

Petroleo Brasileiro S.A., Environment Department Rodovia Presidente Dutra, km 143, 12223900 São José dos Campos [email protected]

Bulgaria Batchvarova, Ekaterina

National Institute of Meteorology and Hydrology Blvd. Tzarigradsko Chaussee 66, 1784 Sofia [email protected]

Prodanova, Maria


Syrakov, Dimiter


Canada Brauer, Michael

The University of British Columbia, School of Occupational and Environmental Hygiene 2206 East Mall, 3rd Floor, V6T1Z3 Vancouver [email protected]

Flagg, David

York University, Department of Earth and Space Science and Engineering 230 Vaughan Road, M6C 2M6 Toronto [email protected]

Gong, Wanmin

Environment Canada, Air Quality Research Division 4905 Dufferin Street, M3H 5T4 Toronto [email protected]


xv

Sloan, James

University of Waterloo, Department of Earth Sciences 200 University Ave. W, N2L3G1 Waterloo [email protected]

Steyn, Douw

The University of British Columbia, Department of Earth and Ocean Sciences 6339 Stores Road, V6T 1Z4 Vancouver [email protected]

Talbot, Donald

Canadian Meteorological Center, Environment Canada 2121 Rte TransCanadienne, H9P 1J3 Dorval [email protected]

Czech Republic Fuka, Vladimir

Charles University, Department of Meteorology and Environment Protection V Holesovickach 2, 18000 Prague [email protected]

Halenka, Tomas


Zemankova, Katerina


Denmark Baklanov, Alexander

Danish Meteorological Institute (DMI), Meteorological Research Lyngbyvej 100, DK-2100 Copenhagen [email protected]

Gryning, Sven-Erik

Risø National Laboratory, Wind Energy Department Frederiksborgueg 399, DK-4000 Roskilde [email protected]

Hedegaard, Gitte

National Environmental Research Institute, Department of Atmospheric Environment Frederiksborgvej 399, 4000 Roskilde [email protected]

xvi


Søren, Thykier-Nielsen

Risø National Laboratory, Wind Energy Department Frederiksborgueg 399, DK-4000 Roskilde [email protected]

Sørensen, Jens Havskov

Danish Meteorological Institute (DMI), Research Department Lyngbyvej 100, DK-2100 Copenhagen [email protected]

Estonia Kaasik, Marko

University of Tartu, Institute of Environmental Physics Ülikoooli 18, 50090 Tartu [email protected]

Prank, Marje

University of Tartu, Institute of Environmental Physics Ülikoooli 18, 50090 Tartu [email protected]

Teinemaa, Erik

Estonian Environmental Research Centre Air Quality Management Marja 4D, 10617 Tallinn [email protected]

Finland Karppinen, Ari

Finnish Meteorological Institute, Research and Development P.O. Box 503, 00101 Helsinki [email protected]

Siljamo, Pilvi

Finnish Meteorological Institute, Meteorological Research P.O. Box 503, 00101 Helsinki [email protected]

Soares, Joana

Finnish Meteorological Institute Air Quality Research Erik Palmenin Aukio 1, 00560 Helsinki [email protected]

Sofiev, Mikhail

Finnish Meteorological Institute, Air Quality Research Erik Palmenin Aukio 1, 00560 Helsinki [email protected]


xvii

France Armand, Patrick

Commissariat à l’Energie Atomique (CEA), DIF/ DASE/ SRCE Bátiment G, 91680 Bruyères-le-Châtel [email protected]

Blot, Romain

University of South – Toulon Var, LSEET-LEPI-ISITV Av. G. Pompidou, BP 62, 83162 La Valette du Var [email protected]

Carvalho, Ana

Laboratoire de Météorologie Dynamique École Polytechnique, 91128 Palaiseau [email protected]

Chaumerliac, Nadine

Université Blaise Pascal, LaMP CNRS 24 avenue des landais, 63170 Aubière [email protected]

Chollet, Jean-Pierre

Université Joseph Fourier Grenoble BP 53, 38041 Grenoble Cedex [email protected]

Didier, Damien

IRSN, DEI/SESUC/BNTA BP17, 92262 Fontenay aux Roses [email protected]

Freve-Nollet, Valerie

Université des Sciences et technologies de Lille, PC2A Bat. C11, 59655 Villeneuve d’Ascq [email protected]

Menut, Laurent

Laboratoire de Météorologie Dynamique, IPSL/CNRS École Polytechnique, 91128 Palaiseau [email protected]

Quelo, Denis

IRSN, DEI/SESUC/BNTA BP17, 92262 Fontenay-aux-Roses [email protected]

xviii


Queguiner, Solen

CEREA (EDF R&D/ENPC), MFEE 6 Quai Watier, F-78240 CHATOU [email protected]

Terrenoire, Etienne

Université des Sciences et technologies de Lille, PC2A Bat. C11, 59655 Villeneuve d’Ascq [email protected]

Vautard, Robert

LSCE/IPSL – Laboratoire CEA/CNRS/UVSQ Orme des Merisiers - Bât. 701, 91191 GIF SUR YVETTE [email protected]

Georgia Tavamaishvili, Ketevan

The National Forensics Bureau, Department of Forensic Chemistry Ilia Chavchavadze ave. # 84, 0162 Tbilisi [email protected]

Germany Aulinger, Armin

GKSS Research Center, Institute for Coastal Research Max-Planck-Str. 1, 21502 Geesthacht [email protected]

Bewersdorff, Ines

GKSS Research Center, Institute for Coastal Research Max-Planck-Str. 1, 21502 Geesthacht [email protected]

Graff, Arno

Umweltbundesamt, Air Quality Assessment Wörlitzer Platz 1, 06844 Dessau [email protected]

Kerschbaumer, Andreas

Freie Universitaet Berlin, Institut fuer Meteorologie Carl-Heinrich-Becker-Weg 6-10, 12165 Berlin [email protected]

Renner, Eberhard

Institute for Tropospheric Research, Modelling Permoserstraße 15, 04318 Leipzig [email protected]


xix

Suppan, Peter

Institute for Meteorology and Climate Research, Atmospheric Environmental Research (IMK-IFU) Kreuzeckbahnstr.19, 82467 Garmisch-Partenkirchen [email protected]

Wolke, Ralf

Institute for Tropospheric Research, Modelling Permoserstraße 15, 04318 Leipzig [email protected]

Greece Astitha, Marina

University of Athens, School of Physics University Campus, Bldg PHYS-5, 15784 Athens [email protected]

Bartzis, John

University of West Macedonia, Engineering and Management of Energy Resources Sialvera and Bakola Str., 50100 Kozani [email protected]

Kallos, George


Kushta, Joni


Israel Kishcha, Pavel

Tel-Aviv University, Department of Geophysics Ramat Aviv, 69978 Tel-Aviv [email protected]

Reisin, Tamir

Soreq Nuclear Research Center, Applied Physics 81800 Yavne [email protected]


xx

Terliuc, Benjamin

Nuclear Research Center Negev, Environmental Researches P.O. Box 9001, 84190 Beer Sheva [email protected]

Italy Anfossi, Domenico

CNR-ISAC, Corso Fiume 4, 10133 Torino [email protected]

Belfiore, Giovanni

ISAC – CNR, Corso Fiume 4, 10010 Torino [email protected]

Carnevale, Claudio

Università degli Studi di Brescia, Dipartimento di Elettronica per l’Automazione Via Branze 38, 25123 Brescia [email protected]

Mircea, Mihaela

Istituto di Scienze dell Via Gobetti 101, Bologna [email protected]

Pirovano, Guido

Cesi Ricera SPA Via Rubattino, 54, 20134 Milano [email protected]

Trini Castelli, Silvia

CNR, National Research Council Corso Fiume 4, 10133 Torino [email protected]

Trozzi, Carlo

Techne Consulting s.r.l. Via G. Ricci Curbastro 34, I00149 Roma [email protected]

Japan Takigawa, Masayuki

Frontier Research Center for Global Change 3173-25 Showa-machi, Kanazawa-ku, 236-0001 Yokohama [email protected]

Niwano, Masanori

FRCGC, JAMSTEC Atmospheric Composition Research Program 3173-25 Syowa-machi, Kanazawa-ku, 236-0001 Yokohama [email protected]


xxi

Ohara, Toshimasa

National Institute for Environmental Studies, Asian Environment Research Group 16-2 Onogawa, 305-8506 Tsukuba [email protected]

Kitada, Toshihiro

Toyohashi University of Technology, Department of Ecological Engineering 1-1 Hibarigaoka, Tempaku-cho, 441-8580 Toyohashi [email protected]

Lithuania Ulevicius, Vidmantas

Institute of Physics, Environmental physics and chemistry Savanoriu 231, LT-02300 Vilnius [email protected]

Norway Denby, Bruce

Norwegian Institute for Air Research (NILU), Urban Environment and Industry Department P.O. BOX 100, 2027 Kjeller [email protected]

Guerreiro, Cristina

Norwegian Institute for Air Research (NILU), Urban Environment and Industry Department P.O. BOX 100, Instituttvn. 18, 2027 Kjeller [email protected]

Saltbones, Jørgen Ingar

Norwegian Meteorological Institute, Meteorological Department P.O. Box 43, Blindern, NO-0313 Oslo [email protected]

Portugal Amorim, Jorge Humberto

Universidade de Aveiro, Departamento de Ambiente e Ordenamento Campus Universitário de Santiago, 3810-193 Aveiro [email protected]

Borrego, Carlos


xxii


Carvalho, Anabela


Caseiro, Alexandre


Costa, Ana Margarida


Coutinho, Miguel

Instituto de Ambiente e Desenvolvimento (IDAD) Campus Universitário de Santiago, 3810-193 Aveiro [email protected]

Ferreira, Joana


Lopes, Myriam


Martins, Helena


Martins, Vera



xxiii

Miranda, Ana Isabel


Monteiro, Alexandra


Moreira, Cármen


Tavares, Richard


Tchepel, Oxana

Instituto Politécnico de Leiria, Escola Superior de Tecnologia e Gestão Morro do Lena, Alto do Vieiro, 2411-901 Leiria [email protected]

Tomé, Mário

Instituto Politécnico de Viana do Castelo, Escola Superior de Tecnologia e Gestão Avenida do Atlântico, 4900-348 Viana do Castelo [email protected]

Valente, Joana


Santos, João


xxiv


Santos, Pedro


Silva, João Vasco


Republic of Macedonia Alcinova Monevska, Suzana

Republic Hydrometeorological Institute, Department for Informatics and Telecommunications Skupi bb, 1000 Skopje [email protected]

Russia Genikhovich, Eugene

Voeikov Main Geophysical Observatory, Department of Monitoring of Air Pollution Karbyshev Street, 7, 194021 St. Petersburg [email protected]

Serbia Rajkovic, Borivoj

Faculty of Physics, Institute of Meteorology Dobracina 16, 11000 Belgrade [email protected]

Vujadinovic, Mirjam

Faculty of Physics, Institute of Meteorology Dobracina 16, 11000 Belgrade [email protected]

Slovenia Mlakar, Primoz

MEIS d.o.o. Mali Vrh pri Smarju 78, SI-1293 Smarje – Sap [email protected]

Boznar, Marija Zlata

MEIS d.o.o. Mali Vrh pri Smarju 78, SI-1293 Smarje – Sap [email protected]


xxv

South Korea Kim, Yong Pyo

Ewha Womans University, Environmental Science and Engineering 11-1 Daehyundong, Seoudaemungu, 120-750 Seoul [email protected]

Roh, Woosub

Yonsei University, Atmospheric Science Department Seodaemongu sinchon-dong, 82 Seoul [email protected]

Spain Alonso, Lucio

University of the Basque Country, Chemical & Environmental Engineering Alameda de Urquijo s/n, 48013 Bilbao [email protected]

Arasa, Raul

University of Barcelona, Astronomy and Meteorology Avinguda Diagonal 647 7º Planta, 08028 Barcelona [email protected]

Arnold, Delia

Technical University of Catalonia, Institute of Energy Technologies Av. Diagonal 647, 08028 Barcelona [email protected]

Baldasano, Jose M.

Barcelona Supercomputing Center (BSC-CNS), Earth Sciences Jordi Girona 31, 08034 Barcelona [email protected]

Casares-Long, Juan

Universidad de Santiago de Compostela, Department of Chemical Engineering Rua Lope Gomez de Marzoa, 15782 Santiago de Compostela [email protected]

xxvi


De Blas Martín, Maite


Durana Jimeno, Nieves


Gangoiti Bengoa, Gotzon


García Fernández, José


Ilardia, Juan Luis


Navazo Muñoz, Marino


Saavedra, Santiago

Universidad de Santiago de Compostela, Ingeniería Química c/Lope Gómez de Marzoa, 15782 Santiago de Compostela [email protected]

Saez de Cámara, Estíbaliz


San José, Roberto

Universidad Politécnica de Madrid, Environmental Software and Modelling Group Campus de Montegancedo, 28660 Boadilla del Monte, Madrid [email protected]


xxvii

Soler, Maria Rosa

University of Barcelona, Astronomy and Meteorology Avinguda Diagonal 647 7º Planta, 080128 Barcelona [email protected]

Souto, Jose A.

University of Santiago de Compostela, Chemical Engineering Lope Gómez de Marzoa – Campus Sur, 15782 Santiago de Compostela [email protected]

Valdenebro Villar, Veronica

University of the Basque Country, Applied Mathematics Alameda de Urquijo s/n, 48013 Bilbao [email protected]

Vargas, Arturo

Technical University of Catalonia, Institute of Energy Technologies Avda. Diagonal 647, 08028 Barcelona [email protected]

Switzerland Andreani, Sebnem

Paul Scherrer Institute, Laboratory of Atmospheric Chemistry PSI, 5232 Villigen PSI [email protected]

The Netherlands Barbu, Alina

Delft University of Technology, Delft Institute of Applied Mathematics Mekelweg 4, 2628 CD Delft [email protected]

Builtjes, Peter

TNO, Air Quality and Climate P.O. Box 342, 7300 AH Apeldoorn [email protected]

Manders, Astrid

RIVM, Laboratory for Environmental Monitoring P.O. Box 1, 3720 BA Bilthoven [email protected]

Schaap, Martijn

TNO, B&O P.O. box 342, 7300 AH Apeldoorn [email protected]


xxviii

Timmermans, Renske

TNO, Air Quality and Climate P.O. Box 342, 7300 AH Apeldoorn [email protected]

Turkey Incecik, Selahattin

Istanbul Technical University øTÜ Ayaza÷a Kampüsü 34469 Maslak-Istanbul [email protected]

Ukraine Bugaieva, Liudmyla

National Technical University of Ukraine, Chemical Technology Processes Cybernetics 37, Peremogy ave., 03056 Kiev [email protected]

United Kingdom Dore, Anthony

Centre for Ecology ad Hydrology, Bush Estate Penicuik, EH26 OQB Midlothian [email protected]

Fisher, Bernard

Environment Agency, Risk and Forecasting Science Kings Meadow Road, RG1 8DQ Reading [email protected]

Hill, Richard

Westlakes Scientific Consulting Ltd, Environmental Science CA24 3LN Moor Row, Cumbria [email protected]

USA Bullock, Russell

National Oceanic and Atmospheric Administration, Air Resources Laboratory, Atmospheric Sciences Modeling Division US EPA Mail Drop E243-03, 27711 Research Triangle Park, NC [email protected]

Carmichael, Gregory

The University of Iowa 424 IATL, 52242 Iowa [email protected]


xxix

Davidson, Paula

National Oceanic and Atmospheric Administration, NWS 1325 E-W Highway, 20910 Silver Spring MD [email protected]

Duvall, Rachelle

Environmental Protection Agency, National Exposure Research Laboratory 109 TW Alexander Drive, 27711 Research Triangle Park [email protected]

Gadgil, Ashok

Lawrence Berkeley National Laboratory, Environmental Energy Technologies Mailstop 90-3058, 1 Cyclotron Road, 94720 Berkley [email protected]

Hanna, Steven

Harvard University, Exposure, Epidemiology and Risk Program 7 Crescent Avenue, 04046-7235 Kennebunkport, Maine [email protected]

Hogrefe, Christian

University at Albany, Atmospheric Sciences Research Center 251 Fuller Road, 12203 NY, Albany [email protected]

Isakov, Vlad

National Oceanic and Atmospheric Administration /EPA, Atmospheric Sciences Modeling Division 109 T.W. Alexander Drive, 27711 Research Triangle Park, NC [email protected]

Luecken, Deborah

National Oceanic and Atmospheric Administration /EPA, Atmospheric Sciences Modeling Division 109 TW Alexander Drive MD E243-03, 27711 Research Triangle Park [email protected]

Mathur, Rohit

National Oceanic and Atmospheric Administration /EPA, Atmospheric Sciences Modeling Division

xxx


109 T.W. Alexander Drive, 27711 Research Triangle Park, NC [email protected] Mobley, David


Napelenok, Sergey


Nolte, Chris

National Oceanic and Atmospheric Administration, Atmospheric Sciences Modeling Division 109 TW Alexander Drive, Mail Code E243-01, 27701 Research Triangle Park, NC [email protected]

Pleim, Jonathan


Porter, P. Steven

University of Idaho, Civil Engineering 1776 Science Center, 83402 Idaho Falls [email protected]

Rao, S. Trivikrama


Sarwar, Golam

National Oceanic and Atmospheric Administration /EPA, Atmospheric Sciences Modeling Division


xxxi

109 TW Alexander Drive MD E243-03, 27711 Research Triangle Park [email protected] Schiermeier, Frank

National Oceanic and Atmospheric Administration 303 Glasgow Road, 27511 Cary, NC [email protected]

Uliasz, Marek

Colorado State University, Department of Atmospheric Science 1371 Campus Delivery, 80523-1371 Fort Collins, Colorado [email protected]

Venkatram, Akula

University of California, Riverside Department of Mechanical Engineering A368 Bourns Hall, 92521 Riverside [email protected]

Watkins, Timothy


Venezuela Diaz, Luis

PDVSA INTEVEP, Department of Air Quality URB. Santa Rosa, 1201 Norte 4 [email protected]

Contents Preface......................................................................................................................vii List of Participants................................................................................................ xiii Chapter 1

Local and urban scale modeling..................................................... 1

1.1 On-line Integrated Meteorological and Chemical Transport Modelling: Advantages and Prospectives.......................................................... 3 Alexander Baklanov and Ulrik Korsholm 1.2 Modelling of the Urban Wind Profile........................................................ 18 Sven-Erik Gryning and Ekaterina Batchvarova 1.3 Development of a Lagrangian Particle Model for Dense Gas Dispersion in Urban Environment ................................................................... 28 G. Tinarelli, D. Anfossi, S. Trini Castelli, A. Albergel, F. Ganci, G. Belfiore and J. Moussafir 1.4 CFD and Mesoscale Air Quality Modelling Integration: Web Application for Las Palmas (Canary Islands, Spain) ....................................... 37 R. San José, J.L. Pérez, J.L. Morant and R.M. González 1.5 On the Suppression of the Urban Heat Island over Mountainous Terrain in Winter ............................................................................................. 46 Charles Chemel, Jean-Pierre Chollet and Eric Chaxel 1.6 Air Quality Management Strategies in Large Cities: Effects of Changing the Vehicle Fleet Composition in Barcelona and Madrid Greater Areas (Spain) by Introducing Natural Gas Vehicles........................... 54 María Gonçalves, Pedro Jiménez-Guerrero and José M. Baldasano 1.7 Evaluation of the Hazard Prediction and Assessment Capability (HPAC) Model with the Oklahoma City Joint Urban 2003 (JU2003) Tracer Observations......................................................................................... 63 Steven Hanna, Joseph Chang, John White and James Bowers 1.8 Origin and Influence of PM10 Concentrations in Urban and in Rural Environments ......................................................................................... 72 Andreas Kerschbaumer, Rainer Stern and Martin Lutz 1.9 Development and Application of MicroRMS Modelling System to Simulate the Flow, Turbulence and Dispersion in the Presence of Buildings.......................................................................................................... 81 S. Trini Castelli, T.G. Reisin and G. Tinarelli 1.10 Numerical Treatment of Urban and Regional Scale Interactions in Chemistry-Transport Modelling .................................................................. 90 R. Wolke, D. Hinneburg, W. Schröder and E. Renner xxxiii

xxxiv

Chapter 2

Contents

Regional and intercontinental modeling ..................................... 99

2.1 Contribution of Biogenic Emissions to Carbonaceous Aerosols in Summer and Winter in Switzerland: A Modelling Study .......................... 101 ù. Andreani-Aksoyo÷lu, J. Keller, M.R. Alfarra, A.S.H. Prévôt, J.J. Sloan and Z. He 2.2 A Regional Air Quality Model over the Kanto Region of Japan: The Effect of the Physics Parameterization on the Meteorological and Chemical Fields ...................................................................................... 109 Masanori Niwano, Masayuki Takigawa, Hajime Akimoto, Masaaki Takahashi and Mitsuhiro Teshiba 2.3 Regional Aerosol Optical Thickness Distribution Derived by CMAQ Model in the Siberian Forest Fire Emission Episode of May 2003................................................................................................... 118 Hee -Jin In, Yong Pyo Kim and Kwon-Ho Lee 2.4 Modelling the Deposition of Reduced Nitrogen at Different Scales in the United Kingdom ....................................................................... 127 Anthony J. Dore, Mark R. Theobald, Maciej Kryza, Massimo Vieno, Sim Y. Tang and Mark A. Sutton 2.5 Long-Term Simulations of Surface Ozone in East Asia During 1980–2020 with CMAQ and REAS Inventory .............................................. 136 Toshimasa Ohara, Kazuyo Yamaji, Itsushi Uno, Hiroshi Tanimoto, Seiji Sugata, Tatsuya Nagashima, Jun-ichi Kurokawa, Nobuhiro Horii and Hajime Akimoto 2.6 The Use of Meso-Scale Atmospheric Circulation Types as a Strategy for Modelling Long-Term Trends in Air Pollution. ................. 145 Douw Steyn, Bruce Ainslie, J.W. Kaminski, J.C. McConnell, Alberto Martilli and L. Neary 2.7 Development and Applications of Biogenic Emission Term as a Basis of Long-Range Transport of Allergenic Pollen ........................... 154 Pilvi Siljamo, Mikhail Sofiev, Tapio Linkosalo, Hanna Ranta and Jaakko Kukkonen 2.8 High Resolution Nested Runs of the AURAMS Model with Comparisons to PrAIRie2005 Field Study Data.................................... 163 Paul A. Makar, Craig Stroud, Brian Wiens, SunHee Cho, Junhua Zhang, Morad Sassi, John Liggio, Michael Moran, Wanmin Gong, Sunling Gong, Shao-Meng Li, Jeff Brook, Kevin Strawbridge, Kurt Anlauf, Chris Mihele and Desiree Toom-Sauntry 2.9 The Effect of Lateral Boundary Values on Atmospheric Mercury Simulations with the CMAQ Model .............................................................. 173 O. Russell Bullock, Jr.

Contents

xxxv

2.10 Air Pollution Modelling with Perturbational Downscaling ................... 182 Eugene Genikhovich, Mikhail Sofiev, Guy Schayes and Irene Gracheva 2.11 Forest Fires Impact on Air Quality over Portugal.................................. 190 A.I. Miranda, A. Monteiro, V. Martins, A. Carvalho, M. Schaap, P. Builtjes and C. Borrego 2.12 The VetMet Veterinary Decision Support System for Airborne Animal Diseases ............................................................................................ 199 Jens Havskov Sørensen, Søren Alexandersen, Poul Astrup, Knud Erik Christensen, Torben Mikkelsen, Sten Mortensen, Torben Strunge Pedersen and Søren Thykier-Nielsen 2.13 Development and Verification of TAPM .............................................. 208 Peter Hurley 2.14 Development of Fire Emissions Inventory Using Satellite Data ........... 217 Biswadev A. Roy, George A. Pouliot, J. David Mobley, Thompson G. Pace, Thomas E. Pierce, Amber J. Soja, James J. Szykman and J. Al-Saadi 2.15 Toward a US National Air Quality Forecast Capability: Current and Planned Capabilities................................................................................ 226 Paula Davidson, Kenneth Schere, Roland Draxler, Shobha Kondragunta, Richard A. Wayland, James F. Meagher and Rohit Mathur 2.16 Two-Way Coupled Meteorology and Air Quality Modeling................. 235 Jonathan Pleim, Jeffrey Young, David Wong, Rob Gilliam, Tanya Otte and Rohit Mathur 2.17 Numerical Simulation of Air Pollution Transport Under Sea/Land Breeze Situation in Jakarta, Indonesia in Dry Season.................... 243 Toshihiro Kitada, Asep Sofyan and Gakuji Kurata 2.18 Synergetic or Non-Linear Effects in PM10 and PM2.5 Scenario Calculations for 2015 in Belgium .................................................................. 252 Clemens Mensink, Felix Deutsch, Jean Vankerkom and Liliane Janssen Chapter 3

Data assimilation and air quality forecasting ........................... 261

3.1 Rapid Data Assimilation in the Indoor Environment: Theory and Examples from Real-Time Interpretation of Indoor Plumes of Airborne Chemical .................................................................................... 263 Ashok Gadgil, Michael Sohn and Priya Sreedharan 3.2 Comparison of Data Assimilation Methods for Assessing PM10 Exceedances on the European Scale .............................................................. 278 Bruce Denby, Martijn Schaap, Arjo Segers, Peter Builtjes and Jan Horálek

xxxvi

Contents

3.3 An Observing System Simulation Experiment (OSSE) for Aerosols ................................................................................................... 287 Renske Timmermans, Martijn Schaap, Arjo Segers, Hendrik Elbern, Richard Siddans, Stephen Tjemkes, Robert Vautard and Peter Builtjes 3.4 Modelling of Benzo(a)pyrene Depositions over North Sea Coastal Areas: Impact of Emissions from Local and Remote Areas ............. 296 Ines Bewersdorff, Armin Aulinger, Volker Matthias and Markus Quante 3.5 Air Quality Forecasting During Summer 2006: Forest Fires as One of Major Pollution Sources in Europe ............................................... 305 Mikhail Sofiev, Pilvi Siljamo, Ari Karppinen and Jaakko Kukkonen 3.6 Comparison of Methods to Generate Meteorological Inputs for Modeling Dispersion in Coastal Urban Areas.......................................... 313 Akula Venkatram, Wenjun Qian, Tao Zhan and Marko Princevac 3.7 Developing a Method for Resolving NOx Emission Inventory Biases Using Discrete Kalman Filter Inversion, Direct Sensitivities, and Satellite-Based NO2 Columns ................................................................. 322 Sergey L Napelenok, Robert W. Pinder, Alice B. Gilliland and Randall V. Marin 3.8 A Suggested Correction to the EMEP Database, Regarding the Location of a Major Industrial Air Pollution Source in Kola Peninsula........................................................................................................ 331 Marko Kaasik, Marje Prank, Jaakko Kukkonen and Mikhail Sofiev 3.9 Fusing Observations and Model Results for Creation of Enhanced Ozone Spatial Fields: Comparison of Three Techniques .............. 339 Edith Gégo, P.S. Porter, V. Garcia, C. Hogrefe and S.T. Rao Chapter 4

Model assessment and verification............................................. 347

4.1 The Effect of Heterogeneous Reactions on Model Performance for Nitrous Acid............................................................................................. 349 Golam Sarwar, Robin L. Dennis and Bernhard Vogel 4.2 Saharan Dust over the Eastern Mediterranean: Model Sensitivity .......... 358 Pavel Kishcha, Slobodan Nickovic, Eliezer Ganor, Levana Kordova and Pinhas Alpert 4.3 Air Quality Ensemble Forecast Coupling ARPEGE and CHIMERE over Western Europe............................................................ 367 Ana C. Carvalho, Laurent Menut, Robert Vautard and Jean Nicolau 4.4 Uncertainty in Air Quality Decision Making........................................... 376 Bernard Fisher

Contents

xxxvii

4.5 Application of Advanced Particulate Matter Source Apportionment Techniques in the Northern Italy Basin ................................ 385 Marco Bedogni, Simone Casadei, Guido Pirovano, Giovanni Sghirlanzoni and Andrea Zanoni 4.6 Has the Performance of Regional-Scale Photochemical Modelling Systems Changed over the Past Decade? ..................................... 394 C. Hogrefe, J.-Y. Ku, G. Sistla, A. Gilliland, J.S. Irwin, P.S. Porter, E. Gégo, P. Kasibhatla and S.T. Rao 4.7 Application of a Regional Atmospheric Emission Inventory to Ozone and PM Modelling over the French North Region: The summer 2006 Heat Wave Case Study. ................................................... 404 E. Terrenoire and V. Fèvre-Nollet 4.8 Evaluating Regional-Scale Air Quality Models....................................... 412 Alice B. Gilliland, James M. Godowitch, Christian Hogrefe and S.T. Rao 4.9 Ozone Modeling over Italy: A Sensitivity Analysis to Precursors Using BOLCHEM Air Quality Model........................................................... 420 Alberto Maurizi, Mihaela Mircea, Massimo D’Isidoro, Lina Vitali, Fabio Monforti, Gabriele Zanini and Francesco Tampieri 4.10 Modelling Evaluation of PM10 Exposure in Northern Italy in the Framework of CityDeltaIII Project...................................................... 426 C. Carnevale, G. Finzi, E. Pisoni and M. Volta 4.11 Comprehensive Surface-Based Performance Evaluation of a Size- and Composition-Resolved Regional Particulate-Matter Model for a One-Year Simulation ................................................................. 434 M.D. Moran, Q. Zheng, M. Samaali, J. Narayan, R. Pavlovic, S. Cousineau, V.S. Bouchet, M. Sassi, P.A. Makar, W. Gong, S. Gong, C. Stroud and A. Duhamel 4.12 Comparison of Six Widely-Used Dense Gas Dispersion Models for Three Actual Railcar Accidents ............................................................... 443 Steven Hanna, Seshu Dharmavaram, John Zhang, Ian Sykes, Henk Witlox, Shah Khajehnajafi and Kay Koslan 4.13 A Statistical Approach for the Spatial Representativeness of Air Quality Monitoring Stations and the Relevance for Model Validation ...................................................................................................... 452 Stijn Janssen, Felix Deutsch, Gerwin Dumont, Frans Fierens and Clemens Mensink 4.14 Estimation of the Modelling Uncertainty Related with Stochastic Processes .............................................................................. 461 Oxana Tchepel, Alexandra Monteiro and Carlos Borrego

xxxviii

Contents

4.15 Development of a New Canadian Operational Air Quality Forecast Model .............................................................................................. 470 D. Talbot, M.D. Moran, V. Bouchet, L.-P. Crevier, S. Ménard, A. Kallaur and the GEM-MACH Team Chapter 5

Aerosols in the atmosphere......................................................... 479

5.1 Predicting Air Quality: Current Status and Future Directions ................. 481 Gregory R. Carmichael, Adrian Sandu, Tianfeng Chai, Dacian N. Daescu, Emil M. Constantinescu and Youhua Tang 5.2 Diagnostic Analysis of the Three-Dimensional Sulfur Distributions over the Eastern United States Using the CMAQ Model and Measurements from the ICARTT Field Experiment ................... 496 Rohit Mathur, Shawn Roselle, George Pouliot and Golam Sarwar 5.3 Heterogeneous Chemical Processes and Their Role on Particulate Matter Formation in the Mediterranean Region .......................... 505 Marina Astitha, George Kallos, Petros Katsafados and Elias Mavromatidis 5.4 Regional Coverage Modelling of Marine Aerosols Concentration in French Mediterranean Coastal Area .......................................................... 514 Romain Blot, Gilles Tedeshi and Jacques Piazzola 5.5 Formation of Secondary Inorganic Aerosols by High Ammonia Emissions Simulated by LM/MUSCAT ........................................................ 522 Eberhard Renner and Ralf Wolke 5.6 The Origins and Formation Mechanisms of Aerosol during a Measurement Campaign in Finnish Lapland, Evaluated Using the Regional Dispersion Model SILAM ........................................................ 530 Marje Prank, Mikhail Sofiev, Marko Kaasik, Taina Ruuskanen, Jaakko Kukkonen and Markku Kulmala 5.7 Modelling Regional Aerosols: Impact of Cloud Processing on Gases and Particles over Eastern North America and in Its Outflow During ICARTT 2004...................................................... 539 W. Gong, J. Zhang, M.D. Moran, P.A. Makar, S.L. Gong, C. Stroud, V.S. Bouchet, S. Cousineau, S. Ménard, M. Samaali, M. Sassi, B. Pabla, R. Leaitch, A.M. Macdonald, K. Anlauf, K. Hayden, D. Toom-Sauntry, A. Leithead and J.W. Strapp 5.8 On the Role of Ammonia in the Formation of PM2.5 ............................... 548 C. Mensink and F. Deutsch Chapter 6

Interactions between air quality and climate change............... 557

6.1 Linking Global and Regional Models to Simulate U.S. Air Quality in the Year 2050................................................................................ 559 Chris Nolte, Alice Gilliland and Christian Hogrefe

Contents

xxxix

6.2 Impacts of Climate Change on Air Pollution Levels in the Northern Hemisphere with Special Focus on Europe and the Arctic................................................................................................. 568 Gitte B. Hedegaard, Jørgen Brandt, Jesper H. Christensen, Lise M. Frohn, Camilla Geels, Kaj M. Hansen and Martin Stendel 6.3 Regional Climate Change Impacts on Air Quality in CECILIA EC 6FP Project .............................................................................................. 577 Tomas Halenka, Peter Huszar and Michal Belda Chapter 7

Air quality and human health .................................................... 587

7.1 Models of Exposure for Use in Epidemiological Studies of Air Pollution Health Impacts ............................................................................... 589 Michael Brauer, Bruce Ainslie, Michael Buzzelli, Sarah Henderson, Tim Larson, Julian Marshall, Elizabeth Nethery, Douw Steyn and Jason Su 7.2 Long-Term Regional Air Quality Modelling in Support of Health Impact Analyses............................................................................. 605 C. Hogrefe, B. Lynn, K. Knowlton, R. Goldberg, C. Rosenzweig and P.L. Kinney 7.3 A Modeling Methodology to Support Evaluation of Public Health Impacts on Air Pollution Reduction Programs................................... 614 Vlad Isakov and Halûk Özkaynak 7.4 Evaluating the Effects of Emission Reductions on Multiple Pollutants Simultaneously ............................................................................. 623 Deborah Luecken, Alan Cimorelli, Cynthia Stahl and Daniel Tong 7.5 Modelling of the Exposure of Urban Populations to PM2.5, NO2 and O3, and Applications in the Helsinki Metropolitan Area in 2002 and 2025 ........................................................................................................ 632 J. Kukkonen, P. Aarnio, A. Kousa, A. Karppinen, K. Riikonen, B. Alaviippola, M. Kauhaniemi, J. Soares, T. Elolähde and T. Koskentalo 7.6 The Importance of Exposure in Addressing Current and Emerging Air Quality Issues................................................................... 640 Tim Watkins, Ron Williams, Alan Vette, Janet Burke, B.J. George and Vlad Isakov Poster Session ....................................................................................................... 649 P1. Local and urban scale modelling ................................................................. 651 P1.1 Finite Volume Microscale Air-Flow Modelling Using the Immersed Boundary Method ................................................................... 651 V. Fuka and J. Brechler

xl

Contents

P1.2 Simplified Models for Integrated Air Quality Management in Urban Areas............................................................................................... 653 B. Sivertsen, A. Dudek and C. Guerreiro P1.3 Assessment of the Breathability in Urban Canyons Through CFD Simulations and Its Application to Sustainable Urban Design ............. 655 Mário Tomé, Ricardo J. Santos, António Martins and Mário Russo P1.4 Inter-Comparison of Gaussian Plume, Street Canyon and CFD Models for Predicting Ambient Concentrations of NOx and NO2 at Urban Road Junctions ................................................................................ 657 Richard Hill, Peter Jenkinson and Emma Lutman P2. Regional and intercontinental modelling .................................................... 659 P2.1 Local to Regional Dilution and Transformation Processes of the Emissions from Road Transport .......................................................... 659 Dimiter Syrakov, Kostadin Ganev, Reneta Dimitrova, Angelina Todorova, Maria Prodanova and Nikolai Miloshev P2.2 Application of Back Trajectories Using Flextra to Identify the Origin of 137Cs Measured in the City of Barcelona.................................. 661 Delia Arnold, Arturo Vargas, Petra Seibert and Xavier Ortega P2.3 The Role of Sea-Salt Emissions in Air Quality Models ........................ 663 Raúl Arasa, Maria R. Soler and Sara Ortega P2.4 SPECIATE – EPA’s Database of Speciated Emission Profiles............. 665 J. David Mobley, Lee L. Beck, Golam Sarwar, Adam Reff and Marc Houyoux P2.5 Regional Transport of Tropospheric Ozone: A Case Study in the Northwest Coast of Iberian Peninsula.................................................. 667 Santiago Saavedra, María R. Méndez, José A. Souto, José L. Bermúdez, Manuel Vellón and Miguel Costoya P2.6 Modelling of Atmospheric Transport of POPs at the European Scale with a 3D Dynamical Model Polair3D-POP ........................................ 669 Solen Quéguiner and Luc Musson-Genon P2.7 Evolution of the Ozone Episodes in Northern Iberia (Cantabric and Pyrenaic regions) Under West European Atlantic Blocking Anticyclones .................................................................................................. 671 V. Valdenebro, G. Gangoiti, A. Albizuri, L. Alonso, M. Navazo, J.A. García and M.M. Millán P2.8 High Temporal Resolution Measurements and Numerical Simulation of Ozone Precursors in a Rural Background ............................... 673 M. Navazo, N. Durana, L. Alonso, J. Iza, J.A, García, J.L. Ilardia, G. Gangoiti and M. De Blas

Contents

xli

P2.9 Nonlinearity in Source-Receptor Relationship for Sulfur and Nitrate in East Asia ................................................................................. 675 Woo-Sub Roh, Seung-Bum Kim and Tae-Young Lee P2.10 Modelling the Impact of Best Available Techniques for Industrial Emissions Control in Air Quality: Setting Up Inventories and Establishing Projections ....................................................... 677 R. Rodriguez, P. Maceira, J.A. Souto, J. Casares, A. Sáez and M. Costoya P2.11 Lake Breezes in Southern Ontario: Observations, Models and Impacts on Air Quality............................................................................ 679 David Flagg, Jeff Brook, David Sills, Paul Makar, Peter Taylor, Geoff Harris, Robert McLaren and Patrick King P2.12 High Time and Space Resolution Ozone Modelling in Regional Air Quality Management of a Complex Mountain Area Using Calgrid 2.44 ................................................................................ 681 Carlo Trozzi, Silvio Villa and Enzo Piscitello P2.13 Analysis of Atmospheric Transport of Radioactive Debris Related to Nuclear Bomb Tests Performed at Novaya Zemlya ..................... 683 Jørgen Saltbones, Jerzy Bartnicki, Tone Bergan, Brit Salbu, Bjørn Røsting and Hilde Haakenstad P2.14 Development and Application of a New Model for the Atmospheric Transport and Surface Exchange of Semi-Volatile Organics Using the CMAQ Model Framework. ................ 685 Fan Meng, Baoning Zhang, Fuquan Yang and James Sloan P2.15 Saharan Dust over Italy: Simulations with Regional Air Quality Model BOLCHEM ........................................................................... 687 Mihaela Mircea, Massimo D’Isidoro, Alberto Maurizi, Francesco Tampieri, Maria Cristina Facchini, Stefano Decesari and Sandro Fuzzi P3. Data assimilation and air quality forecasting............................................. 689 P3.1 Detection of a Possible Source of Air Pollution Using a Combination of Measurements and Inverse Modelling .............................. 689 Borivoj Rajkovic, Zoran Grsic and Mirjam Vujadinovic P3.2 Improving Emission Inventory in Lithuania.......................................... 691 Vidmantas Ulevicius, Vytautas Vebra, Kestutis Senuta and Svetlana Bycenkiene P4. Model assesment and verification................................................................ 693 P4.1 Tropospheric Ozone and Biogenic Emissions in the Czech Republic......................................................................................................... 693 K. Zemankova and J. Brechler

xlii

Contents

P4.2 Air Pollution Dispersion Modelling Arround Thermal Power Plant Trbovlje in Complex Terrain – Model Verification and Regulatory Planning................................................................................ 695 Marija Zlata Božnar, Primož Mlakar, Boštjan Grašiþ and Gianni Tinarelli P4.3 Development of a Quasi-Real-Time Forecasting System over Tokyo............................................................................................................. 697 Masayuki Takigawa, Masanori Niwano, Hajime Akimoto and Masaaki Takahashi P4.4 A Construction and Evaluation of Eulerian Dynamic Core for the Air Quality and Emergency Modelling System SILAM .................... 699 Mikhail Sofiev, Michael Galperin and Eugene Genikhovich P4.5 BOLCHEM Air Quality Model: Performance Evaluation over Italy........................................................................................................ 702 Alberto Maurizi, Mihaela Mircea, Massimo D’Isidoro, Lina Vitali, Fabio Monforti, Gabriele Zanini and Francesco Tampieri P4.6 Evaluation of an Operational Ensemble Prediction System for Ozone Concentrations over Belgium Using the CTM Chimere............... 705 Andy Delcloo and Olivier Brasseur P4.7 The Use of MM5-CMAQ-EMIMO Modelling System (OPANA V4) for Air Quality Impact Assessment: Applications for Combined Cycle Power Plants and Refineries (Spain) ............................ 707 R. San José, J.L. Pérez, J.L. Morant and R.M. González P4.8 Verification of Ship Plumes Modelling and Their Impacts on Air Quality and Climate Change in QUANTIFY EC 6FP Project ........... 709 Tomas Halenka, Peter Huszar and Michal Belda P5. Aerosols in the atmosphere........................................................................... 711 P5.1 Quantifying Source Contribution to Ambient Particulate Matter in Austria with Chemical Mass Balance Receptor Modeling ........................ 711 A. Caseiro, H. Bauer, I. Marr, C. Pio, H. Puxbaum and V. Simeonov P6. Interactions between air quality and climate change ................................ 713 P6.1 On the Effective Indices for Emissions from Road Transport............... 713 Kostadin Ganev, Dimiter Syrakov and Zahari Zlatev P7. Air quality and human health ...................................................................... 715 P7.1 A Multi-Objective Problem to Select Optimal PM10 Control Policies .......................................................................................................... 715 Claudio Carnevale, Enrico Pisoni and Marialuisa Volta

Contents

xliii

P7.2 What Activity-Based Analysis and Personal Sampling Can Do for Assessments of Exposure to Air Pollutants?............................................ 717 Doina Olaru and Jennifer Powell P7.3 Intake Fraction for Benzene Traffic Emissions in Helsinki................... 719 Joana Soares, Ari Karppinen, Leena Kangas, Matti Jantunen and Jaakko Kukkonen P7.4 Source Apportionment of Particulate Matter in the U.S. and Associations with In Vitro and In Vivo Lung Inflammatory Markers.......................................................................................................... 721 Rachelle M. Duvall, Gary A. Norris, Janet M. Burke, John K. McGee, M. Ian Gilmour and Robert B. Devlin P7.5 Air Pollution Assessment in an Alpine Valley ...................................... 723 Peter Suppan, Klaus Schäfer, Stefan Emeis, Renate Forkel, Markus Mast, Johannes Vergeiner and Esther Griesser P7.6 New Approaches on Prediction of Maximum Individual Exposure from Airborne Hazardous Releases ............................................... 725 John G. Bartzis, Athanasios Sfetsos and Spyros Andronopoulos P7.7 The Detroit Exposure and Aerosol Research Study .............................. 727 Ron Williams, Alan Vette, Janet Burke, Gary Norris, Karen Wesson, Madeleine Strum, Tyler Fox, Rachelle Duvall and Timothy Watkins Author Index......................................................................................................... 729 Subject Index…………… ……………………………..…………….............…..735

1.6 Air Quality Management Strategies in Large Cities: Effects of Changing the Vehicle Fleet Composition in Barcelona and Madrid Greater Areas (Spain) by Introducing Natural Gas Vehicles María Gonçalves, Pedro Jiménez-Guerrero and José M. Baldasano

Abstract Air quality modelling involves a strategy to manage air pollution in large cities, where air quality problems presently are mainly related to on-road traffic. Nowadays, one of the strategies to reduce emissions is based on the substitution of vehicles by introducing new technologies (e.g. cleaner fuels, hybrid vehicles, fuel cells, etc.). This work assesses the variation on air quality due to the substitution of specific vehicle fleets by natural gas vehicles in the two largest cities of Spain: Barcelona and Madrid. Six different scenarios are studied, focusing on the total or partial modification of public transportation vehicles (buses, taxis), freight vehicles and private vehicles. One scenario involving a combination of all of them is also studied. Under this perspective, the WRF/HERMES/CMAQ modelling system has been implemented and validated with a high resolution (1 km and 1 hour) in the area thanks to the calculation power of the MareNostrum super-computer of the Barcelona Supercomputing Center (94.21 TFlops peak). Daily average concentrations of NO2, SO2 and PM, both PM10 and PM2.5, and 8-hour average concentration for O3 and 1-hour maximum concentrations for these species are estimated both in Barcelona and Madrid Greater Areas. All the scenarios studied involve a reduction in NO2, SO2 and PM concentrations. Most important changes in air quality are registered when the combined scenario is implemented. Ozone concentrations remain approximately in the same levels as in the base case scenario, except for some VOC-limited areas where the reduction of NOx involves a slight O3 increase (under the 10%). A large reduction in PM concentration is observed for both cities when the 50% of commercial light vehicles is transformed. Results of the simulations for the combined scenario indicate that it is particularly effective in reducing PM10 (up to –43% in maximum hourly concentration at some points) and PM2.5 (up to –36%). Keywords Air quality modelling, alternative fuels, emissions, natural gas vehicles, urban pollution management

C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science +Bus iness Media B.V . 2008

54

Air Quality Management Strategies in Large Cities: Effects of Changing the Vehicle Fleet

55

1. Introduction Air pollution is a major environmental problem in urban areas, where 80% dwellers in Europe live. The EU air quality standards (EU-AQS) are currently exceeded in some urban sites, particularly in terms of NO2 and PM10 (EEA, 2006). In spite of the unitary emissions reduction by vehicle achieved during last years, large contributions to atmospheric pollutants emissions still come from on-road transport (Costa and Baldasano, 1996; Oduyemi and Dadvison, 1998; Colvile et al., 2001; Crabbe et al., 1999; Ghose et al., 2004). Moreover some EU-AQS will be reduced in a near future (2010) and some of the emissions abatement strategies put into practice to reduce traffic emissions does not seem to be effective (Carslaw et al., 2007). Therefore public administrations focus their efforts on assessing the best alternatives to reduce urban on-road transport impacts. Different possibilities are being tested; among others the European Commission promotes the use of alternative fuels, specifically the natural gas as a fuel in the medium term (EC, 2001). The evaluation of the effects on air quality caused by the implementation of emission abatement strategies is essential in order to aid policy-makers and to establish the real efficacy of the different environmental plans. This work assesses the changes on air quality achieved with the introduction of natural gas vehicle fleets in two main cities of Spain: Barcelona and Madrid, using the WRF-ARW/HERMES/CMAQ simulation system. The evaluation is carried out in terms of ozone (O3), nitrogen dioxide (NO2), sulphur dioxide (SO2) and particulate matter (PM10 and PM2.5) concentrations. PM2.5 is assessed, in spite of not being still regulated by EU directives, due to its importance related with human health (Pope and Dockery, 2006).

2. Methods Air quality simulations are performed for an episode of photochemical pollution, selected considering air quality data monitored in Catalonia and Madrid, but also considering that the traffic situation must correspond to a normal day, avoiding weekends or holidays, in order to exclude distorting factors in traffic circulation. The chosen episode (June 17–18, 2004) fits in a typical summertime low-pressure gradient with very high levels of photochemical pollutants (especially O3 and PM10) over the Iberian Peninsula. The emissions scenarios were performed under a realistic approach, changing specific vehicle fleets: (1) transformation of the 100% of the urban buses to natural gas buses; (2) transformation of the 50% of taxis to natural gas vehicles (NGV); (3) transformation of 50% of intercity buses to natural gas buses; (4) transformation of a 50% of light commercial vehicles to NGV; (5) transformation of a 10% of private cars to natural gas cars; (6) transformation of 100% of heavy duty freight transport vehicles to NGV; and (7) combination of all the rest.

M. Gonçalves et al.

56

The Eulerian, three-dimensional model WRF/HERMES/CMAQ was applied with high spatial (1 km2) and temporal (1 hour) resolution. The use of fine scale was imposed by the necessity of assessing the subtle air quality variations in urban areas, as shown in the CityDelta project experience (Cuvelier et al., 2007), and in very complex terrains as Barcelona and Madrid Greater Areas (BGA and MGA) (Jiménez et al., 2005). The calculation was performed in a feasible time thanks to the MareNostrum supercomputer hold by the Barcelona Supercomputing Center (94.21 Tflops peak). The traffic module of HERMES emission model includes an intensive road net description, circulation and mobility data and the vehicle fleet composition for Spain for the year 2004, differentiating the specific composition of Barcelona and Madrid cities. The vehicle fleet is divided in 72 vehicle categories as a function of their age, the cubic capacity of their engine, the weight and the type of fuel they use. The EMEP-CORINAIR/EEA methodology was used in order to obtain the speed dependant emissions factors for each of them. The emission factors for NGV were obtained applying the emission reduction factors provided by the European NGV Association (Table 1) to the diesel Euro III emission factors. Table 1 Emission correction factors for the natural gas vehicles categories provided by the European Natural Gas Vehicles Association (ENGVA). NG category

Emission correction factor

Reference category CO

NMHC

NOx

PM

NG cars and LDVs

Euro III diesel cars and LDVs (7.5 t)

0.58

0.11

0.16

0.12

The 8-hour average O3 and 24-hour average NO2, SO2, PM10 and PM2.5 concentrations and 1-hour maximum O3, NO2, SO2, PM10 and PM2.5 concentrations were calculated for each scenario over the Barcelona and Madrid Greater Areas.

3. Results The emissions variation analysis indicates that the ozone precursors and the primary pollutant emissions decrease in all scenarios, especially in the combined case (scenario 7), when the changes in the vehicle fleet are more pronounced (up to 26.1% of vehicles changed in BGA and up to 23.1% in MGA). The main reductions affect PM and NOx emissions. The simulation results for the base case were validated with air quality data from several stations in the city areas and accomplished the EU directives (1999/30/CE, 2002/3/CE) and US-EPA guidelines related to O3 concentrations developed during the years 1991 and 2005. In the BGA, the station of Barcelona-Example shows a


57

negative MNBE for the primary pollutants (–7.78% for NO2; –23.25% for SO2 and –14.21% for PM10) due to the important street-scale influence of on-road traffic. The O3 is closely overestimated by the model (MNBE 8.94% for tropospheric O3). At the same time, the 1-hour peaks are slightly underestimated (UPA –12.82% for O3; –2.84% for NO2; –18.18% for SO2 and –8.47% for PM10). The model also tends to underestimate the concentration of the different pollutants for the stations of the MGA, as Getafe (MNBE –3.96% for O3; –14.58% for NO2; –20.02% for SO2 and –10.33% for PM10). The results of the evaluation confirm the need for working with fine grids in areas where the influence of on-road traffic is important; it becomes essential for addressing air quality processes in urban and industrial areas. With respect to the results of the base case scenario simulations, Figure 1 indicates that main problems in both Greater Areas are related to NO2 and PM10 concentrations; moreover their maximum concentrations are located over the road net which reflects the influence of traffic emissions on urban air quality. The difference between the 8-hour O3 and 24-hour NO2, SO2, PM10 and PM2.5 average concentrations and 1-hour maximum concentration for these pollutants, in the base case scenario and the different scenarios tested, over BGA and MGA, were estimated. All emissions scenarios exhibit a slight increase in O3 (the maximum hourly concentration rises up to 7.8% at some points of BGA), due to the NOx reduction in VOC-controlled areas; while the NO2, SO2 and PM concentrations decrease. The combined scenario involves the largest reductions in all cases (Figure 2). For instance the NO2 24-hour average concentration decreases up to –23.2%, and the maximum hourly concentration of SO2 up to –20.7%, of PM10 up to – 42.8% and of PM2.5 up to –35.6% at some points of the study areas. PM reductions are especially remarkable, indicating the on-road traffic origin of the largest part of the particulate matter concentrations that are registered in the urban areas. These reductions are larger over MGA, where the traffic has a larger weight as pollutant emissions source than in BGA: on average PM2.5 decreases – 2.2% over the MGA and a –0.2% over the BGA, while the PM10 reductions achieve the –14.9% in MGA and –6.6% in BGA Concerning SO2 reductions in MGA are also larger than in BGA, up to –8.7% average reductions in the combined scenario versus –1.5%. Again, the emission sources composition is an essential factor that affects the emissions abatement strategies effects. On one hand, MGA is characterized by a commercial and tertiary activity, meanwhile BGA presents a more important industrial component, which has a direct effect on the weight of traffic contribution to SO2 emissions and indirectly involves that a traffic-emissions abatement strategy could be more effective in reducing SO2 concentrations in Madrid than in Barcelona. Nevertheless, currently SO2 levels do not represent a problem in urban air quality.

58


Fig. 1 8-hour O3 average concentration and 24-hour NO2, SO2 and PM10 average concentration in the base case scenario over the BGA (up) and the MGA (down)


59

Fig. 2 Difference of 8-hour O3 average concentration and 24-hour NO2, SO2 and PM10 average concentration between the combined and the base case scenario over the BGA (up) and the M GA (down)

60


Concerning individual scenarios tested, introducing a 50% of natural gas commercial light vehicles is the most effective in reducing NO2 concentrations in BGA (up to –2.5% on average); nevertheless in MGA better results are achieved when changing the 10% of the private cars fleet (up to –8.2% on average). These differences reflect the necessity of assessing the best strategies to be applied locally; taking into account the specific characteristics of the study area, in this case the different vehicle fleet composition between both cities (larger percentage of cars in Madrid than in Barcelona) entails this behaviour. In order to reduce SO2 and PM concentrations, the most effective measure involves the substitution of the oldest diesel and petrol light commercial vehicles (Figure 2), which are important contributors to these species emissions, by natural gas light duty vehicles (up to 7.5 t). In this case the PM10 hourly maximum concentration is reduced up to –20.8% at some points of BGA and up to 14.6% at some points of MGA, and the maximum hourly concentration of PM2.5 up to –11.3% at BGA and up to –12.2% at MGA.

4. Summary and Conclusion The NO2, SO2 and PM concentrations decrease in all scenarios, both for BGA and MGA domains. The largest reductions are achieved in the combined scenario (– 23% in NO2, –21% in SO2, –43% in PM10 and –36% in PM2.5 at some points of the study areas). The individual scenarios that prove to be more effective in reducing NO2, SO2 and PM concentrations are the transformation of 50% of commercial light vehicles and the 10% of private cars. Changing the 100% of the urban buses, the 50% of the intercity buses and the 100% of the heavy duty freight transport vehicles do not involve a vehicle fleet change larger than 1.5%, so that emissions (less than 5%) and air quality variation is not remarkable (less than 3%). Large cities are typically VOC-controlled areas; therefore the reduction in NOx concentration causes the increase of O3 maximum hourly and 8-hour average concentrations (less than 10% in all study cases). The most effective individual scenarios in reducing NO2 concentration are the 50% of commercial light vehicles change in Barcelona (up to –18%) and the 10% of private cars change in Madrid (up to –10%), because the vehicle fleet of Madrid is mainly composed of diesel and petrol private cars and taxis (82% versus a 66% in Barcelona). PM and SO2 concentrations reduction is especially noticeable when changing the commercial light vehicles into NGV. The specific local characteristics of the affected areas must be taken into account when assessing the effects of environmental management strategies. This work confirms the importance of having detailed emissions inventories and designing the environmental strategies under a realistic approach to obtain representative results. In the present case the traffic emissions reduction strategies are more effective in improving air quality in MGA than in BGA. The simulation system WRFARW/HERMES/CMAQ can be applied to assess the efficacy of abatement emissions


61

strategies in terms of air quality variation and due to its high resolution (1h-1km2) detects subtle changes in urban areas. Acknowledgments The authors gratefully acknowledge E. López for the implementation of HERMES. Also the authors thank to Natural Gas Corporation and the European Natural Gas Vehicles Association (ENGVA) for the collaboration in the assessment of the emission factors for NGV. This work was funded by the projects CICYT CGL2006-0803 and CICYT CGL2006-11879 of the Spanish Ministry of Education and Science and CALIOPE project 441/2006/3-12.1 of the Spanish Ministry of the Environment.

References Carslaw DC, Beevers SD, Bell MC (2007) Risks of exceeding the hourly EU limit value for nitrogen dioxide resulting from increased road transport emissions of primary nitrogen dioxide, Atmospheric Environment 41, 2073–2082. Colvile RN, Hutchinson EJ, Mindell JS, Warren RF (2001) The transport sector as a source of air pollution, Atmospheric Environment 35, 1537–1565. Costa M, Baldasano JM (1996) Development of a source emission model for atmospheric pollutants in the Barcelona area, Atmospheric Environment 30A, 2, 309–318. Crabbe H, Beaumont R, Norton D (1999) Local air quality management: a practical approach to air quality assessment and emissions audit, The Science of the Total Environment 235, 383–385. Cuvelier C et al. (2007) CityDelta: A model intercomparison study to explore the impact of emission reductions in European cities in 2010, Atmospheric Environment 41, 189–207. EC (2001) Communication of the European Commission of 07/11/2001 on an action plan and two proposals for directives to foster the use of Alternative Fuels for Transport, starting with the regulatory & fiscal promotion of biofuels. Brussels 7.11.2001 COM (2001) 547, 47 pp. EEA (2006) Transport and environment: facing a dilemma. TERM2005: indicators tracking transport and environment in the European Union. EEA Report nº 3/2006, 56 pp. Ghose MK, Paul R, Banerjee SK (2004) Assessment of the impacts of vehicular emissions on urban air quality and its management in Indian context: the case of Kolkata (Calcutta), Environmental Science and Policy 7, 345–351. Jiménez P, Jorba O, Parra R, Baldasano JM (2005) Influence of high-model grid resolution on photochemical modelling in very complex terrains, International Journal of Environment and Pollution 24, 180–200. Oduyemi K, Dadvison B (1998) The impacts of road traffic management on urban air quality, The Science of the Total Environment 218, 59–66. Pope CA, Dockery DW (2006) Health effects of fine particulate air pollution: lines that connect, Journal of Air and Waste Management Association 56, 709–742.

62


Discussion S.T. Rao: Have you also looked at the concentrations of toxic air contaminants (e.g., benzene, acetaldehyde, formaldehyde), not just criteria pollutants (e.g., ozone, CO, NO2) as changed the mobile source emissions from biofuels, electric, etc.? J. Baldasano: We have focused on O 3, NOx and PM, because they represent the most important problems related to air quality presently in the cities of Madrid and Barcelona (Spain).

1.4 CFD and Mesoscale Air Quality Modelling Integration: Web Application for Las Palmas (Canary Islands, Spain) R. San José, J.L. Pérez, J.L. Morant and R.M. González

Abstract The integration of sophisticated mesoscale air quality modelling systems, such as MM5-CMAQ and new generation of Computational Fluid Dynamics (CFD) modelling tools has been developed in this contribution. We have used an advanced and adapted version of the MIMO model (U. Karlsruhe, Germany) which is a sophisticated CFD model, to simulate the air concentrations at urban level with 10 m spatial resolution over the city of Las Palmas (Canary Islands, Spain). The CFD code receives the traffic emission data every second produced by a cellular automata model (CAMO). The integrated CFD model is called MICROSYS. This model is an Eulerian 3D tool which is running in diagnostic mode once every minute. The boundary conditions are obtained from the well-known MM5-CMAQ running over the city in prognostic mode. The MM5-CMAQ (OPANA V4) model is run with 1 km spatial resolution covering a domain of 16 × 16 km over the city. This system is operating in forecasting mode since 2004 and is operated over the Internet. The forecasting information for meteorology and air quality concentrations for the following 72 hours is used by MICROSYS to simulate the expected air concentrations at street level for the next three days. The system operates under daily basis and produces the detailed forecasting information at 6:00 GMT everyday. The Internet service includes a sophisticated VRML (Virtual Reality Modelling Language) tool to visualize in a 3D mode the air concentrations at street level by an Internet client. The VRML tool runs on the client server. We present also some comparative results related to the use of shared 64 bits memory machines and single 32 bits one-processor machines for CFD runs. Keywords Air pollution, CFD, modelling, numerical simulations

1. Introduction The use of urban air quality models based on Computational Fluid Dynamics tools (CFDs) requires receiving frequent metrological and air quality information from the numerical boundaries. The CFDs models are in fact embedded into the mesoscale air quality applications from which they receive the proper boundary C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science +Bus iness Media B.V . 2008

37

38

R. San José et al.

conditions. The advances on the capability of Computational Fluid Dynamics models and Air Quality Modelling Systems during the last decade have been quite substantial. The increase on computer capabilities and on the knowledge of turbulence parameterization and numerical schemes has also been very important during the last ten years. On the other hand, there is a considerable public interest on information related to the “real” pollution they are exposure on when they are walking in the street going to work or even during the period they are driving a car from/to work or other daily activities. At street level the differences in the concentration values at both sides of a street can be important, particularly, for instance, on relation to photochemical production during summer time in Mediterranean regions. In this contribution we have used the CFD model MIMO (U. of Karlsruhe (Germany)) and the mesoscale air quality modelling system MM5-CMAQ-EMIMO (NCEP/EPA /Technical University of Las Palmas de Gran Canaria (Canary Islands, Spain)) to simulate the impact of different emission reduction scenarios in the downtown area of Las Palmas de Gran Canaria (Canary Islands, Spain) City. These complex systems could evaluate the impact of several urban strategic emission reduction measures such as reduction of private traffic, increase of public transportation, impact on introduction of new fuel cell vehicles, etc. Also, they could be used for analysis of pollution concentrations at different heights (buildings) and on different areas of urban neighbourhoods. Air dispersion in urban areas is affected by atmospheric flow changes produced by building-street geometry and aerodynamic effects. The traffic flow, emissions and meteorology are playing also an important role. Microscale air pollution simulations are a complex task since the time scales are compared to the spatial scales (micro) for such a type of simulations. Boundary and initial conditions for such simulations are also critical and essential quantities to influence fundamentally the air dispersion results. Microscale Computational Fluid Dynamical Models (CFDM) are playing an increasing role on air quality impact studies for local applications such as new road and building constructions, emergency toxic dispersion gases at urban and local scale, etc. Microscale air dispersion simulations are applied to predict air-flow and pollution dispersion in urban areas. Pullen et al. (2005). Different combinations and applications appear in the literature such as Pospisil et al. (2004) by integrating a Lagrangian model and a traffic dynamical model into a commercial CFD code, Star-CD to simulate the traffic-induced flow field and turbulence. In this contribution we have applied the microscale dispersion model MIMO (Ehrhard et al., 2000) to create an operational air quality forecasting system based on then web in Las Palmas de Gran Canaria (Canary Islands, Spain). The MIMO CFD code has been adapted and incorporated into a mesoscale air quality modelling system (MM5-CMAQ-EMIMO) to fit into the one-way nesting structure. MM5 is a meteorological mesoscale model developed by Pennsylvania State University (USA) and NCAR (National Centre for Atmospheric Research, USA) (Grell et al., 1994). The CMAQ model is the Community Multiscale Air Quality Modelling System developed by EPA (USA) (Byun et al., 1998) and EMIMO is the Emission Model developed by San José R. et al. (2003). MM5 is a well recognized non-hydrostatic mesoscale meteorological models which uses global

CFD and Mesoscale Air Quality Modelling Integration: Web Application for Las Palmas

39

meteorological data produced by global models such as GFS model (NCEP, USA) to produce high resolution detailed three dimensional fields of wind, temperature and humidity which are used in our case as input for the photochemical dispersion model CMAQ (San José et al., 1997). In addition of MM5 output data, EMIMO model produces for the specific required spatial resolution, hourly emission data for different inorganic pollutants such as particulate matter, sulphur dioxide, nitrogen oxides, carbon monoxide and total volatile organic compounds VOCs. The VOCs are splitted according to SMOKE (Sparse Matrix Operator Kernel Emissions) Williams et al. (2001) and Coats (1995). The CFD and mesoscale models solve the Navier-Stokes equations by using different numerical techniques to obtain fluxes and concentrations at different scales. Mesoscale air quality models cover a wide range of spatial scales from several thousands of kilometres to 1 km or so. In this contribution we have applied the MM5-CMAQ-EMIMO model over Las Palmas de Gran Canaria (Canary Islands, Spain) domain to obtain detailed and accurate results of the pollutant concentrations at this spatial resolution in forecasting mode and the MIMO CFD model over a 1 × 1 km domain with several spatial resolutions ( 2–10 m ) and different vertical resolutions. MM5-CMAQ-EMIMO data serves as initial and boundary conditions for MIMO modelling run. In Figure 1 we observe the spatial architecture for the application of the MM5CMAQ-EMIMO mesoscale air quality modelling system. In Figure 2 we show a detailed diagram of the EMIMO modelling system. EMIMO is currently operating with the so called Version 2 which includes the CLCL2000 with 44 different landuse types with 100 m spatial resolution. EMIMO 2.0 also uses the CIESIN 30’’ (CIESIN, 2004), population database and the Digital Chart of the World 1 km land use database to produce adequate emission data per 1 km grid cell per hour and per pollutant. In this case EMIMO has used the detailed GIS database provide by the City of Las Palmas de Gran Canaria (Canary Islands, Spain). In order to apply the EMIMO CFD model, we need detailed information related to the building structure in the 1 km grid cell. This information is shown in Figure 3 for the total of the Las Palmas de Gran Canaria (Canary Islands, Spain) Community (Spain). The height of the buildings is not included in this file and it has been estimated directly for this experiment. A cellular automata traffic model (CAMO) has been developed. CAMO – which has been included into the EMIMO modelling system – is based on transitional functions defined in a discrete interval t as follows:

s (t 1) p ( s (t ), a (t )) u (t ) v( s (t ))

(1)

where s(t + 1), s(t) is a defined sate, a(t) is an input symbol and u(t) is an output symbol. We have used the Moore neighbourhood with eight different surrounding cells where each cell – representative of a vehicle – can move on. The whole system focusing on the 1 × 1 km urban area in Las Palmas de Gran Canaria (Canary Islands, Spain) downtown is called MICROSYS system. We have selected a

R. San José et al.

40

subdomain of 300 × 300 m with 5 m spatial resolution and 15 vertical layers for this particular experiment (Figure 4). The first 10 layers are equally spaced with 5 m spatial resolution up to 50 m in height and the last five layers are located at 55, 61.55, 68.20, 75.52 and 83.58 m in height.

Fig. 1 MM5-CMAQ-EMIMO architecture for this application

Fig. 2 EMIMO model basic architecture


41

Fig. 3 E00 vector file for Las Palmas de Gran Canaria (Canary Islands, Spain) Community

Fig. 4 Subdomain view for this experiment in Las Palmas de Gran Canaria (Canary Islands, Spain) City downtown area

42

R. San José et al.

2. Results The system is operating under daily basis as follows: (a) the MM5-CMAQEMIMO mesoscale air quality forecasting system is operating under daily basis, starting at 01:00 GMT and running to complete six simulation days; (b) the MICROSYS CFD system is operating 1 minute every hour during the last 72 hours of simulation (future time). This 72 steady state simulations produce pollution concentrations in a 3D domain. Since the system is mounted for seven 1 × 1 km domains, covering the entire city, we have to multiply these 72 simulations times seven which makes 504 steady state simulations. Figure 5 shows a picture of the web site when showing a 1 × 1 km area in the n north of the city of Las Palmas de Gran Canaria (Canary Islands, Spain). On the right side, we observe a full user control panel to select several functions and capabilities to visualize the results. In Figure 6 we observe a detailed view for NO2 over an area of the District Ciudad del Mar (Las Palmas, Canary Islands, Spain) and a time series in a cross of streets in such a district. Finally in Figure 7, we see an example.

Fig. 5 OPANA model including the MICROSYS CFD for Las Palmas (Canary Islands, Spain) as seen in the web site for a 1 × 1 km area in the north of the Municipality


43

Fig. 6 Detailed view of the District: Ciudad del Mar in Las Palmas (Canary Islands, Spain) for NO2 concentrations on June 14, 2007 at 08:00 GMT. Also we see the time series in a cross of streets in such a district

Fig. 7 VRML view on Internet for an area in the North of the city observing the NO2 concentrations in a 3D context. The user can navigate through the streets with different views

44

R. San José et al.

3. Conclusions The MM5-CMAQ-EMIMO modelling system has been used to provide detailed initial and boundary conditions to a system called MICROSYS which is composed by the MIMO CFD microscale dispersion model and CAMO which is a cellular automata traffic model. The results show that the air quality modelling system offers realistic results although no comparison with eddy-correlation measurement system has been performed in the area. The tool can be used for many air quality impact studies but in particular for traffic emission reduction strategies. Acknowledgments We would like to thank Professor N. Moussiopoulos (AUTH, Greece) for providing the MIMO model. Also to EPA/PSU/NCAR for providing the MM5-CMAQ modeling system code.

References Byun DW, Young J, Gipson G, Godowitch J, Binkowsky F, Roselle S, Benjey B, Pleim J, Ching JKS, Novak J, Coats C, Odman T, Hanna A, Alapaty K, Mathur R, McHenry J, Shankar U, Fine S, Xiu A, Lang C (1998) Description of the Models-3 Community Multiscale Air Quality (CMAQ) model. Proceedings of the American Meteorological Society 78th Annual Meeting Phoenix, AZ, January 11–16, 1998, pp. 264–268. CIESIN (2004) Center for International Earth Science Information Network (CIESIN). Global Rural-Urban Mapping Project (GRUMP): Urban/Rural population grids. CIESIN, Columbia University, Palisades, NY. http://sedac.ciesin.columbia.edu/gpw/ Coats CJ, Jr (1995) High Performance Algorithms in the Sparse Matrix Operator Kernel Emissions (SMOKE) Modelling System, Microelectronics Center of North Carolina, Environmental Systems Division, Research Triangle Park, NC, 6 pp. Ehrhard J, Khatib IA, Winkler C, Kunz R, Moussiopoulos N, Ernst G (2000) The microscale modelo MIMO: Development and assessment. Journal of Wind Engineering and Industrial Aerodynamics, 85, 163–176. Grell G, Dudhia J, Stauffer D (1994) A Description of the Fifty Generation Penn State/NCAR Mesoscale Model (MM5). NCAR Tech. Note, TN-398 + STR, 117 pp. Pospisil J, Katolicky J, Jicha M (2004) A comparison of measurements and CFD model predictions for pollutant dispersion in cities. Science of the Total Environment. 334–335; 185–195. Pullen J, Boris JP, Young T, Patnaik G, Iselin J (2005) A comparison of contaminant plume statisticsfrom a Gaussian puff and urban CFD model for two large cities. Atmospheric Environment, 39, 1049–1068.


45

San José R, Prieto JF, Castellanos N, Arranz JM (1997) Sensitivity study of dry deposition fluxes in ANA air quality model over Las Palmas de Gran Canaria (Canary Islands, Spain) mesoscale area, Measurements and Modelling in Environmental Pollution, Ed. San José and Brebbia, pp. 119–130. San José R, Peña JI, Pérez JL, González RM (2003) EMIMO: an emission model, 292–298, Springer-Verlag. ISBN: 3-540-00840-3. Williams A, Caughey M, Huang H-C, Liang X-Z, Kunkel K, Tao Z, Larson S, Wuebbles D (2001) Comparison of emissions processing by EM-S95 and SMOKE over the Midwestern U.S. Preprint of International Emission Inventory Conference: One Atmosphere, One Inventory, Many Challenges. Denver, CO, May 1–3, pp. 1–13.

Discussion A. Baklanov: What was the resolution of the finest grid of MM5 used for boundary conditions for your CFD model runs? For correct downscaling with CFD it should be a city scale. However you use the MM5 version without any urban parameterization. So, your boundary conditions for CFD obstacle-resolved model within a city are not correct, because they don’t consider urban features. R. San José: You are right that we do not use the urbanized version of MM5 (although we have been working with it). The reason for that is the important increase on computer time which is a very sensitive issue for our operational forecasting system. However, the CFD model is running with 5 m resolution for the CFD domain (1 × 1 km) (we have 7 CFD 1 × 1 km domains for covering the whole city). We believe that this is a good approach although the computer limitations do not allow us to run MM5 urban instead of classic MM5. M. Mircea: What emission inventory is used from EU to urban scale? How is done the nesting? R. San José: We use EMEP 2000 50 km spatial resolution. We go down to 1 km by using EMIMO model approach which uses population and km of roads per squared km (mixed top-down and bottom-up approaches). The nesting for this application is done as one-way nesting. Using CAMx, we have used the two-way nesting approach.

1.9 Development and Application of MicroRMS Modelling System to Simulate the Flow, Turbulence and Dispersion in the Presence of Buildings S. Trini Castelli, T.G. Reisin and G. Tinarelli

Abstract A modelling system for the simulation of the flow and dispersion from the mesoscale down to the urban microscale is under development. This modelling system is a microscale version of the regional off-line system RMS (RAMSMIRS-SPRAY) – MicroRMS. A modified version of RAMS6.0 is used, in which a Cartesian grid and the ADaptive Aperture method are implemented for defining the presence of buildings in arbitrarily steep topography and where alternative versions of the k-İ turbulence closure model were incorporated. After RAMS, the Lagrangian stochastic dispersion model MicroSPRAY is applied, specially devoted to simulate accidental gaseous releases at microscales, including the presence of obstacles and buildings. At present, the efforts are focused on the development of a micro-version of the interface code MIRS, calculating the surface and boundary layers’ parameters and the Lagrangian variables. The first step in the project was to harmonize the treatment of buildings between RAM6.0 and MicroSPRAY approaches. Here we present the first tests of MicroRMS prototype, applied to the MUST exercise of Cost732 Action, a flow and dispersion field test carried out in the Great Basin Desert (USA) in 2001, where 120 standard containers were set up in a regular array of obstacles. Keywords Buildings, COST732 Action, flow and dispersion modelling, microscale, MUST experiment

1. Introduction Modelling atmospheric flows and pollutant dispersion in urban areas is a problem of peculiar characteristics, due to the complexity and heterogeneity of the urban site configuration. In particular, small-scale fluid dynamics superposes to the atmospheric larger scale flow and turbulence and the dispersive processes strongly depend on the specific structure of the urban fabric. Advanced computational fluid dynamics (CFD) models are generally applied to simulate the flow structure and the C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science +Bus iness Media B.V . 2008

81

82

S. Trini Castelli et al.

pollutant diffusion around buildings, obstacles or in urban canyons. In recent years, we have been proposing an alternative approach, where the flow was simulated in different cases by using an atmospheric model – RAMS6.0 (Reisin et al., 2007). In this approach, a complete description of the atmospheric processes is ensured, since boundary layer, radiation and moist processes, together with the interaction between the surface and the soil, are included. Moreover, for further developments, it is possible to take advantage of the several capabilities offered by atmospheric models, like data assimilation and nudging. We use the latest version of RAMS6.0, where a Cartesian grid is implemented and the so called ADaptive Aperture method (Walko and Tremback, 2002) is used for defining the presence of buildings and dealing with arbitrarily steep topography, enabling simulation at very high resolution, in the order of metres. To harmonize with the CFD approach and accounting for CFD results, we implemented in RAMS6.0 not only a standard version of the k-İ turbulence closure model (Trini Castelli et al., 2001, 2005) but also its renormalization group (RNG k-İ) version (Reisin et al., 2007). Test-simulations of the flow and turbulence were performed using both closure schemes and some results are discussed here. To simulate the dispersion process, we are developing a modelling system which will be the microscale version of the regional off-line system RMS (RAMS-MIRS-SPRAY, see for instance Trini Castelli et al., 2003), hereafter called MicroRMS, interfacing RAMS6.0 with MicroSPRAY Lagrangian particle dispersion model (Tinarelli et al., 2007). The advantage of the off-line coupling lays in the independency of the meteorological and dispersion models. This enables the flexibility of considering different dispersion scenarios with the same meteorology. At present, our efforts are focused on the development of a micro-version of the interface code MIRS (Trini Castelli and Anfossi, 1997), calculating the surface and boundary layer’s parameters and the Lagrangian variables (variances of the velocity fluctuation, local velocity decorrelation time scales, etc.) and harmonizing the treatment of buildings between RAM6.0 and MicroSPRAY. Having such a preprocessor offers a number of options in calculating the boundary-layer and turbulence parameters (Physick and Trini Castelli, 2005). MicroSPRAY is especially devoted to simulate accidental gaseous releases at microscale. It considers different emission geometries and conditions, different micrometeorological situations and the possible presence of obstacles, such as buildings. New modules treating the physics of non-neutral gases are also under implementation in the model. MicroRMS is conceived so to provide the simulation of the flow and dispersion encompassing all relevant scales, synoptic, mesoscale and down to the urban microscale. In this work, we present the preliminary test of MicroRMS prototype, applied in the frame of the MUST exercise of COST732 Action (http://www.mi.unihamburg.de/cost732).

2. MUST Case in COST732 Action The following description of the MUST case is an extract from a document by courtesy of Dr. Bernd Leitl, who prepared it for the COST732 community. The

Development and Application of MicroRMS Modelling System to Simulate the Flow

83

Mock Urban Setting Test – MUST data set provides flow and dispersion data measured within an idealized urban roughness. The experimental setup is based on an extensive field test carried out on a test site of the US Army in the Great Basin Desert in 2001 (Yee and Biltoft, 2004). A total of 120 standard size shipping containers were set up in a nearly regular array of 10 by 12 obstacles, covering an area of around 200 m by 200 m. The containers were 12.2 m long, 2.54 m high and 2.42 m wide and formed an idealized roughness. The exact location and orientation of each of the individual obstacles were documented with sufficient accuracy. At the centre of the container array, a so-called VIP van was placed, serving as collection point for sampled wind and concentration data. The size of the VIP van differed significantly from the size of the surrounding containers. The terrain of the field site is characterized as ‘flat open terrain’, an ideal horizontally homogenous roughness formed by bushes and grass land with a height of approximately 0.5–1 m. Other orographical structures, like dunes, were assumed to have no significant effect on the approach flow conditions at the test site. The nearest significant mountains are located 12 and 24 km far from the experimental field. The terrain slope is documented to be 0.5 m per km, rising to the south. Wind tunnel (WT) measurements within a scaled model of the MUST configuration were carried out for instance by Bezpalcova and Harms (2005). The laboratory data represent the reference dataset used in the COST732 exercise.

3. Configuration of the Modelling System and Simulations Since we are primarily interested in testing the applicability of our modified version of MicroRMS in real terrain, in this work we kept the geometry and the initial conditions as for the field test case. The preliminary runs were performed for the 0° and –45° (that in the real field was –41°) inflow cases, referring to the available WT dataset sketched in Figure 1, since at present the real field MUST data are under processing and not yet available.

Fig. 1 Sketch of the geometry and of the different flow direction in the MUST WT experiment (Courtesy of Dr. B. Leitl)

84


The measured field mean wind profile for the inflow is available at three levels up to 16 m and the data are processed on the basis of measurements at the south tower, during the days of September 25 and 26, 2001. The observed values are given as normalized over the wind at a reference height of 8 m. Since RAMS needs an input profile higher than the top of the domain, we extrapolated logarithmically a profile from the observed data available. In Figure 2 we plot the profile used as input in RAMS for both 0° and –45° cases, the field observed data and the inflow data used in the wind tunnel experiment, to highlight the differences. The wind tunnel values are normalized with respect to the wind at a reference height of 7.29 m. The same initial speed profile was used for the 0° and for the –45° simulations. In the comparison with the WT data the U-component of the measured wind velocities is always aligned with the mean approach flow wind direction. Differently from the approach in CFD simulations (see, for instance, Milliez and Carissimo, 2007), no initial turbulent kinetic energy (TKE) profile was input and RAMS develops its own TKE field starting from a minimum initial threshold. RAMS simulation domain extends 420 m in the longitudinal dimension with a grid size of 0.6 m and 320 m in the latitudinal direction with a grid size of 1 m. In the vertical there are 35 levels with a resolution of 0.2 m up to a height of 3.3 m, then stretched up to a total height of 25 m. The 120 containers’ locations and sizes were set according to the data provided for the MUST case.

Fig. 2 Input wind profile in RAMS6.0 simulation (solid line), against observed field MUST data (diamonds) and approach flow profile in wind tunnel experiment (solid + dot line)

We recall that RAMS is not built to produce steady-state conditions. However, we set the initial boundary condition so to approach as close as possible a steady state solution and we verified that a quasi-steady flow was reached after 4 minutes of simulation. The time step during the whole simulation was 4 •10-3 s. In these simulations we run RAMS without using the modified boundary conditions at the buildings that we implemented, for which in a so called ‘influence region’ around the buildings a logarithmic interpolation between the values at the building face and the values of the variable in the ambient atmosphere near the buildings is imposed (Reisin et al., 2007).


85

4. Results and Discussion The main goal of these preliminary runs with MicroRMS prototype is to verify the possibility of using this modelling system in such high resolutions and in a configuration typical to CFD studies. The present results need thus to be considered as indicative and qualitative, while a more quantitative analysis will be conducted running final simulations against the real field data. In Figure 3 the streamlines of the wind field produced by RAMS in the –45° case are plotted. The structure of the simulated flow appears to be plausible and its consistency can be verified looking at the variables’ profiles in some points of the domain. Fig. 3 Detail of the streamlines of the horizontal mean flow field at 1m level by RAMS6.0

In Figure 4 a comparison among two RAMS simulations, using k-İ or RNG-k-İ closure models, and the WT measured data in a point in the central part of the building arrays, located between two buildings, is presented. The simulated pattern of the wind profile well captures the values observed in the wind tunnel for both closures, while the TKE shows a maximum at a higher position than the observed one. The k-İ closure produces higher values for the TKE, while the RNG-k-İ tends to underestimate the observed range. The difference between the two closures, k-İ producing higher TKE values than RNG-k-İ, is consistent with many results from literature. In this specific case, with such highly complex configuration of buildings, apparently k-İ closure is better reproducing the turbulent field.

86


Fig. 4 Case 0°. Wind profile in RAMS6.0 simulation at (x,y) = (–45.201 m, 3.577 m) (solid lines) using k-İ and RNG k-İ closures, against measured profiles in wind tunnel experiment at (x,y) = (–44.925 m, 4.05 m) (dots)

Fig. 5 Case –45°. Wind profile in RAMS6.0 simulation at (x,y) = (–26.100 m, –5.50 m) (solid lines) using RNG k-İ closure, against measured profiles in wind tunnel experiment at (x,y) = (–25.875 m, –5.25 m) (dots)

In Figure 5 an analogous comparison is proposed for the –45° case run with the RNG-k-İ model, in a different point than for the 0° case, but always located in the central part of the domain. The simulated wind profile shows higher values than the observed WT data, in particular in the middle heights. The difference may be related to the initial wind profile input in the model, which seems to be better representative for the 0° case than for the –45° case. The TKE comparison confirms the results found for the 0° case. The meteorological fields produced by RAMS were then processed through the prototype micro-version of MIRS so to make the grid structures and the assimilation of the building data compatible between RAMS and MicroSPRAY. In this preliminary test, since only the TKE was made available from RAMS, we made the assumption of isotropy for turbulence and the standard deviation Vi of the wind


87

velocity fluctuations were set as V u V v V w 2k 3 , where k is the TKE. The Lagrangian time scales were then calculated using also the TKE dissipation İ, as T L 2V 2 C 0 H , where C0 is the Lagrangian structure function constant that here was assigned as C0 =4. For this first run, aimed at testing the functionality of the models’ chain, we referred to the MUST experiment named 2681829 (–41°), during which the atmosphere was characterized by a neutral stratification (Monin-Obukhov length L = 28,000 m), the source was located in (x,y) = (–77.46 m, 67.47 m) at 1.8 m height and the emission was continuous, with a release rate of 225 l/minute, lasting 15 minutes. As an example, in Figure 6 a picture of the 3D particle plume is reported, showing the interaction of the plume with the building array. The direction of the plume and its spread appear to be consistent with the expectations. To present the result on the concentration fields, we are repeating the simulation changing the inflow as for the real field case, which is –41°. Fig. 6 Case –45°. MicroSPRAY simulated plume dispersion, case 2681829

5. Conclusions First results using the microscale version of the regional off-line modelling system RMS (RAMS-MIRS-SPRAY) – MicroRMS, were presented. These encouraging results proved the capabilities of MicroRMS to simulate pollutant dispersion in complex urban configurations as in the MUST experiment. At present a new version of the interface parameterisation code, MIRS4.0, is under development, to harmonize the treating of buildings and obstacles between RAMS and MicroSPRAY models and to include parameterisations for the boundary layer and turbulence suited to the urban scale. Further investigations, related to the effect of the buildings’ boundary conditions on the flow, the turbulence and the dispersion and to the possible deflection of the plume induced by the buildings, are planned within the frame of the MUST real field experiment and of the COST732 Action activity. Acknowledgments The authors like to thank Professor M. Schatzmann, Dr. B. Leitl, Dr. J. Franke and all the COST732 community for the precious cooperation

88


and for making available the MUST WT data and the information about MUST field experiment.

References Bezpalcova K, Harms F (2005) EWTL Data Report/Part I. Summarized Test Description Mock Urban Setting Test. Environmental Wind Tunnel Laboratory, Center for Marine and Atmospheric Research, University of Hamburg Milliez M, Carissimo B (2007) numerical simulations of pollutant dispersion in an idalized urban area, for different meteorological conditions, Bound.-Layer Meteorol. 122, 321–342 Physick W, Trini Castelli S (2005) Lagrangian Particle Models. Link with meteorological models, Section 11.2.5. in ‘Air Quality Modelling. Theories, Computational Methods and Available Databases and Software’, vol II – Advanced Topics, Zannetti P Ed., pp. 116–118, EnviroComp Institute and Air & Waste Management Association, Pittsburgh, USA. Reisin T, Altaratz Stollar O, Trini Castelli S (2007) Numerical simulations of microscale urban flow using the RAMS model. Developments in Environmental Science, Vol. 6, 32–44, Borrego C and Renner E Eds., Elsevier, Amsterdam, NL. Tinarelli G, Brusasca G, Oldrini O, Anfossi D, Trini Castelli S, Moussafir J (2007) Micro Swift-Spray (MSS), a new modelling system for the simulation of dispersion at microscale. Air Pollution Modelling and its Applications XVII, 449–458, Borrego C and Norman A Eds., Springer, New York, USA. Trini Castelli S, Anfossi D (1997) Intercomparison of 3D turbulence parameterisations as input to 3D dispersion Lagrangian particle models in complex terrain. Il Nuovo Cimento, 20C(3), 287–313. Trini Castelli S, Ferrero E, Anfossi D (2001) Turbulence closures in neutral boundary layers over complex terrain. Bound.-Layer Meteorol., 100, 405–419. Trini Castelli S, Anfossi D, Ferrero E (2003) Evaluation of the environmental impact of two different heating scenarios in urban area. Int. J. Environ. Pollut., 20, 207–217. Trini Castelli S, Ferrero E, Anfossi D, Ohba R (2005) Turbulence closure models and their application in RAMS. Environ. Fluid Mech., 5, 169–192. Walko R, Tremback C (2002) The Adaptive Aperture (ADAP) Coordinate. 5th RAMS Workshop and Related Applications, Santorini, Greece. Yee E, Biltoft CA (2004) Concentration fluctuation measurements in a plume dispersing through a regular array of obstacles. Bound.-Layer Meteorol. 111, 363–415.

Discussion A. Venkatram: The model underestimates the wind speed and overestimates the TKE. You concluded that you were “happy” with the results? Why? What is the significance of the 30s time scale in the Lagrangian


89

particle model? The predictions are independent of the averaging time? Trini Castelli: These preliminary simulations were mainly aimed at verifying the possibility of using RAMS model for simulations of complex configuration with buildings, at very high resolution, and to check the micro version of our RMS modelling system. Therefore, we were “happy”, because this preliminary test was successful and results were reliable: we are not aware of other works performed with RAMS at this kind of resolution in such complex condition. We tested different turbulence closures at the 18 available measuring points and while in some cases we had underestimation of velocity and overestimation of TKE, in some other cases the results were good, alternately for the different closures. Surely we need to deeper investigate the reason why, for instance, k-İ closure works well in the upflow part of the domain while RNG-k-İ gives much better results in the downflow part of it. Since this is a steady-state case, the predictions of concentration are actually independent on the averaging time. The 30 s time scale for averaging was used to collect a sufficient number of particles to obtain stable concentration fields. Since the steadiness was reached after about 60 s, after this time interval the time scale does not affect concentration fields anymore, and it is used just to smooth the statistical fluctuations due to discretization in particles. Furthermore, this time scale was used for graphical reasons, to produce the plume-dynamics and concentration field animation since the initial time. S. Hanna: Your dispersion results for the MUST case with the mean wind at 45º angle to the obstacle orientation suggested that the plume direction was the same for the particles at heights below and above the obstacle height. The observation showed a significant shear. It will be interesting to see the results of your comparison of the model simulations to the observations. Trini Castelli: We recall that the height of the emission source in the case we considered was 1.8 m and the height of the buildings is 2.54 m. The 3D animation shown does not allow appreciating the shear of the plume because particles at any vertical level are plotted together, but when ground level concentration contours are plotted a deviation from the initial direction of the plume is clearly registered, indicating the shear of the plume below the obstacles’ height. However, further investigation is needed and a comparison between measured and predicted concentrations is under process.

1.3 Development of a Lagrangian Particle Model for Dense Gas Dispersion in Urban Environment G. Tinarelli, D. Anfossi, S. Trini Castelli, A. Albergel, F. Ganci, G. Belfiore and J. Moussafir

Abstract A new version of the Lagrangian particle model MicroSpray is proposed. It simulates the dense gas dispersion in situations characterized by the presence of buildings, other obstacles, complex terrain, and possible occurrence of low wind speed conditions. Its performances are compared to an atmospheric CFD model output and to a field experiment (Thorney Island). Keywords Dense gas dispersion, lagrangian particle model, tracer experiment 1. Introduction Accidental release and dispersion of hazardous material may cause severe environmental problems. Thus, correctly simulating the distribution of these hazardous substances is important. This is mostly accomplished by empirical models or, in some specific cases, by computational fluid dynamics (CFD) models. Another way, here proposed, is offered by Lagrangian particle dispersion (LPD) models. The LPD approach is a compromise between the complexity and CPU time demanding of CFD models and the simpler integral models. Thus, in this work, a new version of the LPD MicroSpray model, especially oriented to deal with dense gas dispersion in urban environment, is described. Then the comparison between the MicroSpray simulations and those of the CFD Mercure in two academic flat terrain cases is shown. Finally, some preliminary comparisons of MicroSpray predictions against the tracer Thorney Island field experiment No 8 are presented.


28

Development of a Lagrangian Particle Model for Dense Gas Dispersion in Urban

29

2. Brief Outline of MicroSpray and Mercure Models

2.1. MicroSpray MicroSpray is part of the model system MSS that includes MicroSWIFT and MicroSpray. MicroSWIFT is an analytically modified mass consistent interpolator over complex terrain. Given topography, meteorological data and buildings, a mass consistent 3D wind field is generated (Moussafir et al., 2004; Brusasca et al., 2005). It may also prescribe diagnostic turbulence parameters to be considered by MicroSpray inside the flow zones modified by obstacles. MicroSPRAY is a LPD model directly derived from SPRAY code, which is based on a 3D form of the Langevin equation for the random velocity (Tinarelli et al., 1994, 2000), and is able to take into account the presence of obstacles. Micro-Swift takes obstacles or buildings into account by setting as impermeable some of the cells of the terrain following grid where meteorological fields are defined. MicroSpray has been extended to deal with dense gas dispersion. We recall that an emitted cloud of hazardous material initially denser than the ambient air, begins to disperse under the action of its own buoyancy and momentum (horizontal, vertical or oblique in any direction). Then, its excess of density reduces as ambient air is entrained. Finally, at some distance downwind, transition to passive dispersion takes place. An important issue is the spread at the ground due to gravity. These effects are simulated into MicroSpray by implementing new algorithms. Concerning the initial phase, the following five conservation equations are integrated for each particle at each time step, based on Glendening et al. (1984): º d ªUp (1) Mass us b2 » E us « dt ¬ U a ¼ Energy

d us b 2 B dt

>

@

Up 2 N us wp b 2 Ua

(2)

Vertical momentum

d ª Up 2 º « u s wp b » dt ¬ Ua ¼

x horizontal momentum

d dt

ªUp º us b2 u p » « U ¬ a ¼

E us ua

(4)

y horizontal momentum

d dt

ªUp º us b2 v p » « U ¬ a ¼

E u s va

(5)

B b 2 us

(3)

2 2 2 where: u s u p 2 v p 2 w p 2 , B g U a U p U a , Ua u a v a wa , u e >D 1 u s D 2 Ua @ and E 2 b u e ; a , p refer to air and plume, respectively, B is the buoyancy, E represents the entrainment rate, b the plume radius and ue is

G. Tinarelli et al.

30

the entrainment velocity. The first three equations were derived by Hurley and Manins (1995) and inserted into the TAPM model (Hurley, 2005). Regarding the gravity spreading, we remind that when a dense plume reaches the ground a horizontal momentum is generated by its weight, which thus tends to spread the plume. We simulate this last process by an empirical method. The module of the horizontal velocity, U g , is computed as (Eidsvik, 1980): (6) U g D1 B H where D 1 2 and H is the mean height of the column above each particle. The two horizontal velocity components U gs , Vgs , due to the gravity spread are: U gs U g cos J and V gs U g sin J .

2.2. Mercure The Mercure model (Carissimo et al., 1997) is the atmospheric adaptation of the CFD code ESTET developed by Electricité. de France (EDF), commonly used for industrial CFD applications at EDF R&D. MERCURE code has undergone extensive validation (see: Riou, 1987; Duijm and Carissimo, 2001; Moon and Albergel, 1997). Relevant aspects of the code include: 3D flow simulation, influence of terrain and obstacles, multiple fluids and full non-hydrostatic formulation. MERCURE solves the classic Navier-Stokes equations system with adaptations for multiple fluids and for passive scalar tracer variables. A conservation relation for thermodynamic energy (enthalpy or virtual potential temperature) is optionally solved. Solving the thermal energy equation implies that thermal buoyancy (or dense) effects are included in the solution. Turbulence closure is by means of supplementary equations for the conservation of turbulent kinetic energy and dissipation using the k-H model. The conservation equations are discretized using a combination of finite difference and finite-volume methods solving separately each type of operator. As default, the inlet boundaries are Dirichlet for all parameters and the outlet boundaries are zero gradient for all parameters. Important aspects of the MERCURE setup for this study include: x Ideal gas equation of state x Boussinesq approximation is used, implying that density variations only affect the flow through buoyancy (or dense) terms x Gravity is the only retained volume force (Coriolis effects are ignored) x Thermal forcing due to radiative flux divergence is negligible

Development of a Lagrangian Particle Model for Dense Gas Dispersion in U rban

31

3. MicroSpray Validation

3.1. Comparison with MERCURE Simulations with MERCURE have been performed in two different flow conditions. The first one refers to a mean wind velocity at 10 m (1.5 m s-1) and the second one to a higher wind at 10 m (5 m s-1). In both simulations, a neutral turbulence has been considered considering a logarithmic wind profile, horizontally homogeneous. In order to verify that the two models behave similarly in a base case, both MERCURE and MicroSpray have firstly considered a continuous emission located 10 m above the ground without dense gas effects. Then the two models simulated an emission two times denser than air. Figure 1 shows ground level concentrations (glc) obtained in the base case by the two models. Maximum values are almost identical and also the overall pattern of the impact at ground is very similar. V=5.0m/s U Air

max = 1.59E-2 kg/kg

U Air

max = 1.64E-2 kg/kg

MERCURE

Spray

V=1.5m/s U Air

U Air

max = 4.32E-3 kg/kg

max = 4.66E-3 kg/kg

Fig. 1 glc obtained by MERCURE (above) and MicroSpray (below) in lower (left) and higher (right) wind speed in the base case (no dense plume)

Figure 2 shows instead the comparison of the dense emission case. In this condition, an emission from a stack having a diameter of 2.17 m and a vertical exit velocity of 1.14 m s-1 has been considered. The source is located at (0, 0, 10 m) and the plume is followed until about 400 m downwind the source.

G. Tinarelli et al.

32 V=5.0m/s U = 2 * UAir

U = 2 * U Air

max = 0.21 kg/kg

max = 0.225 kg/kg

MERCURE

V=1.5m/s U 2 * U Air

max = 0.005 kg/kg

Spray

U = 2 * U Air

max = 0.0078 kg/kg

Fig. 2 glc obtained by MERCURE (above) and MicroSpray (below) in lower (left) and higher (right) wind speed in the dense gas case

In the lower wind speed case, MERCURE shows an evident splitting of the plume at ground, and a large horizontal spread due to the gravity effects. MicroSpray shows qualitatively a similar result, even if less pronounced. Both splitting and plume spread are present and the maximum glc is correctly captured. In the higher wind speed case, both models show a less pronounced spread effect at ground, due to the higher ventilation causing a more efficient entrainment. Maxima glc are still comparable and impact area seems closer to the source in MERCURE simulation.

3.2. Comparison with Thorney Island Exp. 8 The main information on this experiment, needed for this work, were found in two Data Set Reports: Rediphem (Nielsen and Ott, 1996) and MDA (Hanna et al., 1991). A mixture of Freon-12 and Nitrogen (3,958 kg) was instantaneously emitted from a cylinder (diameter = 14 m, height = 13 m), without any initial momentum. 42 samplers, located at different heights (0.4, 2.4, 4.4 and 6.4 m) in the range 70–500 m downwind the source, collected the emitted tracer for 660 s. The initial tracer concentration was 1 mol/mol. Wind speed was 2.4 m s-1 at 10 m and the wind heading was about 18 degrees to the left of the array centerline. Other important data were: friction velocity u* = 0.126 ms-1 and roughness length z0 = 0.012 m. The stability was Pasquill category D. No turbulence data were given. The relative emission density, Ue/Ua, where Ue and Ua, are the emission and ambient densities, was equal to 1.63. A computation domain of 200 × 800 × 200 m was considered. MicroSwift had horizontal grid spacing of 2 m and a stretched grid in the vertical. No obstacles


33

have been included in the MicroSwift simulation, since no exact information of their presence and dimensions were available. A logarithmic wind profile, horizontally homogeneous, was reconstructed on the basis of the above reported values of u* and z0, that were also used to reconstruct the turbulence fields. Twenty thousand particles were released at t = 0 uniformly distributed within the source cylinder centered at × = 200 m and y = 0 m and then their trajectories were calculated.

Fig. 3 Plant view of the smoke emitted after 1s (left) and 10s (right) obtained during the experiment (upper part) and glc obtained by MicroSpray (lower part)

Finally, concentration at sampler locations was computed. Figure 3 shows a qualitative comparison between simulation and experiment. In the upper part, two photographs show a plant view of the behaviour of the emitted puff 1s and 10s after the emission. The smoke is suddenly moved towards the ground, while the horizontal spread takes place emptying the internal part of the column and giving to the puff a

G. Tinarelli et al.

34

circular shape. This circular puff is then transported along the mean flow direction. MicroSpray gives a similar behaviour. Ground level concentration field patterns show a circular form and the horizontal spread is similar to the experimental one. Also the movement of the puff along the mean direction is correctly captured. Table 1 Statistical indexes – Thorney Island Exp. 8 preliminary simulation. CC

NMSE

FB

MG

VG

FA2

FA5

0.69

2.38

0.65

1.40

1.02

0.55

0.86

Table 1 shows some statistical indexes: correlation coefficient (CC), fractional bias (FB), geometric mean bias (MG) and variance (VG), factor of 2 (FA2) and 5 (FA5). We show both FB, NMSE and MG,VG even if, since observed concentrations vary over three orders of magnitude, the logarithmic indexes are more appropriate (Hanna et al., 1991). With reference to this, it is known that a “perfect” simulation would have MG = VG = 1. Table 1 shows that VG is rather good whereas MG (and, obviously, FB as well) is less satisfactory, indicating an overall underestimation. The general behaviour of the tracer experiment is correctly captured (VG = 1.02 and the values of CC and FA2 are satisfactory too). It is worthwhile to recall that no information on the turbulence characteristics of the experiment and of the wind profile were available (only knowing the value of the upwind mean wind at 10 m).

4. Conclusion In this paper, we have presented a new version of the LPD model MicroSpray devoted to simulate the dense gas dispersion. We have compared its prediction both with the CFD model Mercure and with a tracer experiment (Thorney Island Exp.8). Some preliminary results obtained above presented in flat terrain suggest that MicroSpray is able to perform correct simulations of dense gas dispersion both in academic cases and real field situations.

References Brusasca G, Tinarelli G, Oldrini O, Anfossi D, Trini S Castelli, Moussafir J (2005) Micro-Swift-Spray (MSS) a new modelling system for the simulation of dispersion at microscale. General description and validation. In: Air Pollution Modelling and its Applications XVII, C Borrego and D Steyn (eds.), Kluwer/Plenum, New York, in press Carissimo B, Dupont E, Musson-Genon L, Marchand O (1997) Note de Principe du Code MERCURE. Version 3.1, Electricité de France, EDF HE-33/97/001, EDF publications, France


35

Duijm NJ, Carissimo B (2001) Evaluation methodologies for dense gas dispersion models. In: The handbook of hazardous materials spills technology. M Fingas (ed.), McGraw-Hill, New York Eidsvik KJ (1980) A model for heavy gas dispersion in the atmosphere. Atmos. Environ., 14, 767–777 Glendening, JW, Businger, JA, Farber, RJ (1984) Improving plume rise prediction accuracy for stable atmospheres with complex vertical structure. J. Air Pollut. Control Assoc., 34, 1128–1133 Hanna SR, Strimaitis DG, Chang JC (1991) Hazard response modeling uncertainty (a quantitative method). Vol. 2, Evaluation of commonly used hazardous gas dispersion models. Sigma Research Corporation for AFESC, Tyndall AFB, FL, and API, Report Nos. 4545, 4546, and 4547, 338 pp Hurley PJ, Manins PC (1995) Plume rise and enhanced dispersion in LADM. ECRU Technical Note No. 4, CSIRO Division of Atmospheric Research, Australia Hurley PJ (2005) The Air Pollution Model (TAPM) Version 3. Part 1: Technical Description. CSIRO Atmospheric Research Technical Paper No. 71 Moon D, Albergel A (1997) The use of the MERCURE CFD code to deal with an air pollution problem due to building wake effects Journal of Wind Engineering and Industrial Aerodynamics, Volumes 67–68, April–June 1997, pp 781–791, Computational Wind Engineering Moussafir J, Oldrini O, Tinarelli G, Sontowski J, Dougherty C (2004) A new operational approach to deal with dispersion around obstacles: the MSS (MicroSwift-Spray) software suite, 9th International Conference on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes Garmisch 1–4 June 2004 Nielsen M, Ott S (1996) A collection of data from dense gas experiments, Riso Report 845(EN) Riou Y (1987) Comparison between the MERCURE-GL code calculations, wind tunnel measurements and. Thorney lsland field trials, J. Hazardous Mater. 16, 247–265 Tinarelli G, Anfossi D, Brusasca G, Ferrero E, Giostra U, Morselli MG, Moussafir J, Tampieri F, Trombetti F (1994) Lagrangian particle simulation of tracer dispersion in the lee of a schematic two-dimensional hill. J. Appl. Meteorol., 33 (6), 744–756 Tinarelli G, Anfossi D, Bider M, Ferrero E, Trini Castelli S (2000) A new high performance version of the Lagrangian particle dispersion model SPRAY, some case studies. In: Air Pollution Modelling and its Applications XIII, SE Gryning and E Batchvarova (eds.), Kluwer/Plenum, New York, pp 499–507

36

G. Tinarelli et al.

Discussion P. Builtjes: Is my impression correct that MicroSpray is better than CFDMercure, and if the models give similar results, why do you use MicroSpray and not Mercure? D. Anfossi: I would not say that MicroSpray is better than CFD-Mercure, but I certainly say that that the two models give similar results. The advantage of using MicroSpray is related to the computing time: a few minutes against many hours. B. Fisher: The treatment of the boundary condition at the surface is difficult in a Lagrangian particle model. How is this boundary condition treated within the CFD code Mercure? D. Anfossi: In MERCURE the boundary conditions are treated as follows. For the wind: wall law at ground (neutral case here); for the other variables: Neumann condition (null gradient)

1.7 Evaluation of the Hazard Prediction and Assessment Capability (HPAC) Model with the Oklahoma City Joint Urban 2003 (JU2003) Tracer Observations Steven Hanna, Joseph Chang, John White and James Bowers

Abstract Results are presented of an evaluation of the Hazard Prediction and Assessment Capability (HPAC) suite of models in an urban environment using data from the Joint Urban 2003 (JU2003) Field Experiment in Oklahoma City (OKC). JU2003 included 29 separate SF6 tracer continuous releases (of 30-minute duration) on ten days from a point source near ground level in or immediately upwind of the built-up downtown area. The ten Intensive Operating Periods (IOPs) consisted of six daytime periods and four nighttime periods. Tracer was sampled at over 100 locations at distances ranging from 0.1 to 4 km from the source. The current study tests two alternate urban configurations of HPAC and four optional meteorological inputs. The two HPAC configurations were the Urban Dispersion Model (UDM) and the Urban Canopy (UC) options. The four meteorological data options were basic default National Weather Service (NWS) data, a single averaged wind, a single upwind anemometer and radiosonde, and detailed three-dimensional winds from a meteorological model, MM5. The evaluations of the maximum 30-minute averaged concentrations on six downwind distance arcs are summarized in this paper. In most cases, the MM5 meteorological inputs yielded the best HPAC results. Also, in general, the UC urban option produced higher concentrations, by about a factor of two, than the UDM urban option. The UDM urban option performed better during the night IOPs and the UC urban option performed better during the day IOPs. There is an obvious day-night difference in the model biases, with most options overpredicting during the night and most options underpredicting during the day, suggesting that they are overstating the relatively-small observed day-night difference in near-ground urban stability and in tracer concentrations. Keywords Air quality model evaluation, HPAC model, JU2003 field experiment, urban dispersion models


63

64

S. Hanna et al.

1. Introduction and Approach This paper presents the methods and some of the results of an evaluation of the Hazard Assessment and Prediction Capability (HPAC) transport and dispersion modeling system (DTRA, 2004; Sykes et al., 2007) using the Joint Urban 2003 (JU2003) continuous release tracer data. Allwine et al. (2004) summarize the JU2003 field experiment, which took place in Oklahoma City during July 2003. Clawson et al. (2005) describe the JU2003 SF6 observations. The ten IOPs had three different source release locations: Botanical Garden (upwind of the downtown area) for IOPs 03 through 07; Westin Hotel (in the built-up downtown area) for IOPs 01, 02, and 08; and Park Avenue (in a street canyon in the downtown area) for IOPs 09 and 10. IOPs 01–06 took place during the day and IOPs 07–10 took place during the night. During each of the ten JU2003 IOPs, three continuous releases of SF6 of 30minute duration were made at 2-hour intervals, except that only two releases were made during IOP01. Hanna et al. (2007) present the results of a similarity analysis of the JU2003 wind, turbulence, and continuous-release concentration data. Instantaneous (puff) releases of SF6 also took place during each IOP, but these releases are not discussed or analyzed here (see Zhou and Hanna, 2007, for the results of an analysis of the along-wind diffusion of the puffs). Samplers were set out on a rectilinear grid in the built-up downtown area at distances less than 1 km from the source. Samplers were also set out on three concentric arcs, covering an angular range of about 120° at distances 1, 2, and 4 km to the north of the downtown area. The specific sampler locations changed from one IOP to the next, depending on the release location and the wind direction. Figure 1 presents a map of the sampler locations. The averaging time for the samplers was adjustable and was generally set to 5, 15, or 30 minutes. The analysis in this paper uses 30-minute averaged concentrations, C, normalized by the source emission rate, Q. In the downtown area, where the samplers were on a rectangular grid, the authors subjectively assigned each sampler to one of three effective “arc” distances: 0.30, 0.62, and 0.85 km. The data from the sampling arcs at 1, 2, and 4 km were also used. We consider only the street-level samplers. We have found that the first two trials of the daytime IOP05 are more representative of nighttime stable conditions, though the releases took place during the early morning. The IOP05 trials with relatively high observed C/Q were caused by relatively low mixing depths (less than 200 m). Consequently, IOP05 trials 1 and 2 are removed from some parts of the analysis of the daytime IOPs or are included in the nighttime category. Also, in the evaluations reported here, data from IOPs 01 and 02 are not included because of problems with setting up the input data. Previous analyses (see Hanna et al., 2007) of the JU2003 observed 30-minute averaged maximum concentrations on several downwind distance arcs showed that values of C/Q were generally about three times higher during the night IOPs than during the day IOPs. We hypothesized that this relatively small difference is due to the relatively small differences in near-ground stability, ranging from slightly unstable during the day to slightly stable during the night.

Evaluation of the Hazard Prediction and Assessment Capability (HPAC) Model

65

Fig. 1 Map of SF6 sampler locations during JU2003. The samplers that were used in the current paper are marked by solid triangles, and are the Field Research Division - Air Research Laboratory (FRD-ARL) continuous samplers described by Clawson et al. (2005)

This paper considers two urban HPAC options (Urban Dispersion Model (UDM) and Urban Canopy (UC)). The paper also considers four meteorological input options: BDF - Basic National Weather Service (NWS) default MED - Mesoscale Meteorological Model-Version 5 (MM5) MEDOC outputs AVG - Average wind speed and direction from all anem (Hanna et al., 2007) UPWND - Wind speed and direction from DPG Portable Weather Instrumentation Data System (PWIDS) #15 on the Post Office, located just upwind of the downtown area, with observed mixing heights based on the upwind Pacific Northwest National Laboratory (PNNL) radiosonde data.

2. HPAC Options and Inputs HPAC was used to calculate SF6 tracer concentrations at the locations of the bag samplers within the central business district (CBD) and the three outer arcs (at 1, 2, and 4 km). As discussed above, the model runs used two HPAC urban model

S. Hanna et al.

66

configurations and four different sets of meteorological data. Model source parameters included the initial Gaussian spread ıy and ız (set to 0.232 m), the tracer release duration (30 minutes), the release height (1.9 m), and the mass dissemination rate, Q (obtained from the JU2003 database). Cloud cover was set based on the standard hourly NWS observations. Surface moisture was set to “normal” and gridded terrain elevation was used except for the HPAC model runs that used the MM5 gridded data (the MED meteorological option), which already contained the terrain information. With two exceptions, boundary layer calculations were set to default. The gridded MM5 MEDOC output files include the mesoscale model’s boundary surface heat flux and boundary layer depth estimates. In the case of the UPWND meteorological option, the boundary layer depth was estimated by the authors from PNNL radiosonde soundings for each continuous release. The PNNL radiosonde data set was selected because of the close proximity of the balloon release site to PWIDS #15 and its location upwind of the CBD. Default values were used for other HPAC inputs such as the Bowen ratio (2) albedo (0.16), canopy height (30 m, the value used for the Urban 2000 study), and canopy flow index (2). All HPAC runs except those using the MED meteorological option used HPAC’s SWIFT diagnostic wind field model to derive mass-consistent wind fields. The SWIFT default parameters were used in these runs with the exception of the wind field update interval, which was set to 1 hour for the BDF and AVG meteorological options, and to 10 minutes for the UPWND meteorological option. Other HPAC input parameters were: x x x

Conditional averaging time set to 30 minutes No large scale variability Sampling height set to 1.5 m (the height of the bag samplers)

The BOOT statistical model evaluation software (Chang and Hanna, 2004) was used to compare predicted and observed arc maximum 30-minute averaged C/Q paired and unpaired in space and time. The limited results presented here focus on the maximum C/Q on a given downwind arc. The following performance measures were used, where we let X = C/Q: Fractional Bias FB = 2<Xo–Xp>/(<Xo> + <Xp>) Normalized Mean Square Error NMSE = /(<Xo> <Xp>) Fraction of Xp within a factor of two of Xo (FAC2) Geometric Mean MG = exp(–) Geometric Variance VG = exp ( Subscripts p and o refer to predicted and observed, and the symbol < > represents an average. Residual plots were also used in the evaluation, where the ratio of Xp/Xo was plotted versus downwind distance, x. The five lines on the box plot represent the 98th, 84th, 50th, 16nd, and 2nd percentiles, respectively, for the group of data considered.


67

3. Results The BOOT software was used to calculate the performance measures and generate the residual plots as defined above for the eight model configurations (two urban model options and four alternate meteorological inputs). Table 1 presents the performance measures, calculated separately for the day and night IOPs. Figure 2 shows the residual plots for urban options UDM and UC coupled with the MED meteorological option. It was necessary to split the analysis into day and night because the first comparisons showed that most of the eight model configurations had a tendency to overpredict at night and underpredict during the day. Warner et al. (2007) found similar biases in their evaluations of HPAC with the JU2003 data for a wider variety of urban model options and meteorological inputs. The MM5 MEDOC inputs were better able to account for the low mixing depths than the other meteorological options. Some basic conclusions from the performance measures and residual plots are: Table 1 Performance measures for evaluations of HPAC with the JU2003 data. See text for definitions of performance measures, meteorological options, and model options. Note that when FB = –2/3, there is a mean factor of two overprediction, and when FB = +2/3, there is a mean factor of two underprediction. FB = 0 and MG = 1 indicate an unbiased model. IOP03, IOP04, & IOP06 (daytime only, excluding IOP05) Met Options Model Options FB NMSE FAC2 BDF UDM 0.88 3.4 19% BDF UC -0.29 1.6 59% MED UDM 0.91 3.9 50% MED UC -0.38 1.9 70% AVG UDM 1.02 4.9 46% AVG UC 0.52 1.4 63% UPWND UDM 0.44 1.2 74% UPWND UC -0.89 3.9 44%

MG 2.28 1.41 1.81 0.89 2.15 1.59 1.04 0.47

IOP07-IOP10 (nighttime only) Met Options Model Options BDF UDM BDF UC MED UDM MED UC AVG UDM AVG UC UPWND UDM UPWND UC

MG 1.58 1.45 1.02 0.36 0.32 0.11 0.38 0.17

x

x

FB -1.15 -1.56 -0.47 -1.19 -1.35 -1.74 -1.30 -1.62

NMSE 20.1 34.0 7.2 13.0 31.0 85.2 22.8 38.0

FAC2 32% 5% 49% 25% 28% 3% 29% 5%

VG 3 2 2.1 1.6 2.4 1.6 1.5 2.3

VG 1.3E+07 1.3E+10

3.4 22.6 9.8 385.0 6.9 116.0

For most model and meteorological input combinations, there is a tendency to overpredict by an approximate factor of 3 or 4 at night and underpredict by an approximate factor of 2 during the day. For the daytime IOPs (03, 04, and 06), MED-UC and UPWND-UDM tend to have the least bias, lowest scatter and highest FAC2; and little trend with x. MED-UDM has an underprediction tendency of about a factor of 3 or 4 at small x. UDM and UC simulations of C/Q tend to

68

S. Hanna et al.

agree for all meteorological options at x > 1 km (outside of the built-up area). For all meteorological options, the ratio of the concentrations predicted by UC to UDM is about 3 or 4 at small x (inside of the builtup area). UDM simulations are always lower than the observed values (by factor of 2 to 5) at small x for all meteorological options. For the nighttime IOPs (07, 08, 09, and 10), MED-UDM has the least bias, lowest scatter, and highest FAC2; and has little trend of the residuals with x. Unlike the daytime runs, UDM and UC simulations do not agree as well at large x. The same bias occurs at all x. For all meteorological options, the ratio of the concentrations predicted by UC to UDM is about 2 at all x. The AVG and UPWND meteorological options lead to large mean overpredictions of a factor of 3–10. Our HPAC evaluations to date and the evaluations reported by other groups (e.g., Warner et al., 2007) have confirmed that urban HPAC overpredicts during the night and underpredicts (by a smaller amount) during the day. This type of behavior suggests that the model may be using too broad of a diurnal range in stabilities. The analyses by Hanna et al. (2007) of sonic anemometer data (including calculations of surface heat fluxes and Monin Obukhov lengths L) suggest that the stability in the built-up downtown area of OKC is near neutral, and usually slightly unstable, throughout the diurnal cycle. This result can be attributed to the strong mechanical mixing due to the buildings and anthropogenic heat input. At night, the slightlyunstable near-surface urban boundary layer is capped at a height of about 200 m by a more stable layer representative of the upwind boundary layer. It is hypothesized that urban HPAC would do better if it used a more nearly-neutral stability parameterization throughout the diurnal cycle, which would ameliorate the nighttime overpredictions and daytime underpredictions. Currently, the HPAC meteorological preprocessor assumes that the sensible heat flux in the upwind area is also valid in the urban area. (When used with MM5 inputs, HPAC uses the heat fluxes computed by the mesoscale model.) Use of the upwind sensible heat flux in the urban area is probably reasonable during the day, but is not valid at night, which may explain the larger biases at night. Note that the HPAC meteorological preprocessor does modify the friction velocity u* based on the increased roughness in the urban area. The observed arc-maximum 30-minute average normalized concentration data from JU2003 indicate that, on average in the CBD (x < 1 km), the daytime C/Q’s are about a factor of 3 smaller than the nighttime C/Q’s. At 1 km < x < 4 km, the day-night difference in C/Q increases to about a factor of 8, due to the increasing nighttime stability in the suburbs as the plume travels out of the CBD. The factor of 3 difference in CBD C/Q is consistent with an assumption of slightly unstable conditions during the day and very slightly-stable conditions during the night in the CBD. If a dispersion model is going to do well with these data, it must be able to simulate the observed factor of 3 day-night difference in C/Q. But because most HPAC model options showed large overpredictions during the night and large underpredictions during the day, it is concluded that those options are overstating the day-night difference in near-ground urban stability.


69

Finally, we hesitate to overinterpret these results. Just because a model combination seems to do better than another, it could be because that model tends to over or underpredict with respect to the other models and may have a compensating error. Until additional analyses are carried out, such as investigation of the effective wind speeds being used by each meteorological input option, it is difficult to untangle the possible effects of different meteorological inputs and decide which is more realistic. For example, it could be that the model that is currently performing better is using an effective wind speed that is too low or too high (i.e., there are compensating errors).

Fig. 2 Residual plots (Xp/Xo vs x) for HPAC urban options UDM and UC with MM5 MEDOC meteorological inputs, for day (top) and night (bottom) IOPs. The significant lines on the box plots indicate, from bottom to top, the 2nd, 16th, 50th (i.e., the median), 84th, and 98th percentiles of the n Xp/Xo points at that arc distance

Acknowledgments This study was supported by the U.S. Defense Threat Reduction Agency, with Rick Fry as project manager.

70

S. Hanna et al.

References Allwine KJ, Leach M, Stockham L, Shinn J, Hosker R, Bowers J, Pace J (2004) Overview of Joint Urban 2003 – An atmospheric dispersion study in Oklahoma City. Preprints, Symposium on Planning, Nowcasting and Forecasting in the Urban Zone. American Meteorological Society, January 11–15, 2004, Seattle, Washington. Chang JC, Hanna SR (2004) Air quality model performance evaluation, Meteorol. Atmos. Phys. 87, 167–196. Clawson K, Carter R, Lacroix D, Biltoft C, Hukari N, Johnson R, Rich J, Beard S, Strong T (2005) Joint Urban 2003 (JU2003) SF6 Atmospheric Tracer Field Tests, NOAA Tech Memo OAR ARL-254, Air Resources Lab., Silver Spring, MD, 162 pp. + Appendices. DTRA (2004) HPAC Version 4.04.04 (DVD Containing Model and Accompanying Data Files), DTRA, 8725 John J. Kingman Road, MSC 6201, Ft. Belvoir, VA 22060-6201. Hanna S, White J, Zhou Y (2007) Observed winds, turbulence, and dispersion in built-up downtown areas of Oklahoma City and Manhattan. Bound.-Lay. Meteorol. 125, 441–468. Sykes RI, Parker S, Henn D, Chowdhury B (2007) SCIPUFF Version 2.3 Technical Documentation. L-3 Titan Corp, POB 2229, Princeton, NJ 08543-2229, 336 pp. Warner S, Platt N, Urban J, Heagy J (2007) Comparisons of transport and dispersion model predictions of the JU2003 field experiment. J. Appl. Meteorol. Climatol., in press. Zhou Y, Hanna S (2007) Along-wind dispersion of puffs released in a built-up urban area. Bound.-Lay. Meteorol. 125, 469–486.

Discussion B. Denby: Did you come to any conclusion in regard to “suitability for application” from this study? i.e., Does HPAC do what it is designed to do? Did you also run HPAC in operational mode for this campaign? S. Hanna: The HPAC model’s suitability for application has been shown for many types of scenarios prior to this study. Most of these evaluations with field data were carried out by the model developers (Sykes et al., 2007) and have been published in peerreviewed journals. HPAC has also been shown to be suitable for stable conditions in urban areas in prior evaluations using observations from Salt Lake City. The current evaluation involves tracer releases during the day and night in Oklahoma City, using various input meteorology options. The model is seen to do best


71

using its preferred meteorological inputs – mesoscale model simulations, in this case the MM5 model. Thus HPAC does appear to do what it is designed to do. During the field experiment, HPAC was run in operational mode, using MM5 inputs, in order to aid in deciding how much tracer gas to release, where to place the samplers, and other aspects of the experiment. Those runs were made by NCAR and not by the authors of the current paper. A. Venkatram: You described several ways of choosing the meteorological inputs for the model. Is there a basis for selecting the inputs that would be most representative of the “real” situation? S. Hanna: In a “real” situation, HPAC would be run using the meteorological model inputs. The HPAC system is set up so that real-time meteorological simulations by five or six models are continually made and stored on the Meteorological Data Server (MDS) at NCAR. Whenever HPAC must be run, the user can access these model simulations via the internet from anywhere in the world. S.T. Rao: Have you looked at measurements at the roof top level? Does the model capture the concentrations at roof top levels better than those as the street levels? S. Hanna: Yes, we have analyzed the roof top concentration measurements, which were taken on about five tall downtown buildings. However, the current paper uses only the surface concentration measurements. In general though, the rooftop concentrations are significant fractions of the surface measurements (e.g., 0.01–0.5), with the larger ratios found for buildings that are the farthest distance from the source and/or have elevations at the low end of the range.

1.2 Modelling of the Urban Wind Profile Sven-Erik Gryning and Ekaterina Batchvarova

Abstract Analysis of meteorological measurements from tall masts in rural and urban areas show that the height of the boundary layer influences the wind profile even in the lowest hundreds of meters. A parameterization of the wind profile for the entire boundary layer is formulated with emphasis on the lowest 200–300 m and presented here. Results are shown from applying the parameterization of the wind profile on independent measurements from an urban experimental campaign that was carried out in Sofia, Bulgaria in 2003. Keywords Boundary layer height, Sofia experiment, wind profile, urban boundary layer

1. Introduction Analysis of profiles of meteorological measurements from a 160 m high mast at the National Test Site for wind turbines at Høvsøre (rural, Denmark) and at a 250 m high TV tower at Hamburg (urban, Germany) shows that the wind profile based on surface-layer theory and Monin-Obukhov scaling is valid up to a height of 50 to 80 m. At higher levels deviations from the measurements progressively occur, being most pronounced at atmospheric neutral conditions as illustrated in Figure 1. 300 200

Height (m)

100

Hamburg-urban 50 30

-200 N / 2@

ȝ

¦ j 1

ª¬ A j cos( 2ʌ f j t) B j sin( 2ʌ f j t)º¼ ,

(1)

where µ is a constant term, (N/2) is the greatest integer less than or equal to N/2. For one-year air quality measurements with hourly resolution N is equal to 8,760 (365*24) and the highest frequency is f =(N/2)/N = ½ = 0.5, which correspond to a wave period (period = 1/frequency) of 2 hours. Therefore, the measurement data with 1-hour resolution do not allow analysing waves with period less then 2 hours. The fast Fourier transform (FFT) algorithm was used to study the variance in the frequency-domain. The periodogram values (Pj) at frequency fj were calculated as:

Pj

A

2 j

B 2j u N / 2

(2)

The parameter Pj can be interpreted in terms of variance and allow to identify which frequencies give most important contribution to the variability.

2.3. Data filtering Depending on the analysis objective and the properties of the phenomena under study, it could be important to remove some frequencies from the series. For this purpose the data filtering techniques can be used. The Kolmogorov-Zurbenko (KZ) filter commonly applied to the air quality data (Rao et al., 1997; Hogrefe et al., 2006) is used to decompose the time series into deterministic and stochastic components. The KZ filter is a low-pass filter that removes higher frequency variations from the data. The KZ(m,k) filter of the original time series x is computed as a simple moving average of m points applied k times (number of iterations) The filter is designed taking into account the desired separation frequency wc (Rao et al., 1997). The application of the KZ filter allows to decompose the original time series C(t) on baseline (CB) (deterministic) and short-term (CS) (stochastic) components in time t (Rao et al., 1997): C(t)=CB(t)+CS(t). Thus, the output of the filtering process corresponds to the baseline component and the short-term component which is defined as a difference between the original and the filtered data.

3. Photochemical Model Application The air quality was modelled with the CHIMERE chemistry-transport model (Schmidt et al., 2001), forced by the MM5 meteorological fields (Figure 1), and carried out for 2004 year, regarding gaseous and particulate pollutants (Monteiro et al., 2007).

O. Tchepel et al.

464 meteorological model MM5

boundary conditions

emissions

climatological model GOCART

anthropogenic biogenic

CHIMERE gaseous chemistry

aerosols module

44 species, 116 reactions

Fig. 1 The MM5-CHIMERE air quality modelling system

transport

gas deposition/aerosols

horizontal (PPM) vertical diffusion

dry wet

pollutants concentration

pollutants deposition

gases/aerosols

gases/aerosols

In the present application, the model is run at a regional scale over a domain covering all mainland Portugal, with a 10 × 10 km2 grid size with a vertical resolution of eight vertical layers of various thickness extending from ground to 500 hPa. Boundary conditions are provided by a prior large-scale simulation, covering Western Europe with a 50 × 50 km2 resolution. Boundary conditions for the regional simulation are taken from the monthly means of the MOZART and GOCART models, as in Hodzic et al. (2005). Regarding anthropogenic emissions, the most updated annual emission inventory (2003 year) was used, disaggregated at the municipality level (Monteiro et al., 2007). Simpson et al. (1999) methodology was adopted to calculate biogenic emissions with the CHIMERE model. Time disaggregation was obtained by application of monthly, weekly and hourly profiles from the University of Stuttgart (Monteiro et al., 2007).

4. Results and Discussion Figure 2 presents the result of the statistical analysis of the one-year log-transformed observation and modelling time series considered in the analysis. A comparison of the modelling data with the measurements shows underestimation of the NO concentration predicted by the model (median value) and wider spread. It can be seen that the model predicts the same distribution for the urban background station (Chelas) and for the urban traffic station (Benfica), while the measurements reveal significant difference between these two observation points located in proximity. This tendency is also observed for NO2 time series. In general, for NO2 and O3 the model outputs are in good agreement with the observations in terms of median, although the spread of the data in the interquartile range is different.

Estimation of the Modelling Uncertainty Related with Stochastic Processes

465

These log-transformed data were used for the spectrum analysis after removing the linear trend and the subtracting the annual average log concentration. An example of the results of the spectrum analysis for NO2 is presented in Figure 3 for several locations. The plot represents the relative contributions of different frequencies to the variance. Measurement data

Modelling results Modelling results

6

6

4

4

2

2

0

Chamusca_O3

Boavista_O3

Vila N.T._O3

Chamusca_NO2

Chelas_NO2

99% 1% 75% 25% Median

Vila N.T._NO2

Vila N.T._O3

Chamusca_O3

Boavista_O3

Chamusca_NO2

Chelas_NO2

Vila N.T._NO2

-8

Benfica_NO2

-6

-8 Chelas_NO

-4

-6

Benfica_NO

-4

Benfica_NO2

-2

Chelas_NO

-2

Benfica_NO

0

Chamusca_NO

ln(Concentration)

8

Chamusca_NO

ln(Concentration)

Measurement data

8

99% 1% 75% 25% Median

Fig. 2 Box plot of the measurement and CHIMERE modelling data. Concentration in Pg m-3

The variance spectrum of the measured time series reveals that higher frequencies (small period, T = 1/f) have important contribution in areas with direct influence of traffic, while for the rural station seasonal variations are most important. Nevertheless, all stations present clearly identified peaks for the period of T = 12 hours and T = 24 hours (f § 0.08 h-1 and f § 0.04 h-1). For the Benfica urban traffic station, the contribution of the wave with the 12 hours period to the total variance is even higher than day-to-night variations represented in the spectrum by the wave with T = 24 hours. This pattern is clearly related with fluctuations in the traffic emissions. The analysis of the CHIMERE variance spectrum shows that seasonal variations are underestimated in the modelled timeseries. Moreover, one-week variance peak is missing in the model spectrum due to incorrection on working day/weekend emissions disaggregation. However, the 12and 24-hours peaks are well identified and in some cases are even overrated. Based on the spectrum analysis the separation frequency for the KZ filter was defined with the purpose to remove all fluctuations with the period less than 12 hours from the original data. The filter parameters m = 3 and k = 3 were selected, providing the separation frequency wc = 0.0905 h-1. The KZ3,3 filter was applied to the log-transformed measurement and predicted hourly concentration data, thus allowing the deterministic component and short-term noise to be separated. The short-term component is obtained as the filter residual. It is characterised by zero mean and covariance with the baseline of about 2% showing a good effectiveness of the timescales separation for NO2 and O3. The covariance for NO is about 8%. The distribution parameters of the short-term components for different pollutants are presented in Figure 4. The data reveal very good agreement between Ozone modelling data and observations at rural background station (Chamusca) but significant spread for NO and NO2. For the urban stations, both traffic (Benfica)

466

O. Tchepel et al.

and background (Chelas), the model is underestimating NO and overestimating NO2 short-term variability. For O3 short-term component at urban station (Boavista), the model predictions are characterized by larger spread in the 98percentile but narrower in the interquartile range.

Fig. 3 Variance spectrum for the log-transformed hourly NO2 measurements and CHIMERE predictions for urban traffic station (Benfica), urban background station (Chelas), sub-urban background station (Vila N Telha) and regional background station (Chamusca)

Estimation of the Modelling Uncertainty Related with Stochastic Processes Modelling results Modelling results

Measurement data Measurement data

1.5

1.5

1.0

1.0

0.5

0.5

0.0

0.0 ln (Concentration)

-0.5

-1.0

-1.5

-0.5

-1.0

-1.5

-2.0

Chamusca_O3

Boavista_O3

Vila N.T._O3

Chamusca_NO2

Chelas_NO2

Vila_N.T._NO2

Benfica_NO2

Chamusca_NO

Chamusca_O3

Boavista_O3

Vila N.T._NO2

C hamusca_NO2

Chelas_NO2

Vila N.T._NO2

Benfica_NO2

Chelas_NO

Benfica_NO

Chamusca_NO

-2.5 99% 1% 75% 25% Median

Chelas_NO

-2.0

-2.5

Benfica_NO

ln (Concentration)

467

99% 1% 75% 25% Median

Fig. 4 Box plot of the log-transformed stochastic variations

After the filtering, the data were back-transformed to the original concentration units and Root Mean Squared Error (RMSE) and Pearson Correlation coefficient were calculated in order to analyze the relationship between the observations and the model predictions. Table 2 summarizes the statistics and confirms a better agreement between the observed and predicted values after the short-term components being removed. Thus, the analysis of the deterministic components shows a decrease in RMSE and improves the correlation between the two time series in comparison with the same parameters calculated for the original data without filtering. Table 2 Model validation using original and filtered hourly concentration data. Parameter Pollutant Station Benfica Chelas Vila N.T. Chamusca Boavista

NO Orig. 134.8 25.2 20.6 0.9 –

RMSE (Pg m-3) NO2 Det. Orig. Det. 115.6 48.9 45.9 18.1 21.5 18.9 15.7 22.3 20.6 0.7 5.3 4.6 – – –

Orig. – – 30.8 21.4 32.1

O3

Det. – – 27.9 19.8 30.9

Pearson correlation coefficient NO NO2 O3 Orig. Det. Orig. Det. Orig. 0.21 0.25 0.55 0.58 – 0.11 0.15 0.39 0.40 – 0.10 0.10 0.43 0.43 0.40 0.42 0.46 0.35 0.40 0.60 – – – – 0.49

Det. – – 0.43 0.65 0.54

Orig. – original data; Det. – deterministic component after the filtering

5. Conclusions The proposed approach contributes to better understanding of the model prediction uncertainty related with short-term variations of the pollutants concentration. The analysis of the measurement data in frequency domain shows that the contribution of short- and long-term components to the total variability is different at different locations and accentuates the important role of the intra-day ( 100 Pg/m3, the correlation is higher than 0.5 between any two PM10sites. On the contrary, low correlation between PM10 measurements taken at different sites is the evidence of local aerosols. In particular, for days when PM10 < 100 Pg/m3, the correlation is about 0.2. Therefore, for those days without significant dust, i.e. when PM10 < 100 Pg/m3, we could characterize an average level of local aerosols by estimating their mean and standard deviation. The threshold of local PM10 aerosols was determined as their mean plus standard deviation. As estimated, the threshold in Tel-Aviv (59.7 ȝg/m 3 ) is higher than that in Beer-Sheva (56.6 ȝg/m 3) or in Carmiel (48.7 ȝg/m3 ). Only PM10 measurements, which exceeded the threshold of local aerosols, have been compared with model data.

4. Quantitative Comparisons Between Model and PM10 Data As an illustration, Figure 1a shows a comparison between 24-hour model-predicted surface dust concentrations and PM10 measurements taken at Tel Aviv during March 2006. It is clearly seen that for dusty days, when Saharan dust is transported over Israel, dust aerosols dominate PM10 concentrations. With respect to model capabilities to produce dust forecasts, we see that the old DREAM model frequently underestimates PM10 data. DREAM-8 produces more accurate forecasts than the old DREAM. The correspondence between model data and PM10 measurements higher than the threshold of local aerosols over Tel-Aviv, Israel, was evaluated by means of scatter-plots (Figure 1b). The bisector curve, shown in the scatter-plot, indicates ideally accurate forecasts: the points on or close to the bisector represent the best correspondence between the model-simulated data and the PM10 measurements.

Saharan Dust over the Eastern Mediterranean: Model Sensitivity

361

Fig. 1 a – the comparison between PM10 and model data at Tel Aviv during March 2006. The horizontal line corresponds to the level of local aerosols. b – The scatter-plot between the common logarithm of PM10 and model data at Tel Aviv obtained during the period from February to June 2006. The dashed lines show the root-mean-square interval of the original DREAM deviations from the bisector, while the thin lines show the root-mean-square interval of DREAM-8

The root-mean-square interval of deviations of points from the bisector can be used in order to characterize the range of forecast accuracy. One can see that the root-mean-square interval for DREAM-8 is half as wide as that for the original DREAM (Figure 1b). This conclusion is supported by the results of the correlation analysis. A high correlation 0.54 between PM10 and DREAM-8 data was obtained,

362

P. Kishcha et al.

in contrast to that of 0.27 for the old DREAM model. This suggests the advantage of DREAM-8 compared to the original DREAM.

5. A Typical Route for Dust Transport to Israel As described by Alpert and Ziv (1989), a typical route for dust transport into Israel in spring is from the Sahara desert, through Egypt, into the Eastern Mediterranean. Sometimes, typical dust transport into the Eastern Mediterranean ends with an anticlockwise movement of dust around the Eastern Mediterranean, based on DREAM-8 predictions of dust transport and wind distributions between 13 and 14 March 2006. The dust transport over Gulf of Suez, Sinai, Israel to Cyprus and Turkey was associated with a low-pressure system centred over Greece. The SeaWIFS satellite picture corroborates model simulations by displaying dust around the Eastern Mediterranean cost on March 13 (not shown). Note that Ganor (1991) described a similar event of counter-clockwise dust transport around the Eastern Mediterranean, observed on April 29, 1987. For the dust event on 13–14 March 2006, the comparison between PM10 measurements of surface dust concentration, taken in Tel Aviv with 24-hour model-predicted data, shows that both models produced dust forecasts of quite good accuracy.

6. Unusual Clockwise Dust Transport We noticed that, sometimes, Saharan dust could be transported to the Eastern Mediterranean in an unusual long-distance clockwise movement, from the western part of the Sahara desert through Southern Europe towards Alps and then to the Eastern Mediterranean, based on DREAM-8 predictions. This dust transport started on April 5, 2006 (Figure 2a). The synoptic situation was characterized by the specific structure consisting of an extensive high, dominating the whole of North Africa and the western Mediterranean, and an intensive low over the eastern Atlantic, near the Iberian Peninsula. The low significantly affected the western part of the Sahara desert, as revealed by the wind distribution at 3,000 m altitude. A strong airflow with speeds higher than 20 m/s moved from the Western Sahara to the Western Mediterranean. Such a synoptic situation produced favourable conditions for the development of a heavy dust storm accompanied by an intensive dust intrusion into southern Europe. On the following day, April 6, as the low shifted eastward from the Atlantic across Europe, dust was transported through the central Mediterranean, Italy, Greece, into the Black Sea (Figure 2b). On April 7, dust movement shifted eastward in such a way that dust reached Cyprus and, eventually, the Eastern Mediterranean (Figure 2c).


363

a)

b)

c)

d)

e) DREAM-8

250 200

DREAM orig.

150

150

100

100

50

50

0

0

C oncentration, ug/m ^3

200

PM10

07

/0 4 07 /200 /0 4/ 6 0 0 8 2 0 0 2 :0 0 /0 6 4 1 08 /2 00 4:00 /0 6 4/ 0 0 9 2 0 0 2 :0 0 6 /0 1 4 09 /200 4:00 /0 6 4/ 0 1 0 2 0 0 2 :0 0 /0 6 4 1 10 /2 00 4:00 /0 6 4/ 2 0 0 2 :0 06 0 14 :0 0

C oncentration, ug/m^3

250

Fig. 2 a–d are dust loading predicted by DREAM-8 on 5–8 April 2006 respectively, e – is the comparison between PM10 and 24-hour model-predicted data at Tel-Aviv

On April 8, however, another intensive low developed over the Eastern Mediterranean near Turkey, which caused strong wind over Lybia. As the low shifted eastward between 8 and 10 April, the typical counter-clockwise dust transport was observed over the Eastern Mediterranean and Israel (Figure 2d). The aforementioned dust event was unusual for the season in question – the spring. In

364

P. Kishcha et al.

fact, it was a combination of clockwise long-distance transport through southern Europe supplemented in the end, on April 7, 2006, with counter-clockwise transport through Egypt. This situation suggests that both fine dust particles from the Western Sahara and coarse dust particles from the Eastern Sahara could have been presented in the atmosphere. The comparison between PM10 measurements with model-simulated data clearly shows the advantage of DREAM-8 compared to the original DREAM (Figure 2e). In particular, the maximum PM10 concentration is about 240 Pg/m3, while the one of DREAM-8 is about 100 Pg/m3. We also see that the original DREAM model hardly displays dust. This suggests that the use of eight dust-particle-size classes including four less than 1 Pm (DREAM-8) instead of two classes (the original DREAM model) could be significant for long-distance dust transport predictions.

7. Conclusions The quantitative comparison between the PM10 measurements and those, predicted by the DREAM models, showed that the models are capable of giving mainly acceptable forecasts. For DREAM-8, the correlation between model data and PM10 data was found to be higher than that for the original DREAM, indicating that the use of eight-particle size bins in the dust forecasting instead of only two size bins improves dust forecasts. The advantage of DREAM-8 compared to the original DREAM is particularly significant for long-distance dust transport predictions. During spring 2006, Saharan dust was transported into the Eastern Mediterranean not only in the typical route (from the Sahara desert through Egypt into Israel), but also in an unusual clockwise movement from the western part of the Sahara desert through Southern Europe into Israel. Acknowledgments This study was supported by the Israeli Ministry of Environment’s grant, by the GLOWA-Jordan River BMBF-MOST project and also by the BMBF-MOST grant #1946 on climate change. The authors gratefully acknowledge Boris Starobinets for helpful comments and discussion.

References Alpert P, Ziv B (1989) The Sharav cyclone: observations and some theoretical considerations. J. Geophys. Res. 94, 18495–18514. Ganor E (1991) The composition of clay minerals transported to Israel as indicator of Saharan dust emission. Atmos. Environ., 25A, 12, 2657–2664. Kishcha P, Alpert P, Shtivelman A, Krichak S, Joseph JH, Kallos G, Katsafados P, Spyrou C, Gobbi GP, Barnaba F, Nickovic S, Perez C, Baldasano JM (2007) Forecast errors in dust vertical distributions over Rome (Italy): multiple particle size representation and cloud contributions, J. Geophys. Res., 112, D15205, doi:10.1029/2006JD007427.


365

Nickovic S (2005) Distribution of dust mass over particle sizes: Impacts on atmospheric optics, paper presented at Fourth ADEC Workshop: Aeolian Dust Experiment on Climate Impact, Ministry of the Environ., Nagasaki, Japan, 2005. Nickovic S, Kallos G, Papadopoulos A, Kakaliagou O (2001) A model for prediction of desert dust cycle in the atmosphere, J. Geophys. Res., 106, 18113– 18129. Pérez C, Nickovic S, Baldasano JM, Sicard M, Rocadenbosch F, Cachorro VE (2006) A long Saharan dust event over the Western Mediterranean: lidar, sun photometer observations and regional dust modeling. Journal of Geophysical Research, 111, D15214, doi:10.1029/2005JD006579.

Discussion D. Syrakov: What is the threshold in your first figure? What does it mean? Is there concentrations less then it and how have you determined it? P. Kishcha: The dust models, used in this study, produce forecast of Saharan dust, which is transported from the Sahara into the Mediterranean. No local air pollution is included in the models. On the other hand, PM10 measurements include all kids of atmospheric aerosols both natural and anthropogenic. To compare correctly PM10 vs model data, we have to distinguish between mineral dust, transported from the Sahara, and local PM10 aerosols. Without conducting some chemical analysis this is not a simple task. To get around this problem we used the following approach: When aerosol concentrations retrieved from PM10 measurements were highly correlated with each other at different sites, this was the evidence that we deal with aerosols from remote sources, i.e. from the Sahara desert. As estimated, for example, for days with PM10 > 100 Pg/m3, the correlation is higher than 0.5 between any two PM10-sites. On the contrary, low correlation between PM10 measurements at different sites is the evidence of local aerosols. In particular, for days when PM10 < 100 Pg/m3, the correlation is about 0.2. Therefore, for those days without significant dust, i.e. when PM10 < 100 Pg/m3, we could characterize the average level of local aerosols by estimating their mean and standard deviation. The threshold of local PM10 aerosols was determined as the mean plus standard deviation. Only PM10 measurements, which exceeded the threshold of local aerosols, have been compared with model data.

P. Kishcha et al.

366

S. Hanna: In your statistical evaluations (e.g. correlation coefficients) of the observed and modeled concentrations, you applied a threshold to the observed concentrations but not to the modeled concentrations. Shouldn’t you also apply the same threshold to the modeled concentrations? This is, the statistical calculations should use only those data pairs where both observed and modeled concentrations exceed the threshold. P. Kishcha: Two models were evaluated by using PM10 measurements: DREAM-8 and the old DREAM. It was obtained that the old DREAM model underestimated dust concentrations, while DREAM-8 produced more accurate forecast. We could not apply the same threshold to model-simulated data as we did to PM10 measurements: otherwise we could remove low model values produced by the old DREAM model, and thereby artificially improve its forecasts. G. Kallos: In order to use eight size classes in dust particles you need to have an accurate source area classification. Do you have such data available or you use just what was in the model version developed at the University of Athens several years ago (SKIRON project)? P. Kishcha: In DREAM-8 with eight particle size bins we use the same dust sources as in the original DREAM model with four particle size bins (and also in SKIRON). Ref.: Nickovic et al., JGR, 18, 113– 118,129, 2001. Y.P. Kim: What was the dry deposition algorithm in the model? P. Kishcha: The dry deposition of dust particles was parameterized according to the scheme of Georgi JGR, 9794-–9806, 1986.

4.1 The Effect of Heterogeneous Reactions on Model Performance for Nitrous Acid Golam Sarwar, Robin L. Dennis and Bernhard Vogel

Abstract Recent studies suggest that emissions, heterogeneous reactions, and surface photolysis of adsorbed nitric acid may produce additional nitrous acid in the atmosphere. The effects of these sources on nitrous acid formation are evaluated using the Community Multiscale Air Quality modeling system. Predicted nitrous acid with and without these sources are compared with observed data from northeast Philadelphia. The incorporation of these sources greatly improves the model performance for nitrous acid. It also increases the average hydroxyl radical and ozone by 10% and 1.7 ppbv, respectively. Keywords Emissions, heterogeneous reaction, nitrous acid, surface photolysis reaction

1. Introduction The importance of nitrous acid (HONO) chemistry in producing hydroxyl (OH) and hydroperoxy (HO2) radicals is well established. Alicke et al. (2002, 2003) suggested that the photolysis of HONO may provide as much as 34% of daily integrated OH levels. Zhou et al. (2002) reported that the photolysis of HONO may produce up to 24% of the total daily radical production. Using concurrently measured HONO and OH at the Meteorological Observatory Hohenpeissenberg in summer 2002 and 2004, Acker et al. (2006) suggested that the photolysis of HONO produced 42% of the integrated photolytic HOx (OH + HO2) formation. While the effect of HONO on OH is established, the chemical reactions producing HONO are not well understood. Most air quality models, including the Community Multiscale Air Quality (CMAQ) modeling system, employ only homogeneous chemical reactions for HONO. Recent studies suggest that emissions, heterogeneous reactions, and surface photolysis of adsorbed nitric acid may produce additional HONO in the atmosphere (Vogel et al., 2003; Zhou et al., 2003). In this study, the effects of these sources on HONO are evaluated using the CMAQ modeling system.


349

G. Sarwar et al.

350

2. Methodology 2.1. Model description Model simulations were performed using the CMAQ modeling system (version 4.6) (Binkowski and Roselle, 2003; Byun and Schere, 2006). The CMAQ chemical transport model was configured to use the mass continuity scheme to describe advection processes, the Asymmetric Convective Model version 2 (ACM2) (Pleim, 2007) to describe vertical diffusion processes, the multiscale method to describe horizontal diffusion processes, an adaptation of the ACM algorithm for convective cloud mixing, and the Carbon Bond (CB05) mechanism to describe the gas-phase chemical mechanism (Yarwood et al., 2005; Sarwar et al., 2007). Aqueous chemistry, aerosol processes, and dry and wet deposition were also included. The meteorological driver for the CMAQ modeling system was the PSU/NCAR MM5 system version 3.5 (Grell et al., 1994). Predefined clean air vertical profiles for initial and boundary conditions provided in the CMAQ modeling system were used. Model simulations were performed for July 2001; the model was spun up for seven days to minimize the effect of initial conditions on predictions.

2.2. HONO chemistry The CB05 mechanism contains five homogeneous reactions related to HONO (Yarwood et al., 2005). These reactions and their rate constants are shown in Eqs. (1)–(5) (NO = nitric oxide, NO2= nitrogen dioxide, H2O = water vapor, k = expression for rate constant, first order rate constants are in units of second-1, second order rate constants are in units of cm3 molecule-1 second-1, third order rate constants are in units of cm6 molecule-2 sec-1, T = temperature in Kelvin, M = the total pressure in molecules/cm3, photolysis rate for Eq. (3) is at 400 N (typical summer noon). While this study used the CB05 mechanism, the chemical reactions used in other widely used atmospheric chemical mechanisms are also similar. For example, the CB-IV mechanism contains the same five chemical reactions for HONO (Gery et al., 1989). The Statewide Air Pollution Research Center (SAPRC99) mechanism contains four chemical reactions for HONO (Eqs. (2)–(4) and an additional photolytic reaction) (Carter, 2000). NO + NO2 + H2O ĺ 2.0*HONO; k = 5.0 × 10-40 NO + OH ĺ HONO;

k={ko[M]/(1+ko[M]/k)}Fz Z={(1/N)+log10[ko[M]/k]2}-1 ko = 7.0 × 10-31 (T/300)-2.6 k = 3.6 × 10-11 (T/300)-0.1

(1)

The Effect of Heterogeneous Reactions on Model Performance for Nitrous Acid

F = 0.6 and N = 1.0 HONO + hv ĺ NO + OH; photolysis rate = 0.002 sec-1 OH + HONO ĺ NO2 + H2O; k = 1.8 × 10-11 e(-390/T) HONO + HONO ĺ NO + NO2 + H2O; k = 1.0 × 10-20

351

(2) (3) (4) (5)

As shown below, CMAQ predictions with observed data indicate incorporating only the homogeneous reactions results in HONO predictions that are seriously deficient. HONO emissions are currently not included in the CMAQ modeling system. Several investigators have reported HONO emissions from motor vehicles (Winer and Biermann, 1994; Kirchstetter et al., 1996; Kurtenbach et al., 2001). Kirchstetter et al. (1996) measured HONO and NOx emissions from on-road vehicles at Caldecott Tunnel in San Francisco, California and reported a value of 2.9 × 10-3 for the HONO/NOx emissions ratio. Kurtenbach et al. (2001) conducted measurements in the Wuppertal Kiesbergtunnel and reported a value of 8 × 10-3 for the same ratio. Winer and Biermann (1994) also reported a value of 8 × 10-3 for the same ratio. For this study, HONO emissions were estimated using a value of 8 × 10-3 for the HONO/NOx emissions ratio for on-road and off-road vehicles. Several heterogeneous reactions have been suggested to produce HONO formation in the atmosphere (Aumont et al., 2003). However, most of these reactions appear to be not significant for HONO production in the atmosphere. The heterogeneous reaction involving NO2 and H2O has been shown to be important for HONO production in the atmosphere (Vogel et al., 2003): 2NO2 + H2O ĺ HONO + HNO3

(6)

Laboratory studies suggest that reaction (6) is first order in NO2 (Finlayson-Pitts and Pitts, 2000) and can occur on aerosol and ground surfaces. This heterogeneous reaction is implemented into the CMAQ modeling system with a rate constant of 3.0 × 10-3 × S/V m-1 following Kurtenbach et al. (2001) (S/V is the ratio of surface area to volume of air). The reaction can occur on aerosol as well as ground surfaces. Aerosol surface areas are generally smaller than ground surface areas; thus aerosol surface areas are less effective in producing HONO in the atmosphere. Ground surface areas provided by leaves can be estimated using the Leaf Area Index (LAI). Jones (2006) suggests that the use of LAI underestimates leaf surface areas by at least a factor of two since it only accounts for areas on one side of the leaves. Thus, the S/V ratios for leaves were estimated as follows: S/V = 2*LAI / surface layer height

(7)

Buildings and other structures can also enhance ground surface areas in urban environments. However, estimates of these areas are not readily available. In the absence of such information, an ad hoc approach was used to estimate the ground surface areas for buildings and other structures in urban environments. Svensson

G. Sarwar et al.

352

et al. (1987) studied the kinetics of the reaction involving NO2 and H2O and suggested a value of 0.2 m-1 for typical urban environments. They, however, also indicated that some building materials may provide an order of magnitude greater surface area than their simple projected surface areas due to porosity and roughness. For this study, the S/V ratio for buildings and other structures at the grid-cell with the highest urban environment was assigned a value of 0.3 m-1. The S/V ratios for buildings and other structures for other urban environments were linearly scaled to this S/V ratio by assuming that the values are proportional to the percent urban area in any grid-cell. Using this procedure, the total S/V ratio for the grid-cell containing northeast Philadelphia was estimated to be 0.28 m-1, which is almost 5 times lower than the value of 1.3 m-1 used by Cai et al. (2007) for New York. HONO produced via the heterogeneous reaction on ground surfaces was released into the first layer of the model. Several recent studies also suggest the possibility of the production of HONO via surface photolysis (Zhou et al., 2002, 2003; Vogel et al., 2003; Acker et al., 2006). In their study, Vogel et al. (2003) used a hypothetical species to add a source for HONO production in the surface layer of the model via photolysis since the species that may undergo photolysis to produce HONO was not known. Zhou et al. (2003) recently conducted laboratory experiments and suggested that adsorbed nitric acid (HNO3) on surfaces can undergo photolysis to produce HONO and NO2. A photolysis reaction producing HONO and NO2 was added to the CMAQ modeling system (Eq. (8)) by assuming that the adsorbed amount is equal to the HNO3 deposited via dry deposition since the last precipitation (O3P = ground state oxygen atom). HNO3(ads) ĺ 0.5 HONO + 0.5 O3P + 0.5 NO2 + 0.5 OH

(8)

Zhou et al. (2003) reported a photolysis rate of 1.3 × 10-3 minute-1 at noontime tropical condition for the surface photolysis of adsorbed HNO3, which is 24 times greater than the gaseous HNO3 photolysis rate in the CMAQ modeling system. The surface photolysis rate of adsorbed HNO3 was scaled to the photolysis rate of gaseous HNO3 used in the CMAQ modeling system. HONO and NO2 produced via the surface photolytic reaction was released into the first layer of the model only.

2.3. Observed data Model predictions are compared with measurements from the Northeast Oxidant and Particle Study conducted at northeast Philadelphia (http://lidar1.ee.psu.edu) in July 2001. Continuously measured data from the study were converted into hourly averaged data which were used to calculate an average diurnal profile. Average values for night and day were calculated using the average diurnal profile.


353

3. Results Four different model simulations were performed as shown in Table 1. Table 1 Summary of cases investigated. Case A B C D

Emissions Included NOx, SO2, CO, VOC, NH3, aerosol NOx, SO2, CO, VOC, NH3, aerosol, HONO NOx, SO2, CO, VOC, NH3, aerosol, HONO NOx, SO2, CO, VOC, NH3, aerosol, HONO

Chemical mechanism used CB05 CB05 CB05 + reaction 6 CB05 + reaction 6 and 8

The average predicted and observed HONO at northeast Philadelphia are presented in Table 2. Observed HONO mixing ratio was 50% greater at night than that during the day. Predicted HONO mixing ratio for the case A was only 0.01 ppbv at night compared to an observed value of 1.26 ppbv. Contrary to the observed data, predicted HONO mixing ratio was greater during the day by at least a factor of 4 over the mixing ratio at night. Predicted HONO mixing ratio was significantly lower than the observed data both at night and during the day. Table 2 Summary results.

Case A B C D

Night Obs. HONO (ppbv) 1.26 1.26 1.26 1.26

Pred. HONO (ppbv) 0.01 0.11 0.95 0.96

Obs./ pred. (ratio) 126 11 1.3 1.3

Day Obs. HONO (ppbv) 0.85 0.85 0.85 0.85

Pred. HONO (ppbv) 0.04 0.07 0.25 0.60

Obs./ Pred. (ratio) 21 12 3.4 1.4

When HONO emissions were added to the model (case “B”), predicted HONO reached to 0.11 ppbv at night and was slightly higher at night than during the day. Predicted HONO mixing ratio was still lower than the observed data by a large margin both at night and during the day. When the heterogeneous reaction was also added to the model (case “C”), predicted HONO mixing ratio further improved at night and reached to within 30% of the observed data. However, predicted HONO mixing ratio was still lower than the observed data during the day by a factor of 3.4. When the surface photolysis of adsorbed HNO3 was added to the model (case “D”), predicted HONO mixing ratio improved to 0.60 ppbv during the day due to the increased HONO production and reached to within 40% of the observed data. The observed ratio of HONO/HNO3 was 2.9 at night. For the case A, the ratio was only 0.006 and improved to 0.6 for the case D. During the day, the observed ratio was 0.73 compared to a value of only 0.02 for the case A. It improved to 0.2 for the case D. The relative contribution of these production pathways to predicted HONO for the case D is dominated by two pathways. The reaction #6 contributed 56% to the predicted HONO and was the largest contributor. The reaction #8 was the second

354

G. Sarwar et al.

largest contributor and contributed 32% to the predicted HONO. The HONO emissions and the homogeneous reactions (Eqs. (1)–(5)) contributed only 8% and 4% to the predicted HONO, respectively. HONO undergoes photolysis in the atmosphere to produce OH radicals. The additional HONO enhanced the average OH by 10% in the case D compared to that of the case A. The inclusion of the additional HONO sources increased the average O3 by 1.7 ppbv. The increased O3 is a contribution of additional VOC oxidation via enhanced OH as well as the photolysis of additional NO2 generated from the surface photolysis of adsorbed HNO3 (reaction #8).

4. Summary The results of this study suggest that heterogeneous reaction and surface photolysis of adsorbed HNO3 are important sources of HONO in the atmosphere. Most air quality models, however, do not currently account for these sources; thus, models tend to under-predict HONO by a large margin. The incorporation of these sources improves HONO predictions. The improved HONO predictions can increase OH as well as O3. Surface areas of leaves can be estimated using the LAI. Better estimates of the surface areas of buildings and other structures in urban environments are needed to improve HONO predictions via the heterogeneous reaction. However, such information is not readily available; efforts should be directed in determining such values. The surface photolysis of adsorbed HNO3 producing HONO and NO2 during the day is an emerging topic; many related issues are still unknown. However, it appears to provide the missing HONO source during the day without which model predictions remain under-predicted compared to observed data. Thus, this reaction should be further explored in laboratory as well as field studies before it can be confidently used in air quality models. Disclaimer The research presented here was performed under the Memorandum of Understanding between the U.S. Environmental Protection Agency (EPA) and U.S. Department of Commerce’s National Oceanic and Atmospheric Administration (NOAA) and under agreement number DW13921548. This work constitutes a contribution to the NOAA Air Quality Program. Although it has been reviewed by EPA and NOAA and approved for publication, it does not necessarily reflect their policies or views.

References Acker K, Möller D, Wieprecht W, Meixner FX, Bohn B, Gilge S, Plass-Dülmer C, Berresheim H (2006) Strong daytime production of OH from HNO2 at a rural mountain site, Geophys. Res. Lett., 33, L02809, doi:10.1029/2005GL024643.


355

Alicke B, Platt U, Stutz J (2002) Impact of nitrous acid photolysis on the total hydroxyl radical budget during the limitation of oxidant production/Pianura Padana Produzione di Ozono study in Milan, J. Geophys. Res., 107, doi:10.1029/2000JD000075. Alicke B, Geyer A, Hofzumahaus A, Holland F, Konrad S, Pätz HW, Schäfer J, Stutz J, Volz-Thomas A, Platt U (2003) OH formation by HONO photolysis during the BERLIOZ experiment, J. Geophys. Res., 108, 8247, doi:10.1029/ 2001JD000579. Aumont B, Chervier F, Laval S (2003) Contribution of HONO sources to the NOx/HOx/O3 chemistry in the polluted boundary layer, Atmos. Environ., 37, 487–498. Binkowski FS, Roselle SJ (2003) Community Multiscale Air Quality (CMAQ) model aerosol component, I: Model description, J. Geophys. Res., 108(D6): 4183, doi:10.1029/2001JD001409. Byun D, Schere KL (2006) Review of the governing equations, computational algorithms, and other components of the Models-3 Community Multiscale Air Quality (CMAQ) modeling system, Appl. Mech. Rev., 59, 51–77. Cai C, Hogrefe C, Schwab JJ, Katsafados P, Kallos G, Ren X, Brune WH, Zhou X, He Y, Demerjian KL (2007) Performance evaluation of an air quality forecast modeling system for a summer and winter season – Part II: HONO formation processes and their implications for HOx budgets, J. Geophys. Res., in review. Carter, WPL (2000) Implementation of the SAPRC-99 chemical mechanism into the Models-3 Framework, report to the United States Environmental Protection Agency. Available at http://www.cert.ucr.edu/~carter/absts.htm#s99mod3 Finlayson-Pitts BJ, Pitts JN Jr (2000) Chemistry of the Upper Lower Atmosphere, Theory, Experiments and Applications, Academic, San Diego, CA. Gery MW, Whitten GZ, Killus JP, Dodge MC (1989) A photochemical kinetics mechanism for urban and regional scale computer modeling. J. Geophys. Res., 94(D10), 12925–12956. Grell G, Dudhia J, Stauffer D (1994) A description of the fifth-generation Penn State/NCAR Mesoscale model (MM5), NCAR Tech. Note NCAR/TN-398+STR. Jones MR (2006) Ammonia deposition to semi-natural vegetation, PhD dissertation, University of Dundee, Scotland. Kirchstetter TW, Littlejohn D (1996) Measurements of nitrous acid in motor vehicle exhaust, Environ. Sci. Technol., 30, 2843–2849. Kurtenbach R, Becker KH, Gomes JAG, Kleffmann J, Lorzer JC, Spittler M, Wiesen P, Ackermann R, Geyer A, Platt U (2001) Investigations of emissions and heterogeneous formation of HONO in a road traffic tunnel, Atmos. Environ., 35, 3385–3394. Pleim JE (2007) A combined local and nonlocal closure model for the atmospheric boundary layer. part I: model description and testing, J. Appl. Meteor. Clim., 46, 1383–1395. Sarwar G, Luecken V, Yarwood G, Whitten G, Carter WPL (2007) Impact of an updated carbon bond mechanism on predictions from the Community Multiscale Air Quality (CMAQ) modeling system: preliminary assessment, J. Appl. Meteor. Clim., accepted.

356

G. Sarwar et al.

Svensson R, Ljungstrom E, Lindqvist O (1987) Kinetics of the reaction between nitrogen dioxide and water vapour, Atmos. Environ., 21, 1529–1539. Vogel B, Vogel H, Kleffmann J, Kurtenbach R (2003) Measured and simulated vertical profiles of nitrous acid – Part II, model simulations and indications for a photolytic source, Atmos. Environ., 37, 2957–2966. Winer AM, Biermann HW (1994) Long pathlength differential optical absorption spectroscopy (DOAS) measurements of gaseous HONO, NO2 and HCHO in the California South Coast Air Basin. Res. Chem. Intermed., 20, 423–445. Yarwood G, Rao S, Yocke M, Whitten G (2005) Updates to the Carbon Bond Chemical Mechanism: CB05, Final Report to the US EPA, RT-0400675, Available at http://www.camx.com/publ/pdfs/CB05_Final_Report_120805.pdf Zhou X, Civerolo K, Dai H, Huang G, Schwab JJ, Demerjian KL (2002) Summertime nitrous acid chemistry in the atmospheric boundary layer at a rural site in New York State, J. Geophys. Res., 107(D21), 4590, doi:10.1029/ 2001JD001539. Zhou X, Gao H, He Y, Huang G, Bertman SB, Civerolo K, Schwab J (2003) Nitric acid photolysis on surfaces in low-NOx environments: significant atmospheric implications, Geophys. Res. Lett., 30(23), 2217, doi:10.1029/2003GL018620.

Discussion A. Venkatram: How do you adsorb HNO3 in the model? How do you justify assuming that all the day deposited HNO3 is available for photolysis? Does the inclusion of HNO3 photolysis affects HNO3 in the atmosphere? G. Sarwar:

W. Gong: G. Sarwar:

In the model, it is assumed that HNO3 that is removed from the atmosphere via dry deposition gets adsorbed on surfaces. Adsorbed HNO3 can then undergo photolysis to produce HONO during the day. A fraction of the deposited HNO3, indeed, may not be available all day for photolysis. However, the detailed processes that form HONO in the atmosphere are still unknown. In this study, we explored the effects of heterogeneous reaction, homogeneous reactions, emissions, and surface photolysis of HNO3 on HONO. Since we used all available adsorbed HNO3 for photolysis, it does represent the upper limit of HONO formed via the surface photolysis process. The inclusion of surface photolysis of adsorbed HNO3 did not significantly affect HNO3 in the atmosphere. Did the changes involving HONO make any difference in model predictions of O3?


357

The changes involving HONO made some differences in ozone predictions. Predicted ozone increased by up to 2.4 ppbv. The diurnally averaged ozone increased by 1.4 ppbv. The increases started during the morning hours and continued throughout the day. However, the increases in predicted ozone were not adequate to explain the observed early morning rise. M. Brauer: How was HONO measured in the field campaign? Depending on the measurement approach there may be artifacts from other nitrogen species. G. Sarwar: Ambient HONO were measured using ion chromatography technique that has a detection limit of about 0.11 pptv and accuracy of about 5%.

4.4 Uncertainty in Air Quality Decision Making Bernard Fisher

Abstract This paper considers the decisions that should be made arising from a prediction using a model, taking into account the uncertainty associated with the prediction. Sometimes taking account of uncertainty leads to better decisions than just taking a decision on the basis of a single, central value. This is illustrated by examples taken from air quality models. Regulatory models need to be simple, leading to effective decision making, but their use implies accepting greater uncertainty. The paper describes an approach to this dilemma. Keywords Decision analysis, loss function, regret, uncertainty, ozone, nitrogen dioxide, limit value

1. Introduction to Decision Theory There has been considerable interest recently in the uncertainty associated with predictions made by air quality models (see Borrego et al., 2006, 2007). However less attention is paid as to how this leads to a decision. A decision should always be related to an action. Let the parameter ș describe the state of nature, with prior probability p(ș), which describes the uncertainty, and let d denote the action or decision, so that L(d,ș) is the regret or loss. Then the average expected risk or loss is

L (d )

L(d ,T ) p (T ) ¦ T

(1)

The preferred action (Morgan and Henrion, 1990) is the action d = dmin which minimises L(d ) and the risk, or loss, for this action is equal to Lmin

Lmin

min d

L(d ,T ) p(T ) ¦ L(d ¦ T T

min ,T ) p (T )

(2)

In air pollution one goes to great effort to evaluate the concentration ș and its probability distribution, which depends on the uncertainty in the parameters within the model. Often Monte Carlo simulation may be used. However on its own the Monte Carlo analysis does not lead to a decision (see discussion in Fisher and Willows, 2007). Sometimes measured values are used to calibrate the model, in an

C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

376

Uncertainty in Air Quality Decision Making

377

attempt to reduce the uncertainty, but not eliminating it. What is generally neglected is the problem of deciding on the form of L(d,ș). A decision can be made by ignoring the uncertainty in the value of ș. Let us assume that if one ignores the uncertainty, one takes a best estimate of ș, say the mean T . The decision ignoring uncertainty diu would then be based on taking the minimum of L(d ,T ) . The extra loss incurred by ignoring uncertainty is ( L(d ¦ T

iu ,T )

L(d min ,T )) p (T )

(3)

which from the definition of dmin is always positive or equal to zero. In Eqs. (1) and (2) it is assumed that there are a finite number of discrete states ș may occupy. ș could vary continuously over a range, in which case the summation would become an integral. The range over which ș varies could also be replaced by a number of discrete intervals. Later we will consider a simple example of the variation of ș. However the decision itself is always discrete. There may be alternatives, but one is always looking for the preferred option. The development has parallels with hypothesis testing, when the decision-maker in that case has to decide about acceptable errors. In the present case acceptable error is described by the form of L(d,ș).

2. Air Pollution Example: Limit Value Exceedences We shall consider a simple air pollution example, considering whether an air quality limit value, denoted by the concentration șc, is exceeded. There are two possible decisions d0 corresponding to a belief that șc is ‘not exceeded’ and decision d1 that the limit value is ‘exceeded’. There is a loss associated with taking decision d0 no exceedence, when there really is an exceedence, associated with health effects, given by

L ( d 0 ,T )

a0 (T T c ) when ș> șc and zero otherwise.

(4)

Similarly there is a loss associated with decision d1 exceedence, if there really is no exceedence, associated with unnecessary control measures, given by

L(d1 ,T )

a1 (T c T ) when ș< șc and zero otherwise.

(5)

If one ignores uncertainty and the model tells us that ș= T , taken to be greater than șc, then the decision is d1 (exceedence). It is assumed for convenience that the uncertainty is described by a rectangular probability distribution around the mean T , so that p (T )

1 , where T - w/2 d ș d T + w/2 and zero otherwise, w

(6)

B. Fisher

378

where w is the range of uncertainty. A loss occurs, when ignoring uncertainty, if the true concentration is less than the limit value, though the predicted concentration is above. In the above idealised case the loss would be L (d1 )

Tc

³T

w / 2

L(d1 ,T ) p (T )dT

a1 w (T T c ) 2 2 2

(7)

If one took account of uncertainty, the expected loss, for decision d1, would be as above, but for decision d0, would be

L (d0 )

T w/ 2

³T L(d ,T ) p(T )dT 0

c

a0 w (T T c ) 2 2 2

(8)

If L (d1 ) d L (d 0 ) then the preferred decision taking account of uncertainty would be d1, the same as the decision ignoring uncertainty. However if L (d1 ) > L (d 0 ) the preferred decision would be d0. Thus the expected value of including una w certainty is either 0, when L (d1 ) < L (d 0 ) , or L (d1 ) - L (d 0 ) = 1 (T T c ) 2 2 2 a0 w 2 (T T c ) , when L (d1 ) > L (d 0 ) . 2 2 If a0~a1 and T > șc then would generally expect that d1 was the preferred decision, since L (d1 ) - L (d 0 ) v - (T T c ) w a0 near the stricter indicative limit value of 20 ȝg m-3 and uncertainty should be included in an assessment of exceedence. For NO2 the health benefits of emission reductions set by the magnitude of a0 are less easy to evaluate. Generally one should expect that for stricter limit values, more attention should be paid to the uncertainty in the decision.

5. Regulatory vs Research Models An issue that arises for a regulator is when to use a simplified, regulatory model to make decisions, or when to invest in a more complex model, or in improved monitoring. One might decide to choose a simpler model with greater uncertainty, because it is more practical to apply. The decision about choosing one model in preference to another (a research model rather than a regulatory model) does not affect the loss function L(d,ș) associated with taking the wrong decision, but rather the form of the probability p(ș), through the range of uncertainty w. (The consequence of improving models or making more measurements is to reduce w.) The influence of w can be broadly described as follows. When the uncertainty is small (w>1), the expected value of considering uncertainty L (d1 ) – L (d 0 ) will be (roughly) proportional to (a1 a0 ) w 2 > 0 and uncertainty cannot be ignored. Generally the loss from making the wrong decision L (d ) increases, when the range of uncertainty w increases, except if uncertainty can be ignored. The effect of

B. Fisher

382

using a simplified, regulatory model is to increase w. If uncertainty can be ignored in both the research and the regulatory model, then there is no advantage in using the research model for decision making. It is important to consider both the difference between the best estimate and the limit value, T – șc, and the range of uncertainty w in model inter-comparisons. Two recent regulatory examples are briefly mentioned here, both involving the extraction of a simplified model from a complex system. Bounding curves can be used to describe NO2 in industrial plumes. NO2 plume bounding curves have been compared with chemical transport models with full chemistry, and real and idealised meteorology (Yu et al., 2007). At night, the upper bounding curve is given by [ NO2 ] [ NOx ]

[O3 ] ([ NOx ] [O3 ] / 4)

(9)

The ozone background (O3) used for the bounding curve is selected on the basis of results from the CMAQ model. In the day, the lower bounding curve is given by [ NO2 ] [ NOx ]

[O3 ] ([ NOx ] [O3 ])

(10)

Scatter plots based on CMAQ and NAME III model runs and bounding curves (see Yu et al., 2007 for details) for two representative episodes are generally well below the 1-hour mean NO2 limit value of 200 ȝg m-3, and so a regulatory model based on bounding curves could be applied and uncertainty ignored. In the other example the excess downwind ozone concentration averaged over a 24-hour period following the time when an air parcel passes over an industrial VOC source, describing the VOC chemical reactivity, was considered. Ozone formation downwind of an industrial VOC source, based on the chemical reactivity and estimates from the CMAQ model, were compared in a sensitivity experiment (see Yu et al., 2007, for details) carried out for a June 2001 episode over the UK. In this experiment an imaginary VOC source was added to the base case emissions and located in the Thames Estuary area. The imaginary VOC source, containing only ethene, was assumed to have a constant emission strength of 10 t per hour. The differences in the daily maximum 1-hour ozone mixing ratios and the 24-hour averaged O3 mixing ratios from the base case for each day of the simulation period were calculated and compared to estimates based on the chemical reactivity (Derwent and Nelson, 2002). A maximum 24-hour average ozone excess of 30 ȝg m-3 is attained on the 24 June 2001 using CMAQ, which is substantially lower than the 175 ȝg m-3 estimate based on the chemical reactivity. The two estimates straddle the ozone target value of 120 ȝg m-3. Hence in this case further investigation is needed before deciding between the two methods. Caution is also needed because the two calculations are not strictly directly comparable, apart from the rather uncharacteristically high emission rate.

Uncertainty in Air Quality Decision Making

383

6. Conclusions Decisions depending on the exceedence of air quality limit values should take into account the loss, if a wrong decision is made, and the decision can in some circumstances depend on the uncertainty. In other conditions ignoring uncertainty would lead to the same decision as including uncertainty. Thus there are situations when considering uncertainty is important and others when it is not. It depends on how close the best prediction is to the limit value, and this can have implications when applying regulatory models. Acknowledgments The views expressed in this paper are those of the author and are not necessarily those of the Environment Agency.

References Abbott J, Stedman V, Vincent V (2007) Annual audits of the contribution to pollutant concentrations from processes regulated by the Environment Agency: method, development SC030172/SR2, application of method SC030172/SR3. http://publications.environment agency.gov.uk/pdf/SCHO0307BMDS-e-e.pdf http://publications.environment agency.gov.uk/pdf/SCHO0307BMDP-e-e.pdf Borrego C, Miranda AI, Costa AM, Monteiro A, Ferreira J, Martins H, Tchepel O, Carvalho AC (2006) Uncertainties of models and monitoring, Air4EU project report M.2 http://www.air4eu.nl/reports_products.html Borrego C, Monteiro A, Costa AM, Miranda AI, Builtjes P, Kerschbaumer A, Lutz M (2007) Estimation of modelling uncertainty for air quality assessment: the AIR4EU Berlin case. Proceedings of the 6th International Conference on Urban Air Quality, Limassol, Cyprus, 27–29 March 2007. Derwent RG, Nelson N (2002) Development of a reactivity index for the control of the emissions of organic compounds, UK Environment Agency R&D Technical Report P4-105 RC8309. Fisher B, Willows R (2007) Uncertainty in air pollution models used for regulatory and risk assessment purposes. International Technical Meeting on Air Pollution Modelling and its Applications, Leipzig. Developments in Environmental Science, Volume 6, C. Borrego and E. Renner (Editors) pp. 392–401, Elsevier. Morgan MG, Henrion M (1990) A Guide to Dealing with Uncertainty in Quantitative Risk and Policy Analysis, Cambridge University Press, Cambridge, ISBN 0-521-36542-2. Tarrason L et al. (2005) Transboundary Acidification, Eutrophication and Ground Level Ozone in Europe. EMEP Status Report, Norwegian Meteorological Institute, EMEP Report 1/2005. Yu Y, Sokhi RS, Middleton DR (2007) Estimating contributions of Agencyregulated sources to secondary pollutants using CMAQ and NAME III models. Environment Agency Science Report: SC030171.

384

B. Fisher

Discussion D. Steyn: Do you think policy makers are equipped to behave rationally when presented with a complex forecast and uncertainties estimates? B. Fisher: It is unreasonable to expect the policy maker, or lawyer, to be able to make rational judgements when presented with complex forecasts with uncertainty estimates. The modeller or expert should present a summary of the evidence favouring options, using quantitative decision analysis techniques as appropriate, with the full working available for review if required. A. Venkatram: The modeler can provide information on model uncertainty. However, there is little guidance on constructing a realistic loss function. Can you provide an example? B. Fisher: At the ITM we tend to concentrate on obtaining improved estimates of the probability distribution of the model forecast. Therefore consideration of the loss function has only been hinted at in the paper. The loss function involves wider issues relating to the evidence supporting, say, an air quality limit value. It could be a rectangular function of concentration penalising any exceedence of the limit value. This is one idealised case. Given the uncertainty in the limit value one would expect the loss function to be smooth around the limit value and for it to increase with increasing concentrations exceeding the limit value. The slope of the loss function would depend on the estimated economic or health damage caused by concentrations exceeding the limit value.

5.2 Diagnostic Analysis of the Three-Dimensional Sulfur Distributions over the Eastern United States Using the CMAQ Model and Measurements from the ICARTT Field Experiment Rohit Mathur, Shawn Roselle, George Pouliot and Golam Sarwar

Abstract Previous comparisons of air quality modeling results from various forecast models with aircraft measurements of sulfate aerosol collected during the ICARTT field experiment indicated that models that included detailed treatment of gas- and aqueous-phase atmospheric sulfate formation, tended to overestimate airborne SO42- levels. To understand the three-dimensional distributions and fate of atmospheric SO42- and to diagnose the possible reasons for these over-predictions, we perform detailed analysis of modeled SO42- budgets over the eastern U.S. during the summer of 2004 using an instrumented version of the Community Multiscale Air Quality (CMAQ), namely the sulfur-tracking model. Two sets of three-dimensional model calculations are performed using different gas-phase chemical mechanisms: (1) the widely used CBM4 mechanism, and (2) the SAPRC mechanism.

Keywords Aerosols, chemical mechanisms, CMAQ, ICARTT, sulfate

1. Introduction The regional and global distribution of atmospheric sulfur compounds is of interest because of their important impacts on the environment and the climate. A large fraction of tropospheric sulfur oxides originate from SO2 which is emitted into the atmosphere as a result of anthropogenic combustion activities. The formation of atmospheric H2SO4 through rapid oxidation of emitted SO2 via both gas- and aqueous-phase pathways in the atmosphere has been widely studied as deposition of these compounds has led to acidification of lakes and forests. H2SO4 in the atmosphere can nucleate or condense on existing particles to produce SO42- aerosol which constitutes a relatively large fraction of the total ambient fine particulate matter (or PM2.5; particles with diameter less than 2.5 ȝm). Sulfate aerosols can further affect the climate by backscattering solar radiation and by changing the albedo of clouds (Charlson et al., 1992). Consequently, the accurate C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

496

Diagnostic Analysis of the Three-Dimensional Sulfur Distributions

497

characterization of the three-dimension distributions of tropospheric SO42- is of interest. In the eastern U.S., SO42- constitutes a large fraction of the airborne fine particulate matter. While measurements of surface-level SO42- and PM2.5 concentrations are available from a variety of surface networks, similar aloft measurements are only available during infrequent intensive field studies. During July and August 2004, airborne measurements of a variety of trace species were made from extensively instrumented aircrafts deployed as part of the International Consortium for Atmospheric Research on Transport and Transformation (ICARTT) field study (Fehsenfeld et al., 2006; Singh et al., 2006) and provide a unique opportunity to examine the performance of existing atmospheric chemistry-transport models in representing the processes that shape the three-dimensional distribution of airborne pollutants. Comparison of air quality modeling results from various forecast models with aircraft measurements of sulfate aerosol collected during the ICARTT field experiment indicated that models with detailed treatment of gas- and aqueous-phase atmospheric sulfate formation, tended to over-predict airborne SO42- levels (McKeen et al., 2007; Yu et al., 2008). To understand the three-dimensional distributions and fate of atmospheric SO42- and to diagnose the possible reasons for these over-predictions, in this study we perform detailed analysis of SO42- budgets over the eastern U.S. during the summer of 2004 simulated using the Community Multiscale Air Quality (CMAQ) and through comparisons of these model results with surface and aloft measurements.

2. The Modeling System The Community Multiscale Air Quality (CMAQ) model (Byun and Schere, 2006) driven with meteorological fields from the Eta model (Black, 1994) is used to examine the three-dimensional atmospheric chemical conditions during July– August 2004. Details on the linkage between the Eta and CMAQ models can be found in Otte et al. (2005). To be consistent with previous analysis, the input emissions data were constructed in a manner similar to that used in forecast mode. The emission inventories used in the model calculations discussed here were constructed to represent the 2004 period. NOx emissions from point sources were projected to 2004 (relative to a 2001 base inventory) using estimates derived from the annual energy outlook by the Department of Energy (http: //www.eia.doe.gov/oiaf/ aeo/index.html). Mobile emissions were estimated using the least-squares regression approximations to the MOBILE6 model following the approach of Pouliot and Pierce (2003). Area source emissions were based on the 2001 National Emissions Inventory, version 3 (http://www.epa.gov/ttn/chief), while BEIS3.12 (Pierce et al., 2002) was used to estimate the biogenic emission. CMAQ simulations were performed for the July 11–August 18, 2004 period. The aerosol module used in CMAQ is described in Binkowski and Roselle (2003) with updates described in Bhave et al. (2004). The aerosol distribution is modeled

498

R. Mathur et al.

as a superposition of three lognormal modes that correspond nominally to the Aitken (diameter (Dp) < 0.1 ȝm), accumulation (0.1 < Dp < 2.5 ȝm), and coarse (Dp > 2.5 ȝm) modes. The model results for PM2.5 concentrations are obtained by summing species concentrations over the first two modes. The horizontal model domain was discretized using grid cell sizes of 12 km. Twenty-two layers of variable thickness set on a sigma-type coordinate were used to resolve the vertical extent from the surface to 100 hPa. Daily 24-hour duration model simulations were conducted using the meteorological output from the 12 UTC Eta cycle.

3. Results and Discussion The Eta-CMAQ system was deployed during the summer of 2004 to provide developmental fine particulate matter forecasts over the eastern United States (Mathur et al., 2005). That configuration of the modeling system used the CBM4 chemical mechanism (Gery et al., 1989). Comparisons of modeled surface-level daily average PM2.5 compositional characteristics with corresponding measurements from the Speciated Trends Network (STN), indicated a slight high bias in predicted surface-level SO42-. Comparisons of predicted SO42- levels aloft with measurements from the NOAA-WP3 and NASA DC-8 aircraft, however indicated though the model captured the general characteristics of SO42- vertical distribution, it exhibited a systematic and often significant high bias aloft (see Figure 1).

Fig. 1 Comparison of average modeled SO42- vertical profiles with corresponding measurements from the (a) NOAA-WP3 and (b) NASA DC-8 aircrafts. The average profiles are constructed using modeled-observed pairs over all flights during July 15–August, 2004

Since the in-cloud aqueous phase oxidation of S(IV) to S(VI) constitutes a major fraction of the SO42- production, and since the CBM4 mechanism is known to be biased high in its predictions of H2O2 (which is the primary termination pathway for HO2 radicals in the mechanism), it can be hypothesized that in-cloud SO2 oxidation by H2O2 in this model formulation contributes to the noted SO42- overprediction. To further examine the modeled SO42- budgets, an instrumented version


499

of CMAQ was used to analyze sulfate production pathways. This model version, referred to as the CMAQ sulfur-tracking model, tracks sulfate production from gas-phase and aqueous-phase chemical reactions, as well as contributions from emissions and initial and boundary conditions. Five aqueous-phase reactions are individually tracked, including S(IV) oxidation by hydrogen peroxide (H2O2), ozone (O3), methyl-hydrogen peroxide (MHP), peroxyacetic acid (PAA), and catalysis by iron (Fe) and manganese (Mn). Contributions from each pathway are tracked in separate modeled species and are advected, diffused, processed through clouds, and depo-sited (both wet and dry). Figure 2c presents a breakdown of the various modeled SO42- production pathways along the NOAA-WP3 flight track on August 6, 2004 and illustrates that along this particular flight path, with the CBM4 mechanism approximately half of the simulated SO42- was produced through the aqueous oxidation pathways, amongst which the in-cloud H2O2 oxidation was the dominant contributor. Comparisons of predicted H2O2 concentrations with measurements from the NASA DC-8 (not shown) further confirmed the high bias in predicted H2O2 concentrations suggesting that H2O2 biases inherent in the CBM4 mechanism could be magnifying the role of modeled in-cloud SO2 oxidation.

Fig. 2 (a) Flight path of the NOAA WP3 on August 6, 2004; (b) Comparison of modeled and observed SO42- along flight path; Relative contribution of various pathways to modeled SO42- with the (c) CBM4 mechanism and (d) SAPRC mechanism

Additional model calculations with the more detailed and contemporary SAPRC mechanism were performed to further examine and quantify the uncertainties in the relative contributions of gaseous and aqueous oxidation pathways to SO42- formation. Comparisons of SO42- predictions with the CBM4 and SAPRC mechanisms for the August 6, 2004 WP3 flight are shown in Figure 2b; the contributions of the various pathways to the simulated SO42- using the two mechanisms for this case are shown in Figure 2c and d. Interestingly, model configurations with either mechanisms result in similar predictions of SO42- concentrations (Figure 2b), though there

500

R. Mathur et al.

are significant differences in the relative contributions of the gaseous and aqueous pathways (Figure. 2c and d). As expected, aqueous SO42- production from the H2O2 oxidation pathway was considerably reduced (by about half) in the SAPRC mechanism configuration due to improved prediction of ambient H2O2 (compared to NASA DC-8 measurements; not shown). However, the reductions in SO42- production from the aqueous pathway were compensated by a corresponding increase in contribution from the gas-phase OH oxidation pathway (Figure 2d). Figure 3 presents a comparison of modeled H2O2 and OH concentrations predicted using the CBM4 and SAPRC mechanisms along the same flight path. As expected, the H2O2 concentrations modeled with the SAPRC mechanism are significantly lower than those predicted using the CBM4 mechanism. However, the SAPRC OH concentrations are significantly larger than those modeled using the CBM4 mechanism. This combined with the availability of additional SO2 (from reduced aqueous-phase conversion) results in the noted increase in SO42- production from the gas-phase pathway in the SAPRC model configuration.

Fig. 3 Comparison of (a) H2O2 and (b) OH predictions using the CBM4 and SAPRC mechanisms along the August 6, 2004 NOAA WP3 flight path

The model’s ability to simulate the regionally averaged vertical profiles of sulfur species sampled by the NOAA WP3 aircraft campaigns during the study period is illustrated in Figure 4, which presents comparisons of the average composite vertical profiles for SO42- and the SO2/total-sulfur ratio. In constructing these profiles we averaged both the observed and the modeled data within each vertical model layer and over all the flights. These vertical profiles may thus be regarded as representing the mean conditions that occurred over the northeastern U.S. during the study period. As illustrated in the comparisons both the CBM4 and the SAPRC chemical mechanism model configurations over-predict the observed SO42- aloft.


501

Both configurations also significantly underestimate the aloft SO2/total sulfur ratio, suggesting that in both model configurations the S(IV) to S(VI) conversion occurs more efficiently than that suggested by the aircraft measurements. Detailed analysis of the SO42- production pathways in the model however indicate that even though the CBM4 and SAPRC configurations of the model yield similar levels of SO42aloft, the relative importance of the gas and aqueous production pathways is significantly different between the two chemical mechanisms and highlights the uncertainties in these mechanisms especially for aloft conditions.

Fig. 4 Comparisons modeled and observed regionally-averaged vertical profiles of SO42- and the SO2/total Sulfur ratio

In recent years regional air quality models are being applied to study and address increasingly complex multi-pollutant air pollution issues over time scales ranging from episodic to annual cycles. Majority of the chemical mechanisms currently in use in such models have been validated against smog chamber measurements designed to represent surface-level photochemistry. Our results indicate that the extrapolation of these mechanisms to represent chemistry associated with multi-day transport and free-tropospheric conditions need to be scrutinized in more detail. Acknowledgments The research presented here was performed under the Memorandum of Understanding between the U.S. Environmental Protection Agency (EPA) and the U.S. Department of Commerce’s National Oceanic and Atmospheric Administration (NOAA) and under agreement number DW13921548. This work constitutes a contribution to the NOAA Air Quality Program. Although it has been reviewed by EPA and NOAA and approved for publication, it does not necessarily reflect their policies or views.

502

R. Mathur et al.

References Bhave PV, Roselle SJ, Binkowski FS, Nolte CG, Yu SC, Gipson GL, Schere KL (2004) CMAQ aerosol module development Recent enhancements and future plans, Proc. of the 2004 Models-3/CMAQ Conference, October 18–20, 2004, Chapel Hill, NC, available at:http://www.cmascenter.org/conference/2004/ abstracts/Model%20Development/bhave_abstract.pdf Binkowski FS, Roselle SJ (2003) Models-3 Community Multi-scale Air Quality (CMAQ) model aerosol component:1. Model description, J. Geophys. Res., 108(D6), 4183, doi:10.1029/2001JD001409. Black T (1994) The new NMC mesoscale Eta Model: description and forecast examples. Weather Forecast., 9, 265–278. Byun DW, Schere KL (2006) Review of governing equations, computational algorithms, and other components of the models-3 Community Multiscale Air Quality (CMAQ) modeling system, Appl. Mech. Rev., 59, 51–77. Charlson et al. (1992) Climate forcing by anthropogenic aerosols, Science, 255, 423–430. Fehsenfeld FC et al. (2006) International Consortium for Atmospheric Research on Transport and Transformation (ICARTT): North America to Europe – Overview of the 2004 summer field study, J. Geophys. Res., 111, D23S01, doi:10.1029/ 2006JD007829. Gery MW, Whitten GZ, Killus JP, Dodge MC (1989) A photochemical kinetics mechanism for urban and regional scale computer modeling. J. Geophys. Res., 94, 12,925–12,956. Mathur R, Kang D, Yu S, Schere KL, Pleim J, Young J, Pouliot G, Otte T (2005) Particulate matter forecasts with the Eta-CMAQ modeling system: towards development of a real-time system and assessment of model performance, Proc. of the 2005 Models-3/CMAQ Conference, September 26–28, 2005, Chapel Hill, NC, Available at: http://www.cmascenter.org/conference/2005/ppt/1_11.pdf. McKeen et al. (2007) Evaluation of several PM2.5 forecast models using data collected during the ICARTT/NEAQS 2004 field study, J. Geophys. Res., 112, D10S20, doi:10.1029/2006JD007608. Otte TL et al. (2005) Linking the Eta Model with the Community Multiscale Air Quality (CMAQ) Modeling System to Build a National Air Quality Forecasting System, Weather Forecast., 20, 367–384. Pierce T, Geron C, Pouliot G, Kinnee E, Vukovich J (2002), Integration of the Biogenic Emission Inventory System (BEIS3) into the Community Multiscale Air Quality Modeling System, Preprints, 12th Joint Conf. on the Apps. of Air Pollut. Meteor. with the A&WMA, Amer. Meteor. Soc., Norfolk, VA, 20–24 May 2002, J85–J86. Pouliot G, Pierce T (2003) Emissions processing for an air quality forecasting model, 12th Intl. Conf. on Emission Inventories, San Diego, CA, April 28–May 1, 2003.


503

Singh HB, Brune WH, Crawford JH, Jacob DJ, Russell PB (2006) Overview of the summer 2004 Intercontinental Chemical Transport Experiment – North America (INTEX-A), J. Geophys. Res., 111, D24S01, doi:10.1029/2006JD007905. Yu S, Mathur R, Schere K, Kang D, Pleim J, Young J, Tong D, Pouliot G, McKeen SA, Rao ST (2008) Evaluation of real-time PM2.5 forecasts and process analysis for PM2.5 formation over the eastern U.S. using the Eta-CMAQ forecast model during the 2004 ICARTT study, J. Geophys. Res., 113, D06204, doi:10.1029/2007/ JD009226.

Discussion D. Steyn: You mention hourly versus weekly averaging time as a possible source of error. Is it possible that differences are due to move than the statistics of hourly versus longer averaging but rather reflect different chemical processes? R. Mathur: In the comparisons presented, the surface measurements represent a 24-hour average where as the aircraft measurements were instantaneous. Certainly, averaging of data over longer time periods removes the inherent variability and can result in better agreement between the model and the measured values. You are correct that the averaging process implicitly represents different processes, with larger averaging times representing cumulative effects of several processes. For instance, in this case, surface-level 24-hour average SO42- concentrations at a location represent the cumulative affects both gas- and aqueous phase sulphate formation and well as transport and turbulent mixing processes. Ideally, to have greatest confidence in model results it would be desirable that the model captures temporal variability at all resolvable scales. B. Fuher: Are you able to put this very interesting model validation in to context? For example have there been other studies, either routine monitoring or field experiments, which demonstrate similar over prediction of sulphate? R. Mathur: Other than the model inter-comparison of McKeen et al. (J. Geophys. Res., 112, D10S20, doi:10.1029/2006JD007608, 2007) and our previous analysis with ICARTT data (Yu et al., J. Geophys. Res., in press), we are not aware of any previous studies that have identified similar systematic over-predictions in sulphate or have attempted to diagnose the reasons for the noted over-estimation. It should be noted that the summer of 2004 (period of the ICARTT study) was characterized by unusually wet conditions with widespread cloudiness. Consequently, the noted overestimation of sulphate by the models could not only be related to representation of

504

R. Mathur et al.

chemical processing of S(IV) to S(VI) but also to representation of the clouds and their impacts on sulphate production. Our analyses in this study has focussed on diagnosing aspects related to chemical processing; additional work underway is investigating the possible effects related to representation of clouds and associated missing processes on ambient sulphate levels.

5.5 Formation of Secondary Inorganic Aerosols by High Ammonia Emissions Simulated by LM/MUSCAT Eberhard Renner and Ralf Wolke

Abstract Ammonia (NH3) is the most abundant gaseous base and responsible for neutralizing a large fraction of acidic gases promoting the formation of atmospheric particles. Therefore, the contribution of ammonia to the formation of secondary particles (PM2.5 and PM10) in a regional scale is examined. The aerosols result from SO2 and NOx via sulfuric and nitric acid formation in the gas- and the liquid-phase and following subsequent reactions with ammonia. A period in May and a period in August/September 2006 were simulated by a nested application of the model system LM-MUSCAT.

Keywords Ammonia, ammonium sulfate, ammonium nitrate, chemical transport

1. Introduction Atmospheric particulate matter (PM) in ambient air has been associated with human health effects (Dockery et al., 1993; Pope et al., 1995; Brunekreef, 1997; Hoek et al., 2002). Since particles with aerodynamic diameters smaller than 10 Pm (PM10) are able to pass the larynx, the European air quality standards prescribe a limit for PM10. Currently, the focus of environmental sciences and politics includes besides primary particulate emissions also the formation and growth processes of secondary particles (e.g., Andreani-Aksoyoglu et al., 2004). In order to observe the fate of high ammonia emissions especially from agriculture and livestock husbandry in terms of particulate matter, the formation of secondary inorganic particles (PM2.5 and PM10) in a regional scale is investigated. The dominant contribution to the particle mass is given by heterogeneous condensation of gaseous compounds on pre-existing aerosol particles. Especially the generation of sulfuric and nitric acids by the precursor species SO2 and NOx and their reactions with ammonia leading to ammonium sulfate and ammonium nitrate are relevant. The aerosol-forming, photolytic and gas-chemical reactions together with the transport and diffusion processes are integral part of the chemistry-transport model MUSCAT, which was on-line coupled to the Local Model (LM) of German Weather Service. A regime for long-term simulations was tested. The horizontal C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

522

Formation of Secondary Inorganic Aerosols by High Ammonia Emissions Simulated

523

resolution of the outer region (Europe) was chosen to be 15 km and of the inner nested region (Germany and bordering regions of central Europe) 7.5 km. The present study estimates, to which extent the gaseous and particulate emissions of ammonia, in a region of high ammonia emissions, especially from agriculture and livestock husbandry, contribute to the formation of secondary particulate matter. For this purpose, long-term real-weather simulations were performed for a spring and a late summer period.

2. Model Description 2.1. Meteorological-chemical model system LM-MUSCAT The three-dimensional non-hydrostatic model LM (Local Model) is the operational meteorological model of the German Weather Service (Doms and Schättler, 1999). The governing equations are formulated in terrain-following coordinates. Radiative processes as well as cloud microphysics and turbulence are treated in a parameterized manner. The model technique is capable of self-nesting and fourdimensional data assimilation, where the model GME (Global Model of the Earth) by German Weather Service serves as the global driver model. In that way, the horizontal grid spacing is reduced from 60 km (GME) down to 7.5 km for the innermost region. In both models 50 layers are used for the vertical discretization. The three-dimensional chemistry transport model MUSCAT (Multi-Scale Atmospheric Transport Model) performs the transformation of the gaseous and particulate species together with the transport processes advection, turbulent diffusion, dry/wet deposition, and sedimentation (Knoth and Wolke, 1998; Wolke and Knoth, 2000). Due to the online coupling with LM, these calculations are involved in the meteorological model, thus exploiting the current atmospheric conditions. Regardless, the implicit-explicit procedures of the time integration scheme of MUSCAT are widely independent of the meteorological model and allow for autonomous time steps and different horizontal grid resolutions in selected areas. For this purpose, the three-dimensional LM fields of wind, temperature, humidity, density, pressure, exchange coefficients, etc., are interpolated temporally and spatially without flux divergences. The chemical part of MUSCAT contains the gas phase mechanism RACM (Stockwell et al., 1997) considering 76 reactive gas species, 217 chemical and 22 photolytic reactions. Particle formation and appropriate interactions with the gas phase are included. The radiation activity and the biogenic emission are also calculated in MUSCAT, where the informations on cloud cover, temperature and other meteorological parameters are taken from LM. Anthropogenic emissions are accounted for by point, line, and area sources in different layers above the surface. The annual emission intensities are disaggregated in time and space.

E. Renner, R. Wolke

524

2.2. Aerosol model The aerosol model, used for these simulations, is mass-based and quiet similar to that of the EMEP Eulerian model (Simpson et al., 2003). The formation of ammonium sulfate and ammonium nitrate is calculated by an equilibrium approach. The dominant contribution to mass accretion is treated as heterogeneous condensation of gaseous compounds on pre-existing aerosols. Here, the particulate matter is formed by the reactions between ammonia and sulfuric or nitric acid, which are generated from the gas phase precursor species SO2 and NOx. In the following, the relevant sources and sinks of particulate matter are specified in detail. Finally, the different aerosol species are subsumed to derive PM10. 1. Primary particulate matter (PPM): PPM is that part of PM, which is directly emitted by the various point and area sources (industry, traffic, etc.). PPM10 is the sum of PPM2.5 (particles with aerodynamic diameters smaller than 2.5 ȝm) and PPM10–2.5 (particles with aerodynamic diameters between 2.5 and 10 ȝm). 2. Formation of ammonium sulfate: The precursor gas SO2 can be oxidized in the gas phase by OH radicals to form sulfuric acid: SO2 + OH + O2 + H2O ĺ ... ĺ H2SO4 + HO2

(1)

In addition to the gas phase reaction an oxidation pathway in clouds with a simple first order reaction constant Rk is considered. Rk is calculated as a function of relative humidity (%) and cloud cover (H): Rk = 8,3e-5(1 + 2H) (min–1) for RH < 90% Rk = 8,3e-5(1 + 2H)>1,0 + 0,1(RH–90.0@ (min–1) for RH t 90%. This parameterisation is described in Schaap et al. (2004). It enhances the oxidation rate under cool and humid conditions. If ammonia is available, the sulfuric acid reacts fast and irreversible with ammonia resulting in ammonium sulfate on the particle surface: H2SO4 + 2 NH3 ĺ {(NH4)2SO4}a } ĺ {(NH4)1.5SO4}a H2SO4 + NH3 ĺ {NH4HSO4}a

(2)

Because of the different stoichiometric weights in the products, a mean stoichiometric value of 1.5 is used for ammonium sulfate (Ackermann, 1997).


525

In the investigated case 5% of SO2-emisisions are emitted primarily as ‘free’ sulfate ions (Simpson et al., 2003). This is a further effective pathway of the formation of ammonium sulfate by the direct reaction of this sulfate with ammonia in the condensed phase. The reaction is similar to Eq. (2). Any “excess” of sulfuric acid or sulfate ions, which could not be neutralized by ammonia, is assumed to exist in the condensed phase and therefore to participate in the formation of particle mass, too. 3. Formation of ammonium nitrate: In general, the atmospheric ammonia is neutralized first of all by the sulfuric acid (see above). The remaining ammonia can then combine with nitric acid to form ammonium nitrate. The nitric acid is mainly yielded by the following reaction chains in the gas phase and partly in the liquid phase: Day:

NO2 + OH ĺ HNO3

Night: NO2 ĺ .. ( NO3 , N2O5 ) ... ĺ HNO3

(3) (4)

In contrast to the fast and irreversible formation of ammonium sulfate, ammonium nitrate is a semi-volatile compound which forms on the aerosol surface in equilibrium with its gaseous precursors nitric acid and ammonia (Mozurkewich, 1993; Nenes et al., 1998; Zhang et al., 2000): HNO3 + NH3 ļ {NH4NO3}a

(5)

This equilibrium is dependent on the ambient atmospheric temperature and humidity. It is shifting to the gas phase under dry and warm conditions. 4. Aerosol sinks: The dry deposition velocity is determined by means of a resistance approach accounting for the atmospheric turbulence state, the kinetic viscosity, and the gravitational settling in dependence on the particle size. An essential role in removing particulate matter from the atmosphere is played by wet deposition, which is distinguished between in-cloud and sub-cloud aerosol scavenging. These processes are parameterized in dependence on the scavenging and collection efficiency (Tsyro and Erdman, 2000). 5. Total PM10: The total particulate mass PM10 is composed of the various primary and secondary constituents described in the foregoing points: PM10 = PPM2.5 + PPM10–2.5 + {(NH4)1.5SO4}a + {SO42-}a + {NH4NO3}a

(6)

526

E. Renner, R. Wolke

3. Emissions The inventories of the anthropogenic emissions are based on the EMEP/CORINAIR data (see CORINAIR), improved by TNO (Builtjes et al., 2003), using the SNAP-codes (Selected Nomenclature of Air Pollution) for characterizing the different source types (e.g., energy transformation, industrial combustion, road transport, agriculture). The considered list of chemical species includes the main pollutants CO, NOx, SO2, NH3, PPM2.5, PPM10–2.5, methane, and non-methane volatile organic compounds (NMVOC). The temporal variation of the emissions is represented by time functions, which break down the annual totals into hourly values in dependence on the source category, and the month and day of emissions. The differenttiation of the NMVOC follows the VOC-split by Winiwarter and Züger (1996). The biogenic VOC and NO emissions are calculated within the chemistrytransport model MUSCAT for each time step and grid cell in dependence on the land-use type, temperature and radiation. Consequently, the meteorological situation has a direct influence on the emission amount. There are numerous volatile organic compounds, which are emitted from plants. In this study, only the emissions of isoprene and terpenes by forests are taken into account. According to the parameterization of Guenther et al. (1993), the emission rates are formulated as products of plant specific standard emission rates with a temperature and/or a light dependent function. Agriculturally used areas may be significant sources of NO caused by microbiological processes. The biogenic NO emissions are parameterized using empirical relationships, which depend on the land use type, season, and soil temperature. The functional dependence follows Williams et al. (1992) and Stohl et al. (1996).

4. Results In order to identify the influence of different meteorological situations on the formation of secondary particulate matter the simulated concentrations for two days in May are presented. The discussion is focussed exemplarily on the 7th of May in 2006, a sunny day with moderate winds from East and the 26th of May, a day with westerly winds. The figures show the concentration distributions of the relevant gaseous and particulate species only in the N1-region in the lowest modelling layer. In Figures 1 and 2 the concentrations of the gaseous precursors SO2, NO2 and NH3, the secondarily formed ammonium sulfate and ammonium nitrate and PM10 are displayed. Additionally the temperature and the relative humidity are presented. At the 7th of May the PM10 concentrations in Germany are about 10 ȝg/m3. The ammonium sulfate concentrations are about 5 ȝg/m3 whereas no ammonium nitrate was formatted at this time in this region. The reason for this is the circumstance that we have at this situation extreme dry and hot air in this region. So, the equilibrium between the gaseous precursors nitric acid and ammonia and ammonium nitrate as a semi-volatile compound, which is formed on the aerosol surface, is shifted to the gas phase. At the 26th of May we found a different situation. At the westerly winds we have in the northern part of Germany air masses, which are cool and wet. As a


527

result we see comparable ammonium sulfate concentrations of about 5 ȝg/m3. Contrary to the former situation we have additionally ammonium nitrate concentrations of about 10 ȝg/m3 at this time in this region. The issue of the simulations was finally to get the additional PM10 burden caused by the NH3 emissions in this region and the later formation of secondary ammonium sulfate and ammonium nitrate. One can summery that at situations with low PM10 concentrations 50–70% of the PM10 concentrations within the determined region can be secondary formed. By situations with higher and very high PM10 burden this fraction will decline more and more. Further studies have shown (not shown here), that the availability of SO2 is the limited factor of the formation of secondary aerosol mass.

Fig. 1 Concentrations, relative humidity and temperature at 7th of May

Fig. 2 Concentrations, relative humidity temperature at 26th of May

528

E. Renner, R. Wolke

References Ackermann I (1997) MADE: Entwicklung und Anwendung eines AerosolDynamikmodells für dreidimensionale Chemie-Transport-Simulationen in der Troposphäre. Mitteilungen aus dem Institut für Geophysik und Meteorologie der Universität Köln, Heft 115, 1–153. Andreani-Aksoyoglu S, Prévôt ASH, Baltensperger U, Keller J, Dommen J (2004) Modeling of formation and distribution of secondary aerosols in the Milan area (Italy). Journal of Geophysical Research 109 (D5), D05306, doi:10.1029/ 2003JD004231. Builtjes PJH, van Loon M, Schaap M, Teeuwisse S, Visschediijk AJH, Bloos JP, (2003) TNO-report R 2003/166. Brunekreef B (1997) Air pollution and life expectancy: is there a relation? Journal of Occupational and Environmental Medicine 54, 781–784. CORINAIR (Co-ordinated Information on the Environment in the European Community – AIR). Web address: http://reports.eea.eu.int/EMEPCORINAIR3/en Dockery DW, Pope CA, Xu X, Spengler JD, Ware JH, Fay ME, Ferris BG, Speizer FE (1993) An association between air pollution and mortality in six U.S. cities. New England Journal of Medicine 329, 1753–1759. Doms G, Schättler U (1999) The Nonhydrostatic Limited-Area Model LM (LokalModell) of DWD: Part I: Scientific Documentation (Version LM-F90 1.35). German Weather Service, Offenbach. Guenther AB, Zimmerman PR, Harley PC, Monson RK, Fall R (1993) Isoprene and monoterpene emission rate variability: model evaluations and sensitivity analyses. Journal of Geophysical Research 98 (D7), 12609–12617. Hoek B, Brunekreef G, Goldbohm S, Fischer P, van den Brandt PA (2002) Association between mortality and indicators of traffic-related air pollution in the Netherlands: a cohort study, The Lancet 360, 1203–1209. Knoth O, Wolke R (1998) An explicit-implicit numerical approach for atmospheric chemistry-transport modelling. Atmospheric Environment 32, 1785–1797. Mozurkewich M (1993) The dissociation constant of ammonium nitrate and its dependence on temperature, relative humidity and particle size. Atmospheric Environment 27 A, 261–270. Nenes A, Pilinis C, Pandis SN (1998) Isorropia: a new thermodynamic model for multiphase multicomponent inorganic aerosols. Aquatic Geochemistry 4, 123– 152. Pope CA, Thun MJ, Namboodiri MM, Dockery DW, Evans JS, Speizer FE, Heath CW (1995) Particulate air pollution as a predictor of mortality in a prospective study of U.S. adults. American Journal of Respiratory and Critical Care Medicine 151, 669–674. Schaap M, van Loon M, ten Brink HM, Dentener FJ, Builtjes PJH (2004) Secondary inorganic aerosol simulations for Europe with special attention to nitrate. Atmospheric Chemistry and Physics, 4, 857–874. Simpson D, Fagerli H, Jonson JE, Tsyro S, Wind P (2003) Transboundary Acidification, Eutrophication and Ground Level Ozone in Europe. PART I,


529

Unified EMEP Model Description. EMEP/MSC-W: EMEP Status Report 2003, ISSN 0806-4520. Stockwell WR, Kirchner F, Kuhn M, Seefeld S (1997) A new mechanism for regional atmospheric chemistry modeling. Journal of Geophysical Research 102 (D22), 25847–25879. Stohl A, Williams E, Wotawa G, Kromp-Kolb H (1996) A European inventory of soil nitric oxide emissions and the effect of these emissions on the photochemical formation of ozone. Atmospheric Environment 30, 3741–3755. Tsyro S, Erdman L (2000) Parameterization of aerosol deposition processes in EMEP MSC-E and MSC-W transport models. EMEP/MSC-E & MCS-W Note 7/00, Norwegian Meteorological Institute, Oslo. Williams EJ, Guenther A, Fehsenfeld FC (1992) An inventory of nitric oxide emissions from soils in the United States. Journal of Geophysical Research 97 (D7), 7511–7519. Winiwarter W, Züger J (1996) Pannonisches Ozon Projekt, Teilbericht Emissionen. Endbericht, Seibersdorf Report OEFZS-A-3817. Wolke R, Knoth O (2000) Implicit-explicit Runge-Kutta methods applied to atmospheric chemistry-transport modelling. Environmental Modelling and Software 15, 711–719. Zhang Y, Seigneur C, Seinfeld JH, Jacobsen M, Clegg SL, Binkowski FS (2000) A comparative review of inorganic aerosol thermodynamic equilibrium modules: differences, and their likely causes. Atmospheric Environment 34, 117–137.

5.3 Heterogeneous Chemical Processes and Their Role on Particulate Matter Formation in the Mediterranean Region Marina Astitha, George Kallos, Petros Katsafados and Elias Mavromatidis

Abstract The impact of particulate matter on air quality and the environment is an important subject for areas like the Greater Mediterranean Region, mostly due to the coexistence of major anthropogenic and natural sources. Such coexistence can create air quality conditions that exceed the imposed air quality limit values. Particulate matter formation and the factors enhancing or reducing such formation in the Mediterranean Region will be the primary focus of the work presented herein. Natural particulate matter appears mainly in the form of desert dust, sea salt and pollen among others and anthropogenic particulate matter appears as particulate sulfate and nitrate. The processes affecting the formation of new types of aerosols are based on the heterogeneous uptake of gases onto dust particles. New model development will be presented referring to the implementation of sea salt production and heterogeneous chemical processes leading to new aerosol formation in the photochemical model CAMx. Results from these simulations showed reasonable agreement with the available measurements. These results also revealed interesting effects of the coexistence of natural and anthropogenic particulate matter concerning the direct and indirect impacts on air quality and the environment.

Keywords Heterogeneous processes, modelling, Mediterranean, particulate matter

1. Introduction The direct and indirect impacts of aerosols on the environment and climate are the subject of scientific investigation during the last years and have gained the interest of the public since the IPCC 3rd Assessment Report (Penner et al., 2001). Focusing on the Mediterranean Region, the coexistence of natural and anthropogenic type of sources is a factor enhancing the air quality degradation without the capability to ensure the exact causes of such degradation. In this work, natural particulate matter appears in the form of desert dust and sea salt and anthropogenic particulate matter appears as particulate sulfate and nitrate. This categorization is due to the fact that Sahara is the desert responsible for many severe dust outbreaks that influence the C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

505

506

M. Astitha et al.

area and the Mediterranean Sea is considered an important source for sea salt and sulfur. The first step towards the identification of the feedback effects of excessive dust load on air quality was done by studying the direct shading effect of dust particles on photochemical processes. This was accomplished by implementing dust optical depth in the calculation of photolysis rates (Astitha et al., 2006) during severe dust transport episodes. The shading effect of dust particles caused a decrease in ozone and sulfate concentration at the surface, considered of minor importance (ozone and particulate sulfate decrease was maximum 2–3% of the initial concentration). The second step is the investigation of the processes affecting the formation of new types of aerosols based on the heterogeneous uptake of gases onto dust particles (Levin et al., 2005; Alastuey et al., 2005). These processes require extensive investigation since they are complex and uncertain and model development is needed to properly include the relevant physicochemical processes. For this purpose, advanced atmospheric and photochemical models are implemented with the aid of air pollutant measurements from stations in the region. The models used are the SKIRON/Eta atmospheric modeling system with the implementation of the dust module and the CAMx photochemical model. New model development will be presented referring to the implementation of sea salt production and heterogeneous chemical processes leading to new aerosol formation in the photochemical model.

2. Model Development A short description of the modeling systems used for performing simulations is provided in this section: The SKIRON/ETA is a modeling system developed at the University of Athens from the Atmospheric Modeling and Weather Forecasting Group (Kallos, 1997; Kallos et al., 2006; Nickovic et al., 2001). It has enhanced capabilities with the unique one to simulate the dust cycle (uptake, transport, deposition). The CAMx model (Environ, 2006) is an Eulerian photochemical model that allows for integrated assessment of air-pollution over many scales ranging from urban to super-regional (http://www.camx.com).

2.1. Particulate matter of natural origin Natural species like sea salt and desert dust are implemented inside the photochemical model, allowing the interaction between species of different origin (natural, anthropogenic). Desert dust surface fluxes were derived from SKIRON/Eta model in four size sections (centered diameters of 1.5, 12, 36, 76 ȝm). Only the first two sizes were implemented in the emissions processor, producing emission fields of crustal material to be imported in the photochemical model. Sea salt production was directly developed inside the air quality model. The production of sea salt was based on the work of Gong (2003), ǽhang et al. (2005)

Heterogeneous Chemical Processes and their Role on Particulate Matter Formation

507

for the open ocean function and the work of de Leeuw et al. (2000) for the surf zone function. The coding implementation was similar to that of CMAQ model (Shankar et al., 2005 and personal communication). Sea salt was speciated into sodium, chloride and sulfate aerosol using the factors fNA = 0.3856, fCl = 0.5389, fSO4 = 0.0755 respectively.

2.2. Heterogeneous uptake of gases on dust particles Air pollutants in the atmosphere can be categorized as natural or anthropogenic depending on their origin. Depending on their formation mechanism, they can also be categorized as of 1st generation (primary) and 2nd generation (secondary). In this work we apply a new terminology considering the 3rd generation species as the ones formed by the synthesis of the 1st and the 2nd generation. The schematic diagram in Figure 1 explains the synthesis described above. The synthesis of the 3rd generation pollutants is driven by the heterogeneous reactions occurring in the surface of a wet particle (i.e. desert dust for this work). A first-order kinetic relationship is used to represent the change in gases and aerosols concentration due to uptake on dust particles as shown in Eq. (1):

dC

g

k g ,r C

dt

g

(1)

where Cg is the gas concentration and kg,r (s-1) is the removal rate of species g to a particle surface with radius r. Following the formulation by Fuchs and Sutugin (1970) and Saylor (1997) the removal rate is calculated as shown in Eq. (2): k g ,r

4 S D g rN r 1 f (K n, J )K n

1.333 0.71K n1 4(1 J ) 1 K n1 3J

f (Kn , J )

(2)

where Kn is the Knudsen number, r the radius of the particle, Dg the gas-phase molecular diffusion coefficient in air, Nr is the particle number density and Ȗ is the uptake coefficient (different for each gas). The reactions chosen for this work are the following: dust

1.

SO

2.

NO 2 ( gas ) o DNO 3 ( aerosol )

3.

HNO3 ( gas ) o DNO3 b( aerosol )

2

( gas ) o DSO

4

( aerosol

)

Ȗ = 1.0*10-4

dust

Ȗ = 1.0*10-4

dust

Ȗ = 0.01

dust

4.

O3 ( gas ) o O2

Ȗ = 5.0*10-5

M. Astitha et al.

508

ANTHROPOGENIC EMISSIONS

Fig. 1 Schematic diagram of the synthesis between primary and secondary pollutants leading to new particle formation (3rd generation pollutants)

GASES

PARTICULATE MATTER

SO2, NO, NO2, VOC CO, NH3, CO2

PM2.5 PM10

2nd GENERATION POLLUTANTS SECONDARY POLLUTANTS O3, SO4, NO3, NH4

NATURAL EMISSIONS

AEROSOLS 1st GENERATION POLLUTANTS

3rd GENERATION POLLUTANTS

DUST SEA SALT (NaCl)

PARTICLES COVERED WITH GASES OR AEROSOLS

The reactions are considered irreversible concerning the uptake of each gas on aerosol surface and the dust particles are assumed to be of spherical shape.

3. Results and Discussion The new processes of sea salt production and heterogeneous uptake on dust particles are implemented in CAMx air quality model, providing the ability to study the 3rd generation aerosol formation. Two periods were chosen for simulation: August 1–20, 2001 and April 1–20, 2003. The main reason for choosing the month of April is that severe desert dust episodes occur in the Mediterranean Region during the transient seasons and photochemical processes are considerable, while August is a month where photochemical processes are at their peak along with the transport of desert dust in the region. The two long-term simulations revealed the ability of the model to describe the aforementioned processes as well as the shortcomings, uncertainties and future work followed by the new development. Meteorological fields and desert dust fluxes were derived from SKIRON/Eta modeling system for the periods mentioned above. Particulate matter is treated using the multi-sectional approach in CAMx model. Three size sections are selected, each having diameters in ranges: 0.03–0.1, 0.1–2.5, 2.5–10.0 ȝm. Results will be presented for the generation and the transport of the new aerosols formed on dust particles as described by chemical reaction 1 in the previous section. CRST_x is the dust particle, PSO4_x is the particulate sulfate of anthropogenic origin with the contribution from sea salt and DSO4_x is the particulate sulfate formed on dust particle via heterogeneous reaction 1, where x denotes the size section (1–3). Figure 2 presents the evaluation of the sea salt production, performed by comparing time series (3h average) of bulk sodium aerosol measurements from a coastal station (Finokalia station, ECPL chemistry, professor N. Mihalopoulos) and model output for the period 1–11 August 2001. The scatter plot diagram shown in Figure 3a refers to the same measuring station at Finokalia, for total sea salt concentration and the sum of the modeled sodium and chloride concentration. Figure 3b shows the comparison of model output with daily measurements from several stations in


509

Spain (X. Querol, 2007, personal communication) for coarse sodium aerosol. The model seems to slightly underpredict the sodium aerosol, and that can be attributed to the horizontal resolution (dx = 0.235q, dy = 0.18q) which is not adequate to resolve correctly the coastal zone. Another reason can be the fact that, the wind direction in the coastal zone is not taken into consideration in the production of sea salt, so the emissions and consequently the concentrations are rather smoothed as shown in Figure 2. SODIUM AEROSOL (ȝg/m 3) 8,000

ȝg/m 3

7,000 6,000 5,000 4,000 3,000 2,000 1,000

11/8/2001 6:00-9:00

11/8/2001 15:00-18:00

10/8/2001 21:00-0:00

10/8/2001 3:00-6:00

9/8/2001 18:00-21:00

10/8/2001 12:00-15:00

9/8/2001 0:00-3:00

9/8/2001 9:00-12:00

8/8/2001 6:00-9:00

7/8/2001 21:00-0:00

8/8/2001 15:20-18:00

7/8/2001 3:00-6:00

7/8/2001 12:25-15:00

6/8/2001 0:00-3:00

6/8/2001 9:00-12:00

6/8/2001 18:00-21:00

5/8/2001 6:00-9:00

5/8/2001 15:00-18:00

4/8/2001 3:00-6:00

4/8/2001 21:00-0:00

3/8/2001 18:00-21:00

4/8/2001 12:00-15:30

3/8/2001 3:00-6:00

3/8/2001 10:30-12:00

2/8/2001 9:00-12:00

2/8/2001 18:00-21:00

2/8/2001 00:00 - 3:00

1/8/2001 9:00 - 10:30

1/8/2001 15:00 - 18:00

1/8/2001 00:00 - 3:00

0,000

date (UTC) CAMx-NAtotal

MEASUREMENTS-bulk Na

Fig. 2 Comparison of sodium aerosol bulk measurements at Finokalia station (Crete, Greece) with CAMx model output for 1–11 August 2001 (3h average concentrations) a

b

MEASURED vs MODELED VALUES OF SODIUM CHLORIDE AEROSOL (ȝg/m3) AT FINOKALIA 2-11 AUGUST 2001

Sodium Aerosol(PM10) Obs vs CAMx for sites in SPAIN 1-13 August 2001 (Daily average)

3,00

7,00 MODEL (NACL=NA+CL)

y = 0,623x - 0,1676 R2 = 0,6034 r=0,77

2,50

3

8,00

MODEL OUTPUTS (ȝg/m )

9,00

6,00 5,00 4,00 3,00 2,00

y = 0,6185x + 0,5674

1,00

R = 0,7148 r=0,845

2,00 1,50 1,00 0,50

2

0,00 0,00

2,00

4,00

6,00

Measurem ents (NACL=1,8NA)

8,00

10,00

0,00 0,00

0,50

1,00

1,50

2,00

2,50

3,00

3

OBSERVATIONS (ȝg/m )

Fig. 3 (a) Comparison of sodium chloride measurements at Finokalia station (3h average, Crete, Greece) with CAMx model output, (b) Comparison of sodium aerosol (daily average PM10) from several stations in SPAIN for August 2001

During the second simulation for April 2003, the model performance for sea salt production was equally reliable. Figure 4a presents the comparison of modeled versus observed sodium and chloride aerosols (PM10) for the period 1–18 April 2003. The correlation coefficient reached high values coinciding with the August simulation. In Figure 4b the comparison of sulfate aerosol measurements from Finokalia during April 2003 with model output is quite good as well. The measurements were two to four days in duration and the modeled values were averaged for the same time period. In this figure modeled sulfate concentration is taken to be the sum of PSO4 and DSO4.

M. Astitha et al.

510 Modeled vs Observed Aerosol Concentration (PM10, ȝg/m3 ) Finokalia station, Crete 1-18 April 2003

Modeled vs Observed Aerosol Sulfate 3

Concentration (PM10, ȝg/m ) Finokalia station, Crete 1-18 April 2003

5,0 4,5

y = 0,8775x - 0,6155 R2 = 0,8076 r=0,898

4,0

6

y = 0,5798x + 0,0348 R2 = 0,7731 r=0,879

y = 0,8072x + 0,0618 R2 = 0,6099 r=0,78

5

3,0

Model Output

Model Output

3,5

2,5 2,0 1,5 1,0

4 3 2 1

0,5

0

0,0 0,0

0,5

1,0

1,5

2,0

2,5

3,0

3,5

4,0

4,5

0

5,0

1

2

CL-Aerosol

Na-Aerosol

Linear (Na-Aerosol)

3

4

5

6

Observations

Observations Linear (CL-Aerosol)

SO4-Aerosol

Linear (SO4-Aerosol)

Fig. 4 Comparison of modeled versus observed concentrations of (a) chloride and sodium, and (b) sulfate aerosol at Finokalia station for 1–18 April 2003

In general, the simulations showed rather small amounts of particulate sulfate produced from heterogeneous reactions on dust particles (DSO4) in cases where small amounts if dust and SO2 were available. Such result is evident in the time series of Figure 5a, where the modeled values for PSO4 and DSO4 are plotted for Erdemli, Turkey for 1–9 April 2003. This was not the case when there was significant amount of SO2 and available amounts of dust particles. In Figure 5b, DSO4 concentration (grey line) can be greater than PSO4 (dotted line), changing completely the general picture of total sulfate concentration. It should be noted that the discussed production of 3rd generation SO4 varies also with altitude, due to the transport of dust particles in higher vertical layers. The column mass load for the fifth day of the simulation for April 2003 is presented in Figure 6. The column mass load is plotted for each aerosol (g/m2) in the middle (0– 4 km) and the upper troposphere (4–8 km) showing the differences in all three pollutant loads. Significant role in these results plays the value of the gamma (Ȗ) uptake coefficient, which is considered 10-4 in this work (Zhang and Carmichael, 1999; Usher et al., 2002; Bauer and Koch, 2005; among others) and the weather conditions in the study area. Due to lack of data in the areas with high dust and SO2 concentrations, the comparison with the model results (for example Figure 4b), did not reveal the above discussed patterns in the daily average values. As part of an on-going work, the results presented herein await for more evaluating data. 2h Average CAMx Sulfate Concentration at Erdemli, Turkey Layer 1: 0-50m 1-9 April 2003

3,5E+01

2h average CAMx Sulfate Concentration at Saudi Arabia (near Red Sea) Layer 1: 0-50m 1-10 April 2003

7,00 3,0E+01

6,00 2,5E+01

ȝ g /m 3

ȝ g /m 3

5,00 4,00

2,0E+01

1,5E+01

3,00

1,0E+01

2,00 1,00

5,0E+00

0,00

0,0E+00

1/4/03 1/4/03 2/4/03 2/4/03 3/4/03 3/4/03 4/4/03 4/4/03 5/4/03 5/4/03 6/4/03 6/4/03 7/4/03 7/4/03 8/4/03 8/4/03 9/4/03 9/4/03 0:00- 12:00- 0:00- 12:00- 0:00- 12:00- 0:00- 12:00- 0:00- 12:00- 0:00- 12:00- 0:00- 12:00- 0:00- 12:00- 0:00- 12:002:00 14:00 2:00 14:00 2:00 14:00 2:00 14:00 2:00 14:00 2:00 14:00 2:00 14:00 2:00 14:00 2:00 14:00

PSO4total

DSO4total

PSO4+DSO4

1/4/03 1/4/03 2/4/03 2/4/03 3/4/03 3/4/03 4/4/03 5/4/03 5/4/03 6/4/03 6/4/03 7/4/03 8/4/03 8/4/03 9/4/03 9/4/03 10/4/03 10/4/03 0:00- 14:00- 4:00- 18:00- 8:00- 22:00- 12:00- 2:00- 16:00- 6:00- 20:00- 10:00- 0:00- 14:00- 4:00- 18:00- 8:00- 22:002:00 16:00 6:00 20:00 10:00 24:00 14:00 4:00 18:00 8:00 22:00 12:00 2:00 16:00 6:00 20:00 10:00 24:00

PSO4_2

DSO4_2

PSO4+DSO4

Date (UTC)

Fig. 5 Two hour average modeled sulfate concentration (PSO4 and DSO4) for two grid cells in the domain affected by desert dust transport. Left plot: Erdemli, Turkey. Right plot: Saudi Arabia


511

Fig. 6 Column integrated dust aerosol CRST (left), particulate sulfate PSO4 (middle) and particulate sulfate formed on dust DSO4 (right), for April 5, 2003, 12 UTC (g/m2). Size range: 0.1–2.5ȝm diameter. The upper plots are for the vertical column 0–4 km and the lower plots are for 4–8 km

4. Conclusions The preliminary results presented in this work emphasized on the sea salt particle production and the heterogeneous uptake of sulfur dioxide on dust particles using air quality modeling techniques. The generation of new aerosols on dust surfaces can be significant for both the middle and the upper troposphere, not because of the high amounts of produced species, but due to the different properties of such generation. This new 3rd generation aerosol can be higher than the sulfate produced from anthropogenic sources, depending on the weather and air quality conditions. Dust particles acting as reactive surfaces in a wet environment might lead to new climate modifiers for “desert dust-sensitive areas” like the Mediterranean Region. Overall, the new development presented in the previous sections, revealed the capability of an air quality model like CAMx to handle the complex processes of heterogeneous reactions on dust particles. Also revealed the uncertainty of such processes based on the assumptions during the implementation of the relevant coefficients and algorithms. This work is still ongoing, mostly the evaluation of the heterogeneous reactions and provides the capability for future improvement of the physicochemical processes treated by the model. Acknowledgments This work is realized in the framework of Act. 8.3 of the Operational Program Competitiveness (3rd Community Support Program) and is co-funded by 75% from the European Social Fund (EU), 25% from the Greek General Secretariat for Research and Technology (GSRT) under the project PENED 2003. The authors would like to acknowledge the contribution from Professor N. Mihalopoulos and X. Querol for providing air quality measurements.

512

M. Astitha et al.

References Alastuey et al. (2005) Characterisation of TSP and PM2.5 at Izana and Sta Cruz de Tenerife (Canary Islands, Spain) during a Saharan dust episode. Atmos. Environ., 39, 4715–4728. Astitha M, Kallos G, Katsafados P, Pytharoulis I, Mihalopoulos N (2006) “Radiative effects of natural PMs on photochemical processes in the Mediterranean Region”. 28th NATO/CCMS ITM, May 2006, Leipzig, Germany. Bauer SE, Koch D (2005) Impact of heterogeneous sulfate formation at mineral dust surfaces on aerosol loads and radiative forcing in the Goddard Institute for Space Studies general circulation model, J. Geophys. Res., 110(D17). de Leeuw, G., F.P. Neele, M. Hill, M.H. Smith, E. Vignati (2000) Production of sea spray aerosol in the surf zone, J. Geophys. Res., 105(D24), 29397–29409. Environ (2006) User’s Guide to the Comprehensive Air Quality Model with Extensions (CAMx). Version 4.31. ENVIRON Inter. Corp., Novato, CA. Fuchs NA, Sutugin AG (1970) Highly Dispersed Aerosols. Ann Arbor Science Publishers, Ann Arbor, MI. Gong SL (2003) A parameterization of sea-salt aerosol source function for suband super- micron particles. Glob. Biog. Cycles, 17(4), 1097. Kallos G (1997) The regional weather forecasting system SKIRON: an overview. Proceedings of the symposium on regional weather prediction on parallel computer environments, University of Athens, Greece, pp. 109–122. Kallos GA, Papadopoulos P, Katsafados, Nickovic S (2006) Trans-Atlantic Saharan dust transport: Model simulation and results. J. Geophys. Res., 111. Levin Z, Teller A, Ganor E, Yin Y (2005) On the interactions of mineral dust, sea-salt particles and clouds: A measurement and modelling study from the Mediterranean Israeli Dust Experiment campaign. J. Geophys. Res., 110, D09204. Nickovic S, Kallos G, Papadopoulos A, Kakaliagou O (2001) A model for prediction of desert dust cycle in the atmosphere. J. Geophys. Res., 106, D16. Penner JE et al. (2001) Aerosols, their direct and indirect effects, in Climate Change 2001: The Scientific Basis: Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change, edited by J.T. Houghton et al., pp. 291–348, Cambridge University Press, New York. Saylor RD (1997) An estimate of the potential significance of heterogeneous loss to aerosols as an additional sink for HO radicals in the troposphere. Atmos. Environ., 21, 3653–3658. Shankar U, Bhave PV, Vukovich JM, Roselle SJ (2005) Implementation and initial applications of sea salt aerosol emissions and chemistry algorithms in the CMAQ v4.5-AERO4 module. 4th Ann. CMAS Models-3 Users’ Conference. Usher CR, Al-Hosney H, Carlos-Cuellar S, Grassian VH (2002) A laboratory study of the heterogeneous uptake and oxidation of sulfur dioxide on mineral dust particles, J. Geophys. Res.-A, 107, 4713, doi:10.1029/2002JD002051. Zhang Y, Carmichael GR (1999) The role of mineral aerosol in tropospheric chemistry in East Asia-A model study, J. Appl. Met., 38, 353–366.


513

Zhang KM, Knipping EM, Wexler AS, Bhave PV, Tonnesen GS (2005) Size distribution of sea-salt emissions as a function of relative humidity. Atmos. Environ., 39, 3373–3379.

Discussion G. Sarwar: Please explain the reason for the decrease in ozone concentration. M. Astitha: The decrease in Ozone concentration is related to shading effects (effects of dust on photolysis rates) and heterogeneous processes (ozone with dust). M. Mircea: How many dust bins have your model? Did you study the impact of the number of size bins (of their size/on the gas uptake)? M. Astitha: At the present version of the dust flux model, we used four size bins but we used only the first two with particle size less than 10 ȝm. No we did not. Our new version of dust model includes eight particle sizes (up to PM10). I found your question interesting and I will look for such an impact in my future work. A. Venkatram:

M. Astitha:

Did you examine the sensitivity of your results on O3 and sulfate to the uptake coefficients for O3 and SO2? The inclusion of the heterogeneous conversion of SO2 to SO4 appears to have led to overestimation of SO4. Are you worried about these results? The uptake coefficients for O3 and sulphate on dust particles were chosen based on the literature. I haven’t done sensitivity experiments changing O3 and SO2 uptake coefficients but I intend to do this in the future. The amount of SO4 produced from the uptake of SO2 on dust is in general much smaller than the anthropogenic sulphate, except for some specific locations. We have seen the overestimation and we plan to carry out a series of sensitivity tests to see how the situation will be improved.

B. Fisher: How are the uptake coefficients determined? M. Astitha: The uptake coefficients (Ȗ values) of gases onto the particle surface are determined mostly through laboratory and field experiments.

5.7 Modelling Regional Aerosols: Impact of Cloud Processing on Gases and Particles over Eastern North America and in Its Outflow During ICARTT 2004 W. Gong1, J. Zhang1, M.D. Moran1, P.A. Makar1, S.L. Gong1, C. Stroud1, V.S. Bouchet2, S. Cousineau2, S. Ménard2, M. Samaali2, M. Sassi2, B. Pabla1, R. Leaitch1, A.M. Macdonald1, K. Anlauf1, K. Hayden1, D. Toom-Sauntry1, A. Leithead1 and J.W. Strapp3

Abstract A regional aerosol model, AURAMS (A Unified Regional Air-quality Modelling System), is used to simulate gases and aerosols over eastern North America for the ICARTT field campaign period during summer 2004. The model performance is evaluated against both ground-based and airborne observations during the field campaign. A model sensitivity study is used to assess the impact of cloud processing on the aerosol characteristics in the air masses over eastern North America and its outflow to the North Atlantic during the study period.

Keywords Air-quality modelling, cloud processing of gas and aerosol, regional aerosols

1. Introduction During the summer of 2004, several coordinated field campaigns were conducted over North America, the North Atlantic, and western Europe as part of the International Consortium for Atmospheric Research on Transport and Transformation (ICARTT). These field programs were designed to study the emission of aerosol and ozone precursors, their chemical transformations and removal during transport to and over the North Atlantic, and their impact downwind on the European continent (Fehsenfeld et al., 2006). 1

Environment Canada, 4905 Dufferin Street, Toronto, ON, M3H 5T4, Canada. Environment Canada, 2121, Voie de Service Nord No. 404 Route Transcanadienne, Montreal, QC, H3P 1J3, Canada. 3 Environment Canada, 4905 Dufferin Street, Toronto, ON, M3H 5T4, Canada. 2


539

540

W. Gong et al.

One of the campaigns, an aircraft study conducted by the Environment Canada scientists using the National Research Council of Canada (NRCC) Convair 580, focused on the chemical transformation of gases and aerosols by clouds. The summer of 2004 in eastern North America was characterized as being cooler and wetter than normal, with relatively more frequent southwesterly flow (Fehsenfeld et al., 2006). The observations at New England and Nova Scotia coastal sites and over the Gulf of Maine showed that the fine particulate matter (PM2.5) in the plumes from eastern North America was composed mainly of acidic sulphate and highly oxidized organics, an indication of significant atmospheric processing. An important question is how much of that PM2.5 was contributed by cloud processing. In this study we address the above question through the application of a regional aerosol model (AURAMS) to the ICARTT field campaign period over eastern North America. The model’s performance is examined against surface ozone and PM measurements from the AIRNow network, the speciated PM measurements from the IMPROVE network, and observations from the Convair 580 flights. The impact of cloud processing on aerosols over eastern North America and its outflow to the North Atlantic is then assessed with a model sensitivity study.

2. Simulation Setup AURAMS is a multi-pollutant, regional air-quality modelling system with size segregated and chemically speciated representation of aerosols. It has been described elsewhere (e.g., Gong et al., 2006; Tarasick et al., 2007). Simulation with AURAMS version 1.3.2 was carried out on an 85 × 105 polar-stereographic grid over eastern North America with a horizontal resolution of 42 km (true at 60q N). The simulation period is July 7–August 19, 2004. The first seven days are counted as model spin-up. An older version of AURAMS was run in real-time during the ICARTT field campaign to provide guidance for flight planning (Bouchet et al., 2004) and to participate in a multi-model inter-comparison and evaluation effort (McKeen et al., 2005, 2007). The real-time AURAMS run was also evaluated against ozonesonde data by Tarasick et al. (2007). Some of the important updates since the real-time run include changes in anthropogenic emission inventory, biogenic emission processing and vegetation database, and chemical lateral boundary conditions. For this simulation, the anthropogenic emissions files were prepared from the 2000 Canadian, 2001 U.S., and 1999 Mexican national emissions inventories with version 2.2 of the SMOKE emissions processing system (http://www.smokemodel.org/index.cfm). State-specific adjustments to the NOx emission from major point sources were also incorporated to account for the emission changes due to the NOx State Implementation Plan (SIP) Call by the U.S. EPA (U.S. EPA, 2004) which came into effect in 2003. The biogenic emission fields were processed inline using BEIS v3.09. Being an off-line model, AURAMS is driven by a regional weather forecast model simulation (GEM; Côte et al., 1998). For this study, GEM version 3.2.0, with the additional recent parameterization for anthropogenic heat islands (Makar

Modelling Regional Aerosols: Impact of Cloud Processing on Gases and Particles

541

et al., 2006), was used in a regional configuration with a 15 km resolution in its uniform “core” cantered over North America. Time-independent chemical lateral boundary conditions were used for this simulation. The gridded monthly ozone climatology of Logan (1998) was used for the initial and boundary condition for O3; the monthly MOPITT data (http://neo.sci. gsfc.nasa.gov/Search.html?group=35) was used for CO; and for the rest of the gases species and particle components the initial and boundary conditions were constructed from an existing AURAMS annual simulation (Moran et al., 2007) and observational data (e.g., Macdonald et al., 2006).

3. Model Evaluation 3.1. Surface ozone and PM2.5 Model predicted ground-level ozone and PM2.5 mass concentrations are compared to the hourly observations from the AIRNow network (http://airnow.gov/index. cfm). The AIRNOW observations are reported as hourly averages, whereas the model outputs are hourly, valid at the top of the hour. No attempt is made here to process model output in order to obtain equivalent hourly averages. Standard statistical measures are calculated at each AIRNow site (without spatial interpolation) within the area as in McKeen et al. (2005, 2007) over the 36-day simulation period (excluding the first seven-day spin-up). The spatial distribution of the correlation coefficient (r) and the mean bias (MB) are shown in Figure 1 for daily 1-hour maximum ozone and daily average PM2.5. The mean and median of these statistical measures are also included in the figure.

Fig. 1 Spatial distribution of (a) correlation coefficient (r), (b) mean bias (MB) for 1-hour daily maximum ozone, (c) r and (d) MB for daily mean PM2.5 over the 36-day simulation period

The model correlated relatively well with the observation of O3 over southern Ontario and the U.S. eastern seaboard, but the O3 comparison was relatively poor over the Ohio River Valley and the Appalachians. The mean correlation coefficient of 0.54 is a marginal improvement from 0.51 from the real-time run for the same period. There is a significant positive bias that averages at 13.5 ppbv. This amounts to a considerable increase from the real-time run mainly due to the change in biogenic emission processing and vegetation data. The simple representation of the U.S. SIP Call’s NOx control resulted in an average of 2 ppbv reduction in the ozone mean bias.

W. Gong et al.

542

As for PM2.5, the mean correlation coefficient of 0.51 is a noticeable improvement from the previous 0.47 from the real-time run. The overall bias is small. The spatial distribution of model performance (particularly in terms of r) is similar to that for ozone.

3.2. Speciated PM2.5

7/16/04 30

7/26/04

8/5/04

20

10

0 7/16/04

7/26/04

8/5/04

Quaker City, OH

Observed

10

0 7/26/04

8/5/04

30

10

0 7/16/04

8/15/04

7/26/04

8/5/04

8/15/04

Model (base case) 20

10

0

8/15/04

20

7/16/04

30

Model (base case)

20

8/15/04

Connecticut Hill, NY Observed

30

PM2.5, SO42.5, NO32.5, OM2.5 (ug m)

0

30

PM2.5, SO42.5, NO32.5, OM2.5 (ug m-3)

PM2.5, SO42.5, NO32.5, OM2.5 (ug m-3)

Observed

10

PM2.5, SO42.5, NO32.5, OM2.5 (ug m-3)

PM2.5, SO42.5, NO32.5, OM2.5 (ug

-3

)

Cape Cod, MA 20

PM2.5, SO42.5, NO32.5, OM2.5 (ug m-3)

Measurements of speciated PM2.5 at 23 sites from the IMPROVE network (http:// vista.cira.colostate.edu/views/Web/Data/DataWizard.aspx) are available for compareson with the model results. The measurements at these sites are 24-hour averages (midnight-to-midnight, local daylight saving time) every three days. The hourly model outputs are therefore averaged over the same 24-hour periods and also sampled every three days accordingly to be paired up with the observations. The model-observation comparison of speciated PM2.5 at three sites is shown in Figure 2.

7/16/04

7/26/04

8/5/04

8/15/04

Model (base case)

20

10

0 7/16/04

7/26/04

8/5/04

8/15/04

Fig. 2 Comparison with speciated PM2.5 measurements (left column – measurement, right column – model) at three sites from the IMPROVE network: (top) Cape Cod, (middle) Connecticut Hill, and (bottom) Quaker City; red – sulphate, blue – nitrate, and green – organics; pluses denote total PM2.5 mass

In general, the model tends to overpredict sulphate close to the sources (e.g. Quaker City, OH). The overprediction decreases moving away from source regions (e.g. Connecticut Hill, NY), and over the east coast region the model in general under-predicts sulphate (e.g. Cape Cod, MA). Nitrate2.5 is generally overpredicted by the model. There is a significant underprediction of the organic aerosol component by the model. This can be largely attributed to the fact that the implementation of Jiang’s SOA scheme (Jiang, 2003) in this study only considered the existing secondary organic aerosol as “seed” particles for the partitioning. A recent sensitivity simulation with several SOA parameterization updates (partitioning of SOA products to both primary and secondary OA, updates to partitioning and stoichio-metric coefficients, and inclusion of isoprene as a precursor to SOA) yielded an order of magnitude increase in SOA. Figure 3 shows the model-vs-observation scatter-plots of total PM2.5, sulphate2.5, nitrate2.5, and sulphate-to-PM2.5 ratio for all available sites in terms of averages over the simulation period (or, 12 24-hour samples). In this case, the correlations


543

between model and observation are good for both PM2.5 and sulphate2.5. The modelled and observed mean values are also comparable (particularly in the case of sulphate; a bias of ~ –1.3 ȝg m-3 for PM2.5). As mentioned above, nitrate is overpredicted by the model and this is reflected here also. The overestimate of sulphate fraction by the model is often a result of the significant underprediction of organic component. 12

12 Y = 1.47 * X - 6.39 # of data points = 22 Average X = 10.84 Average Y = 9.56 R2 = 0.71

8 4 0

Model

Model

16

8

Y = 1.53 * X - 2.47 # of data points = 23 Average X = 4.46 Average Y = 4.33 R2 = 0.71

4 0

0

4

8 12 16 20 24 Observation

0

4 8 12 Observation

16

Nitrate2.5 (ug m-3)

2

(b)

Y = 0.88 * X - 1.45 # of data points = 23 Average X = 4.46 Average Y = 2.45 R2 = 0.70

0.7

(c)

Sulphate2.5/PM2.5

(d)

0.6

1.6

0.5

1.2 Y = 2.48 * X + 0.11 # of data points = 23 Average X = 0.22 Average Y = 0.66 R2 = 0.17

0.8 0.4

Model

Sulphate2.5 (ug m-3)

16

(a)

20

Model

PM2.5 (ug m-3)

24

Y = 0.77 * X + 0.18 # of data points = 19 Average X = 0.40 Average Y = 0.49 R2 = 0.25

0.4 0.3 0.2

Acadia

Cape Cod Martha's Vineyard

0.1

0 0

0.4

0.8 1.2 1.6 Observation

2

0.1 0.2 0.3 0.4 0.5 0.6 0.7 Observation

Fig. 3 Model-vs-observation scatter-plots of (a) PM2.5, (b) sulfate2.5, (c) nitrates2.5, and (d) sulphate fraction. Both model results and observations are averaged over the simulation period. Also plotted in (b) are results from a sensitivity run with in-cloud oxidation turned off (blue)

3.3. Comparison with aircraft observations During ICARTT 2004, 23 research flights were flown with the NRCC Convair 580 based at Cleveland, Ohio. Measurements onboard the aircraft included trace gases, aerosol particle size distribution and chemistry, and cloud microphysics and chemistry (see Hayden et al., 2007; Leithead et al., 2007; Zhang et al., 2007) for details on the measurement and instrumentation). Complementary to the surface-based observations, these airborne measurements provide information on both primary and secondary pollutants aloft which is valuable for improving our knowledge of the atmospheric processes and for evaluating model representations of these processes.

Flt 17 Flt 16

(a)

Flt 16

Flt 17

(b)

Fig. 4 Model predicted (a) O3 and (b) SO2 at 1,500 m for 21 Z August 10, 2004, overlaid with aircraft-observed O3 and SO2 along the flight tracks from Flt 16 and 17

Here we focus on two of the flights, Flt 16 and 17, conducted on August 10, 2004, between Lake Chicago and Lake Erie in the air mass ahead of an advancing cold front in an attempt to study cloud processing of industrial and urban plumes down-wind Chicago. Figure 4 shows the modelled ozone and SO2 at 1,500 m (close to the altitude for the in-cloud flight legs) overlaid with the in-situ measurements. As seen the model predicted ozone is considerably higher than the aircraft observations. On the other hand, the model is capturing the SO2 plumes reasonably

W. Gong et al.

544

well particularly for Flt 16. The plumes observed by the aircraft are more intense than the model, which is reasonable given that the model resolution is at 42 km. 4000

4000

3000

3000

3000

3000

3000

1000 0

0.01 0.1 1 10 100 SO2 (ppbv)

Altitude (m)

4000 3000 2000 1000

aircraft obs. model Altitude (m)

(b)

0

1000 0

0.01 0.1 1 10 100 NO2 (ppbv)

2000 1000 0

50

150 250 350 CO (ppbv)

1000 0

0

20 40 60 80 O3 (ppbv)

0

4000

4000

4000

3000

3000

3000

3000

2000 1000

2000 1000 0

0.01 0.1 1 10 100 NO2 (ppbv)

2000 1000

150 250 350 CO (ppbv)

4 8 12 SO4= (ug/m3)

10-min. PILS bulk CW samples (air equiv.) model

2000 1000 0

0 50

10-min PILS bulk CW samples (air equiv.) model

2000

4000

0 0.01 0.1 1 10 100 SO2 (ppbv)

2000

Altitude (m)

0

2000

Altitude (m)

1000

Altitude (m)

2000

Altitude (m)

4000

Altitude (m)

aircraft obs. 4000 model Altitude (m)

4000

Altitude (m)

Altitude (m)

(a)

0

20 40 60 80 O3 (ppbv)

0

4 8 12 SO4= (ug/m3)

Fig. 5 In-situ comparison of observed and modelled gaseous species and particle sulphate (a) fro Flt 16 and (b) for Flt 17. Measurements of the gaseous species (SO2, NO2, CO, and O3) are at 1-s frequency; measurements of inorganic ions are from the 10-minute integrated PILS; also included is air equivalent cloud-water sulphate from bulk cloud water samples (varying in sampling duration)

A series of in-situ comparisons between modelled and observed gaseous species (SO2, NO2, CO, and O3) and particle sulphate is presented in Figure 5. The comparison is done by plotting the model results and the observations along the flight track against height in a form of “vertical profiles”. This is in an attempt to get a sense of whether the model comes close to the aircraft observation over the flight area in terms of the range and vertical structure, since a true point-to-point in-situ comparison is not appropriate given the model resolution. The sampling through the modelled fields is done given the grid locations along the flight path over the entire flight period. The comparison is surprisingly good for SO2, NO2 and, to a lesser degree, CO. The excursions to higher values from the observations reflect the encountering of plumes during horizontal legs at two different altitudes. The 1-s (or ~100 m) observations are able to resolve much sharper peaks than the model at 42km resolution. Consistent with the evaluation above, the modelled ozone shows a significant positive bias (more pronounced at lower levels), although the observed vertical structure seems to be captured by the model. The model overpredicted particle sulphate in comparison to the 10-minute integrated measurement from PILS (Particle-in-Liquid-Sampler) particularly at lower levels (though the PILS measurements for Flt 17 are uncertain due to solution problems on that flight). Note that no screening for in-cloud sampling of PILS is done here, and the PILS samples in clouds are more difficult to interpret. Also included in the plot are air-equivalent cloud-water sulphate measurements in air-equivalent concentration from the bulk


545

cloud-water samples with varying durations. Much higher sulphate concentrations are observed in clouds. If these clouds remained non-precipitating the sulphate would remain in the atmosphere after the evaporation of the clouds.

4. Impact of In-Cloud Oxidation The model simulation was repeated with the in-cloud oxidation turned off to assess its impact on aerosols over eastern North America and its outflow to the North Atlantic. Figure 6a shows the model predicted sulphate2.5 at 1,810 m for July 23, 2004 at 20 Z from the base case simulation and Figure 6b the difference (delta) between the base case (i.e., with in-cloud oxidation) and the run without in-cloud oxidation in modelled particle sulphate. The period of July 22–24, 2004 corresponded to an outflow event when, as seen from Figure 6a, the plumes from the Ohio valley and the New York-Boston area are merging and being transported to the north Atlantic under south-westerly to westerly flow. Based on Figure 6, for this snapshot, the in-cloud oxidation is responsible for just under 50% the particle sulphate over eastern North America. (a)

(b)

(c)

(d)

Fig. 6 (a) Model predicted sulpahte2.5 at 1,810 m overlaid with wind at the same level for 20Z July 23, 2004; (b) the corresponding difference in model predicted sulphate2.5 with vs without incloud oxidation; (c) modelled sulphate2.5 column loading (ȝg m-2) averaged over July 14–August 18, 2004; (d) difference in the modelled average sulphate column loading with vs without in-cloud oxidation

To assess the overall impact of in-cloud oxidation for the field study period, column loadings of particle mass (total and speciated components) were computed based on the hourly model output and averaged over the 36-day simulation period. Figure 6c presents the average sulphate2.5 column loading over the simulation period from the base case and Figure 6d the difference in the column loading when in-cloud oxidation is turned off. Similar to the case of the single snapshot above, the in-cloud oxidation contributes to about half of the average column loading during the ICARTT period. Not shown here, the impact of in-cloud oxidation on modelled PM2.5 is somewhat smaller due to the increase in nitrate (i.e., more nitrate will partition to particle phase when sulphate is reduced when everything else remains the same). The results from the run without the in-cloud oxidation are also included in Figure 3b. It is seen that, without the in-cloud oxidation, there is a negative mean bias of ~2 ug m-3 (or roughly 50%) in modelled sulphate, although the slope is closer to 1 than the base case.

546

W. Gong et al.

Acknowledgments The authors would like to acknowledge the IMPROVE network and the AIRNow group and cooperating agencies in U.S. and Canada for making their data available. The National Research Council of Canada Institute for Aerospace Research Lab and the Meteorological Service of Canada Aircraft facility are also acknowledged for aircraft operation and support.

References Bouchet VS, Crevier L-P, Cousineau S, Ménard S, Moffet R, Gong W, Makar P, Moran M, Pabla B (2004) Realtime regional air quality modelling in support of the ICARTT 2004 campaign. Proc. 27th NATO/CCMS ITM on Air Pollution Modelling and Its Application, Banff, Canada, October 25–29. Côté J, Desmarais J-G, Gravel S, Méthot A, Patoine A, Roch M, Staniforth A (1998) The operational CMC/MRB Global Environmental Multiscale (GEM) model. Part 1: Design considerations and formulation, Mon. Wea. Rev., 126, 1373–1395. Fehsenfeld FC, Ancellet G, Bates T, Goldstein A, Hardesty M, Honrath R, Law K, Lewis A, Leaitch R, McKeen S, Meagher J, Parrish DD, Pszenny A, Russell P, Schlager H, Seinfeld J, Trainer M, Talbot R (2006) International Consortium for Atmospheric Research on Transport and Transformation (ICARTT): North America to Europe: Overview of the 2004 summer field study, J. Geophys. Res., 111, D23S01, doi:10.1029/2006JD007829. Gong W, Dastoor AP, Bouchet VS, Gong S, Makar PA, Moran MD, Pabla B, Ménard S, Crevier L-P, Cousineau S, Venkatesh S (2006) Cloud processing of gases and aerosols in a regional air quality model (AURAMS), Atmos. Res., 82, 248–275. Hayden KL, Macdonald AM, Gong W, Toom-Sauntry D, Anlauf KG, Leithead A, Li S-M, Leaitch WR, Noone K (2007) Cloud Processing of Nitrate, J. Geophys. Res. (submitted). Jiang W (2003) Instantaneous secondary organic aerosol yields and their comparison with overall aerosol yields for aromatic and biogenic hydrocarbons, Atmos. Environ., 37, 5439–5444. Leithead A, Macdonald AM, Li SM, Gong W, Anlauf KG, Toom-Sauntry D, Hayden KL, Leaitch WR (2006) Investigation of carbonyls in cloud-water during ICARTT, J. Geophys. Res. (submitted). Logan JA (1998) An analysis of ozonesonde data for the troposphere: Recommendations for testing 3D models, and development of a gridded climatology for tropospheric ozone, J. Geophys. Res., 104, 16115–116149. Macdonald AM, Anlauf KG, Leaitch WR, Liu P (2006) “Multi-year chemistry of particles and selected trace gases at the Whistler high elevation site”, Eos. Trans. AGU, 87(52), Fall Meet. Suppl., Abstract A53B-0179. Makar PA, Gravel S, Chirkov V, Strawbridge KB, Froude F, Arnold J, Brook J (2006) Atmos. Environ., 40, 2750–2766.


547

McKeen S, Wilczak J, Grell G, Djalalova I, Peckham S, Hsie E-Y, Gong W, Bouchet V, Ménard S, Moffet R, McHenry J, McQueen J, Tang Y, Carmichael GR, Pagowski M, Chan A, Dye T (2005) Assessment of an ensemble of seven real-time ozone forecasts over Eastern North America during the summer of 2004, J. Geophys. Res., 110, D21307, doi:10.1029/2005JD005858, 16 pp. McKeen S, Chung SH, Wilczak J, Grell G, Djalalova I, Peckham S, Gong W, Bouchet V, Moffet R, Tang Y, Carmichael GR, Mathur R, Yu S (2007) Evaluation of several real-time PM2.5 forecast models using data collected during the ICARTT/NEAQS 2004 field study, J. Geophys. Res., 112, D10S20, doi:10.1029/2006JD007608, 20 pp. Moran MD, Zheng Q, Samaali M (2007) Long-term multi-species performance evaluation of AURAMS for first 2002 annual run. EC internal report, Toronto, Ontario (in preparation). (Available from first author: email [email protected]) Tarasick DW, Moran MD, Thompson AM, Carey-Smith T, Rochon Y, Bouchet VS, Gong W, Makar PA, Stroud C, Ménard S, Crevier L-P, Cousineau S, Pudykiewicz JA, Kallaur A, Moffet R, Ménard R, Robichaud A, Cooper OR, Oltmans SJ, Witte JC, Forbes G, Johnson BJ, Merrill J, Moody JL, Morris G, Newchurch MJ, Schmidlin FJ, Joseph E (2007) Comparison of Canadian air quality forecast models with tropospheric ozone profile measurements above mid-latitude North America during the IONS/ICARTT campaign: evidence for stratospheric input, J. Geophys. Res., 112, D10S20, doi:10.1029/2006JD007782 (In press). U.S. EPA (2004) NOx Budget Trading Program: 2003 Progress and Compliance Report, Clean Air Markets Program, Office of Air and Radiation, U.S. Environmental Protection Agency, Washington, DC, August, 28 pp. + App. (Available from http://www.epa.gov/airmarkets/progress/progress-reports.html: viewed 11 June 2007) Zhang J, Gong W, Leaitch WR, Strapp JW (2007) Evaluation of modeled cloud properties against aircraft observations for air quality applications, J. Geophys. Res., 112, D10S16, doi:10.1029/2006JD007596.

Discussion P. Builtjes: Is the overestimation of nitrates due to problems in the chemical, and the dependence on temperature? W. Gong: The nitrate over-prediction may partly be attributable to overprediction of gas-phase nitric acid, but the uncertainty in the gasparticle partition of total nitrate (via inorganic heterogeneous chemistry, which is temperature dependent) is a major contributing factor. For example, it has been shown that whether or not treating particles as metastable aqueous droplets in the equilibrium can have an important impact on the partitioning. This is definitely an area needing more investigation.

5.8 On the Role of Ammonia in the Formation of PM2.5 C. Mensink and F. Deutsch

Abstract We studied the formation and composition of PM2.5 using the EUROS model. This model contains comprehensive modules (CACM, MADRID) for the formation of secondary atmospheric aerosols and their precursors. Some spatial and temporal patterns in which ammonia emissions can be associated with elevated PM2.5 and PM10 concentrations are analysed. Especially the episode of 15–16 April 2007 revealed some interesting features, e.g. the importance of the impact of temperature, relative humidity and hygroscopic water on PM2.5 and PM10 concentrations. A hypothesis is formulated in which it is stressed that ammonia can be a provider of an abundant amount of condensation nuclei in the form of ammonium nitrate and ammonium sulphate which, under favourable meteorological conditions, attract hygroscopic water, leading to rapid increase in the PM2.5 mass fraction.

Keywords Ammonia, ammonium, Belgium, hygroscopic water, nitrate, PM2.5 1. Introduction In the proposal for a new directive of the European Parliament and of the Council on ambient air quality and cleaner air for Europe, a target value of 20 µg/m³ to be met by 1 January 2010 and a limit value of 20 µg/m³ to be met by 1 January 2015 are proposed for PM2.5. This is the first time in Europe that exposure to ambient air concentrations of PM2.5 is considered legally. From measurements and model studies it is known that the contribution to PM2.5 of secondary inorganic compounds like ammonium, nitrate and sulphate is considerable, with a mean value of 40% for 6 European cities (Sillanpää et al., 2006). Van Grieken et al. (2003) showed that 40% of the PM2.5 measured at five different locations in Belgium consists of Secondary Inorganic Aerosols (SIA). About 20% of the PM2.5 was found to be elementary carbon (EC) and of organic origin (OA). They also concluded that the other 40% could not be identified, e.g. potentially being soil dust components, sea salt, particles of biogenic origin and water. Maenhaut et al. (2006) showed however, that sea salt and soil dust are rather found in the PM10–PM2.5 fraction, at least for a busy road in Antwerp, where they performed their measurements. They found large amounts of nitrate, sulphate and ammonium in the PM2.5 fraction (contributing more than 50%). These SIA compounds were found to C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

548

On the Role of Ammonia in the Formation of PM2.5

549

be more present in winter time than in summer time. This was especially true for nitrate. In this paper we study the role of ammonia in the formation of PM2.5 and PM10 in Belgium. We analyse its spatial and temporal patterns, by using the extended EUROS model and by making comparisons with observations. The Belgian version of the EUROS model (Delobbe et al., 2002) has been extended to model fine particulate matter (PM) by implementing the Caltech Atmospheric Chemistry Mechanism (CACM, Griffin et al., 2002) and the Model of Aerosol Dynamics, Reaction, Ionization and Dissolution (MADRID 2, Zhang et al., 2004). Currently, the modelling system is able to model mass and chemical composition of aerosols in two size fractions (PM2.5 and PM10–2.5) and seven components: ammonium, nitrate, sulphate, elementary carbon, primary inorganic compounds, primary organic compounds and secondary organic compounds (SOA). The model has been applied and validated for Belgium (Deutsch et al., 2008a, b, c). Once emitted in the air, ammonia (NH3) is converted to ammonium sulphate and ammonium nitrate, contributing to the PM2.5 (and the PM10) fraction: and

2NH3 + H2SO4 Æ (NH4)2SO4 or NH3 + H2SO4 Æ NH4 HSO4 NH3 + HNO3 ÅÆ NH4 NO3

(1) (2)

Since sulphuric acid (H2SO4) has a low vapour pressure, the sulphate will be available in the particle phase and will easily be neutralised to form ammonium sulphate or ammonium bisulphate (1). Ammonia can also react in the gas phase with nitric acid (HNO3) to form ammonium nitrate (2), but this is an equilibrium reaction and its formation depends strongly on temperature, humidity and the individual concentrations. As reported by Erisman and Schaap (2004), the amount of nitrate will increase with decreasing temperatures, provided that there is more ammonium available than needed to neutralise the sulphate. Or as stated by Pinder et al. (2007): in terms of emission control strategies, a reduction in SO2 and hence a reduction in sulphate will increase the available free NH3 and a portion of this free NH3 will react to form ammonium nitrate. In this way a part of the ammonium sulphate is replaced by ammonium nitrate. The relation between NH3 emissions and the formation of PM2.5 and PM10 is studied in Section 2, where we briefly address spatial patterns of NH3 emissions and particle concentrations in Belgium. In Section 3 we study a remarkable event occurring on the evening and the night of 15–16 April 2007, showing a sharp increase in PM2.5 and PM10 concentrations in Brussels, attaining levels of 150 µg/m³ within a couple of hours. In order to analyse this event in more detail we performed a sensitivity calculation using the EUROS model and increasing the ammonia emissions by a factor of 5. Results show that the formation mechanism expressed by (1) and (2) can only partially explain the enormous rise in PM2.5 concentration observed in the monitoring station. This will be further discussed in Section 4.

C. Mensink, F. Deutsch

550

2. Spatial Distribution of NH3 and PM10 over Belgium Figure 1 shows the modelled NH3 concentrations for April 2007. It gives a typical picture of the spatial distribution of the NH3 emissions and concentrations over Belgium. More than 90% of the NH3 emissions originate from agricultural activities (cattle breeding, pig farms and manure spreading). These activities are concentrated in West Flanders, a province in the north western part of Belgium. Fig. 1 Ammonia (NH3) concentrations over Belgium, northern France, western Germany and the south of the Netherlands as modelled by the EUROS model for the period 1–20 April 2007

Legend [µg/m³]

Legend PM10 [µg/m³]

error

error

1 - 10

1 - 18

11 - 12

19 - 21

13 - 14

22 - 23

15 - 17

24 - 25

18 - 20

26 - 27

21 - 23

28 - 29

24 - 26

30 - 31

27 - 29

32 - 34

30 - 33

35 - 38

> 33

> 38

Fig. 2 Maps of the observed annual mean NO2 (left) and annual mean PM10 (right) for Belgium in 2002

Fig. 3 Annual mean PM10-concentration in Belgium as modelled (left) and measured (right) in 2003 (www.irceline.be)

On the Role of Ammonia in the Formation of PM2.5

551

Concentration [µg/m³] 120

2,5

ML01

ML02

R841

Model

100

80

60

40

20

0 1/01/2003 1/02/2003 1/03/2003 1/04/2003 1/05/2003 1/06/2003 1/07/2003 1/08/2003 1/09/2003 1/10/2003 1/11/2003 1/12/2003

Fig. 4 Modelled and the measured PM2.5 concentrations in 2003 for three monitoring stations in Mechelen

Referring to mechanisms (1) and (2) in the previous section, these high emissions and concentrations of NH3 are expected to lead to the formations of ammonium sulphate and ammonium nitrate. For the formation of these SIA, both SO2 and NO2 are available from long range transport as well as from local sources. Figure 2 shows a map of the measured NO2 and measured PM10 concentrations over Belgium in 2002. The NO2 map clearly reflects the traffic patterns and industrial activities over the country, where one can immediately point out the urban areas of Brussels, Antwerp, Ghent, Charleroi and Liège. In contrast, the PM10 map shows remarkable elevated concentrations in West Flanders. These high concentrations can be associated with the high agricultural activities leading to high NH3 emissions (Figure 1). Through mechanisms (1) and (2), elevated PM2.5 and PM10 concentrations are expected. Figure 3 shows a comparison of the modelled and the measured PM10 concentrations in 2003, clearly showing the elevated PM10 concentrations in West Flanders. Unfortunately PM2.5 concentration maps are not yet available, because of lack of sufficient monitoring stations for PM2.5. Figure 4 shows a comparison of the modelled and the measured PM2.5 concentrations in 2003 for three monitoring stations in the area of Mechelen, some 30 km north of Brussels. The figure shows a good comparison between the modelled and measured PM2.5 concentrations, although peak values are underestimated.

3. Temporal Patterns of PM2.5 on 15–16 April 2007 In the evening and night of 15 and 16 April 2007, a sharp increase in PM2.5 and PM10 concentrations was observed between 18h00 and 3h00 CEST in several monitoring stations in Brussels. Figure 5 shows both PM2.5 and PM10

552

C. Mensink, F. Deutsch

Fig. 5 Observed PM2.5 and PM10 concentrations and relative humidity (RH) in Sint-JansMolenbeek (Brussels), for 15–16 April 2007

concentrations as observed by TEOM FDMS equipment in the station Sint-JansMolenbeek (41R001). It can be seen that PM2.5 and PM10 concentrations are about equal, indicating that the increase concerns the smaller fraction of the PM mass and possibly a sudden increase of SIA. During this episode the strong ratio of increase in PM2.5 coincided with a strong ratio of increase in relative humidity (RH). Likewise the temperature dropped sharply from 25°C at 18h00 CEST to 10°C at 2h00 CEST. During this period the wind was coming from westerly directions (260°–310°). Wind speeds were rather low (h ( y ) y obs 2

>

@

T

>

@

B 1 y 0 y b0

@T O 1 >h ( y )

y obs

1 > p p b @T P 1 > p p b @ 2

(1)

@

where B , P , and O are error covariance matrices in discrete spaces for back0 ground initial values yb , parameters pb , and observations yobs , respectively. h is a projection operator, used to calculate the observation vector y obs (t ) from the model state variables. The optimal solution depends on the uncertainty of both observations and control variables, which are represented by the error covariance matrices, B , P , and O . As observational errors are often uncorrelated with each other, O is

G. Carmichael et al.

486

typically assumed as a diagonal matrix. The complexity of P depends on the choices of control parameters. The background error covariance matrix B is often correlated in space and between different species. An accurate estimation of the background error covariance matrix is difficult to provide and, given its huge dimensionality, simplifying approximations are required for the practical implementation. Information on the error statistics may be obtained using differrences between forecasts with different initialization time (NMC method) or ensemble methods based on a perturbed forecast-analysis system (Fisher, 2003). The data assimilation problem is then formulated as an optimization problem

min < p, y 0

for

p PADMISSIBLE, y 0 YADMISSIBLE

(2)

The minimization process is computationally demanding, but can be efficiently n implemented using adjoint modeling to compute the gradients o < and y m p < of the cost functional. Illustrative results on the application of this data assimilation capability using the ICARTT ozone measurements from different sources to improve the predicted ozone distributions are presented in (Chai et al., 2007). In this paper we demonstrated how the modeling and measurement activities of ICARTT can be used in the data assimilation framework. The ICARTT experiments produced comprehensive observation data sets and intense modeling applications upon which to study important aspects of data assimilation. The data used included in-situ ozone from the DC8 and P3, lidar observations from the DC3 and DC8, ozonsodes, MOZAIC profiles, and surface observations (AiRNow and AIRMAP). The model error correlation was constructed using the NMC approach. It is implemented into a 4DVar regional chemical data assimilation system with a truncated SVD regularization method is introduced. The observational (Hollingworth-Lonnberg) method was used to calculate the weighting between observations and model backgrounds in 4D-Var. It should be noted the increase of the computational time is minimal using the current approach, compared to using a diagonal matrix for the background error covariance. The weighting between the model and observations in determining the final optimal analysis depends on the both the background and observational error covariance matrices, which are objectively approximated in the current application. Ozone observations by different platforms during the ICARTT field experiment were assimilated into the regional CTM. It is found with little exception that assimilating observations from each individual platform improves the model predictions against the withheld observations. For example shown in Figure 3 are the domainaveraged vertical profiles (with standard deviation) constructed using the observations and the corresponding predictions for Case 9 (all the data from the sources listed in Figure 3 are used in the assimilation) and the base case (with out data assimilation). It clearly shows that the model biases both below and above 3,000 m

Predicting Air Quality: Current Status and Future Directions

487

Fig. 3 Example of the impact of data assimilation on predicted ozone distributions. Shown are forecasted (base-case) ozone fields for July 20, 2004. The results after the assimilation of ozone from aircraft (in-situ and remote-sensed), surface-based, and ozonesondes are also shown (case-9). Domain averaged vertical profiles and quantile plots are shown for comparison

are substantially reduced for Case 9. The predicted values for Case 9, now show a negative bias at low to mid- altitudes and a positive bias at high altitudes. Also shown are the quantile-quantile (q-q) plots of the ozone observations versus the corresponding predictions, for the base case and Case 9. Each point in a quantilequantile plot shows the values from two data sets that have the same quantile, i.e. the fraction of data points that fall below the given value. The q-q plot of the base case clearly shows the predictions are biased high overall. After assimilation, Case 9 generates a predicted ozone field that has a very similar population distribution as the observations. The q-q plot of Case 9 also indicates that the model has difficulty to generate low ozone concentrations (i @ = M p, yak 1 >i @ , 1 d i d N

(3)

The mean and the covariance of the forecast PDF are approximated by the ensemble statistics:

y kf =

1 N

N

¦ y kf >i@, Pfk i =1

i, j

=

1 N k ¦ y f >i@ y kf N 1 i =1

y >i@ k f

y kf

T

(4)

k

An ensemble of observation vectors { yobs [i ]}1di d N is constructed by adding to k the most recent observation vector yobs perturbations drawn from a normal distribution with zero mean and covariance Ok . Each member of the ensemble is k assimilated using the EKF to obtain the ensemble of analyzed states { ya [i ]}1di d N :

yak >i @ = y kf >i @ Pfk H kT Ok H k Pfk H kT

y >i @ H y >i @ 1

k obs

k

k f

(5)

The ensemble mean and covariance describe the PDF of the assimilated field. The cost of updating the covariance matrix is that of N model evaluations. The ensemble implicitly describes a density function that can be non-Gaussian. Experience gained in numerical weather prediction indicates that relatively small ensembles (50–100 members) are sufficient to accurately capture this density function. Extensions of this approach proposed in the literature include the Ensemble Kalman Smoother (Evensen and Leeuwen, 2000), the 4D-EnKF method, the Ensemble Transform Kalman Filter, the hybrid approach and ensemble nonlinear filters. The application of EnKF presents several challenges: (1) the rank of estimated covariance matrix is (much) smaller than its dimension; (2) the random errors in the statistically estimated covariance decrease only by the square-root of the ensemble size; (3) the subspace spanned by random vectors for explaining forecast error is not optimal; and (4) the estimation and correct treatment of model errors is possible but difficult. In addition, a careful implementation is required for efficiency. In spite of these challenges, EnKF has many attractive features including: (1) it is able to propagate the PDFs through highly nonlinear systems; (2) it does not require additional modeling efforts such as the construction of tangent linear model and its adjoint; and (3) the method is highly parallelizable. The performance of EnKF applied to chemical data assimilation has recently been reported (Sandu et al., 2005; Constantinescu et al., 2007a, b) for the ICARTT study discussed above (Figure 3). The observations used for data assimilation are ground-level ozone (O 3 ) measurements taken by the 340 EPA AirNow stations. These observations are available hourly in the assimilation window (0–23 EDT, July 20, 2004). After assimilation the model is allowed to evolve in forecast mode for another 24 hours. Results for the EnKF are shown in Table 1. For these calculations the “perturbed observations” implementation of the filter was employed. EnKF adjusts the concentration fields of 66 “control” chemical species in each grid point of the domain every hour in the assimilation window. The ensemble size was chosen to be 50 members to provide a good balance between accuracy and computational efficiency. An autoregressive model of background errors was used, which accounted for spatial correlations, distance decay,

490


and chemical lifetime. The “textbook application” of EnKF to this particular scenario lead to filter divergence and a decreasing ability towards the end of the assimilation window. We explored several ways to “inflate” the ensemble covariance in order to prevent filter divergence were investigated. These included: additive inflation (addition of uncorrelated noise to model results), multiplicative inflation (each member’s deviation from the ensemble mean is multiplied by a constant), and model-specific inflation (obtained through perturbing key model parameters like the wind field velocities, boundary conditions, and emissions). The results found that model-specific inflation best preserves the correlations between various chemical species. Illustrative results are shown in Table 1. Shown are results in the analysis window for initial conditions and for joint state (initial and boundary conditions and emissions) analysis. The performance of each data assimilation experiment is measured by the R 2 correlation factor. The correlation between the observations 2 and the model solution in the assimilation window is R = 0.24 for the non2 assimilated solution, R = 0.52 for 4D-Var (results not shown), and 2 R | 0.8 0.9 for EnKF (with various forms of covariance inflation and localization). The time evolution of ozone concentrations at selected ground stations (not shown) shows how the assimilated ozone series follow the observations much closer than the non-assimilated ones in the analysis window. The impact of data assimilation on the forecast skill is also shown in Table 1. The period from 24 to 48 hours represents the forecast. The forecast skill is increased; R2 increased from 0.28 for no-assimilation to 0.34–0.42 for the assimilation cases. The effect of assimilation of surface ozone on forecast improvements at particular sites is mixed. At some stations the effects are significant, while at others the effects are slight. This is due in part to the fact that only ground level observations are assimilated and the vertical profiles are not constrained at all. This is also due to the fact that near surface ozone levels are strongly dependent on chemical production/destruction processes involving a variety of precursor species. Table 1 Model-observations agreement (R2 and RMS [ppbv]) for the EnKF data assimilation of only the initial state and of joint state (ST), emissions (EM) and lateral boundary conditions (BC) parameters. Visible improvements in both analysis and forecasts are obtained.


491

4. Further Improvements in Forecasting One of the main differences between weather and chemical-weather forecasting is the strong local forcing due to emissions, which as discussed earlier are typically the largest source of uncertainty in the predictions. Thus an increase in air quality prediction skill requires better estimates of emissions. This is challenging due to the complexity and uncertainties associated with the bottom-up emission estimates (i.e., inventories built on activity information, e.g., emission factors, fuel use and type, control technologies, and their regional variation), the transient nature of some emissions (fires and dust storms), and their ever changing nature (e.g., trends due to policy and/or technology changes) (Streets et al., 2006). Improvements in emissions also will come from the closer integration of observations and models. The same data assimilation techniques discussed above can encompass emission estimates. (Pan et al., 2007). Another direction that is promising is to incorporate emissions as control variables, along with initial conditions and with boundary conditions, in the assimilation cycle. Unlike the traditional inverse modeling approach where the emissions are adjusted and then used in subsequent model runs, in this manner the emissions are adjusted each assimilation cycle. The importance of including the emission estimates as a control variable in order to improve prediction skill in air quality has been reported by Elbern et al. 2000, where they show a marked improvement in air quality forecast skill when emissions and initial conditions are simultaneously treated as controls. As we have discussed throughout this paper, improved predictions require a closer integration of measurements with models. The weather forecast system is supported by a comprehensive observing system designed to improve forecasting skill. No such system exists to support air quality forecasts. The chemical observations presently available were designed largely for environmental compliance and not to enhance predictive skill. However that opens the question as to what chemical data is needed to improve the predictions? The chemical data assimilation techniques can be used to help address this issue as well (Daescu and Carmichael, 2003; Daescu and Navon, 2004).

5. The Road Forward The importance of air quality prediction in the management of our environment continues to grow. The recent developments in atmospheric chemical observations and modeling are also leading to more effective linkages of air pollution issues on different scales from urban to global. For example it is now recognized that in urban air pollution studies it is important to consider also the regional and global contributions, and in global air pollution, the effects of megacities and hemispheric transport. In addition, the emergence of chemical weather forecasting as an important activity places greater importance on linking pollution detection and prediction

492


capabilities. The close integration of observational data is recognized as essential in weather/climate analysis and forecast activities, and this is accomplished by a mature experience/infrastructure in meteorological data assimilation. Borrowing lessons learned from the evolution of numerical weather prediction (NWP) models, improving air quality predictions through the assimilation of chemical data holds significant promise. As more atmospheric chemical observations become available chemical data assimilation is expected to play an essential role in air quality forecasting, similar to the role it has in NWPs (and may benefit weather forecasting as well). Advances in our predictive capabilities will require a better matching of the observational capabilities of the community with chemical weather forecast needs. This will require closer interactions between the observing and the modeling communities. One important activity will be the use of chemical data assimilation systems to help design the observing systems needed to produce better forecasts. We need to rigorously quantify the value-added to a forecast by: adding observations of additional species; extending surface coverage; including observations above the surface; and enhancing observations from satellites. Advances will also require a growth in activities related to chemical data assimilation techniques and algorithms. While there is much to build upon from the expertise and experiences in assimilation of weather, there are significant differences and challenges related to chemical weather. As we have illustrated in this paper 4DVar and EnKF are powerful techniques, and there are exciting possibilities in combining their strengths in hybrid data assimilation methods. However, there is relatively little experience in applying modern data assimilation techniques to real atmospheric chemistry problems, and much work needs to be done before their true impact on air quality prediction is felt. Furthermore, feedbacks between the meteorological and air quality components – which have mostly been studied as separate systems – are also critical to improve AQ forecasting. Many challenging and important questions remain to be addressed; including: What is the relationship between mixing depth heights and near surface concentrations? What is the role of ambient aerosols in influencing the surface energy budgets and in altering the moisture fields via cloud interactions? How do these feedbacks impact weather and AQ forecasts? To what extent will the assimilation of chemical data lead to improvements in weather forecasting? Sensitivity studies are needed to quantify these feedbacks, which in turn can help prioritize future research efforts. However, this will require a closer integration of meteorological and air quality models, and ultimately the evolution to a tightly coupled combined meteorological and air quality forecasting and data assimilation systems. These aspects are being explored in projects such as the European Union Global and Regional Earth-System (Atmosphere) Monitoring using Satellite and in-situ data (GEMS) project (reference). The integration of enhanced observing systems with modeling tools for use in air quality and climate change is a priority area within the Global Earth Observing System of Systems (GEOSS) and the Integrated Global Atmospheric Chemistry Observations (IGACO) (reference) frameworks. IGACO is a highly focused strategy for bringing together ground-based, aircraft


493

and satellite observations using atmospheric forecast models that assimilate not only meteorological observations but also chemical constituents. Finally, the requirements to use CTMs for urban to global scale applications, to couple chemistry with weather and climate, and to incorporate data assimilation, place even more demands for computational efficiency and accuracy. The growing importance of chemical weather forecasting to society should help stimulate significant advances in the field over the next decade. Acknowledgments This work has been supported by NSF though the award ITR AP&IM 0205198, and by grants from NOAA Global Change program and NASA. The work of A. Sandu and E.M. Constantinescu has also been supported in part by NSF CAREER ACI-0413872, NSF CCF-0515170, and by the Houston Advanced Research Center (HARC) awards H45C and H59.

References Chai T, Carmichael GR, Sandu A, Tang YH, Daescu DN (2006) Chemical data assimilation of transport and chemical evolution over the pacific (TRACE-P) aircraft measurements. J. Geophys. Res., 111(D02301), doi:10.1029/ 2005JD005883. Chai T, Carmichael GR, Sandu A, Hardesty M, Pilewskie P, Whitlow S, Browell E V, Avery MA, Thouret V, Nedelec P, Merrill JT, Thompson AM (2007) Four dimensional data assimilation experiments with ICARTT (International Consortium for Atmospheric Research on Transport and Transformation) ozone measurements. J. Geophys. Res., 112, D12515, doi:10.1029/2006JD007763. Constantinescu EM, Sandu A, Chai T, Carmichael GR (2007a) Ensemble-based chemical data assimilation. I: general approach. Quart. J. Roy. Met. Soc., 133, 1229–1243. Constantinescu EM, Sandu A, Chai T, Carmichael GR (2007b) Ensemble-based chemical data assimilation. II: covariance localization. Quart. J. Roy. Met. Soc., 133, 1245–1256. Dabberdt WF, Carroll MA, Baumgardner D, Carmichael G, Cohen R, Dye T, Ellis J, Grell G, Grimmond S, Hanna S, Irwin J, Lamb B, Madronich S, McQueen J, Meagher J, Odman T, Pleim J, Schmid HP, Westphal DL (2004) Meteorological research needs for improved air quality forecasting – report of the 11th prospectus development team of the us weather research program. Bull. Amer. Meteorol. Soc., 85(4):563+. Daescu DN, Carmichael GR (2003) An adjoint sensitivity method for the adaptive location of the observations in air quality modeling. J. Atmos. Sci., 60(2):434– 450. Daescu DN, Navon IM (2004) Adaptive observations in the context of 4D-Var data assimilation. Meteorol. Atmos. Phys., 85(4):205–226.

494


Elbern H, Schmidt H, Talagrand O, Ebel A (2000) 4d-variational data assimilation with an adjoint air quality model for emission analysis. Environ. Mod. Software, 15:539–548. Evensen G, van Leeuwen PJ (2000) An ensemble kalman smoother for nonlinear dynamics. Mon. Wea. Rev., 128(6):1852–1867. Fisher M (2003) Background error covariance modelling. In Proceedings of the ECMWF Workshop on Recent Developments in Data Assimilation for Atmosphere and Ocean, Reading, UK. Grell GA, Peckham SE, Schmitz R, McKeen SA, Frost G, Skamarock WC, Eder B, Petron G, Granier C, Khattatov B, Yudin V, Lamarque JF, Emmons L, Gille J, Edwards DP (2005) Fully coupled “online” chemistry within the wrf model. Geophys. Res. Lett., 39(37):6957–6975. Jacob DJ, Crawford JH, Kleb MM, Connors VS, Bendura RJ, Raper JL, Sachse G W, Gille JC, Emmons L, Heald CL (2003) Transport and chemical evolution over the pacific (trace-p) aircraft mission: design, execution, and first results. J. Geophys. Res., 108(D20):1–19. Kalman RE (1960) A new approach to linear filtering and prediction problems. Trans. ASME, Ser. D: J. Basic Eng., 82:35–45. Kalnay E (2003) Atmospheric modeling, data assimilation, and predictability. Cambridge University Press, Cambridge/New York. McKeen SA, Wilczak J, Grell G, Djalalova I, Peckham S, Hsie E, Gong W, Bouchet V, Menard S, Moffet R, McHenry J, McQueen J, Tang Y, Carmichael GR, Pagowski M, Chan V, Dye T, Frost V, Lee, P, Mathur R (2005) Assessment of an ensemble of seven real-time ozone forecasts over eastern North America during the summer of 2004. J. Geophys. Res., 110(D21):Art. No. D21307, December 2005. Menut L, Vautard R, Beekmann M, Honore C (2000) Sensitivity of photochemical pollution using the adjoint of a simplified chemistry-transport model. J. Geophys. Res., 105(D12):15379–15402. Pan L, Chai T, Carmichael GR, Tang Y, Streets D, Woo J, Friedli HR, Radke LF, Top-down estimate of mercury emissions in China using four-dimensional variational data assimilation (4D-Var). Atmos. Environ. (in review). Sandu A, Constantinescu EM, Liao WY, Carmichael GR, Chai TF, Seinfeld JH, Daescu D (2005) Ensemble-based data assimilation for atmospheric chemical transport models. In Computational Science – ICCS 2005, Pt 2, volume 3515 of Lecture Notes in Computer Science, pp. 648–655. Springer-Verlag Berlin,. Streets DG, Zhang Q, Wang L, He K, Hao J, Wu Y, Tang Y, Carmichael G (2006) Revisting China’s CO emissions after transport and chemical evolution over the pacific (TRACE-P): synthesis of inventories, atmospheric modeling and observations. J. Geophys. Res., 111(D14):Art. No. D14306. Tang YH, Carmichael GR, Thongboonchoo N, Chai T, Horowitz LW, Poerce RB, Al-Saadi JA, Pfister G, Vukovich M, Avery MA, Sachse GW, Ryerson TB, Holloway JS, Atlas EL, Flocke FM, Weber RJ, Huey LG, Dibb JE, Streets DG, Brune WH (2006) The influence of lateral and top boundary conditions on regional air quality prediction: a multi-scale study coupling regional and global chemical transport models. J. Geophys. Res. (in review).


495

Discussion A. Baklanov: What are your future plans with data-assimilation and when would it be operational? G. Carmichael: We are continuing to develop various data assimilation tools. We are continuing with the 4dvar approach, where we are focusing on in-situ and satellite observations for both forecast and inverse modeling of emission applications. We are working to provide ad joints for CMAQ as well as WRF/Chem applications. A first version of the CMAQ adjoint is now available and the WRF/Chem adjoint we hope to have in one to two years. We are also working on ensemble Kalman filter methods as well as hybrid methods.

5.4 Regional Coverage Modelling of Marine Aerosols Concentration in French Mediterranean Coastal Area Romain Blot, Gilles Tedeshi and Jacques Piazzola

Abstract The present study focuses on the extension of the predictions of the Mediterranean coastal aerosol model (based on the Navy Aerosol Model) to a regional scale, using a mesoscale meteorological model to better take into account the details of the topography and shoreline of the coast for calculations of the wind field and the fetch. The aerosol and the meteorological models are coupled and the spatial variation of the aerosol size distributions is determined in the whole study area. The results show a non-homogeneous spatial coverage of the aerosol concentrations over the northern Mediterranean, with wake due to the shoreline for a continental wind. Another dataset recorded in the Mediterranean in 1995 is used to validate the coupling for steady conditions. The results are found in correct agreement. A discussion is then held about the influence of unsteady conditions on the aerosol concentration is coastal zone.

Keywords: Aerosols, coastal area, fetch, rams, sea salt

1. Introduction The coastal area induces particular specificities for aerosol properties. Inshore and offshore aerosol sources, sinks and surface properties are quite different. At either side of the shoreline the aerosol properties will change gradually due to the mixture of continental and marine aerosol and depending on the direction the wind is blowing from, the fetch length, the sea state…. The aerosol concentration in the atmosphere is thus very variable both in time and in space. Very few relevant models for the aerosol size distributions were published during the last decades. One of the most used is the Navy Aerosol Model, NAM (Gathman, 1983) which provides the particle size distribution at a height of 10 m above sea level. Although this model gives reasonable predictions over the open ocean, experimental evidences show that it is often less reliable in coastal regions (Piazzola et al., 2000). To include coastal effects in the model for the prediction of aerosol concentrations, (Piazzola et al., 2003) proposed an extension of the Navy


514

Regional Coverage Modelling of Marine Aerosols Concentration in French Mediterranean

515

Aerosol Model to coastal areas. It was based on an extensive series of measurements in a Mediterranean coastal zone. The present paper focuses on the extension of the predictions of the Mediterranean coastal aerosol model to a regional scale, using a mesoscale meteorological model, RAMS (Cotton et al., 2003). It allowed to better take into account the details of the topography and shoreline of the coast for calculations of the wind field and the fetch, which are important inputs of the aerosol model. First, this paper presents the coupling between RAMS and the aerosol model into the study area. The spatial variation of the modelled aerosol size distributions is then determined in the whole study area. Another dataset recorded in the Mediterranean in 1995 (Piazzola and Despiau, 1998) is used to validate the coupling for steady conditions. As the coupling provides least performance for unsteady conditions compared to steady cases, a discussion is held about the influence of such conditions.

2. Field Site and Instrumentation The study area is the Toulon-Hyères bay (Figure 1) located on the French Riviera. The region is exposed to air masses coming from the open sea and to air masses coming from the European mainland, which case corresponds to continentally polluted conditions.

Fig. 1 Detailed view of the study area (the opened circles represent the locations of the aerosol concentrations measurements of the 1995 dataset)

The Mediterranean coastal aerosol model is based on an extensive series of measurements which were recorded on the island of Porquerolles during the 2000 and 2001 years. The measurement station was located west of the Porquerolles Island (Figure 1). The stations were equipped with meteorological sensors which were fixed at the top of a 10 m height mast and recorded wind speed and direction, air temperature, relative humidity and pressure. In addition, optical counters allow aerosol size distribution measurements. Size distributions in the 0.21–42.5 µm range were obtained using two Particle Measuring Systems: the classical scattering spectrometer probes CSASP-200 and CSASP-100HV. For validation of the model, a second dataset was used which was previously recorded on a French Navy ship “Albacore” in different locations of the Toulon bay in 1995 (see Figure 1) by Piazzola and Despiau (1998). The aerosol concentrations in the 0.1–20 µm size ranges were recorded using two Particle Measuring Systems, the active scattering spectrometer probe (ASASP) and the classical scattering spectrometer probe (CSASP).

516

R. Blot et al.

3. Aerosol and Meteorological Models As said previously, the Mediterranean coastal aerosol model is based on measurements recorded during 2000 and 2001. This extended period allowed observation of a large variety of aerosol size distributions for different meteorological conditions. The experimental data from Porquerolles were statistically analyzed to develop an empirical coastal aerosol model. The Mediterranean coastal aerosol model is a modification of the Navy Aerosol Model (NAM). The particle size distribution dN ( r ) dr is calculated as the sum of modified lognormal functions, but the coefficients of the various modes are parameterized as functions of not only the wind speed but also the fetch (Piazzola et al., 2000). As stated above, the wind speed and the fetch are the main inputs of the coastal aerosol model (with the relative humidity). The wave field growth depends on the duration of the wind and the fetch. In practice, the waves conditions are either fetch limited or duration limited. Steady state of the wave field correspond to fetch limited conditions for which we assume that the wind have blown constantly long enough in the same direction for wave heights at the end of the fetch to reach equilibrium (Hsu, 1986). These conditions include constant wind speed and direction, i.e., a steady state of the meteorological data. The occurrence of unsteady conditions and their implications on the predictions of the model are specifically commented in the paragraph 5. RAMS is a meteorological model used for numerical simulation of atmospheric processes. It employs multiple grid nesting and resolves (among others) the equations of motion, heat, moisture and mass using a staggered Arakawa-C grid and a terrain-following coordinate system. The kind of soil, of surface and variable SST are taken into account. For boundary conditions a 4DDA is used, allowing the atmospheric fields to be nudged toward large-scale data, as well as to local stations or radiosonde data. The topography and the vegetation cover (variable surface characteristics) used are both issued from the United States Geological Survey (USGS) 30” high resolution model and fitted to the different grids. RAMS has been already used for the simulation of local winds in the study area (Pezzoli et al., 2004; Guenard et al., 2006) and the results were found in agreement with experimental data. The model RAMS has been then implemented in the study area, and topography and surface type were fitted to the grids. Two two-way nested grids have been used, with 4 km (grid 1) and 1 km (grid 2) horizontal resolutions, covering respectively a 250 × 202 km and a 600 × 320 km over the French Mediterranean coast. A nudging was made every 6 hours using the ECMWF reanalyzed pressure levels data as well as the experimental recordings acquired on the Porquerolles Island to account for the very local coastal influence. Time steps for numerical integration were 4 s for grid 1 and 1 s for grid 2. As an example, the Figure 2 shows the wind field (10 m above surface level) simulated by the meteorological model at 1200 UTC 17 November 2000 for the finest grid. For a clearer view, only 5% of the vectors have been plotted. This case corresponds to a high intensity wind blowing from northwest to west direction on


517

the study area, which induces short fetch conditions for the aerosol measurements recorded in Porquerolles.

Fig. 2 Wind field simulated by RAMS for the northwest wind episode, 1200 UTC 17 November 2000

The validation of the simulation was made with wind data acquired at the Porquerolles station. As these data have been already used for the nudging, this is not a “‘real” validation (which is not the aim of the study), but rather a check that the meteorological field can be used accurately for input in the coastal aerosol model. 20

360 340

Exps RAM S

16 14 12

320 Wind direction ( °)

Wind speed (m/s)

18

10 8 6 4 2 0 17/11/2000 0000 UTC

300 280 260

RAMS Exps

240 220 200

17/11/2000 1200 UTC

18/11/2000 0000 UTC Time

18/11/2000 1200 UTC

19/11/2000 0000 UTC

180 17/11/2000 0000 UTC

17/11/2000 1200 UTC

18/11/2000 0000 UTC

18/11/2000 1200 UTC

19/11/2000 0000 UTC

T ime

Fig. 3 Comparison between simulated and experimental wind speed and direction at Porquerolles station location (17–18 November 2000)

4. Simulation of the Aerosol Concentration One of the interests of coupling the meteorological model with the aerosol model is to be able to re-calculate the inputs (wind and fetch) of the aerosol model for each cell of the grid, i.e. to obtain a spatial distribution. This is particularly interesting in a view of the determination of a relevant spatial cover of aerosols at mesoscale. The meteorological model also allows determination of the fetch conditions (duration limited, fetch limited conditions or fully developed sea), as various fetches can be found for different wind directions depending on the location of the study area. Two cases are presented here for simulation/experiment comparisons. The first one (at Porquerolles Island, 0000 UTC 18 November 2000) is part of the dataset used for the development of the Mediterranean coastal aerosol model. Even if the dataset covers a several months period and thus the model is not especially fitted to this short time case, it has been found preferable to use a second test case. This

R. Blot et al.

518

latter corresponds to measurements made for two locations in the Toulon bay, the 16 May 1995. 1,E+04

1,E+04

1,E+03

1,E+03 Exps Model

1,E+02

1,E+01 -1

1,E+00 1,E-01

-3

1,E-01

Exps Model

1,E+02

dN/dr(cm Pm )

-1

1,E+00

-3

dN/dr(cm P m )

1,E+01

1,E-02 1,E-03 1,E-04

1,E-02 1,E-03 1,E-04

1,E-05

1,E-05

1,E-06

1,E-06

1,E-07

1,E-07

1,E-08

1,E-08

0,1

1,0

10,0 D (Pm)

100,0

0,1

1

10

100

D(Pm)

Fig. 4 Comparison between simulated and measured aerosol size distributions in Porquerolles island (18 November 2000, left side) for a 25 km fetch and for a wind speed of 10 m/s and at two different locations in the Toulon bay (16 May 1995, right side) during a northwest wind of 7 m/s

The Figure 5 shows the simulation of the aerosol concentration in the study area for a constant wind blowing from the northwest direction at 1400 UTC 17 November 2000. The analysis was focused on particles of 5 and 1 µm diameter. These sizes were chosen because the behavior of these particles is representative for the marine and continental contributions to the aerosols, respectively. In the present analysis they can be used as tracers for the influence of the fetch on the production of marine aerosol and on the deposition of the continental particles.

Fig. 5 Spatial coverage of the 5 µm (left) and 1 µm (right) particle concentrations for a wind from NW direction and a velocity of 15 m/s at the Porquerolles station, 1400 UTC 17 November 2000

We can notice a non-homogeneous concentration field which is the consequence of the non-homogeneous wind field near the Mediterranean coast and more especially the consequence of the variation of the fetch length along the southern coast. This induces a spatial gradient of aerosol concentrations along the coasts. In particular, we can see the “sheltering” character of the coasts resulting in the occurrence of a “wake” which corresponds to lower aerosol concentrations.


519

5. Influence of Unsteady Conditions The present modelling is limited by the occurrence of unsteady conditions of the wave field and its effect on the production of marine aerosols. The wave field is in steady conditions when equilibrium is reached with the wind input. Before (or after) this state of equilibrium, the amplification (or attenuation) of the wave energy corresponds to unsteady conditions. In this case, the whitecap coverage, and hence the sea spray production, can be small even for high wind speeds before the equilibrium between waves and wind is reached. In the same way, the aerosol generation can be large for low wind speeds during a period of wave attenuation. The time for the waves to reach equilibrium with the wind (fetch limited conditions) can be estimated using the criterion for duration-limited growth proposed by Hsu (1986):

g t / U10

2 68.8 gX U10

23

(1)

where t is the minimum duration for the waves to be in equilibrium with the wind, X the fetch length and U10 the wind speed measured at a 10 m height. In order to analyse the error due to the unsteady conditions, a comparison is made between the temporal variation of the measured and the computed aerosol concentrations. On Figure 6 is plotted the error of the model for the aerosol concentrations (5 µm particles) recorded at Porquerolles island just after a change in the wind direction and intensity. The fetch length is 25 km. 25

Error (%)

20 15 10 5 0 17/11/00 17/11/00 17/11/00 17/11/00 17/11/00 17/11/00 17/11/00 18/11/00 18/11/00 10:00 UTC 12:00 UTC 14:00 UTC 16:00 UTC 18:00 UTC 20:00 UTC 22:00 UTC 0:00 UTC 2:00 UTC Time

Fig. 6 relative errors between computed and recorded aerosol concentrations measured at Porquerolles island the 17 November 2000

Three temporal periods can be identified (see also Figure 3): – Between 0000 UTC 17 November 2000 and 1200 UTC, the wind direction is changing from NW to W while the wind speed is increasing. The relative error is high (more than 20%) around 1100 UTC and decreasing continuously until 1400 UTC, as the atmospheric conditions are becoming more stable (see below). – Between 1200 UTC and 1700 UTC, the atmospheric conditions are roughly stable: the direction is W and the speed has reached its maximum. This induces a minimal error (less than 10%) with a 2 hours shift (around 1400– 1800 UTC).

R. Blot et al.

520

–

After 1600 UTC, the direction is changing again towards NW while the wind speed is decreasing continuously. The error is then increasing with a relative maximum (15%) 3 hours later (around 1900–2000 UTC). We can notice that the time to reach the minimal error is in accordance with the time for the waves to be in fetch limited conditions (3 hours for U10 | 15m / s ), as reported in Eq. (1). A more statistical study has been held. For five different modes (see Table 1) corresponding to diameters between 0.2 and 20 Pm, the accuracy of the model has been checked for steady-state cases only and for all cases (including unsteady conditions). For that, the parameter corresponding to a 68% confidence interval:

1

n

V

(2)

m

has been computed, where n is the number of records for a specific condition, m the mean value of records and V the standard-deviation. Table 1 Model accuracy (Eq. (2)) for a 68% confidence interval and for different particle sizes: comparison between steady conditions and all conditions (including unsteady). D (µm) 0.2 0.5 5 15 20

Steady conditions only 2.15 1.48 1.31 1.42 1.86

All conditions 2.80 1.70 1.49 1.50 2.00

The model accuracy is better for the second and the third modes (0.5 and 5 Pm particles) for the two cases and the largest error is found for the first mode (0.2 Pm particles). For steady-state cases, the model predicts the concentration of 0.5 µm and 5 µm particles within a factor of 1.5 and 1.3, respectively (with a 68% confidence level), which can be considered as a very valuable performance. As expected, when taking unsteady conditions into account the model performance is decreasing, especially for the smaller diameters for which the aerosols travel in the atmosphere is long. Their concentration depends then on atmospheric conditions as they were several hours sooner, that can be different from actual (at recording time) conditions. On the contrary, large particles do not stay a long time in the atmosphere and are less influenced by unsteady conditions.

6. Summary and Conclusion The present paper deals with the extension of the modelling of the Mediterranean coastal aerosol model to a regional scale. It has been coupled with a mesoscale meteorological model (RAMS) to take into account the details of the topography of the coast and the shoreline. The simulations have been validated using two different datasets recorded in the Mediterranean. The results show a good agreement between the modelled and the recorded aerosol concentrations at different locations


521

of the study area, under steady-state conditions. A non-homogeneous spatial coverage of the aerosol concentration has been found. In particular, a “sheltering effect” of the coast has been noticed. The present work, resulting in a 2D-space simulation of the aerosol concentration is of great interest for electro-optical apparatus performance. The least agreement observed between the modelled and the recorded aerosol concentration is mainly due to the occurrence of unsteady conditions of the wave field and its effect on the production of marine aerosols. Calculation of the model accuracy (for a 68% confidence interval) shows that the model predicts the concentration of the 5 µm diameter particles within a factor of 1.3 if only the data recorded during steady conditions are selected, whereas this factor is 1.5 for the whole dataset (including unsteady conditions). This confirms the steady-state character of the Mediterranean coastal aerosol model. It has been seen that the time to reach the minimal error between the modelled and the recorded data corresponds to the time for the waves to reach an equilibrium with the wind, and then to the time necessary for the aerosol concentrations to be in equilibrium. The relevance of the criterion proposed by Hsu for the determination of this time delay is confirmed by our data. One of the future objectives of this study will be to determine if it is possible to take unsteady conditions into account.

References Cotton WR et al. (2003) RAMS 2001: current status and future directions, Meteor. Atmos. Phys., 82, 5–29. Gathman SG (1983) Optical properties of the marine aerosol as predicted by the Navy aerosol model, Opt. Eng., 22, 57–62. Guenard V, Drobinski P, Caccia JL, Tedeschi G, Currier P (2006) Dynamics of the MAP IOP-15 severe Mistral event: observations and high-resolution numerical simulations, Quart. J. R. Met. Soc., 132, 757–777. Hsu SA (1986) A mechanism for the increase of wind stress (drag) coefficient With wind speed over water surfaces: a parametric model, J. Phys. Oceanogr. 16, 144–150. Pezzoli A, Tedeschi G, Resch F (2004) Numerical simulation of strong wind situations near the Mediterranean French coast: comparison with FETCH data, Appl. Meteorol., 43, 7, 997–1015. Piazzola J, Despiau S (1998) The vertical variation of extinction and atmospheric transmission due to aerosol particles close above the sea surface in Mediterranean coastal zone, Opt. Eng., 22, 57–62. Piazzola J, Van Eijk AMJ, De Leeuw G (2000) An extension of the Navy Aerosol Model to coastal areas, Opt. Eng., 39, 1620–1631. Piazzola J, Bouchara F, Van Eijk AMJ, De Leeuw C (2003) Development of the Mediterranean extinction code MEDEX, Opt. Eng., 42, 4, 912–924.

5.6 The Origins and Formation Mechanisms of Aerosol during a Measurement Campaign in Finnish Lapland, Evaluated Using the Regional Dispersion Model SILAM Marje Prank, Mikhail Sofiev, Marko Kaasik, Taina Ruuskanen, Jaakko Kukkonen and Markku Kulmala

Abstract This paper is intended to clarify the geographical extent of processes leading to a nucleation event and the role of atmospheric transport in it. The study is based on the inverse (adjoint) runs of atmospheric advection-diffusion model SILAM and general knowledge on basic mechanisms and time scales of nanometer particle formation in the atmosphere. Results of an aerosol measurement campaign carried out in Värriö, Finland, Eastern Lapland, April–May 2003, were used as sensitivity source data for backward tracing. The footprint areas of three observed nucleation events suggest that (1) spatial scale of a nucleation event may reach about 1,000 km, and (2) impact of atmospheric transport to the aerosol processes recorded by a Eulerian (ground-based) observer may be significant. Formation of an intense event over extensive forested areas supports the theory on the role of biogenic VOC emissions. Need for coupling the models of atmospheric transport and aerosol dynamics was stressed.

Keywords Adjoint model, atmospheric transport, backward tracing, nucleation burst

1. Introduction Most of observations and modelling studies of aerosol formation are based on homogeneous volume assumption, i.e., it is supposed that properties of the passing air mass do not vary substantially during the observation of a certain event at fixed measurement point. However, the meteorological conditions can substantially change on a time-scale of hours, or in specific cases, on a time-scale of minutes. The chemical composition of air is also changed when it is transported over different source and sink areas, and the pollutants can be scavenged by precipitation. Based only on limited experimental data, it is commonly not possible to deduce, whether such changes have possibly occurred recently. This can potentially lead to serious


530

The Origins and Formation Mechanisms of Aerosol during a Measurement Campaign

531

misinterpretations of aerosol dynamics. Thus, there is an obvious need to include detailed atmospheric dynamics into aerosol studies. Although atmospheric nucleation is obviously almost permanently present in the atmosphere, specific conditions and presence of necessary aerosol precursors are needed for nucleation bursts (e.g. Kulmala et al., 1998), when a large number of nanometer-size particles is produced during a short time period. After that newly formed particles grow due to condensation of vapours and reach the size of Aitken mode (30–100 nm) within a few hours. Although precise mechanism and all conditions for the nucleation burst are not known, a low concentration of large aerosol particles (i.e. low condensation sink) is typically necessary (see e.g. Kulmala et al., 2005). Thus, nucleation events occur predominantly during dry and sunny weather in presence of a clean air mass and sufficient solar radiation. Due to biogenic emissions the boreal forest is an important source of fine aerosol (e.g. Tunved et al., 2006). It is found that nucleation events in boreal forest occur even in late winter with snow cover, as life activities in evergreen tree crowns appear with sufficient solar radiation (Kulmala et al., 2004). This paper is intended to clarify the geographical extent of processes leading to nucleation burst and the role of atmospheric transport in it. The basic tools for that are (1) an atmospheric advection-diffusion model, and (2) results of an aerosol measurement campaign as the input data for that model. rw

ay

Murmansk Nikel Monchegorsk Apatity Värriö

en

No

Sw

ed

Russia

Finland

St. Petersburg

Helsinki Tallinn Estonia

Fig. 1 Location map: empty circles – major cities; small filled circles – major industrial pollution sources (metal smelters); cross – measurement location

We have applied the source apportionment modelling techniques based on adjoint formalism in order to trace the air masses backward from the receptor point and to follow the dispersion of potentially aerosol-forming species from anthropogenic and natural sources. The adjoint modelling techniques have substantial advantages compared to the previously more widely applied trajectory analyses. Atmospheric dispersion and removal processes can be properly taken into account – this also

532

M. Prank et al.

facilitates the separate modelling of various aerosol size modes. Solution of the adjoint dispersion equation provides a prediction of the geographic areas, where the sources responsible for the observed concentrations are located, and quantitative estimates of their contributions to the measured concentrations. In case of trajectory analyses, merely a qualitative picture of the transport path of a single air parcel can be modelled, as driven by prevailing airflows. The trajectories for various particle size modes would be identical. Trajectory analyses also do not yield any information on the most probable source areas as a function of distance along the trajectory.

2. Materials and Methods The field experiments were carried out at the SMEAR I site (Station for Measuring Forest Ecosystem – Atmosphere Relation, 67q 46’ N, 29q 35’ E), located in Värriö nature park in eastern Lapland, less than 10 km from the border of Russia, 100–200 km far from major pollution sources at Kola Peninsula (Figure 1). Campaign included measurements of aerosol particle size distributions with EAS (electric aerosol spectrometer, Tammet et al., 2002) from April 28 to May 11, 2003. A description of campaign design and results (records of particle size distributions) is given by Ruuskanen et al. (2007). The EAS was used to measure the aerosol size distribution in the aerodynamic diameter range of 3 nm–10 Pm with spectral resolution of eight fractions per decade (i.e., 28 fractions in total) and time resolution of 10 minutes. This study is focused on the nucleation (3–24 nm, based on exact EAS fractions) and the Aitken mode (24–100 nm) particles. The SILAM model version 3.7 used in this study is based on a Lagrangian particle Monte-Carlo dynamics. The model has been developed at the Finnish Meteorological Institute and it has been validated against the ETEX meso-scale dispersion experiment, the results of various field measurement campaigns, and the long-term air quality observations of EMEP (Sofiev et al., 2006a). For aerosol, the SILAM considers advective and turbulent transport, and size-dependent dry (including sedimentation) and wet deposition. The SILAM model has two modes of operation: forward and adjoint (Sofiev et al., 2006b). In the forward mode, the input data contains the emissions from specified sources, meteorological fields produced by numerical weather prediction models, and land use. The output of the forward simulations consists of the 3D spatial concentration patterns developing in time and 2D dry and wet deposition fields. In the adjoint mode, the model input in the specific case of this study contains measured aerosol concentrations (the so-called sensitivity source function), and the meteorological fields produced by the ECMWF numerical weather prediction model. The output (the so-called sensitivity distribution) is a 4D probability field for the sources of observed concentrations. This output specifies the probability that the measured concentration is originated from a specific location or region; it could also be called the footprint of the observations.


533

We have performed the adjoint simulations separately for each size mode. The dry and wet deposition processes are included; however, the current model setup does not include adjoint aerosol dynamics or chemistry. Generally, a predicted positive sensitivity distribution value can be interpreted as the presence of an aerosol or its precursor (if chemical transformation is taken into account) in the specific volume of air at the specific time moment that has contributed to the corresponding observed concentration. Pollution sources or concentrations that are not predicted as positive sensitivity distribution values have not contributed to the observed concentration peaks. Forward model runs were also performed in order to investigate the contribution of anthropogenic and natural emissions of sulphur, primary particulate matter (PM2.5) and sea salt to the aerosol detected at Värriö. Anthropogenic emission data is based on the EMEP database. The horizontal resolution of model output was 20 km, and the concentration and sensitivity distribution fields were saved each 15 minutes. Output in the vertical direction consisted of five layers up to the height of 3,150 m above the ground surface. For the numerical results presented here, the output fields were averaged over the two lowest layers (in-total from the ground level up to a height of 150 m). The wind vectors overlaid in the concentration maps are also averaged over the two lowest layers, corresponding to a layer from the ground level to approximately a height of 150 m.

3. Results The measured mass concentrations of the nucleation mode and Aitken nuclei are presented in Figure 2. The nucleation-mode peaks occurred on 30th of April (event 1), and on 5th and 9th of May (events 2 and 3, respectively). The highest peak of Aitken nuclei (also, coarser particles) was observed at April 30, just after event 1.

Fig. 2 Hourly average mass concentrations of aerosol in nucleation and Aitken modes measured by the EAS during the Värriö campaign in April–May 2003

534

M. Prank et al.

Before event 1, the concentrations of all size modes were anomalously low and, after a few hours, highly polluted air masses were observed with large amount of Aitken nuclei and remarkable amount of larger particles. Also, a remarkable amount of nucleation mode particles appeared again a few hours after change in air mass (Figure 3). The first nucleation event is interesting due to the dramatic change of air masses at Värriö that interrupted its observation. It is rather complicated, on the basis of the performed SILAM run, to highlight the origin of its precursors than in the case of first event. We can expect either emissions from the Arctic Ocean or biogenic emissions from coniferous forests of Lapland during a few hours of transport over land areas before reaching the measurement site. Although the area was still covered with snow and the temperatures were slightly below zero, solar heating of tree crowns might induce some vegetation activity. As the air masses advected very lose to the Nikel metallurgy factory (Russia), we cannot exclude some triggering influence of gaseous emissions from there. According to the forward run with a local correction in EMEP-based database (Kaasik et al., 2007), the polluted air mass (incl. the peak of Aitken nuclei) originated from the Nikel area.

Fig. 3 Plot of fractional number concentrations of aerosol particles in Värriö, April 30

The second nucleation event started on 5th of May at 11 a.m. GMT, when a large number of nanometer particles grew to Aitken sizes in late evening. Inverse computations show that the air masses spent two previous days over the continental areas of northern Sweden and central Finland (most of time over boreal forest) and were transported over the Botnian Bay (Figure 4). When the nucleation event began at Värriö, a well-defined footprint was formed that extended over the Finnish Lapland from north-east to south-west (Figure 4b). During and shortly after this nucleation event the forward model computations did not show any advection of substantial concentrations of sulphate over Värriö. Sunny and relatively warm weather during these days conditioned accumulation of biogenic aerosol precursors in the air mass. Considering the typical time scale of nucleation, the observed nano-particles were formed during the morning hours of May 5 over the southern Lapland.


535

The nucleation event 2nd of May 5 (Figure 4) was followed by a long “tail” of Aitken nuclei. When the event started, the source area of Aitken nuclei extended along the western coast of Finland to the Baltic Sea and reached the coast of Latvia (Figure 5). The maximum of Aitken mode particle concentration at Värriö occurred a few hours later than that of the nucleation mode particles, in the late evening of May 5 with remarkable concentrations of Aitken nuclei observed until the evening of May 6. Thus, we suppose that these Aitken nuclei were the aged particles from the nucleation event that took place during May 5 all over the western Finland and was observed at Värriö as two sequential peaks of concentration – first of the nucleation-mode particles and then of Aitken ones. Considering the part of “tail” over the Baltic Sea, some contribution of sea salt is possible. a)

c)

b)

+

+

V ä r r iö

+

V ä r r iö

V ä r r iö

Fig. 4 Time evolution of the sensitivity distribution (footprint) of the nucleation mode particle mass concentration for the nucleation event 2: (a) May 4, 19 p.m. GMT, (b) May 5, 11 a.m., and (c) May 6, 2 a.m. a)

c)

b)

+

V ä r r iö

+

V ä r r iö

+

V ä r r iö

Fig. 5 Sensitivity distributions of Aitken mode particle mass concentration, nucleation event 2 (Figure 2): (a) May 5, 12 a.m. GMT, nucleation started at Värriö, Aitken “cloud” approaching; (b) May 6, 0 a.m., nucleation mode is vanishing, long “tail” of Aitken mode footprint is approaching Värriö; (c) May 6, 18 p.m., end of the elevated concentration of Aitken-mode particles

M. Prank et al.

536

Looking at fractional aerosol concentrations during the nucleation event 2, it is evident that uniform condensation growth of particles is disturbed (Figure 6). Continuous growth of particles is seen in size range of 3–10 nm, but a small secondary maximum, appearing about 3.5 hours after the main one, becomes equal to first one for particles about 20 nm in diameter. Both maximums propagate through fractions very fast, with seeming growth rate up to 27 nm/h (Kaasik et al., 2006). Neither such a structure nor growth rate can be explained by condensation growth only. A likely explanation is an advective effect: condensation growth in the air mass approaching later might start earlier; the gap in time series may appear due to air mass that had no significant nucleation due to some reason. a)

b)

Fig. 6 Evolution of number concentration in fraction 1 (3.2–4.2 nm) and fraction 7 (18–23 nm) during the nucleation event 2: data points and polynomial fit (used for determining the growth rate of particles, see Kaasik et al., 2006). Local winter time: GMT + 2 hours

In the case of nucleation event on 8th of May, the air masses were transported from the Norwegian Sea. A well-defined footprint was formed that extended over the Finnish and Norwegian Lapland from north-west to south-east. Thus, the leading role of marine emissions is expected.

4. Conclusions Detailed inverse dispersion modelling provides new insights and improves the understanding of processes related to new particle formation (nucleation): it localises the areas contributing to emission of both aerosol particles and their precursors, and provides a detailed time schedule of the new particle formation event. The nucleation process, where the biogenic VOC are present cannot be assumed to take place at a single place: the nucleation and particle-growth event can extend over hundreds and sometimes thousands of kilometres. The need for incorporating the aerosol formation and development processes into an atmospheric transport model is recognised and corresponding cooperation project between the Finnish Meteorological Institute, University of Tartu and University of Helsinki is initiated.


537

Acknowledgments This study was supported by Nordic Research Board (NordForsk) – cooperation network NetFAM, Maj and Tor Nessling Foundation (Finland) and the Estonian Science Foundation, research grant 7005.

References Kaasik M, Sofiev M, Prank M, Ruuskanen T, Kukkonen J, Kulmala M (2006) Model-delineated origin and growth of particles during the nucleation events observed in Värriö campaign in 2003. In: Proceedings of BACCI, NEC and FcoE activities in 2005 Report Series in Aerosol Science, 81, 221–226. Kaasik M, Prank M, Kukkonen J, Sofiev M (2007) A suggested correction to the EMEP database, regarding the location of a major industrial air pollution source in Kola Peninsula (in this volume). Kulmala M, Toivonen A, Mäkelä JM, Laaksonen A (1998) Analysis of the growth of nucleation mode observed in Boreal forest. Tellus B, 50B, 449–462, 1998. Kulmala M, Boy M, Suni T, Gaman A, Raivonen M, Aaltonen V, Adler H, Anttila T, Fiedler V, Grönholm T, Hellen H, Herrmann E, Jalonen R, Jussila M, Komppula M, Kosmale M, Plauškaite K, Reis R, Savola N, Soini P, Virtanen S, Aalto P, Dal Maso M, Hakola H, Keronen P, Vehkamäki H, Rannik Ü, Lehtinen K E J, Hari P (2004) Aerosols in boreal forest: wintertime relations between formation events and bio-geo-chemical activity. Boreal Environment Research, 9, 63–74. Kulmala M, Petäjä T, Mönkkönen P, Koponen IK, Dal Maso M, Aalto PP, Lehtinen KEJ, Kerminen V-M (2005) On the growth of nucleation mode particles: source rates of condensable vapour in polluted and clean environments. Atmospheric Chemistry and Physics, 5, 409–416. Ruuskanen TM, Kaasik M, Aalto PP, Hõrrak U, Vana M, Mårtensson EM, Yoon YJ, Keronen, P, Mordas G, Ceburnis D, Nilsson ED, O’Dowd C, Noppel M, Alliksaar T, Ivask J, Sofiev M, Prank M, Kulmala M (2007) Concentrations and fluxes of aerosol particles during the LAPBIAT measurement campaign in Värriö field station. Atmospheric Chemistry and Physics, 7, 3683–3700. Sofiev M, Siljamo P, Valkama I, Ilvonen M, Kukkonen J (2006a) A dispersion modelling system SILAM and its validation against ETEX data. Atmospheric Environment 40, 674–685. Sofiev M, Siljamo P, Ranta H, Rantio-Lehtimäki A (2006b) Towards numerical forecasting of long-range air transport of birch pollen: theoretical considerations and a feasibility study. International Journal on Biometeorology, doi:10 1007/s00484-006-0027-x, 50, 392–402. Tammet H, Mirme A, Tamm E (2002) Electrical aerosol spectrometer of Tartu University. Atmospheric Research 62, 315–324. Tunved P, Hansson H-C, Kerminen V-M, Ström J, Dal Maso M Lihavainen HY, Viisanen Y, Aalto PP, Komppula M, Kulmala M (2006) High natural aerosol loading over boreal forests. Science, 312, 5771, 261–263, doi:10.1126/science.1123052.

538

M. Prank et al.

Discussion Gryning: The events sub-parts footprints. You said they are about the same, but they look very different. Please comment. M. Prank: The footprints shown are for the same moment in time, while the subparts of the event happen at different times, so latest one has several hours more to reach the measurement station. Also, as the landuse in all the areas is the same, the minor differences in the subfootprints are not enough to explain the huge drop in the measured concentrations. B. Fisher: Could you tell us what the biogenic nucleation material consists of? M. Prank: By biogenic emissions I mean precursors, biogenic VOCs like terpenes and Į-pinene

6.2 Impacts of Climate Change on Air Pollution Levels in the Northern Hemisphere with Special Focus on Europe and the Arctic Gitte B. Hedegaard, Jørgen Brandt, Jesper H. Christensen, Lise M. Frohn, Camilla Geels, Kaj M. Hansen and Martin Stendel

Abstract The evolution in air pollution levels and spatial distribution in the 21st century is investigated with respect to climate change. The coupled atmosphereocean general circulation model ECHAM4-OPYC3 is providing meteorological fields for two time slices (1990s and 2090s) to the chemical long-range transport model DEHM-REGINA. The dominating impacts from climate change on a large number of the chemical species are related to the predicted temperature increase since most of the reaction rates of the involved species are temperature dependent. The ECHAM4-OPYC3 projects a global mean temperature increase of 3 K with local maxima up to 11 K in the Arctic. As a consequence of this temperature increase, the temperature dependent biogenic emission of isoprene is predicted to increase significantly over land by the DEHM-REGINA model simulation. This leads to an increase in the ozone production and together with an increase in water vapour to an increase in the number of free OH radicals. Furthermore an increase in the number of radicals contributes to a significant change in the typical life times of many species, since hydroxyl radicals are participating in a large number of chemical reactions.

Keywords Air pollution, biogenic emissions, chemical transport model, climate change, coupled models, isoprene

1. Introduction Recently, there has been a growing interest in the effects of climate change on future air pollution levels. It is well known that the composition of the atmosphere will change due to changes in anthropogenic emissions. According to the newly released IPCC report (Solomon et al., 2007) some meteorological parameters will change in the future due to the man-made changes of the composition of the atomsphere. E.g. a general temperature increase will affect many if not all other


568

Impacts of Climate Change on Air Pollution Leve

569

meteorological parameters and since the distribution of air pollution is highly dependent on the meteorology, it is hypothesized, that the air pollution levels and spatial distribution even with unchanged emissions will be changed in a warmer climate. Here the hemispheric chemical long-range transport model DEHM-REGINA (Danish Eulerian Hemispheric Model – REGIonal high resolutioN Air pollution model) is used to investigate the future air pollution levels and distribution in the northern hemisphere.

2. Experimental Design The general circulation model ECHAM4-OPYC3 (see Roeckner et al., 1996, 1999; Stendel et al., 2002 for model descriptions) is providing meteorology for the 21st century and part of the 20th century based on the IPCC SRES A2 scenario (Nakicenovic et al., 2000). The meteorology is saved every 6 hours and here after used as input to the chemical long-range transport model DEHM-REGINA (see Christensen, 1997; Frohn et al., 2002, 2003; Frohn, 2004 for a full model description). In order to save computing time the experiment is carried out as a timesliced experiment instead of simulating the 21st century in one continuous run. The time-slices are the two decades 1990s and the 2090s. In Figure 1 the model setup is sketched. The six-hourly climate data is used as a one-way input to the DEHM-REGINA model. The DEHM-REGINA model also needs an emission input. In the experiment carried out here the anthropogenic emissions are conserved at a 1990 emission level. The emissions consist of a combined data set from the EDGAR, GEIA and EMEP data bases (cf. Hedegaard, 2007). The chemical transport model keeps track of the transport, chemistry, depositions and emissions of 63 chemical species and the model includes 120 of the most important chemical reactions between these species. Horizontally the model has a resolution of 150 × 150 km. The ECHAM4 model on the other hand is defined in a spectral grid with truncation T42 (T42 corresponding roughly to a 2.8˚ × 2.8˚ transformed grid). Vertically ECHAM4 is defined in a hybrid sigma-pressure coordinate system and divided into 19 layers extending from the surface of the earth to the 10 hPa pressure level. The DEHM-REGINA is defined in sigmapressure coordinate system and is divided into 20 irregularly distributed layers extending from the earth surface to 100 hPa pressure level. Therefore in order to use the meteorological fields from ECHAM4-OPYC3 as input data to the DEHMREGINA model a transformation has been carried out (cf. Hedegaard, 2007; Hedegaard et al., 2007).

G. Hedegaard et al.

570

ECHAM4 Sea ice surface/ mixed layer

A2 Emissions EMEP-GAIA-EDGAR

OPYC3

interior

3D advection 3D dispersion Chemistry

Emissions (nat.+ant)

DEHMREGINA

Wet and dry deposition

Fig. 1 Off-line setup of the ECHAM4-OPYC3 and the DEHM-REGINA model. In this setup the emissions are kept constant at a 1990 level in order to separate out the effect of climate change. The meteorological input originates from the climate model ECHAM4-OPYC3 which is a coupled atmosphere-ocean model and in the simulations used here forced with the IPCC A2 emission scenario

3. Results and Discussion The chemical species analyzed are, sulphur (SO2), sulphate (SO4), ozone (O3), nitrogen dioxide (NO2), PM10 (particular matter with diameters below 10 ȝm), sea salt, hydroxyl radical (OH) and isoprene (C5H8). Also ammonium (NH4), ammonia (NH3), nitrate (NO3), nitrogen oxides (NOx), TSP (total suspended particles) and PM2.5 (particular matter with diameter below 2.5 ȝm) have been treated in this analysis. We will discuss a few of these species here. The result of the total analysis is given in Hedegaard (2007).

3.1. Changes in ozone and isoprene In Figure 2 (left figure) the difference in the average 2 m temperature of the two time slices (2090s -1990s) is shown. The temperature is increasing everywhere in the domain. This general temperature increase with local hot spots over Southern Europe and the Arctic is similar to other model results (Stendel et al., 2002). The difference in the ten-year average ozone concentration of the two decades is displayed in Figure 3 (left plot). A latitudinal dependence is evident in this figure. Ozone concentrations increase in the future and the increase gets stronger with increasing latitude. North of approximately 30ºN the increase is highly significant (cf. right plot of Figure 3). South of 30ºN difference in ozone concentration changes and in the equatorial areas the ozone concentrations levels tends to


571

decrease significantly between the two decades. However, also a blurred landocean contrast in the ozone increase is evident and the ozone concentration generally increases less over the ocean.

Fig. 2 Left: 2 m temperature difference between the mean values of the two decades 2090s -1990s

in K˚. Right: The statistical significance of the changes of mean values between the two decades according to the t-test. The threshold value for significance is chosen to be within the 0.95 fractile corresponding to the 10% significance level (which is the same as the 1,734 percentile value in the plot). Dark colours (positive percentile values) indicate a significant increase and light shaded colours (negative percentile values) indicate a significant decrease. The projected temperature increase is highly significant everywhere in the domain (black everywhere)

Fig. 3 Left: The difference in ozone concentration between the mean values of the two decades 2090s 1990s in percent. Right: The statistical significance of the changes of mean values between the two decades according to the t-test. Threshold values as in Figure 2

572

G. Hedegaard et al.

The ozone production is strongly dependent on the presence of the precursors NOx and Volatile Organic Compounds (VOC). In the experiment analyzed here, the anthropogenic emissions are kept constant. However, VOCs also have biogenic emitters, which can alter their emissions due to changes in meteorology. The only natural VOC emitter included in the DEHM-REGINA model is isoprene. Isoprene is through participation in chemical reactions with OH acting as a sink for hydroxyl radicals (For further details see Hedegaard, 2007; Hedegaard et al., 2007). In DEHM-REGINA, the submodel BEIS (Biogenic Emissions Inventory System) is included to account for the biogenic isoprene emissions (Guenther et al., 1995). Isoprene is emitted from trees and other plants and therefore primarily existent over land. From Figure 4 (left plot) it can be seen that the concentrations of isoprene is expected to increase where there are emitters present and this increase is highly significant (right plot). The general increase in isoprene concentration over land due to the temperature increase can contribute to explain the increase in ozone, which posses a blurred land-sea contrast in the distribution field. The projected observed level of isoprene will alter the ozone production in a positive direction and thereby enhancing the ozone level.

Fig. 4 Left: the difference in isoprene concentration between the mean values of the two decades 2090s–1990s in ppbV. Right: the statistical significance of the changes of mean values between the two decades according to the t-test. Threshold values as in Figure 2

Langner et al. (2004) used the regional chemistry/transport/deposition model MATCH to simulate the distribution of surface ozone in the future. The Rossby Centre regional atmospheric climate model (RCA) version 1 provided the projected meteorology in these simulations. Langner et al. (2004) found a general increase in the surface ozone concentration over southern and Central Europe. They calculated the domain-total emission of isoprene to increase with 59% due to the predicted temperature increase. This is generally consistent with the results found in the current experiment.


573

Langner et al. (2004) state that the predicted changes found in surface ozone concentrations are substantial and if the climate scenario (IPCC, IS92a) is representative for the future climate, the increase in surface ozone due to the predicted warming would be significant compared to the expected reductions resulting from the emission reduction protocols currently in force. This is in line with Tuovinen et al. (2002) who made a sensitivity analysis of which factors will effect the surface ozone concentration in Europe. They found that the increased biogenic VOC emissions significantly will counteract the effects of reduced anthropogenic emissions. Murazaki and Hess (2006) have also studied the contribution from climate change on the future ozone levels and spatial distribution with the global chemical transport model MOZART-2. Substantially different from the simulations carried out here, Murazaki and Hess (2006) kept both the anthropogenic and biogenic emissions constant (Personal e-mail correspondence with P. Hess, 2006). Murazaki and Hess (2006) divide the surface ozone into two contributions; locally produced ozone and background ozone. In the high emission areas the local increase in ozone is expected to exceed the decrease in background ozone resulting in a net increase. On the contrary a net decrease in ozone is predicted away from these high-emission zones. In the work carried out here the ozone concentration is predicted to increase everywhere over the United States (cf. Figure 3). This difference relative to results of Murazaki and Hess (2006) is probably due to the lack of the temperature dependence of the biogenic emissions in the experiment carried out by Murazaki and Hess (2006). These emitters are as earlier mentioned ozone precursors and by the results of this work they contribute to a relatively large increase in ozone concentration over land.

3.2. Changes in typical life times In this simulation the specific humidity (not shown) is predicted to increase and thereby increasing the number of H2O molecules in the atmosphere. When ozone already is present, more H2O molecules will lead to more hydroxyl radicals through the process shown in the chemical reaction scheme Eq. (1).

O3 hQ o O2 O O H 2 O o 2OH

(1)

It is predicted that the temperature, specific humidity and ozone concentration (only north of approximately 30º N) will increase from the 1990s and to the 2090s. By the reaction scheme Eq. (1) these projected increases must lead to an increase in the number of hydroxyl radicals. The predicted increase by the reasoning above, are confirmed from the concentrations plots of hydroxyl radicals (not shown) (cf. Hedegaard, 2007); Hedegaard et al., 2007). By the projected general increase in the ozone, it can be concluded that DEHM-REGINA predicts an increase in the

574

G. Hedegaard et al.

reaction rates in a great number of chemical reactions over sea and at higher altitudes due to the resulting increase in hydroxyl radicals (Hedegaard et al., 2007) which will have a great influence on the life times of many chemical species. In Hedegaard (2007) and Hedegaard et al. (2007) it is shown that the life times of for example nitrogen dioxide will be reduced and lead to an increased level in nitrate (NO3) and nitric acid (HNO3). Also the sulphate production through the conversion of sulphur will increase in the future.

4. Conclusions The predicted general temperature increase throughout the 21st century results in an increase in the biogenic emissions of isoprene. Isoprene is an important ozone precursor and the ozone production is projected to increase significantly by these simulations. This increase in ozone together with an increase in specific humidity is found to enhance the chemical reaction rates of a great number of chemical reactions. The humidity and ozone increase results in an increase in the number of hydroxyl radicals, which are the activating agent in many chemical processes. For example an indication of a enhanced sulphur to sulphate conversion and decreased life times of some primary species were found.

References Christensen JH (1993) Testing Advection Schemes in a Three-Dimensional Air Pollution Model, Mathematical and Computational Modelling 18 (2), 75–88. Christensen JH (1997) The Danish Eulerian Hemispheric Model – A threedimensional air pollution model used for the Arctic, Atmospheric Environment 31 (24), 4169–4191. Frohn LM (2004) A study of long-term high-resolution air pollution modelling, PhD thesis, University of Copenhagen and National Environmental Research Institute, Aarhus University, Denmark, pp 203. Frohn LM, Christensen JH, Brandt J (2002) Development of a High-Resolution Nested Air Pollution Model. The Numerical Approach, Journal of Computational Physics 179 (1), 68–94. Frohn LM, Christensen JH, Brandt J, Geels C, Hansen KM (2003) Air pollution modelling using a 3D hemispheric nested model, Atmospheric Chemistry and Physics Discussions 3, 3543–3588. Guenther A, Hewitt CN, Erickson D, Fall R, Geron C, Graedel T, Harley P, Klinger L, Lerdau M, McKay WA, Pierce T, Scholes B, Steinbrecher R, Tallamraju R, Taylor J, Zimmerman P (1995) A global-model of natural volatile organiccompound emissions, Journal of Geophysical Research 100, 8873–8892. Hedegaard GB (2007) Impacts of climate change on air pollution levels in the Northern Hemisphere, Technical Report 240, National Environmental research


575

Institute, Aarhus University, Frederiksborgvej 399, P.O. Box 358, 4000 Roskilde, Denmark. Hedegaard GB, Brandt J, Christensen JH, Frohn LM, Geels C, Stendel M (2007) Impacts of climate change on air pollution levels in the Northern Hemisphere with special focus on Europe and the Arctic, Atmospheric Chemistry and Physics Discussions, submitted October 2007. Langner J, Bergstrøm R, Foltescu V (2004) Impact of climate change on surface ozone and deposition of sulphur and nitrogen in Europe, Atmospheric Environment 39, 1129–1141. Murazaki K, Hess P (2006) How does climate change contribute to surface ozone change over the United States, Journal of Geophysical Research 111 (D05301). Nakicenovic N, Alcamo J, Davis G, de Vries B, Fenhann J, Gaffin S, Gregory K, Grübler A, Jung TY, Kram T, La Rovere EL, Michaelis L, Mori S, Morita T, Pepper W, Pitcher H, Price L, Riahi K, Roehrl A, Rogner HH, Sankovski A, Schlesinger A, Shukla P, Smith S, Swart R, van Rooijen S, Victor N, Dadi Z (2000) Special Report on Emission Scenarios, Cambridge University Press, Cambridge, United Kingdom/New York, , pp. 570. Roeckner E, Arpe K, Bengtsson L, Christoph M, Clausen M, Dümenil L, Giorgetta M, Schlese U, Schulzweida U (1996) The atmospheric general circulation model ECHAM-4: Model description and simulation of present-day climate 218, pp. 167, Max-Planck-Institut für Meteorologie, Hamburg, Germany. Roeckner E, Bengtsson L, Feichter J (1999) Transient climate change simulations with a coupled atmosphere-ocean GCM including the tropospheric sulfur cycle, Journal of Climate 12, 3004–3032, Max-Planck-Institut für Meteorologie, Hamburg, Germany. Solomon S, Qin D, Manning M, Chen Z, Marquis M, Averyt KB, Tignor M, Miller HL (2007) Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, pp. 996, Cambridge University Press, Cambridge, United Kingdom/New York. Stendel M, Schmidt T, Roeckner E, Cubasch U (2002) The Climate of the 21st century: transient simulations with a coupled atmosphere-ocean circulation model, Danish climate centre (02-1), pp. 50, Copenhagen, Denmark. Tuovinen JP, Simpson D, Mayerhofer P, Lindfors V, Laurila T (2002) Surface ozone exposures in Northern Europe in changing environmental conditions. In: Hjort J, Raes F, Angeletti G (eds), A Changing Atmosphere: Proceeding of the 8th European Symposium on the Physico-Chemical Behaviour of the Atmospheric Pollutants, European Commission, DG Research, Joint Research Centre, CD-ROM, Paper Ap6, pp. 6.

576

G. Hedegaard et al.

Discussion D. Stein: Does the climate model capture correctly the frequency of occurrence of weather types conducting to degraded air quality? G. Hedegaard: So far only the average weather parameters predicted by the climate model have been validated against average weather from the MM5 meteorological model. We have not yet investigated the frequency of any particular air-quality-degrading weather type. However, this is one of the aims in the future validation process. P. Kishcha: Some peculiarity around Novilski has been mentioned in the presentation. Could you provide some more details about the effect? G. Hedegaard: The Siberian industrial city, Norilsk is a city based on metal extraction and production and a very large amount of sulphur dioxide is emitted from this production every year. In the simulations here a general increase in sulphate concentration is found in the surroundings of Norilsk due to an increase in the conversion rate of sulphur dioxide to sulphate (the OH concentration is predicted to increase). This increase differs from the Siberian and Arctic areas in general where the overall trend is observed to be a decrease in the concentration of sulphur dioxide and sulphate. P. Builtjes: Did you consider changes in land use in the future; the forest might not be anymore where it is now. G. Hedegaard: It is true that the forest and vegetation distribution in general might change in the future and changes like these are not accounted for in the current simulations. T. Dore: Does the model assume increases in emissions of SO2 from shipping? G. Hedegaard: No. All anthropogenic emissions are kept constant at a 1990 level in order to separate out the signal from climate change in the future air pollution levels.

6.1 Linking Global and Regional Models to Simulate U.S. Air Quality in the Year 2050 Chris Nolte, Alice Gilliland and Christian Hogrefe

Abstract The potential impact of global climate change on future air quality in the United States is investigated with global and regional-scale models. Regional climate model scenarios are developed by dynamically downscaling the outputs from a global chemistry and climate model and are then used by the Community Multiscale Air Quality (CMAQ) model to simulate climatological air quality. The CMAQ model is first applied to a five-year period representing current climate and evaluated by comparison against measurements of chemically speciated fine particulate matter (PM2.5) concentrations in the U.S. Next, the model is applied to a simulated climate for the year 2050 based on the A1B scenario developed by the Intergovernmental Panel on Climate Change (IPCC). Two five-year future simulations are conducted, one with anthropogenic emissions held at 2001 levels, and one with anthropogenic emissions reduced to emulate the A1B scenario for the developed world. In both future simulations, biogenic and other climate-sensitive emissions are varied with the simulated climate. Results for the future simulation with current emissions indicate modest decreases of 1–2 Pg m-3 PM2.5 in most of the eastern U.S., but large decreases exceeding 10 Pg m-3 PM2.5 are predicted for the future reduced emissions case.

Keywords Climate change, CMAQ, particulate matter 1. Introduction Currently, regional-scale air quality models are being used to test proposed emission controls for management of air quality without regard to interannual meteorological variability or the possibility of climate change. In cases where emission controls are implemented over several decades, (e.g., U.S. Clean Air Interstate Rule), taking climate change into account could potentially lead to a different conclusion as to an optimal control strategy. Recently, a number of studies have been conducted exploring the impact of climate change on future air quality (Hogrefe et al., 2004; Stevenson et al., 2006; Liao et al., 2006; Racherla and Adams, 2006; Cooter et al., 2007; Wu et al., 2007). Nolte et al. (2008) described a study in which downscaled regional climate scenarios are created from outputs of a global climate and chemistry model and are used by the CMAQ model to simulate air quality over the U.S. under both current and future (ca. 2050) climatologies. In Nolte et al. (2008), C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

559

C. Nolte et al.

560

modeled current ozone concentrations are evaluated against current observations and compared with predicted future concentrations. As a follow up, this study adopts the same approach in comparing modeled current particulate matter concentrations with observations and with predicted future concentrations.

2. Modeling System The modeling system used for this study is comprised of a global climate model (GCM) linked to regional-scale climate and air quality models. Each component of this modeling system is briefly described below. 2.1. Climatological meteorology

The GCM used is derived from the GISS 2’ model as described by Mickley et al. (2004), coupled to the Harvard tropospheric ozone-NOx model as in Mickley et al. (1999). The GCM has a horizontal resolution of 4° latitude and 5° longitude and nine vertical layers in a sigma coordinate system extending from the surface to 10 hPa. The global climate simulation covers the period 1950–2055, with greenhouse gas concentrations updated annually using observations for 1950–2000 (Hansen et al., 2002) and the A1B scenario from the IPCC for 2000–2055 (IPCC, 2000). The radiation scheme assumes present-day climatological values for ozone and aerosol concentrations and has no feedbacks due to future pollutant concentration changes. A regional climate model based on the Penn State/NCAR Mesoscale Model (MM5) was used to downscale the GCM outputs to a 36 km grid for 1995–2005 and for 2045–2055 (Leung and Gustafson, 2005). Lateral boundary conditions from the GCM outputs were applied at 6 hours intervals without assimilation of observational data. The regional climate model outputs were archived hourly and used to provide meteorological conditions for both emissions and air quality models. 2.2. Emissions

The Sparse Matrix Operator Kernel Emissions (SMOKE) modeling system was used to prepare emissions inputs consistent with the simulated meteorology, as both evaporative emissions and plume rise are functions of temperature. Meteorologically-driven biogenic emissions were computed using the Biogenic Emissions Inventory System (BEIS; Hanna et al., 2005). Three five-year sets of daily emissions inputs were prepared, as listed in Table 1. In the first set, Table 1 Description of air quality simulations. Simulation name

Modeling period

Anthropogenic emissions

CURR

1999–2003

2001

FUT1

2048–2052

2001

FUT2

2048–2052

2050

CURR, anthropogenic emissions were based on the U.S. Environmental Protection Agency National Emission Inventory for 2001, modulated by the climatology

Linking Global and Regional Models to Simulate U.S. Air Quality in the Year 2050

561

simulated for the current period, and were merged with biogenic emissions computed using the same simulated current climatology. For the second set, FUT1, the same underlying anthropogenic emissions were used, though they were modulated by the simulated future climatology and merged with biogenic emissions computed for that future period. For the third set, FUT2, the same biogenic emissions were used as in FUT1, but anthropogenic emissions were scaled according to the A1B 2050 projections by the Asian Pacific Integrated Model (AIM) for the developed world (see Table 2). Table 2 Scaling factors applied for all anthropogenic emission sectors in simulation FUT2, relative to FUT1. Species

Factor

NOx

0.52

SO2

0.37

VOCs

0.79

CO

1.5

PM

1

NH3

1

2.3. Air quality simulations Air quality simulations were performed with the Community Multiscale Air Quality (CMAQ) model (Byun and Schere, 2006) version 4.5. A continuous fiveyear CMAQ simulation was run for each of the three emissions scenarios listed in Table 1. Chemical boundary conditions for ozone, NOx, and related VOCs were taken from monthly averaged outputs of the Harvard tropospheric chemistry module coupled to the GISS GCM. For each time period, mean aerosol boundary conditions were computed from outputs of a related simulation conducted with the modeling system of Liao et al. (2003).

3. Results and Discussion 3.1. Current period evaluation

Total mass and speciated 24-hour measurements of PM2.5 concentrations are collected every third day at sites in the Interagency Monitoring of Protected Visual Environments network (IMPROVE; see http://vista.cira.colostate.edu/improve). Summer and winter differences between modeled (CURR) average PM2.5 concentrations and 2000–2004 observations are shown in Figure 1. For the summer, modeled concentrations at sites in the Pacific Northwest and in the central U.S. are positively biased by 2–6 Pg m-3, while most sites in the northeast exhibit an equivalent negative bias. For the winter, however, positive biases exist at nearly every site, exceeding 6 Pg m-3 at most sites in the central and eastern U.S.

562

C. Nolte et al.

Fig. 1 Differences between average modeled PM2.5 concentrations and measurements at IMPROVE monitoring sites for summer (left) and winter (right)

Similarly, difference plots between modeled and measured sulfate concentrations are shown in Figure 2. During the summer, predicted sulfate concentrations exhibit a positive bias greater than 1.5 Pg m-3 in much of the central U.S. and somewhat weaker negative biases in the northeast and in California. During the winter, sulfate predictions are unbiased (within 0.5 Pg m-3) at most monitoring sites.

Fig. 2 Differences between average modeled sulfate concentrations and measurements at IMPROVE monitoring sites for summer (left) and winter (right)

Measured “soil” concentrations (a derived quantity based on measurements of certain trace elements) and modeled “other unspeciated PM” concentrations are shown in Figure 3. During both summer and winter, a large positive bias in modeled “other PM” concentrations is evident at nearly every monitoring site, making it the largest contributor to the errors in total PM2.5. Preliminary investigations into this bias suggests that it is due to unrealistically high levels of dust (summer average 2 Pg m-3; winter average 8 Pg m-3) in the global model coming into the CMAQ modeling domain via the northern boundary.


563

Fig. 3 Differences between average modeled “soil” concentrations and measurements at IMPROVE monitoring sites for summer (left) and winter (right)

3.2. Current-future differences

Differences between the average PM2.5 concentrations for the two future period simulations and the CURR simulation are shown in Figure 4. For FUT1, summer average PM2.5 concentrations decrease by 1–3 Pg m-3 throughout most of the central and eastern U.S. and in California. In the winter, PM2.5 decreases by 2–5 Pg m-3 in the northern part of the domain. The decrease is even more substantial for the A1B-scaled emissions case FUT2, with average decreases from 3 to 9 Pg m-3 in the eastern third of the U.S during both summer and winter.

Fig. 4 Changes from CURR in five-year-average summer (left) and winter (right) PM2.5 concentrations for FUT1 (top) and FUT2 (bottom)

564

C. Nolte et al.

Fig. 5 Changes from CURR in five-year-average summer (left) and winter (right) sulfate concentrations for FUT1 (top) and FUT2 (bottom)

Differences between average sulfate concentrations for the two future period simulations and the CURR simulation are shown in Figure 5. For FUT1, summer average sulfate concentrations decrease by 0.5–1.5 Pg m-3 throughout most of the central and southern U.S., while there is a slight increase of 0.2–0.4 Pg m-3 in a portion of the Midwest. The decrease is larger for the A1B-scaled emissions case FUT2, with average sulfate decreases from 3 to 5 Pg m-3 in the eastern third of the U.S during summer and 1–2 Pg m-3 during the winter.

Fig. 6 Changes from CURR in five-year-average summer (left) and winter (right) “other PM” concentrations for FUT1

Differences between the summer and winter average “other PM” concentration for FUT1 and CURR simulation are shown in Figure 6. Dust emissions and boundary conditions were the same for FUT2 as for FUT1, so FUT2 “other PM” concentrations are virtually identical to those for FUT1 and are not shown. Relative to CURR, future summer average “other PM” concentrations decrease by 0.5–1.0 Pg m-3 in the western U.S. and in parts of the central U.S., while large decreases exceeding 2 Pg m-3 are evident during the winter. Since “other PM” is a chemically


565

inert species in CMAQ, the large decrease in future concentrations is due to changes in boundary conditions from the global model; these were unrealistically high for the current period as noted in Section 3.1.

4. Summary Three sets of five-year air quality simulations for the continental U.S. have been conducted using downscaled meteorology and chemical boundary conditions from a global climate model. Model predictions for current period PM2.5 concentrations are in reasonable agreement with recent observations, with the error dominated by overpredictions in dust concentrations from the global model. Comparison of model results for the current period with those for the future period with current anthropogenic emissions shows decreases in sulfate of 0.5–1.5 Pg m-3 during the summer, while summer sulfate concentrations decrease 3–5 Pg m-3 in the reduced emissions case. Future work for this study will explore the relative impact of changing meteorological variables and changes in chemical boundary conditions on PM2.5 concentrations. A coupled climate and air quality model is under development, which will integrate feedbacks from pollutants on radiative forcing to better understand the relationships between climate change and air quality. Acknowledgments The authors wish to thank Loretta Mickley and Pavan Racherla for conducting the GCM simulations and Ruby Leung of the Pacific Northwest National Laboratory for performing the MM5 downscaling of the GCM data. William Benjey, Robert Gilliam, and Steven Howard of NOAA’s Atmospheric Sciences Modeling Division and Allan Beidler and Ruen Tang of the Computer Sciences Corporation assisted with processing emissions and meteorological data and with the air quality simulations. Disclaimer The research presented here was performed under the Memorandum of Understanding between the U.S. Environmental Protection Agency (EPA) and U.S. Department of Commerce’s National Oceanic and Atmospheric Administration (NOAA) and under agreement number DW13921548. This work constitutes a contribution to the NOAA Air Quality Program. Although it has been reviewed by EPA and NOAA and approved for publication, it does not necessarily reflect their policies or views.

References Byun D, Schere KL (2006) Science algorithms of the EPA Models-3 Community Multiscale Air Quality (CMAQ) Modeling System, Applied Mechanics Reviews 59, 51–77. Cooter EJ, Gilliam R, Benjey W, Nolte C, Swall J, Gilliland A (2007) Examining the impact of changing climate on regional air quality over the U.S. In:

566

C. Nolte et al.

Developments in Environmental Science, vol. 6, C. Borrego and E. Renner, eds. Elsevier, Amsterdam. Hanna SR, Russell AG, Wilkinson JG, Vukovich J, Hansen DA (2005) Monte Carlo estimation of uncertainties in BEIS3 emission outputs and their effects on uncertainties in chemical transport model predictions, Journal of Geophysical Research 110, D01302. Hansen J et al. (2002) Climate forcings in Goddard Institute for Space Studies SI2000 simulations, Journal of Geophysical Research 107 (D18), 4347. Hogrefe C, Lynn B, Civerolo K, Ku JY, Rosenthal J, Rosenzweig C, Goldberg R, Faffin S, Knowlton K, Kinney PL (2004) Simulating changes in regional air pollution over the eastern United States due to changes in global and regional climate and emissions, Journal of Geophysical Research 109, D22301. Intergovernmental Panel on Climate Change (2000), Special Report on Emissions Scenarios, N. Nacenovic and R. Swart, eds., Cambridge University Press, New York. Available on the Web at http://www.grida.no/climate/ipcc/emission Leung LR, Gustafson WI Jr (2005) Potential regional climate change and implications to U.S. air quality, Geophysical Research Letters 32, L16711. Liao H, Adams PJ, Chung SH, Seinfeld JH, Mickley LJ, Jacob DJ (2003) Interactions between tropospheric chemistry and aerosols in a unified general circulation model, Journal of Geophysical Research 108 (D1), 4001. Liao H, Chen WT, Seinfeld JH (2006) Role of climate change in global predictions of future tropospheric ozone and aerosols, Journal of Geophysical Research 111 (D12), D12304. Mickley, LJ, Murti PP, Jacob DJ, Logan JA, Koch DM, Rind D (1999) Radiative forcing from tropospheric ozone calculated with a unified chemistry-climate model, Journal of Geophysical Research 104 (D23), 30153–30172. Mickley, LJ, Jacob DJ, Field BD, Rind D (2004) Effects of future climate change on regional air pollution episodes in the United States, Geophysical Research Letters 31, L24103. Nolte CG, Gilliland AB, Hogrefe C, Mickley LJ (2008) Linking global to regional models to assess future climate impacts on surface ozone concentrations in the United States, Journal of Geophysical Research, submitted. Racherla PN, Adams PJ (2006) Sensitivity of global tropospheric ozone and fine particulate matter concentrations to climate change, Journal of Geophysical Research 111, D24103. Stevenson, DS, Johnson CE, Collins WJ, Derwent RG, Edwards JM (2006) Multimodel ensemble simulations of present-day and near-future tropospheric ozone, Journal of Geophysical Research 111, D08301. Wu S, Mickley LJ, Leibensperger EM, Jacob DJ, Rind D, Streets DG (2008) Effects of 2000-2050 global change on ozone air quality in the United States, Journal of Geophysical Research, doi:10.1029/2007JD008917, 113, D06302.


567

Discussion J. Baldasano: What were the reasons why you chose the year 2050? C. Nolte: This study is part of a group of studies investigating the impact of future climate change on air quality. One of the reasons for agreeing on a common time period among these studies is to provide an ensemble of air quality projections. The year 2050 was chosen based on consensus of several groups. Climate signals at 2030 could be too small to detect within interannual variability. Climate modelers preferred 2100 for the same reason, but emission scenarios for 2100 are so uncertain as to be untenable. Hence, 2050 was a good compromise. P. Kishcha: Were future land use changes considered in the model predictions under discussion? C. Nolte: No. Future land use categories were assumed to be unchanged. This represents an important uncertainty in our modeling system. A. Aulinger: Did you compute statistics on peak concentrations of PM and O3 in order to assess the number of days with increased health risks due to climate change or changes in precursor concentrations? C. Nolte: Yes. We have computed the number of days per year at each site where the maximum 8-hour average ozone and PM2.5 exceeded threshold values of 80 ppb and 35 Pg m-3. The spatial pattern of change in the number of exceedances generally follows the pattern of the changes in the means. However, our ozone concentration predictions under current climate conditions are positively biased by 10–15 ppb in parts of the U.S., which hinders our ability to predict with accuracy exceedances above a given threshold.

6.3 Regional Climate Change Impacts on Air Quality in CECILIA EC 6FP Project Tomas Halenka, Peter Huszar and Michal Belda

Abstract Recent studies show considerable effect of atmospheric chemistry and aerosols on climate on regional and local scale. For the purpose of qualifying and quantifying the magnitude of climate forcing due to atmospheric chemistry/aerosols on regional scale, the development of coupling of regional climate model and chemistry/aerosol model has been started recently on the Department of Meteorology and Environmental Protection, Faculty of Mathematics and Physics, Charles University in Prague, for the EC 6FP Project QUANTIFY and finally for EC 6FP Project CECILIA. One of the project objectives, aiming to study climate change impacts in Central and Eastern Europe based on very high resolution simulations using regional climate models (RCM) in 10 km grid, is dealing with climate change impacts on and interaction to air quality. For this coupling, existing regional climate model and chemistry transport model are used. Climate is calculated using model RegCM and ALADIN-Climate while chemistry is solved by model CAMx. Climate change impacts on large urban and industrial areas modulated by topographical and land-use effects which can be resolved at the 10 km scale, are investigated by CECILIA as well. Meteorological fields generated by RCM drive CAMx transport, chemistry and a dry/wet deposition. A preprocessor utility was developed for transforming RegCM provided fields to CAMx input fields and format. As the first step, the distribution of pollutants can be simulated off-line for long period in the model couple. There is critical issue of the emission inventories available both for present and scenarios runs as well as cross-boundary transport for regional simulations. The next step is the inclusion of the radiative active agents from CAMx into RCM radiative transfer scheme to calculate the changes of heating rates. Only the modification of radiative transfer due to atmospheric chemistry/ aerosols is taken into account first, the indirect effect of aerosols will be studied later. Ten years time slices for present, control and scenarios runs for mid- and end of century are supposed in framework of the project. Some sensitivity runs will be run in present climate.

Keywords Air-pollution, air-quality, climate change, air pollution modelling


577

578

T. Halenka et al.

1. Introduction In decision making process there is significant problem arising from the weak link between global climate change information and impact studies necessarily based on real local conditions. Global Circulation Models (GCMs) can reproduce reasonably well climate features on large scales (global and continental), but their accuracy decreases when proceeding from continental to regional and local scales because of the lack of resolution. This is especially true for surface fields, such as precipitation, surface air temperature and their extremes, which are critically affected by topography and land use. However, in many applications, particularly related to the assessment of climate-change impacts, the information on surface climate change at regional to local scale is fundamental. To bridge the gap between the climate information provided by GCMs and that needed in impact studies, especially when aiming the interactions of climate and air-quality issues, dynamical downscaling, i.e., nesting of a fine scale limited area model (or Regional Climate Model, RCM) within the GCM is the most convenient tool. In the region of Central and Eastern Europe the need for high resolution studies is particularly important. This region is characterized by the northern flanks of the Alps, the long arc of the Carpathians, and smaller mountain chains and highlands in the Czech Republic, Slovakia, Romania and Bulgaria that significantly affect the local climate conditions. A resolution sufficient to capture the effects of these topographical and associated land-use features is necessary. That is why 10 km resolution has been introduced in the project CECILIA of EC FP6. The main aim of the project dealing with climate change impacts and vulnerability assessment in targeted areas of Central and Eastern Europe is the application of regional climate modelling studies at a resolution of 10 km for local impact studies in key sectors of the region. The project is covering studies on hydrology, water quality, and water management (focusing at medium-sized river catchments and the Black Sea coast), agriculture (crop yield, pests and diseases, carbon cycle), and forestry (management, carbon cycle), as well as air quality issues in urban and industrialized areas (e.g. Black Triangle – a polluted region around the common borders of the Czech Republic, Poland and Germany). Climate change impacts on large urban and Industrial areas modulated by topographical and land-use effects which can be resolved at the 10 km scale are investigated by CECILIA as well. The concentration of air pollutants depends on both anthropogenic and climate factors. A main issue is the quantity of emissions of primary pollutants as well as of precursors of secondary pollutants. Long range transport to the target regions will be taken into account by simulation for the whole Europe, driven by RCM with a grid resolution of 50 × 50 km. These simulations will be used to constrain nested higher resolution runs (10 × 10 km) for a smaller domain focusing in CEE both for present and future climate. The key species will be ozone, sulphur and nitrogen as well as PM, which have a central role in tropospheric chemistry as well as the strong health impacts.

Regional Climate Change Impacts on Air Quality in CECILIA EC 6FP Project

579

2. Models Involved It is now well established that climatically important (so called radiatively active) gases and aerosols can have substantial climatic impact trough their direct and indirect effects on radiation, especially on regional scales (Qian and Giorgi, 2000; Qian et al., 2001, Giorgi et al., 2002). The study of these effects requires coupling of regional climate models with atmospheric chemistry/aerosols to assess the climate forcing to the chemical composition of the atmosphere and its feedback to the radiation, eventually other components of the climate system. For this coupling, existing regional climate model and chemistry transport model are used. At our Department climate is calculated using model RegCM while chemistry is solved by model CAMx, for the projects the attempt is done to develop and to use the couple ALADIN-Climate and CAMx as well. The model RegCM used here was originally developed by Giorgi et al. (1993a, b) and then has undergone a number of improvements described in Giorgi et al. (1999), and, finally, Pal et al. (2005). The dynamical core of the RegCM is equivalent to the hydrostatic version of the mesoscale model MM5. Surface processes are represented via the Biosphere-Atmosphere Transfer Scheme (BATS) and boundary layer physics is formulated following a non-local vertical diffusion scheme (Giorgi et al., 1993a). Resolvable scale precipitation is represented via the scheme of Pal et al. (2000), which includes a prognostic equation for cloud water and allows for fractional grid box cloudiness, accretion and re-evaporation of falling precipitation. Convective precipitation is represented using a mass flux convective scheme (Giorgi et al., 1993b) while radiative transfer is computed using the radiation package of the NCAR Community Climate Model, version CCM3 (Giorgi et al., 1999). This scheme describes the effect of different greenhouse gases, cloud water, cloud ice and atmospheric aerosols. Cloud radiation is calculated in terms of cloud fractional cover and cloud water content, and the fraction of cloud ice is diagnosed by the scheme as a function of temperature. For more details on the use of the model see Elguindi et al. (2006). CAMx is an Eulerian photochemical dispersion model developed by ENVIRON Int. Corp. (Environ, 2006). Currently in version 4.40 CAMx is used for air quality modeling in more than 20 countries by government agencies, academic and research institutions, and private consultants for regulatory assessments and general research. It is available for free in the form of the source code with various supporting programs. CAMx can use environmental input fields from a number of meteorological models (e.g., MM5, RAMS, CALMET) and emission inputs from many emissions processors. CAMx includes the options of two-way grid nesting, multiple gas phase chemistry mechanism options (CB-IV, SAPRC99), evolving multisectional or static two-mode particle size treatments, wet deposition of gases and particles, plume-in-grid (PiG) module for sub-grid treatment of selected point sources, Ozone and Particulate Source Apportionment Technology, mass conservative and consistent transport numerics, parallel processing. It allows for integrated “one-atmosphere” assessments of gaseous and particulate air pollution (ozone, PM2.5, PM10, air toxics) over many scales ranging from sub-urban to continental.

580

T. Halenka et al.

CAMx simulates the emission, dispersion, chemical reaction, and removal of pollutants in the troposphere by solving the pollutant (Eulerian) continuity equation for each chemical species on a system of nested three-dimensional grids. These processes are strongly dependent on the meteorological conditions, therefore CAMx requires meteorological input from a NWP model or RCM for successful run.

Fig. 1 Average concentration of NO2 (upper left), O3 (upper right) and SO2 (bottom panel) for year 2000 for 10 × 10 km resolution central Europe domain in ppbv

3. Preprocessor and Settings Meteorological fields generated by RegCM drive CAMx transport and dry/wet deposition. A preprocessor utility was developed for transforming RegCM fields to CAMx input fields and formats. For fields not provided by the meteorological model the diagnostic formulas are used, cloud/rain water content and cloud optical depth are gained from the rain rates and the vertical profile of water vapour content and temperature. Vertical diffusion coefficients are calculated following O’Brien (1970). As the first step, the distribution of pollutants can be simulated for long period in the model couple. There are problems with the anthropogenic emission inventories available, at this stage emissions from EMEP 50 × 50 km database are interpolated. We are testing VOC speciation technique, biogenic emissions of isoprene and monoterpenes calculated as a function of 2 m temperature, global radiation and landuse by Guenther et al. (1993, 1994). We use 23 vertical ı-levels reaching up to


581

70 hPa, with time step of 150 s, at 45 km resolution in preliminary experiments for RegCM configuration, the same horizontal grid for CAMx. Initial and boundary conditions are set to CAMx’s top concentrations (independent of time) (Simpson et al., 2003) for 45 km resolution run, the results are used for driving the same couple of RegCM-CAMx in 10 km resolution on smaller “CECILIA” region. In our setting CB-IV chemistry mechanism is used (Gery et al., 1989).

Fig. 2 Comparison of simulated ten-days running average concentration of NO2 for selected stations in year 2000 (ppbv). Grey line for 45 km resolution, black line for 10 km resolution

4. Preliminary Results Some examples of the high resolution integration for year 2000 are presented in Figure 1 for selected species. There is much more local features seen in this simulation compare to less resolution run (not shown), especially for O3 the effect of high resolution land use which provides basis for biogenic emission computation is well pronounced in the concentration fields, even more in summer (with respect to limited extent not shown again). More interesting comparison of the driving 45 km

582

T. Halenka et al.

resolution run with 10 km high resolution run for selected stations can be seen in Figure 2 in terms of time series of NO2 simulated concentrations. Reasonable agreement can be seen during spring and summer, some decreases appear in winter and autumn seasons of high resolution simulation. Mainly in cold season the values of high resolution runs are significantly lower in Bohemian region on Kosetice and Svratouch station. More reliable analysis can be done by comparison of both simulations of O3 concentration with real data in Figure 3.

Fig. 3 Comparison of simulated and measured ten-days running average concentration of O3 for selected stations in year 2000 (ȝg/m3). Grey line for 45 km resolution, black line for 10 km resolution, light grey for measurement

Underestimation of the ozone concentration by the model especially during warm season appears for some stations of the Central Europe whereas overestimation is presented in comparison for Ispra mainly in cold period of the year. Basically, high resolution runs brings slight improvement of the results for selected stations.


583

5. Future Outlooks To produce more reliable conclusions longer runs are essential. At present we are preparing the longer experiment in very high resolution of 10 km driven by 50 km run based on off-line simulation using CAMx with the meteorological data from ICTP RegCM simulation driven by ERA40 reanalyses done for EC FP6 IP ENSEMBLES. Period of ten years simulation covering years 1991–2000 will provide the comparison on reliable data both on input (emission databases) and measurement side. Further in framework of the CECILIA Project three time slices of ten years are supposed to be completed in high resolution of 10 km with the couple RegCM-CAMx: control (1991–2000), middle of the century (2041–2050) and end of century (2091–2100), using A1B scenario. The next step of the couple development will be the inclusion of the radiative active agents from CAMx into RegCM radiative transfer scheme to calculate the changes of heating rates. Only the modification of radiative transfer due to atmospheric chemistry/aerosols will be taken into account first, the indirect effect of aerosols will be taken into account later, there are still many uncertainties in understanding of this issue and possibility of inclusion of appropriate processes into the model. The feedback of chemistry/aerosols on climate will be studied in terms of monthly and yearly averages of 2 m temperatures and of the top-of-the-atmosphere (TOA) radiative forcing, the results will provide the estimate of the effect of interactive atmospheric chemistry and aerosols on climate in regional and local scales. Acknowledgments This work is supported in framework of EC FP6 STREP CECILIA (GOCE 037005), partially by EC FP6 Integrated project QUANTIFY (GOCE 003893) as well as under local support of the grant of Programme Informacni spolecnost, No. 1ET400300414 and Research Plan of MSMT under No. MSM 0021620860.

References Elguindi N, Bi X, Giorgi F, Nagarajan B, Pal J, Solmon F, Rauscher S, Zakey A, (2006) RegCM Version 3.1 User’s Guide. PWCG Abdus Salam ICTP. ENVIRON Corporation (2006) CAMx Users’ Guide, version 4.40 Gery MW, Whitten GZ, Killus JP, Dodge MC (1989) A photochemical kinetics mechanism for urban and regional scale computer modeling. J. Geophys. Res., 94, 925–956. Giorgi F, Marinucci MR, Bates GT (1993a) Development of a second generation regional climate model (RegCM2). Part I: boundary layer and radiative transfer processes. Mon. Wea. Rev., 121, 2794–2813.

584

T. Halenka et al.

Giorgi F, Marinucci MR, Bates GT, DeCanio G (1993b) Development of a second generation regional climate model (RegCM2). Part II: convective processes and assimilation of lateral boundary conditions. Mon. Wea. Rev., 121, 2814–2832. Giorgi F, Huang Y, Nishizawa K, Fu C (1999) A seasonal cycle simulation over eastern Asia and its sensitivity to radiative transfer and surface processes. Journal of Geophysical Research, 104, 6403–6423. Giorgi F, Bi X, Qian Y (2002) Direct radiative forcing and regional climatic effects of anthropogenic aerosols over East Asia: a regional coupled climate-chemistry/ aerosol model study. J. Geophys. Res., 107, 4439, doi:10.1029/2001JD001066. Guenther AB, Zimmerman PR, Harley PC, Monson RK, Fall R (1993) Isoprene and monoterpene rate variability: model evaluations and sensitivity analyses. Journal of Geophysical Research, 98, No. D7, 12609–12617. Guenther A, Zimmerman P, Wildermuth M (1994) Natural volatile organic compound emission rate estimates for U.S. woodland landscapes. Atmospheric Environment, 28, 1197–1210. O’Brien JJ (1970) A note on the vertical structure of the eddy exchange coefficient in the planetary boundary layer. Journal of Atmospheric Science, 27, 1213–1215. Pal JS, Small EE, Eltahir EA (2000) Simulation of regional-scale water and energy budgets: Representation of subgrid cloud and precipitation processes within RegCM. Journal of Geophysical Research, 105, 29579–29594. Pal JS, Giorgi F, Bi X, Elguindi N, Solmon F, Grimm A, Sloan L, Syed F, Zakey A, (2005) The ICTP Regional Climate Model version 3 (RegCM3). Benchmark simulations over tropical regions. Submitted to the Bull. Amer. Meteorol. Soc. QianY, Giorgi F (2000) Regional climatic effects of anthropogenic aerosols? The case of Southwestern China. Geophysical Research Letters, 27(21), 3521–3524, 10.1029/2000GL011942. Qian Y, Giorgi F, Huang Y, Chameides WL, Luo C (2001) Simulation of anthropogenic sulfur over East Asia with a regional coupled chemistry/climate model. Tellus, Series B, 53, 171–191. Simpson D, Fagerli H, Jonson J, Tsyro S, Wind P (2003) Transboundary Acidification, Eutrophication and Ground Level Ozone in Europe PART I, Norwegian Meteorological Institute.

Discussion G. Kallos: How did you perform your regional scale model simulations? They are always biases in surface parameterization (ex soil moisture) Heat strongly affects meteorology (ex latent heat flux/sensible heat and therefore cloud and precipitation processes heat effect air quality and deposition. how you managed these issues? T. Halenka: Of course there are always biases of many kinds in any simulation. Especially, this might be of great importance when using these


585

results for driving air-quality simulations. However, when developing the tools for estimation of future development under different scenarios of climate change, there is no other way than to cope with these uncertainties. In presented results for experiment with one year simulations there is not so much material for systematic analysis, but in framework of our CECILIA project ten years period is supposed to be run in framework of “perfect” boundary conditions (based on reanalysis) as well as the control simulation driven by global model before the time slices for the middle and the end of this century. Careful statistical analysis of mainly reanalysis run and comparison to control both in air-quality results and climate ones can reveal the role of individual sources of biases and their impact on air-quality simulations.

7.3 A Modeling Methodology to Support Evaluation of Public Health Impacts on Air Pollution Reduction Programs Vlad Isakov and Halûk Özkaynak

Abstract Environmental public health protection requires a good understanding of the types and locations of pollutant emissions of health concern and their relationship to environmental public health indicators. Therefore, it is necessary to develop the methodologies, data sources, and tools for assessing the public health impact of air pollution reduction programs, also referred to as accountability analysis. Since air quality models are among the main tools that can be used to evaluate the impacts from emissions changes, either due to growth or implementtation of source control strategies, these approaches play a vital role in most air accountability studies. In this study, we present a modeling methodology to estimate concentrations for multiple pollutants that include both local features (hot spots) and regional transport. The local impacts from mobile sources and significant stationary sources are estimated using a dispersion model (AERMOD). These local details are combined with regional background estimates computed by a photochemical grid model (CMAQ) in a “hybrid” approach to derive total concentrations required for the subsequent human exposure analysis. We demonstrate an application of this methodology in New Haven, Connecticut. The city of New Haven has implemented a comprehensive Clean Air Initiative, which includes a number of federally mandated and voluntary air pollution programs. This project is a collaborative effort with state and local agencies including government, academia, and the New Haven community, to apply and evaluate air quality and human exposure models that can be used with health data and to assess the feasibility of using this information to conduct an air accountability study. Although this study is based in one city, the methodologies developed through this project can have broad application to other areas within the United States and internationally.

Keywords Air quality, exposure, modeling

1. Introduction Air quality has improved substantially in the United States in recent decades, in large part due to increasingly stringent federal and state air quality regulations. While C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

614

A Modeling Methodology to Support Evaluation of Public Health impacts on Air Pollution

615

many studies have documented links between better air quality and improvements in a variety of human health metrics, direct evidence is lacking about the extent to which specific control measures have improved health. Over the past decade, various epidemiological studies have examined the relationships between acute and chronic health outcomes and measured ambient particulate matter, ozone and other co-pollutant concentrations. In the context of air pollution health effects, results from recent epidemiological studies indicate the importance of determining the key sources and constituents of indoor, outdoor pollution, and personal exposures to PM, ozone and other air pollutants. However, understanding the magnitude and nature of human exposure is clearly the first step in assessing the occurrence of adverse effects that could follow upon contact with environmental pollutants. One of the ways to access the human exposure is through the use of exposure models such as EPA’s Hazardous Air Pollutant Exposure Model (HAPEM), the Air Pollutant Exposure Model (APEX), or Stochastic Human Exposure and Dose Simulation (SHEDS). Since predicted concentrations from air quality models are key drivers for human exposure models, it is essential to improve the accuracy and precision of spatial and temporal characterization of results from these models. However, complex interactions between interventions over time can make it difficult to isolate the environmental impacts and associated health effects of any one regulation. For example, a regulatory action may have varying effects on emissions depending upon compliance and the real-world effectiveness of the interventions applied. Often, the connection between emissions and ambient air quality depends on complex atomspheric and chemical transformations. Environmental public health protection requires a good understanding of the types and locations of pollutant emissions of health concern and their relationship to environmental public health indicators. Therefore, it is necessary to develop the methodologies, data sources, and tools for assessing the public health impact of air pollution reduction programs, and accounttability analysis. Since air quality models are the principal predictive tool for assessing the impacts of potential emissions control strategies on future-year concentrations, we describe here a modeling approach to support air accountability studies.

2. Air Quality Modeling Approach Environmental health studies require detailed information on air quality. Therefore, air quality modeling should include local-scale features, long-range transport, and photochemistry to provide the best estimates of air concentrations. There are several available modeling approaches capable of assessing pollutant concentration gradients at a fine resolution (Touma et al., 2006) and these can be categorized into two major types of air quality models: source-based dispersion models and Eulerian grid-based chemical transport models. Chemical transport models, such as the Community Multi-scale Air Quality (CMAQ, Byun and Schere, 2006), are used to simulate the transport and formation of ozone, acid rain, particulate matter (PM)

616

V. Isakov, H. Özkaynak

and other pollutants formed by chemical reactions among precursor species that are emitted from hundreds or thousands of emission sources. Such models may be set up to apply to a wide range of scales ranging from global to urban. However, regional-scale grid-based models can address photochemistry effects, but not locallevel gradients. CMAQ provides volume-average, hourly concentration values for each grid cell in the modeling domain. Emissions are assumed to be instantaneously well-mixed. While grid-models are the model platform of choice for simulation of chemicallyreactive airborne pollutants, source-based dispersion models such as AERMOD (Cimorelli et al., 2005) that have been developed to simulate pollutant concentrations within a few hundred meters or a few kilometers from the source are typically used for local scales. These models generally do not take into account atmospheric chemical reactions or they do so using simplified representations such as first-order pollutant decay. They provide detailed resolution of the spatial variations in hourly-average concentrations. It would be desirable to combine the capabilities of grid-models and dispersion models into one model, but this is a yet evolving area of research and development. One option is a hybrid approach (Isakov et al., 2007), where a regional grid model and a local plume model are run independently. To illustrate how air quality models can be used to provide inputs to human exposure models, we focus on a 20 by 20 km area encompassing New Haven, Connecticut that includes many stationary sources emitting toxic pollutants and several major roadways as indicated in Figure 1. The city of New Haven, with population of approximately 125,000, is a recipient of one of EPA’s nationally funded Community Air Toxics projects. Through this project, New Haven has implemented a comprehensive Clean Air Initiative, which includes a number of voluntary air pollution programs. Along with local and state efforts, there are also several Federal regulations that have either recently been or soon will be implemented (e.g., Clean Air Interstate Rule). This project presents an opportunity to assess the feasibility of using air quality and human exposure models that can be used with health data to conduct an air accountability study. Resolving fine scale pollutant gradients to identify local concentration hot spots from both stationary and mobile sources is critical for exposure assessments. For example, individuals who spend more time near busy highways are likely to be exposed to higher levels of air pollution. To account for this near-road exposure we modeled ambient air quality concentrations for multiple pollutants resulting from roadway emissions. There are multiple modeling techniques to simulate near-road dispersion from mobile sources (Borrego et al., 2006; Cook et al., 2006). In this study, we used the AERMOD dispersion model which treats individual road links as area sources to simulate hourly concentrations of various pollutants near the road. AERMOD also simulates near-source impacts from stationary sources. Contributions to photochemical interactions are provided as a background concentration level from CMAQ, a regional grid model.


617

Fig. 1 Modeling domain showing locations of emission sources and model receptors

A hybrid approach (Isakov et al., 2007) is a logical and efficient way to combine regional grid and local plume models. Results of both model simulations are combined to provide the total ambient air pollutant concentrations. The hybrid approach uses the appropriate modeling tools to describe different types of sources, making its application computationally efficient. Furthermore, since local dispersion models are not resource intensive, this methodology allows the study of local concentration variability due to changes in several model inputs and physical parameters, helping to gain confidence in the simulation results by encompassing a range of model outcomes. This constitutes a clear advantage of the hybrid approach, since performing a local concentration variability estimation using a nested grid model alone would be an impractical task, especially over larger urban areas. A schematic of the hybrid approach is shown in Figure 2, where the CMAQ model was used to estimate regional background concentrations and AERMOD was used to estimate local-scale details for stationary and mobile sources.

618


Fig. 2 Application of the hybrid approach in New Haven, Connecticut

In this study, CMAQ was used to simulate ambient concentrations of several air toxics (Luecken et al., 2006). The CMAQ modeling system was run for an annual period in a nested mode at 36 and 12 km horizontal grid dimensions using the 1999 National Emission Inventory and meteorological outputs from 2001 using the MM5 meteorological model. The CMAQ results were extracted for the New Haven modeling domain to provide regional background concentration values. This regional background was combined with local concentrations predicted by the AERMOD dispersion model. The application of this hybrid approach is illustrated in Figure 3 which displays modeled annual average outdoor carbon monoxide concentrations in New Haven, Connecticut. Predicted CMAQ concentrations at 12 by 12 km resolution are combined with estimates at 200 m receptor resolution from AERMOD.


619

Fig. 3 Annual average CO concentrations in New Haven, Connecticut: (a) impact of stationary sources; (b) impact of mobile sources; (c) regional background; (d) combined concentrations using hybrid approach

3. Modeling Approach for Accountability Studies The methodologies developed under this project can be applied to future projects in other areas to simulate air quality impacts for various controls scenarios. For example: (1) what happens if emissions from some specific stationary sources are reduced by “x” percent? (2) what happens if emissions from mobile sources could be reduced by “y” percent? (3) what is the impact of local controls? (4) what is the impact of regional/national controls resulting in reduction of regional background? Figure 4 provides a hypothetical example of the relative impacts of various control strategies on ambient concentrations. These examples include: reducing emissions from mobile sources, controlling emissions from stationary sources, and reducing impact of the regional background. This example helps determine which control options are most effective in reducing ambient concentrations. In order to link these air quality estimates to health effects associated with human exposures to environmental pollutants, we will use exposure modeling. When combined with exposure models, the control strategies can be assessed to optimize the impacted population, or population subset, such as children, people with respiratory problems, etc. In this study, we are evaluating alternative techniques for estimating cumulative exposures to selected air toxics, PM, and ozone using probabilistic cumulative exposure models: HAPEM 6 and SHEDS-Air Toxics, time-series based models, using human activity pattern data, modeled/measured concentrations, and exposure factors.

620


Fig. 4 Example of modeling application to investigate the effect of various control strategies

4. Summary At the present, most Federal and State air quality implementation plans rely heavily on ambient modeling study results for targeting emissions reductions. However, the complexity in the spatial and microenvironmental variation of exposures among the different population subgroups, especially in the context inter- and intra-urban analysis of air pollution health effects, could pose several challenges. Thus, integrated air quality – human exposure modeling provides the means to evaluate the potential health risks from air pollution exposures and the basis to determine optimum risk management strategies, while considering scientific, social and economic factors. Ideally, emission control strategies not only aim at reducing the emissions from principal sources of targeted pollutants but also to identify those sources and microenvironments that contribute to greatest portion of personal or population exposures. Recent advances in exposure modeling tools and better information on time-activity, commuting and exposure factors data provide unique opportunities for improving the assignment of exposures during the course of future accountability and community health studies. Moreover, the combination of sophisticated air quality and exposure models will improve the accuracy of present air quality and exposure forecasts, and help us better quantify the health and economic benefits of emissions reductions programs, as part of air accountability studies.


621

Acknowledgments and Disclaimer The research presented here was performed under the Memorandum of Understanding between the U.S. Environmental Protection Agency (EPA) and the U.S. Department of Commerce’s National Oceanic and Atmospheric Administration (NOAA) and under agreement number DW13921548. This work constitutes a contribution to the NOAA Air Quality Program. It does not necessarily reflect Agency policies or views.

References Borrego CA, Tchepel O, Costa AM, Martins H, Ferreira J, Miranda AI (2006) Traffic-related particulate air pollution exposure in urban areas. Atmospheric Environment 40, 7205–7214. Byun DW, Schere K (2006) Review of the governing equations, computational algorithms, and other components of the Models-3 Community Multiscale Air Quality (CMAQ) modeling system. Applied Mechanics Review 59, 51–77. Cimorelli AJ, Perry SG, Venkatram A, Weil JC, Paine RJ, Wilson RB, Lee RF, Peters WD, Brode RW (2005) AERMOD: a dispersion model for industrial source applications. Journal of Applied Meteorology 44, 682–693. Cook R, Touma JS, Beidler A, Strum M (2006) Preparing highway emissions inventories for urban scale modeling: a case study in Philadelphia. Transportation Research Part D: Transport and Environment 11, 396–407. Isakov V, Irwin JS, Ching J (2007) Using CMAQ for exposure modeling and characterizing the sub-grid variability for exposure estimates. Journal of Applied Meteorology and Climatology 46, 1354–1371. Luecken DJ, Hutzell WT, Gipson GJ (2006) Development and analysis of air quality modeling simulations for hazardous air pollutants. Atmospheric Environment 40, 5087–5096. Touma JS, Isakov V, Ching J, Seigneur C (2006) Air quality modeling of hazardous pollutants: current status and future directions. Journal of Air and Waste Management Association 56, 547–558.

Discussion B. Fisher: Do you tell your users about uncertainty in the high resolution concentration fields, which can arise from errors in emissions models etc? V. Isakov: Yes, assessing uncertainties is an integral part of the health risk assessment process. It is, therefore, desirable to incorporate some treatment of uncertainties in the entire modeling process: emissions and meteorological inputs, model formulation, monitoring data, and

622


exposure and risk. In this study, a major contribution to the uncertainty in the model simulation results originated from the model inputs rather than from the model formulation. Therefore, in order to reduce uncertainty in the high resolution concentration fields, it is important to improve spatial allocation of emissions. For mobile sources, we have developed a practical, readily adaptable methodology to create spatially-resolved, link-based highway vehicle emission inventory. This methodology takes advantage of geographic information system (GIS) software to improve the spatial accuracy of the activity information obtained from a Travel Demand Model. An example of application of this methodology in New Haven, CT is shown in Cook et al. (2008), Journal of Air and Waste Management Association 58, 451–461.

7.4 Evaluating the Effects of Emission Reductions on Multiple Pollutants Simultaneously Deborah Luecken, Alan Cimorelli, Cynthia Stahl and Daniel Tong

Abstract Modeling studies over the Philadelphia metropolitan area have examined how emission control strategies might affect several types of air pollutants simultaneously. NOx reductions in July are predicted to increase ozone in the urban core and decrease it elsewhere, decrease PM2.5 and formaldehyde, and slightly increase acetaldehyde and 1,3-butadiene. In January, NOx reductions increase ozone, formaldehyde and acetaldehyde everywhere. VOC reductions decrease aldehydes but have little effect on ozone in this domain. A combination of VOC and NOx reductions reflects the cumulative behavior of each of the emission reductions separately, and minimizes disbenefits for both HAPs and ozone. A comparison of these changes in terms of their effect on health shows that differing behavior of PM2.5 and ozone can counterbalance each other to some extent. While changes in HAPs are affected by changes to reduce ozone and PM2.5, their effect on health impacts is smaller than PM2.5 and ozone. This study supports considering effects of multiple pollutants in determining optimum pollution control strategies. Keywords Emission control, HAPs, multipollutant, ozone 1. Introduction Many areas around the world have air quality problems with simultaneously high concentrations of one or more pollutants, including ozone (O3), particulate matter (PM2.5), oxides of nitrogen (NOx) and/or hazardous air pollutants (HAPs). There is a growing awareness that pollution control should be considered for its overall benefit to pollutants, rather than on a single pollutant basis (Scheffe et al., 2007). There has been some discussion on the potential effect of volatile organic hydrocarbon (VOCs) and NOx control on species other than ozone, such as other oxygenated nitrogen species and secondary organic aerosol (National Research Council, 1991; Russell et al., 1988; Blanchard et al., 2007), but few comprehensive, multi-pollutant studies, especially considering effects on HAPs. Ozone and the secondarily-produced portion of PM2.5 and HAPs are interrelated through complex atmospheric photochemistry, so control strategies for PM2.5 and ozone might C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

623

D. Luecken et al.

624

decrease or increase HAP concentrations. There are ongoing efforts to reduce concentrations of important pollutants to meet health standards for ozone, NOx and particles, but these reductions will affect the concentrations of radicals and VOCs that produce and destroy HAPs. We need to understand the effect of these controls on HAP concentrations in order to calculate the full economic benefit of control strategies or compare alternative strategies. In this study, we use a three-dimensional air quality model, the Community MultiScale Air Quality (CMAQ) modeling system, to examine the effect of emission control strategies on concentrations of ozone, PM2.5, and four important HAPs: formaldehyde, acetaldehyde, 1,3-butadiene and benzene. The objective of this paper is to begin to address the question of how control strategies formulated for pollutants such as ozone and PM might benefit or disbenefit other pollutants.

2. Model Formulation and Application Model simulations are centered on the Philadelphia metropolitan area, with a 4-km horizontal grid size, (76 by 82 cells), and 15 vertical layers, nested within a continental-scale simulation with a 36-km horizontal grid resolution (Figure 1). We used CMAQ v4.5 (Byun and Schere, 2003; Community Modeling and Analysis System, 2006). Within this domain, we selected four grids representing different chemical characteristics, noted as A, B, C and D. Grid A is located in urban central Philadelphia, B is upwind of the urban area, C and D are on the same latitude as A, but more rural. Emissions for 1999 were used, with a refined mobile source inventory for Philadelphia. The MM5 v3.6.1 model provided meteorological fields (MM5 Community Model, 2007). We performed base case model simulations for January and July, 2001. The SAPRC-99 chemical mechanism (Carter, 1990), modified to include 26 explicit air toxics (Luecken et al., 2006) characterizes the chemistry. 50

45

-76

-74

40

C

35

B

A D

40

39

30

25 -120

-110

-100

-90

-80

-70

Fig. 1 Simulation domain with nested area and location of grid cells for detailed analysis

38

Evaluating the Effects of Emission Reductions on Multiple Pollutants Simultaneously

625

To examine how pollutant concentrations might be affected differently by emission reductions, we performed sensitivity studies with across-the-board anthropogenic emission reductions. The studies simulated two periods, July 14–25, 2001 and January 10–21, 2001. While these reductions do not reflect “realistic” control scenarios, they mimic potential future emission reductions. The three scenarios are (1) 50% reduction in nitric oxide and nitrogen dioxide emissions (NOx-only); (2) 20% reduction in VOC emissions (VOC-only); and (3) 50% reduction in NOx and 20% VOC reduction in VOC emissions (NOx + VOC). Biogenic emissions constitute a significant portion of VOC emissions, so the overall cut in the VOC-only scenario is less than 20%. To account for reduction of pollutant transport, we reduced NOx and VOCs at the boundaries by 50% or 20% of the anthropogenic portion. We compare strategies at the four different grid points shown in Figure 1.

3. Results

3.1. Concentrations Changes in ozone and PM2.5 for the July simulation are shown in Figure 2. Ozone values are 12-day averages of the daily maximum 8-hour average concentrations, and PM2.5 values are calculated as 12-day averages. This domain has an urban corridor (represented by grid A) that is largely VOC-sensitive but surrounding areas that are NOx-sensitive. Ozone in the urban corridor is predicted to increase when NOx is reduced, while other grids show ozone decreases in the NOx-only scenario in July. The VOC-only scenario has a small effect on decreasing ozone, and the NOx +VOC simulations show larger benefit and less disbenefit than the NOx-only simulations. 3 50% NOx

20% VOC

50%NOx +20%VOC

50% NOx

20% VOC

50%NOx +20%VOC

8

2

4

1

0

0 PM2.5 12-day avg

-4

-8

-12

Ozone avg of 1-hr max

Site A (urban) Site B (upwind) Site C (west of city) Site D (east of city)

-1

-2

Concentration change (base minus control),Pg/m3

Concentration change (base minus control), ppb

12

-3

Fig. 2 Change in concentrations of ozone and PM2.5 for each sensitivity scenario during July

D. Luecken et al.

626

The NOx-only reductions have a small effect on the PM2.5 concentrations (4–12% reductions) due to decrease in formation of aerosol nitrate. Nitrate comprises a small portion of the total PM2.5 (1–17%) and we predict the largest PM2.5 decreases where the fraction of nitrate in PM2.5 is largest (grid C, where ammonia emissions are large) and the smallest effect in grid D, where nitrate is low. There is a small change in PM2.5 because changes in ozone and OH affect formation of organic aerosol (increases at grid A, decreases elsewhere) and aerosol sulfate. For the January episode, the ozone increases in the NOx-only scenario for all grids, slightly less so in the VOC + NOx scenario. The PM2.5 changes are less than 0.5%. Figure 3 shows changes in the 12-day average concentrations of formaldehyde, acetaldehyde, 1,3-butadiene and benzene for each of the emission reduction scenarios, calculated as (base – control). Formaldehyde concentrations are reduced by a small amount and acetaldehyde concentrations are increased on average when NOx emissions are reduced, but the variability of the data is also large. The 1,3butadiene increases in C and D and decreases in A and B. The aldehydes are primarily produced in the atmosphere from other VOCs, so changes in ozone, OH, hydroperoxy radical and organic radicals resulting from emission cuts affect concentrations of formaldehyde and acetaldehyde in a complex manner. Although all four HAPs that we examine are VOCs, the concentrations of aldehydes do not change linearly with the VOC reductions. Benzene, on the other hand, shows an approximately linear reduction with VOC emission reductions. In January, both formaldehyde and acetaldehyde increase with the NOx-only scenario but decrease in all others, benzene has negligible change, and 1,3-butadiene decreases for all. Site A (urban) Site B (upwind) Site C (west of city) Site D (east of city)

Concentration change (Base - Control), Pg/m3

0.7 0.6

50% NOx

50%NOx 50% NOx + 20% VOC

20% VOC

50%NOx 50% NOx + 20% VOC 20% VOC

50%NOx 50%NOx 50% NOx + 20% VOC 20% VOC + 20% VOC 20% VOC

0.5 0.4 0.3 0.2 0.1 0

benzene

-0.1 -0.2 -0.3

formaldehyde

acetaldehyde

1,3-butadiene

Fig. 3 Change in concentrations of four HAPs for each sensitivity scenarios during July


627

3.2. Comparison of changes in terms of health days lost Differing responses to pollutant reductions by ozone, PM2.5 and HAPs make it difficult to determine an overall, optimum control strategy. What types of strategies should be pursued if a strategy increases some pollutants but decreases others? To compare relative changes for different types of pollutants, we attempt to normalize them by converting into health-related effects. There are several different ways to characterize health effects. Exposure modeling, as described in Georgopoulos et al. (2005), provides one of the most accepted and complete ways to estimate overall health effects by accounting for indoor and outdoor pollutant sources, movement of people, local peaks, etc. While these models are appropriate for small-scale studies with finely-resolved concentration gradients, they are more difficult to apply and require fine-scale inputs. Another option for estimating changes in health endpoints is concentrationresponse functions derived from relationships between ambient concentrations and health effects (US EPA, 1999). These are useful for analyzing regional-scale control strategies where only ambient sources are controlled. Because the goal of this study is to estimate relative changes in PM2.5, ozone and HAP concentrations, we use concentration-response functions to quantify the direction and approximate changes in health effects due to changes in each pollutant. While not as comprehensive as exposure models, the results can screen potential control scenarios which can be followed up with more detailed exposure modeling. To include effects in both summer and winter, we summed January and July responses. For ozone and PM2.5, we follow the method used for ozone by Tong et al. (2006). Short-term mortality, hospital admissions, and ER visits for respiratory conditions were included for ozone. For PM2.5, long-term mortality for adults 30 years and older was included. For ozone and PM2.5 concentration changes, 'Ci, we calculate change in the base rates of each health endpoint, 'Hi, using a log-linear equation:

'Hi

>expEi * 'Ci @

(1)

where E = 0.00052 for daily ozone short-term mortality (Bell et al., 2006), E = 0.00631 for hospital visits (Burnett et al., 2001), and E =0.0035 for emergency room visits for asthma (Stieb et al., 1996). For PM2.5, we use E = 0.006 for annually-averaged PM2.5 (Pope et al., 2002) for adults older than 30 years. We use a 25 ppb threshold value for ozone, and no threshold for PM2.5. We report the total value in terms of change in health days, HDi from the base rate for each population age group and the days lost for each mortality, Dm, with a median lifetime of 77 years:

HDi

Hbase , i * 'Hi 1 * population * Dm

(2)

There is not a consensus on how to compare health effects from HAPs with those from ozone and PM2.5 because the health effects differ. For HAPs examined in this

D. Luecken et al.

628

study, carcinogenic effects are of most concern. To calculate health days lost due to mortality from carcinogenic effects of HAPs, we use linear equations of the form

'Hi HDi

Ei * 'Ci / 70

(3)

'Hi * population * Dm * 0.44

(4)

Values of E are Unit Risk Estimates reported in the US EPA Integrated Risk System (US EPA, 2007), with E = 1.3E-5 for formaldehyde, E = 2.2E-6 for acetaldehyde, E = 3.0E-5 for 1,3-butadiene and E = 5.0E-6 for benzene. The changes in health days are based on a 70 year lifetime, consistent with National Air Toxic Assessment (US EPA, 2006), a 44% fatality rate for all cancers (American Cancer Society, 2001) and the base mortality distribution. Figure 4 shows the overall change in health days for each scenario at the four grids for ozone, PM2.5, the sum of HAPs and the overall sum. At some grids for some scenarios, the PM2.5 and ozone have opposite effects on health days, and the overall sum reflects those counteracting effects. The HAPs examined in this study show small contributions to the changes in health days. Because only four out of hundreds of recognized HAPs were included in this analysis, the total effects of all HAPs, especially with other high risk HAPs such as diesel PM and acrolein, would be larger than the values displayed here. Since some HAPs decrease under the NOx-controlled scenarios and some increase, the health effects balance out somewhat when computing the sum of effects from HAPs. Grid A

Change in Number of Health Days

8000

4000

4000

0

0

-4000

-4000 50% NOx 20% VOC 50% NOx+20%VOC

-8000

O3 (24-hr)

PM2.5

HAPs

50% NOx 20% VOC 50% NOx+20%VOC

-8000

O3 (24-hr)

Sum

Grid C

8000 Change in Number of Health Days

Grid B

8000

PM2.5

4000

0

0

Sum

Grid D

8000

4000

HAPs

-4000

-4000 50% NOx 20% VOC 50% NOx+20%VOC

-8000

O3 (24-hr)

PM2.5

HAPs

Sum

50% NOx 20% VOC 50% NOx+20%VOC

-8000

O3 (24-hr)

PM2.5

HAPs

Sum

Fig. 4 Change in number of health days for three emission change scenarios at four grids


629

4. Summary and Conclusion Modeling studies over the Philadelphia metropolitan area have examined the potential for emission control strategies to affect several types of air pollutants simultaneously, through both direct and indirect effects of emission reductions. Over this domain, a 50% NOx reduction in July increase ozones in the urban core and decreases ozone elsewhere, decreases PM2.5 and formaldehyde, and slightly increases acetaldehyde. In January, NOx-only reductions increase ozone, formaldehyde and acetaldehyde everywhere, to a significant fractional extent. When considering VOC-only reductions, we predict that a 20% reduction in VOCs decreases aldehyde concentrations everywhere, although the decreases are less than 20%, but has little effect on ozone in this domain. A combination of VOC and NOx reductions reflects the cumulative behavior of each of the emission reductions separately, and minimizes disbenefits for both HAPs and ozone. Comparing these changes in terms of their effect on health allows us to initially rank emissions change scenarios and to compare different scenarios in terms of their overall potential effect on health. The differing behavior of species supports the need to consider effects of multiple pollutants in determining optimum pollution control strategies. While changes in HAPs, including secondarily-produced ones, are affected by changes to reduce ozone and PM2.5, their effect on health impacts is smaller than PM2.5 and ozone. We note that uncertainties in concentrationresponse functions, in HAPs risk estimates, and in base rates of mortality could change the conclusions, and future work should be done to explore the effect of these uncertainties on the identification of optimum control strategies. Acknowledgments We gratefully acknowledge technical assistance of Bill Hutzell, and the support of EPA’s Regional and Applied Research (RARE) Program. Disclaimer The research presented here was performed under the Memorandum of Understanding between the U.S. Environmental Protection Agency (EPA) and the U.S. Department of Commerce’s National Oceanic and Atmospheric Administration (NOAA) and under agreement number DW13921548. Although it has been reviewed by EPA and NOAA and approved for publication, it does not necessarily reflect their policies or views.

References American Cancer Society (2001) Cancer Facts and Figures, 2001. http://www. cancer.org/downloads/STT/F&F2001.pdf Bell ML, McDermott A, Zeger SL, Samet JM, Dominici F (2004) Ozone and shortterm mortality in 95 US urban communities, 1987–2000. Journal of the American Medical Association, 292, 2372–2378. Blanchard C, Tanenbaum, S, Hidy GM (2007) Effects of sulfur dioxide and oxides of nitrogen emission reductions on fine particulate matter mass concentrations:

630

D. Luecken et al.

regional comparisons, Journal of the Air and Waste Management Association, 57, 1337–1350. Burnett RT, Smith-Doiron M, Stieb D, Raizenne ME, Brook JR, Dales RE, Leech JA, Cakmak S. Krewski D (2001) Association between ozone and hospitallization for acute respiratory diseases in children less than 2 years of age, American Journal of Epidemiology, 153 (5), 444–452. Byun D, Schere KL (2003) Review of the governing equations, computational algorithms, and other components of the Models-3 Community Multiscale Air Quality (CMAQ) modeling system, Applied Mechanics Reviews, 59, 51–77. Carter WPL (2000) Implementation of the SAPRC-99 Chemical Mechanism into the Models-3 Framework; Report to the United States Environmental Protection Agency, http://www.cert.ucr.edu/~carter/absts.htm#s99mod3 Community Modeling and Analysis System (2006) University of North Carolina, http://www.cmascenter.org/help/documentation.cfm?temp_id=99999 Georgopoulos PG, Wang S-W, Vikram M, Vyas VM, Sun Q, Burke J, Vedantham R, Mccurdy T, Ozkaynak H (2006) A source-to-dose assessment of population exposures to fine PM and ozone in Philadelphia, PA, during a summer 1999 episode, Journal of Exposure Analysis and Environmental Epidemiology, 15, 439– 457. Luecken DJ, Hutzell WT, Gipson G (2006) Development and Analysis of Air Quality Modeling Simulations for Hazardous Air Pollutants, Atmospheric Environment, 40, 5087–5096. MM5 Community Model (2007) Pennsylvania State University and National Center for Atmospheric Research, http://box.mmm.ucar.edu/mm5 National Research Council (1991) Rethinking the Ozone Problem in Urban and Regional Air Pollution, National Academy Press, Washington, DC. Pope C.A, Burnett RT, Thun MJ, Calle EE, Krewski D, Ito K, Thurston GD (2002) Lung cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air pollution, Journal of the American Medical Association 287(9), 1132– 1141. Russell AG, McCue KF, Cass GR (1988) Mathematical modeling of the formation of nitrogen-containing pollutants. 2. Evaluation of the effect of emission controls, Environmental Science and Technology 22, 1336–1347. Scheffe R, Hubbell B, Fox T, Rao V, Pennell W (2007) The rationale for a multipollutant, multimedia air quality management framework, Environmental Manager, May 2007, 14–20. Stieb, DM, Burnett RT, Beveridge RC, Brook JR (1996) Association between ozone and asthma emergency department visits in Saint John, New Brunswick, Canada, Environmental Health Perspectives 104(12), 1354–1360. Tong D, Muler NZ, Mauzerall DL, Mendelsohn RO (2006) Integrated assessment of the spatial variability of ozone impacts from emissions of nitrogen oxides, Environmental Science and Technology, 40, 1395–1400. US EPA (2007) Integrated Risk Information System, http://www.epa.gov/iris US EPA (2006) 1999 National Scale Air Toxics Assessment, http://www.epa.gov/ ttn/atw/nata1999/nsata99.html


631

US EPA (1999) The Benefits and Costs of the Clean Air Act 1990 to 2010, EPA Report to Congress, http://www.epa.gov/oar/sect812/prospective1.html

Discussion S.T. Rao: In addition to the concentration-response functions, are you looking at the exposure-response functions is assessing the impact of emission reduction strategies? D. Luecken: We have used concentration-response functions to perform our initial analyses because we are looking at relative changes in health effects. However, our plans for follow-up work on the next project include using exposure modelling as a more detailed way to compare overall health impacts in the Baltimore, MD area.

7.2 Long-Term Regional Air Quality Modelling in Support of Health Impact Analyses C. Hogrefe, B. Lynn, K. Knowlton, R. Goldberg, C. Rosenzweig and P.L. Kinney

Abstract This paper investigates the use of long-term regional scale meteorological and air quality simulations for tracking changes in air quality and for supporting public health assessments. For this purpose, year-round simulations with the MM5/CMAQ modelling system have been performed over the northeastern United States for 1988–2000. Emission inputs for the CMAQ simulations were prepared with the SMOKE processing system and were based on 1990 and 1996– 2000 inventories. During this period, significant reductions in anthropogenic emissions of VOC, NOx, and SO2 have occurred in the point source and mobile source sectors. Model evaluation results show that the modelling system performs better in capturing temporal than spatial patterns. Moreover, the modelling system captured the effects of SO2 emission reductions on SO4 concentrations in ambient air and rain water. Finally, examples are provided for how these CMAQ simulations could be used in combination with observations to study the link between air quality and public health.

Keywords Acid deposition, emission trends, health impacts, regional-scale air quality modelling

1. Introduction To date, most studies investigating the link between air pollution and human health have relied on monitoring data to characterize ambient pollutant concentrations (e.g. Bell et al., 2004; Samoli et al., 2005, and references therein). More recently, studies have begun to explore the potential benefits of incorporating concentration fields simulated by grid-based air quality modelling systems into health impact assessments (Bell, 2006). In this paper, we present results from a study aimed at performing long-term air quality simulation over the northeastern portion of the U.S. and assessing the usefulness of these simulations for health impact studies. We describe the design of the model simulations in Section 2, present evaluation results for these simulations in Section 3, and finally provide examples for the potential use of these model simulations and discuss next steps in Section 4.


605

C. Hogrefe et al.

606

2. Description of Modelling System and Databases 2.1. Modelling system Meteorological conditions for the time period from January 1, 1988–December 31, 2000 were simulated with the MM5 meteorological model (Grell et al., 1994). The simulations were performed on two nested grids with 36 and 12 km horizontal grid spacing, respectively. Among the physics options chosen for the simulation are the ETA scheme for representing the planetary boundary layer, the Kain-Fritsch cloud scheme, and the RRTM radiation scheme. Throughout the model simulation, MM5 was nudged towards NCEP reanalysis fields using four-dimensional data assimilation. Annual anthropogenic emission inventories for area, nonroad and point sources were obtained from the EPA National Emission Trends database for 1990 and 1996–2000 (U.S. EPA, 2005). The 1990 emissions were used for the simulation of 1988–1990, while emissions from 1991–1995 were estimated by interpolation between 1990 and 1996 for these source categories. Onroad mobile source emissions were estimated with the MOBILE6 model using annual county-level vehicle miles travelled (VMT) and MM5 temperatures from 1988 to 2000. Biogenic sources for 1988–2000 were estimated with the BEIS3.12 model taking into account MM5 temperature, radiation, and precipitation. All emissions processing including mobile sources and biogenic sources was performed within the SMOKE system (Houyoux et al., 2000). Table 1 provides a summary of domain-total annual anthropogenic emissions for 1990 and 2000. It is evident that significant emission reductions for all pollutants have occurred between 1990 and 2000. An analysis of the time series of the annual total anthropogenic emissions from 1988 to 2000 utilized in these simulations reveals a gradual decline of NOx emissions and a the steep reduction of SO2 emissions in the mid-1990s. This reduction is due to Title IV of the Clean Air Act Amendments and its impact on observed and simulated SO4 concentrations is discussed in Section 3. Using the meteorological and emission fields described above, hourly gridded fields of concentrations, wet deposition, and dry deposition for a large number of gas phase and aerosol species were simulated with the Community Multiscale Air Quality (CMAQ) model (Byun and Schere, 2006), version 4.5.1. The simulations were performed with two nested grids of 36 and 12 km, corresponding to the MM5 grids except for a ring of buffer cells. The boundary conditions for the 36 km grid Table 1 Annual total anthropogenic emissions over the 12 km modelling domain processed by SMOKE for 1990 and 2000. All emissions are shown in kilotons. Area + nonroad Mobile Point

1990 2000 1990 2000 1990 2000

NOx

VOC

CO

SO2

NH3

PM2.5

1,933 1,757 2,895 2,180 3,443 1,884

3,580 3,353 2,877 1,461 1,209 334

10,379 11,627 31,924 18,190 7,743 1,190

1,018 691 67 65 8,725 5,088

502 513 42 68 15 11

790 661 77 41 172 328

Long-Term Regional Air Quality Modelling in Support of Health Impact Analyses

607

correspond to climatological values, while the boundary conditions for the 12 km grid were derived from the 36 km simulation. Gas phase chemistry was represented by the CB-IV mechanism while aerosol chemistry was simulated with the aero3 module. For all subsequent analyses, only simulations from the 12 km CMAQ grid were utilized.

2.2. Observations Observed daily maximum 8-hour ozone concentrations for the period 1988–2000 were determined from hourly ozone observations at surface monitors from the U.S. EPA’s AQS database. In order to be included in the analysis, monitors had to be (a) located within the 12 km CMAQ domain, and (b) have at least 40% non-missing days during each year between 1988 and 2000. The application of these screening criteria resulted in the selection of 112 monitors. For the evaluation of SO4 aerosol concentrations, observed weekly-average air concentrations were extracted at eight monitors located within the 12 km CMAQ domain from the Clean Air Status and Trends Network (CASTNet) database. Furthermore, weekly-average precipitationweighted rainwater SO4 concentrations were extracted from the National Acid Deposition Program (NADP) database for 25 monitors located within the 12 km CMAQ domain.

3. Model Evaluation Figure 1a, b show time series of model performance statistics for daily maximum 8-hour ozone calculated for each ozone season (May–October) from 1988–2000. For each ozone season, all daily observations – model pairs located in the modelling domain were utilized to compute these statistics, therefore, the metrics measure the model’s ability to capture the total temporal and spatial variability of ozone observations. Figure 1a indicates that there is year-to-year variability but no obvious trend in correlation coefficients, model bias, and the ratio of simulated to observed standard deviations. On the other hand, the root mean square error (RMSE) shows a downward trend throughout the analysis time period from about 18 ppb for the 1988 simulation to about 16 ppb for the 2000 simulation. Following the definition by Willmot (1982), Figure 1b shows the decomposition of the total RMSE into its systematic and unsystematic components and reveals that most of the reduction is driven by the unsystematic RMSE with smaller changes in the systematic RMSE.

608

C. Hogrefe et al.

While the analyses shown above measured the total spatial and temporal variability and indicated that model performance for daily maximum 8-hour ozone for all summers was within typical ranges found for regional-scale photochemical models (Hogrefe et al., 2007), it is of interest to separately investigate model performance in time and space. The rationale is that in health impact applications, models can potentially be used to estimate pollutant concentrations both during unmonitored time intervals and at unmonitored locations, but the skill of the model between these two tasks may differ. Figure 2 shows distributions of correlation coefficients, total RMSE, systematic RMSE, and unsystematic RMSE for daily maximum 8-hour ozone for both time series and spatial patterns. The time series distributions for a given quantity (e.g. total RMSE) were constructed by calculating this quantity for each observed and simulated time series for each station and each ozone season, therefore, the time series distributions consist of 1,456 datapoints (13 years times 112 stations). In contrast, the spatial pattern distributions for a given quantity were constructed by calculating this quantity for each observed and

Fig 1 (a) Time series of correlation coefficients, bias, ratio of simulated to observed standard deviations, and total RMSE calculated from observation and model predictions for each May 1– October 31 time period for each year. (b) Same as in (a), but for total, systematic, and un systematic RMSE

Fig 2 Distributions of correlation coefficients (upper left), total RMSE (upper right), systematic RMSE (lower left), and unsystematic RMSE (lower right) for simulated time series (solid lines) and spatial patterns (dotted lines)


609

simulated map of daily maximum 8-hour ozone for each ozone season day and each year, therefore, the time series distribution consists of 2,379 datapoints (13 years times 183 days). Results show that correlation coefficients between observed and simulated time series are substantially higher and have a much narrower distribution than correlation coefficients between observed and simulated spatial patterns. For total RMSE, both the time series and spatial pattern distributions show similar medians but a larger spread for the spatial pattern distribution. While most of the temporal variability in observations and model predictions typically is due to synoptic and seasonal scale fluctuations (Rao et al., 1997), changes in anthropogenic emissions such as those reported in the NEI and shown in Table 1 also play a role. Moreover, since these changes are typically caused by control programs designed to improve public health, the model’s ability to capture the effects of such emission control program is an important consideration when applying models for health impact studies. Here, we investigate the impacts of the steep reductions in SO2 emissions in 1995 due to the implementation of Title IV of the Clean Air Act Amendments. Other studies have shown that these emission reductions led to decreases in acid deposition over the Eastern U.S. (e.g. Lynch et al., 2000). To assess whether CMAQ captures this signal, Figure 3 shows time series of observed and CMAQ aerosol SO4 concentrations in ambient air averaged over eight CASTNet monitors in the modelling domain and precipitation-weighted SO4 and NO3 concentrations measured in rainwater averaged over 25 NADP monitors. For the CASTNet time series, a moving average window of 15 days iterated five times was applied to illustrate seasonal patterns and trends. For the NADP time series, observed weekly samples and corresponding CMAQ values were weighted by the weekly precipitation amount and were then aggregated to compute annual average concentrations for each year from 1988 to 2000 at each site; results were averaged over all sites for presentation in this figure. The CASTNet results show that CMAQ general captures the magnitude and seasonal fluctuations of observed aerosol SO4 concentrations as well as the decrease after 1995. The comparison of the observed and simulated annual-average rainwater SO4 concentrations at the NADP sites shows that CMAQ captures the reduction of concentrations in the mid 1990s and also tends to capture interannual variability in observed concentrations.

Fig. 3 Time series of observed and CMAQ aerosol SO4 concentrations in ambient air averaged over 8 CASTNet monitors in the modelling domain (left) and precipitation-weighted SO4 and NO3 concentrations measured in rainwater averaged over 25 NADP monitors (right). Note, a moving average window of 15 days iterated five times was applied to the CASTNet time series to illustrate seasonal patterns and trends, while annual average values are shown for the NADP time series

610

C. Hogrefe et al.

Separate analysis shows that the change between the 1989–1994 average and the 1995–2000 average concentration is –0.53 mg/l (–23%) for the observations and –0.63 mg/l (–24%) for CMAQ. Moreover, this separate analysis shows that CMAQ not only captures the overall magnitude of the change but also its spatial pattern.

4. Discussion and Next Steps As discussed above, two of the motivations for using photochemical modelling systems such as CMAQ to enhance the characterization of air quality in health impact studies are the potential ability to estimate concentrations in unmonitored locations and to estimate concentrations during unmonitored time periods and for unmonitored species. In this section, we provide examples of both applications. For estimating values at unmonitored locations, it is clear that the model errors at monitored locations discussed above need to be taken into account. One possible approach to achieve this objective is to estimate differences between observed and CMAQ concentrations at each monitor for each day, spatially interpolate these differences using inverse-distance weighting, and then add the interpolated differrence field to the gridded CMAQ output. This approach is illustrated in Figure 4 for daily maximum 8-hour ozone concentrations for July 12, 1999. It can be seen that the overpredictions of the original CMAQ output in the eastern part of the modelling domain are corrected in the combined surface. Furthermore, as intended the combined surface provides estimates of ozone concentrations in unmonitored regions. To evaluate this methodology for 8-hour daily maximum ozone, leave-one-out cross validation was performed at each site for each day during the ozone seasons of 1988–2000, and the average absolute cross-validation error was 0.2 ppb. Other potential methods for accomplishing the same objective would be to construct weighted averages of observation and model simulations or employ a hierarchial Bayesian approach as described in McMillan et al. (2007). The second potential application for CMAQ in health impact studies is to estimate total and speciated PM2.5 during time periods when measurements are not available. In the U.S., such time periods include the years prior to 1999 in which hardly any PM2.5 measurements were performed in the U.S., and the estimation of PM2.5 on days when filter samplers which typically follow a one-in-three days schedule are not operating. Figure 5 shows the CMAQ simulated composition of PM2.5 over the entire simulation time period at the location of CASTNet monitors as stacked time series. It can be seen that SO4 is the major component of PM2.5 in this modelling domain but that its importance has decreased as a result of the SO2 emission reductions discussed earlier. There also is a clear seasonal variation in the composition of PM2.5. While the accuracy of the simulated species concentrations shown in these figures cannot be evaluated due to the lack of detailed speciated PM2.5 for most of the simulation time period, the limited analysis of aerosol SO4 concentrations from CASTNet and SO4 concentrations in rainwater from the NADP network in the previous section indicates that the seasonal-scale fluctuations and trends of this important constituent are well captured by CMAQ, building


611

25

20

15

10

OM EC NH4 NO3 SO4

5

0 Jan- Jan- Jan- Jan- Jan- Jan- Jan- Jan- Jan- Jan- Jan- Jan- Jan88 89 90 91 92 93 94 95 96 97 98 99 00

Fig. 4 Example illustration for combining observations and CMAQ simulations. The maps show daily maximum 8-hour ozone concentrations for July 12, 1999 for observations (upper left), CMAQ (upper right), spatially interpolated differences (lower left), and the combined observation/CMAQ surface (lower right)

Fig. 5 Example for estimating total and speciated PM2.5 concentrations for unmonitored time periods. The figure shows a stacked time series of CMAQsimulated PM2.5 spcecies concentrations (bottomto-top: SO4, NO3, NH4, elemental carbon – EC, and organic mass – OM) for January 1, 1988 – December 31, 2000 at the location of CASTNet monitors after the application of a 15-day moving average filter iterated five times

confidence in the use of these estimates for health impact studies. However, it is also necessary to point out that the evaluation of NO3 and organic aerosols simulated for more recent time periods by various studies revealed larger uncertainties compared to SO4 (e.g. Eder and Yu, 2006). Therefore, once CMAQ simulations are completed for more recent time periods covered by speciated observations, methods will be explored to determine species-specific bias adjustments approaches that take into account seasonal and spatial (urban vs rural) fluctuations in model performance. Despite these uncertainties, it is expected that the information about PM2.5 mass and speciation derived from CMAQ (i.e. based upon our current understanding of the interactions between emissions, meteorology, and atmospheric processes) for unmonitored time period will help to study the link between PM2.5 concentrations and health impacts. In the future, daily maps of ambient ozone and PM2.5 concentrations constructed from observations and CMAQ simulations will be used to study the link between air quality and health. More importantly, these results will be compared to health impact studies that relied solely on ambient measurements of air pollution (e.g. Bell et al., 2004; Samoli et al., 2005) to assess the utility of long-term CMAQ simulations in enhancing the characterization of ambient air quality for health impact studies. Furthermore, it will of interest to develop methodologies that take into account uncertainties in the estimated air pollution surfaces during the health analysis. Finally, the eventual goal is to create a framework in which the impact of emission control programs is quantified both in terms of changes in ambient pollutant concentrations and in terms of changes in health outcomes through the integrated use of observations and photochemical models.

612

C. Hogrefe et al.

Acknowledgments The work presented here was supported by the National Oceanic and Atmospheric Administration under award NAO40AR4310185185, but it has not been subjected to its required peer and policy review. Therefore, the statements, findings, conclusions, and recommendations are those of the authors and do not necessarily reflect the views of the sponsoring agency and no official endorsement should be inferred. Christian Hogrefe also gratefully acknowledges partial support for this work through a research fellowship from the Oak Ridge Institute for Science and Education (ORISE).

References Bell ML (2006) The use of ambient air quality modeling to estimate individual and population exposure for human health research: a case study of ozone in the Northern Georgia Region of the United States. Environ. Intern. 32, 586–593, doi:10.1016/j.envint.2006.01.005 Bell ML, McDermott A, Zeger SL, Samet JM, Dominici F (2004) Ozone and shortterm mortality in 95 US urban communities, 1987–2000. J. Am. Med. Assoc. 292, 2372–2378. Byun DW, Schere KL (2006) Review of the governing equations, computational algorithms, and other components of the Models-3 Community Multiscale Air Quality (CMAQ) modeling system. Appl. Mech. Rev., 59, 51–77. Eder B, Yu S (2006) A performance evaluation of the 2004 release of Models-3 CMAQ. Atmos. Environ., 40, 4811–4824 Grell GA, Dudhia J, Stauffer D (1994) A description of the fifth-generation Penn State/NCAR Mesoscale Model (MM5), NCAR Technical Note, NCAR/TN-398 + STR. Hogrefe C, Ku J-Y, Sistla G, Gilliland A, Irwin JS, Porter PS, Gégo E, Kasibhatla P, Rao ST (2007) Has the performance of regional-scale photochemical modelling systems changed over the past decade?, preprints, NATO 28th ITM, Aveiro, Portugal, September 25–29, 2007. Houyoux, MR, JM Vukovich, CJ Coats, Jr, NJM. Wheeler, P. Kasibhatla (2000) Emission inventory development and processing for the seasonal model for regional air quality. J. Geophys. Res., 105, 9079–9090. Lynch JA, Bowersox VC, Grimm JW (2000) Changes in sulfate deposition in eastern USA following implementation of Phase I of Title IV of the Clean Air Act Amendments of 1990. Atmos. Environ., 34, 1665–1680. McMillan, NJ, Holland DM, Morara M (2007) Combining Different Sources of Fine Particulate Data Using Bayesian Space-Time Modeling, in preparation. Rao ST, Zurbenko IG, Neagu R, Porter PS, Ku JY, Henry RF (1997) Space and time scales in ambient ozone data. Bull. Amer. Meteor. Soc., 78, 2153–2166. Samoli E, Analitis A, Touloumi G, Schwartz J, Anderson HR, Sunyer J (2005) Estimating the exposure–response relationships between particulate matter and mortality within the APHEA multicity project. Environ. Health Perspect., 113, 88–95.


613

U.S. EPA (2005) Criteria pollutant emissions summary files extracted from the national emissions inventory (NEI) database, available online at http://www. epa. gov/ttn/chief/net/critsummary.html Willmott CJ (1982) Some Comments on the Evaluation of Model Performance. Bull. Amer. Meteor. Soc., 63, 1309–1313.

Discussion S.T. Rao: With regard to model’s ability to better capture the temporal variability than the spatial variability, would the spatial homogeneity in the monitors (more monitors in urban than in rural areas) contribute to the lower correlation in space? Also urban influences may have to be accounted for incoming models and observed variables? C. Hogrefe: The uneven spacing of monitors between urban and rural areas may contribute to the lower correlations of observed-vs-simulated spatial maps of daily maximum 8-hour ozone compared to the correlations of observed-vs-simulated time series of daily maximum 8-hour ozone. However, the lower correlations persist even when spatial and temporal moving average windows are applied to the raw data, thereby smoothing out most of the urban/rural differences. Therefore, I do believe that most of the lower correlations for spatial patterns are due to errors in the placement of plumes caused by uncertainties in wind fields and emission patterns, while the higher correlations for time series are at least partially attributable to the use of four-dimensional data assimilation in the meteorological model that ensures the correct timing of synoptic events.

7.5 Modelling of the Exposure of Urban Populations to PM2.5, NO2 and O3, and Applications in the Helsinki Metropolitan Area in 2002 and 2025 J. Kukkonen, P. Aarnio, A. Kousa, A. Karppinen, K. Riikonen, B. Alaviippola, M. Kauhaniemi, J. Soares, T. Elolähde and T. Koskentalo

Abstract A mathematical model EXPAND (EXposure model for Particulate matter And Nitrogen oxiDes) can be used to evaluate human exposure to air pollution in an urban area. The model combines the predicted concentrations and the information on the time use of the population on an hourly basis. The model allows for the exposure in residences, workplaces and traffic, and partly also in other activities, such as recreational facilities. The model has been integrated to an urban dispersion modelling system of the Finnish Meteorological Institute. The computed results are processed and visualised using the GIS system MapInfo. The numerical results contain the predicted hourly spatial concentration, time activity and exposure distributions of PM2.5, NO2 and O3 in 2002 and 2025 in the Helsinki Metropolitan Area.

1. Introduction Ambient air pollution has been associated to excess mortality and morbidity at the current urban levels (e.g. Pope and Dockery, 2006; WHO, 2006). Air pollution is an additional risk factor that increases the statistical probability of death and other adverse health effects caused primarily by cardio-vascular and respiratory diseases. Most of the epidemiological studies have been based on air pollution concentrations at fixed ambient air quality monitoring sites. However, the measurement data from these stations does not necessarily represent areas beyond their immediate vicinity, as the concentrations of pollutants in urban areas may vary by orders of magnitude on spatial scales varying from tens to hundreds of metres. Therefore there is a need to model the population exposures to pollutants. In order to evaluate comprehensively the adverse health effects, it would be ideal to evaluate both the exposure of selected individual citizens, and the spatial and temporal variations of the exposures of the whole urban population. However, this is not commonly possible, due either to the lack of relevant input data, or the physical and computational limitations of exposure models, or both reasons. Exposure can be modelled by deterministic or probabilistic models (the latter have C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

632

Modelling of the Exposure of Urban Populations to PM2.5, NO2 and O3

633

been discussed by, e.g., Hänninen et al., 2003). The probabilistic models predict the exposure distributions; however, the results cannot be presented spatially, using GIS techniques. Jensen (1999) and Gulliver and Briggs (2005) have developed exposure models which can combine time-activity, dispersion modelling, and GIS techniques. In both of these models, the exposure of a few individuals is evaluated in different microenvironments during the day; this approach does not therefore allow for the evaluation of the population exposure. We have developed a primarily deterministic mathematical model, EXPAND (EXposure model for Particulate matter And Nitrogen oxiDes), for the determination of human exposure to ambient air pollution in an urban area (Kousa et al., 2002). The model can be used to evaluate the spatial and temporal variation of the average exposure of the urban population to ambient air pollution in different microenvironments. This paper describes a new model version, in which the time activity sub-model has been refined. The refined model includes a detailed treatment of the time use of population in various traffic modes, including cars and buses, trains, trams, metro, pedestrians and cyclists. The model has also been extended to contain a simple treatment of the infiltration of pollutants from outdoor to indoor, and the model can treat separately various population sub-groups. The emission and atmospheric dispersion modules of the system have been refined to include an improved treatment of fine particulate matter (PM2.5).

2. Methodology 2.1. Emission modelling We have modelled the traffic flows in the street network of the Helsinki Metropolitan Area using the EMME/2 interactive transportation planning package. The model generates traffic demand on the basis of given scenarios, and allocates the activity over the links (i.e. segments of road or street) of this network, according to predetermined set of rules and individual link characteristics (Laurikko et al., 2003). According to the link characteristics and number of vehicles, the software computes the average speed of vehicular traffic for each link on given time of the day. Furthermore, both weekly and seasonal variations of the traffic density are taken into account. For modelling purposes, the profiles of vehicle speed and vehicle numbers are then computed for each link for each hour of the day (separately for weekdays, Saturdays and Sundays), and further aggregated over the year. Emissions are computed for each link using average speed-dependent functions, determined separately for each vehicle category. Vehicle categories for 14 different vehicle types are passenger cars, divided to petrol cars with or without a catalytic converter, and diesel-fuelled vehicles, as well as busses and heavy duty vehicles. The division of the vehicles within the passenger car category is based on the registration statistics. The traffic demand generated by the model is governed by the assumed socioeconomic urban structure and location of the main activities, such as residential

634

J. Kukkonen et al.

areas and workplaces, as well as the usage rate of public transport. Local urban bus routes are directly integrated in the model, but incoming/outgoing coach traffic is generated separately. The activities at two major ports in the city of Helsinki also enhance heavy duty vehicle traffic into the arteries in the road and street network.

2.2. Dispersion modelling Dispersion modelling is based on the combined application of the Urban Dispersion Modelling system (UDM-FMI) and the road network dispersion model (CAR-FMI), developed at the Finnish Meteorological Institute (FMI) (e.g., Karppinen et al., 2000, 2004; Kousa et al., 2001). Clearly, the main limitation of such so-called second generation Gaussian dispersion models is that they do not allow for the detailed structure of buildings and obstacles. Approximately 5,000 road and street links were included in the computations. The model uses meteorological parameters from the FMI database and computes ambient air pollution concentrations for each hour over the whole year. The concentrations were computed in an adjustable grid, in which the spatial resolution is dependent on the distance of receptor points to the nearest traffic sources; these range from 5 m in the immediate vicinity of major roads to 1 km in the least polluted areas.

2.3. Activity modelling We obtained the information on the location of the population from the data set that is collected annually by the municipalities of the Helsinki metropolitan area. This data set contains data on the dwelling houses, enterprises and agencies located in the area. The data set provides geographic information on the total number and age distribution of people living in a particular building or the total number of people working at a particular workplace. The information on the number and location of people in shops, restaurants and other recreational activities is also based on this data set. The location of people in traffic was evaluated using the computed traffic flow information; this information is available separately for buses, cars, trains, trams, metro, pedestrians, and cyclists for each street and rail section on an hourly basis. However, this information does not identify individual persons. The time-microenvironment activity data is based on the time use survey by Statistics Finland. The time activity data were collected from 813 randomly selected over ten-year old inhabitants in the Helsinki metropolitan area. For our model the time-activity of the population was divided into four main categories: home, workplace, traffic, and other activities.


635

2.4. Exposure modelling The residential co-ordinates are combined with the information on the number of inhabitants at each building and the time spent at home during the day. Correspondingly, for the workplace co-ordinates, the number and age distribution of the personnel, and the time spent at the workplace are combined. The population activities at other locations (shops, cinemas, theatres, opera, libraries, restaurants, cafes, pubs, etc.) are also evaluated using statistical information of leisure time (City of Helsinki Urban Facts, 2003). The number of persons in traffic is evaluated based on the predicted traffic flows. In the case of buses, trains, metro, trams and also pedestrians and cyclists, the number of persons and the time they spend in each street or rail section is estimated using the traffic-planning model EMME/2. In the case of private cars, the EMME/2 model predicts the number of cars; we assumed that the number of passengers in each car is equal to the average value in the area, i.e., 1.33 (Hellman, 2004). The ratio of pollutant concentrations in indoor and outdoor air (I/O) is also included in the model. The I/O ratio data is based on the results of EXPOLIS study (Hänninen et al., 2004). The I/O ratios of 0.59 for PM2.5 and 0.71 for NO2 were used for buildings, and 1.0 for traffic (both for PM2.5 and NO2). Clearly, this is a very simple approach; for a more detailed evaluation of indoor concentrations, one would need reliable data on the I/O ratios in terms of the building characteristics, and especially for various traffic modes. In the model, the air pollutant concentrations are interpolated on to a rectangular grid. The data regarding population activities (number of persons * hour) is also transformed to the same grid. The GIS system MapInfo is subsequently utilised in the post-processing and visualisation of this information.

3. Results and Discussion 3.1. Predicted concentrations and their comparison with measured data The concentrations of PM2.5, NO2 and O3 were predicted for 2002, and also for selected scenarios for 2025. The model predictions have also been compared with the data measured at urban monitoring stations. Both the daily and hourly averaged predicted PM2.5 concentrations agreed fairly well with the measured data both at an urban roadside and urban background station. For instance, the correlation coefficient squared between the measured and predicted sequential daily averaged PM2.5 concentrations were 0.54 and 0.60 for these two stations. For a more detailed description of the concentration computations regarding PM2.5, and their comparison to measured data, the reader is referred to Kauhaniemi et al. (2007).

636

J. Kukkonen et al.

3.2. Definition of the scenarios for the future The scenario for 2025 is based on the so-called transport system plan that has been projected by the Helsinki Metropolitan Area Council. The transport system plan takes into consideration all aspects of the transport system within the metropolitan area. It defines common objectives and focal points for the long-term development of the transport system, allowing also for the conceivable development of the regional transport policies. For instance, it is assumed that there will be 1,133,500 inhabitants, 662,400 working places and approximately 456 cars per 1,000 inhabitants in the Helsinki Metropolitan Area in 2025. The corresponding figures for 2000 are 928,950 inhabitants, 540,401 working places and 348 cars per 1,000 inhabitants. This plan also estimates, how land use and transport will evolve up to 2025 (e.g., it assumes new underground and railway lines, and new dwelling areas).

Fig. 1 a, b The predicted number of persons that is exposed to the concentration of nitrogen dioxide exceeding 30 µg/m3 during the morning rush hour in 2002 in the Helsinki Metropolitan Area, on a characteristic winter day (upper figure), and during a day with a prevailing surface temperature inversion (lower figure)


637

3.3. Examples of the spatial and temporal variation of exposure

The number of working age population

The spatial distribution of population that is exposed to a NO2 concentration exceeding a specified value during a morning rush hour on a characteristic day in winter has been presented in Figure 1a. The ‘characteristic’ winter day was selected to represent meteorological dispersion conditions that are common in winter, especially regarding the values of temperature, wind speed and stability. As expected, during the morning rush hour the working age population is mostly located on major traffic routes and in the residential areas. e.g., during the morning rush hour, two thirds of the working age population was in the environments where they exposed to the nitrogen dioxide concentrations less than 20 µg/m3. These environments were mostly homes or workplaces. People were exposed to the highest concentrations in traffic, near the busy traffic lanes and in the city centre. For comparison purposes, the corresponding results have been presented in Figure 1b for an episodic day, during which high concentrations were mainly caused by the local vehicular emissions due to a ground-based temperature inversion. There are substantial differences both in the magnitude and the spatial distribution of the exposures of the working age population for these two cases. For instance, during the episodic case, the number of population exposed to nitrogen dioxide concentration larger than 30 µg/m3 is more than 2.2 times higher than that during the typical winter day. The distributions of the number of persons in terms of concentrations are presented in Figure 2. 200000 180000 160000

Winter

140000

Inversion

120000 100000 80000 60000 40000 20000 0 50

Fig. 2 The distributions of the number of the working age population exposed to the selected ranges of the NO2 concentrations in the Helsinki Metropolitan Area during the morning rush hour in 2002 for two cases: (i) a characteristic winter day (legend “Winter”), and (ii) an episodic day with a prevailing ground-based temperature inversion (legend “Inversion”)

3.4. Predicted results for the scenario for 2025 The transport system plan contains several alternative scenarios for the future. We have applied here only the basic “business-as-usual” –type scenario. The working age population was predicted to be exposed to considerably lower concentrations

638

J. Kukkonen et al.

in 2025, compared with the corresponding results for 2002. According to the basic 2025 scenario, over 80% of the working age population would be exposed to the nitrogen dioxide concentrations less than 10 µg/m3, whereas the corresponding percentage for the working age population in 2002 was approximately 25%. The substantially lower exposures of NO2 in 2025 are mainly caused by the projected lower NOx emissions from local vehicular traffic.

4. Conclusions The numerical results contain the predicted hourly spatial concentration, time activity and exposure distributions of PM2.5, NO2 and O3 in 2002 and 2025 in the Helsinki Metropolitan Area. The results illustrate the most problematic areas and time periods with concurrent high population activity and concentration values. The highest population exposures occur especially in the centre of Helsinki, and along the major traffic routes. As expected, the population exposures are also temporally highly variable diurnally, seasonally and in terms of specific meteorological conditions. The methodologies developed, and the EXPAND model itself, are available to be utilised also in other European urban areas (e.g., Baklanov et al., 2007), and within other integrated modelling systems (e.g., Sokhi et al., 2007). The model, including the GIS-based methodology, could also be extended on a regional scale. It would be especially important to evaluate the exposure of children; however, the data on their time use is scarce. The projections of the time activities for the future also involve major uncertainties. For instance, it is conceivable that both the magnitude and types of time activities during the leisure time will change dramatically in the future. Acknowledgments The study was supported by the EU-funded FUMAPEX and OSCAR projects, and it is part of the CLEAR cluster of air quality projects. We also wish to acknowledge the funding of the Academy of Finland (TERVE, project no 53246).

References Baklanov A, Hänninen O, Slørdal LH, Kukkonen J, Bjergene N, Fay B, Finardi S, Hoe SC, Jantunen M, Karppinen A, Rasmussen A, Skouloudis A, Sokhi RS, Sørensen JH, Ødegaard V (2007) Integrated systems for forecasting urban meteorology, air pollution and population exposure. Atmos. Chem. Phys., 7, 855–874, www.atmos-chem-phys.net/7/855/2007/ City of Helsinki Urban Facts (2003) Statistical Yearbook of the City of Helsinki. Gummerrus Kirjapaino Oy, Jyväskylä. Gulliver J, Briggs D (2005) Time-space modelling of journey-time exposure to traffic-related air pollution using GIS. Ph.D. thesis. Environmental Research 2005, 97 (1), 10–95.


639

Hellman T (2004) The average burden of the private vehicles in Helsinki in 2004. City Planning Department, City of Helsinki. Memo 21.6.2004. Helsinki (in Finnish). Hänninen O, Kruize H, Lebret E, Jantunen M (2003) EXPOLIS simulation model: PM2.5 application and comparison with measurements in Helsinki. J. Exp. Anal. Environ. Epidemiol. 13, 74–85. Hänninen O, Lebret E, Ilacqua V, Katsouyanni K, Künzli N, Sram R, Jantunen M (2004) Infiltration of ambient PM2.5 and levels of indoor generated non-ETS PM2.5 in residences of four European cities. Atmos. Environ. 38, 6411–6423 Jensen SS (1999) A Geographic Approach to Modelling Human Exposure to Traffic Air Pollution using GIS. Ph.D. thesis. National Environmental Research Institute, Denmark. Karppinen A, Kukkonen J, Elolähde T, Konttinen M, Koskentalo T, Rantakrans E (2000) A modelling system for predicting urban air pollution, Model description and applications in the Helsinki metropolitan area. Atmos. Environ. 34–22, pp. 3723–3733. Karppinen A, Härkönen J, Kukkonen J, Aarnio P, Koskentalo T (2004) Statistical model for assessing the portion of fine particulate matter transported regionally and long-range to urban air. Scand. J. Work Environ. Health, 30 suppl. 2: 47–53. Kauhaniemi M, Karppinen A, Härkönen J, Kousa A, Koskentalo T, Aarnio P, Kukkonen J (2007) Refinement and statistical evaluation of a modelling system for predicting fine particle concentrations in urban areas. In: R.S. Sokhi, M. Neophytou (eds), Proceedings of the 6th International Conference on Urban Air Quality, Limassol, Cyprus, 27–29 March 2007, CD-disk: ISBN 978-1-90531346-4, University of Hertfordshire and University of Cyprus (pp. 68–71). Kousa A, Kukkonen J, Karppinen A, Aarnio P, Koskentalo T (2001) Statistical and diagnostic evaluation of a new-generation urban dispersion modelling system against an extensive dataset in the Helsinki Area. Atmos. Environ., Vol. 35/27, pp. 4617–4628. Kousa A, Kukkonen J, Karppinen A, Aarnio P, Koskentalo T (2002) A model for evaluating the population exposure to ambient air pollution in an urban area. Atmos. Environ. 36, 2109–2119. Laurikko J, Kukkonen J, Koistinen K, Koskentalo T (2003) Integrated modelling system for the evaluation of the impact of tranport-related measures to urban air quality. In: Proceedings of the 12th symposium “Transport and Air Pollution”, 16–18 June 2003, Avignon, France. Pope CA, Dockery DW (2006) Health effects of fine particulate air pollution: lines that connect. J. Air Waste Manage. Assoc. 56, 709–742. Sokhi, R.S, Hongjun Mao, Srinivas TG, Srimath, Shiyuan Fan, Nutthida Kitwiroon, Lakhumal Luhana, Jaakko Kukkonen, Mervi Haakana, K Dick van den Hout, Paul Boulter, Ian S McCrae, Steinar Larssen, Karl I Gjerstad, Roberto San Jose, John Bartzis, Panos Neofytou, Peter van den Breemer, Steve Neville, Anu Kousa, Blanca M Cortes, Ari Karppinen and Ingrid Myrtveit (2007). An integrated multi-model approach for air quality assessment: development and evaluation of the OSCAR air quality Assessment system. Environ. Mod. Software (in print). WHO (2006) Air quality guidelines for particulate matter, ozone, nitrogen dioxide and sulphur dioxide. Global update 2005. Summary of risk assessment.

7.1 Models of Exposure for Use in Epidemiological Studies of Air Pollution Health Impacts Michael Brauer, Bruce Ainslie, Michael Buzzelli, Sarah Henderson, Tim Larson, Julian Marshall, Elizabeth Nethery, Douw Steyn and Jason Su

Abstract Observational epidemiological studies have had an important role in understanding the public health impacts of air pollution. In such studies, accurate assessment of exposure remains a major challenges, especially in studies involving large populations. Here we review state-of-the-art approaches to assessment of population exposure in epidemiological studies with a focus on approaches applied in the Border Air Quality Study (www.cher.ubc.ca\baqs.htm). The strengths and limitations of these methods are discussed and future research needs identified.

Keywords Air quality, epidemiology, exposure assessment, health effects, land use regression, vehicle emissions, wood smoke

1. Introduction Observational epidemiological studies have assumed an important role in understanding the public health impacts of air pollution. Although laboratory toxicologycal studies and controlled human exposure experiments continue to provide new knowledge, especially regarding biological mechanisms, epidemiological studies are increasingly the main measure upon which policies, standards, guide-lines and regulations are based. Exposure assessment is a major challenge in epidemiological studies of air pollution health effects, especially those involving large study populations from which population health inferences can be made. While measurement of environmental concentrations of air pollutants is reasonably routine, especially for common air pollutants, assessing exposure – the intersection over time and space between humans and concentrations in air – incorporates additional layers of complexity. Importantly, in addition to spatial and temporal variability in ambient concentrations, exposure contains a behavioural component that is also complex and variable (Nieuwenhuijsen, 2003). Nieuwenhuijsen, in an adaptation from an earlier National Research Council Report, describes general approaches to exposure assessment for epidemiological studies which have been modified here (Figure 1) to describe air pollution exposure assessment:


589

590

M. Brauer et al.

Fig. 1 Approaches to assess air pollution exposure for epidemiological studies

Although continuous measurements of personal exposure for all study participants throughout the period of observation provides the best measure of exposure, and has been accomplished in several intensive studies of health impacts of shortterm exposure, this is seldom possible due to financial and logistical constraints and not realistic for studies of long term exposures. Further, the very act of wearing a personal air pollution monitoring device has been shown to modify individual activities and therefore likely to lead to biased measures of exposure. In contrast, in most urban areas, ambient monitoring networks provide continuous measurements of air quality for a small number of air pollutants at one or more locations, providing a ready-made and easy to exploit source of potential exposure information. In the following sections, we describe the main approaches to assess exposure for common air pollution epidemiological study designs. We emphasize, in particular, the approaches used in and developed for the Border Air Quality Study (www.cher.ubc.ca\baqs.htm).

2. Components of Exposure To simplify estimation of exposure, researchers often sub-divide exposure into different components. For example, exposure has both temporal and spatial components both of which may be sub-divided into increasing levels of resolution depending upon, for example, the health impact of interest or the spatial and temporal variability of a particular pollutant. Although outdoor air pollution is generally the area of interest from a policy perspective, an understanding of the time-activity patterns of individuals has led to an increased emphasis on exposures experienced in specific indoor microenvironments (home, work/school, in-transit) and the extent of infiltration of ambient pollutants into these environments. The microenvironment with greatest overall impact on exposure is usually the home environment, although in some cases exposures (to ambient pollutants) encountered while in transit or while outdoors may be important contributors.

Models of Exposure for Use in Epidemiological Studies of Air Pollution Health Impacts

591

3. Epidemiologic Study Designs Acute (short-term exposures) Time series studies have been conducted in hundreds of urban areas throughout the world. The general approach is to collect daily regulatory ambient air quality monitoring data for all available pollutants and to assess associations with counts of specific health outcomes, such as daily mortality, or hospital admissions. In this design the ambient monitoring data are typically averaged for the entire study area or, less frequently, counts are apportioned to data from individual monitors. Exposure on any given day is assumed to be the same for the entire study population. Although it is well-known that this assumption is poor, the relevant issue for the purposes of interpretation of the epidemiologic studies is the extent to which exposures are correlated to the overall average exposure. A number of studies have demonstrated that, for particulate matter, reasonably high correlations over time are found between ambient concentrations measured at central (urban background) locations and measured personal exposures (Ebelt et al., 2000; Janssen et al., 1999). These findings generally support the use of ambient monitoring network data in such time series studies of the short-term impacts of air pollution. For ozone, correlations are lower (Sarnat et al., 2000). In both cases the level of the correlations does appear to vary across individuals and is related to the ventilation properties of the indoor environment. Several studies have used additional information from personal monitoring to “correct” for the impact of non-ambient exposures (Ebelt et al., 2005; Koenig et al., 2005; Strand et al., 2006) and have shown that associations with health outcomes are more closely related to the component of exposure derived from ambient source pollution, compared to total measured exposure which is composed of particulate matter of indoor and outdoor origin or compared to particulate matter of indoor origin. While these studies show the importance of accurately accounting for indoor sources of exposure and the ventilation properties of indoor environments, new approaches need to be developed to account for these factors in studies of large populations where such data are not readily available. One promising area of current research is the development of models of infiltration (Allen et al., 2003). Using these models, information on building characteristics that may be readily available from property assessment data can be used to calculate building-specific factors to characterize the indoor infiltration of pollutants (Setton et al., 2005). Adjustment factors can then be applied to studies of long-term exposure if data on subject mobility (residential history, work/school locations) is known. Chronic (long-term) exposure While the health impacts of short term exposures have been repeatedly demonstrated in numerous epidemiological studies and in many cases supported by controlled human exposure studies, the population health impacts of long-term exposures are generally believed to have greater public health impacts (Kunzli et al., 2001). The study of health impacts associated with long term exposures relies mainly upon spatial comparisons between areas of differing air pollutant concentrations (Pope et al., 2002). Until recently most comparisons were

M. Brauer et al.

592

based upon comparisons between cities where health impacts and other risk factors (e.g. diet, smoking, age, occupation, etc.) are measured at the level of the individual while air pollution exposure has been based upon ambient monitoring network data (single monitors or an average of all available monitors in the study area). While this approach is reasonable in that it compares concentrations between cities (regional and local sources) as surrogates of exposure, it masks any within-city differrences in concentrations (local sources) that may exist. Moving from an assignment of subjects to the monitor nearest to their residence was reported to increase the measurement of risk associated with air pollution (Miller et al., 2007). Interpolation of available monitors to produce individual-specific estimates of exposure has also enabled studies of within-city contrasts in concentrations. In one example, estimates of risk based upon within-city contrasts were three times larger than those based upon between-city comparisons in concentrations (Jerrett et al., 2005b). Increasingly, measurement studies highlighting spatial differences in air pollutant concentrations (Gilbert et al., 2003; Zhang et al., 2004), combined with citizen concern regarding neighborhood sources of air pollution such as traffic, have led to an increase in studies designed to estimate the impact of spatial contrasts in concentrations on health effects. Given that most regional air pollution is wellcharacterized by monitoring networks, local air pollution concentration differrences require alternate approaches for characterization (Jerrett et al., 2005a). Identification of within-city contrasts in concentrations of air pollutants also has implications for the design of new measurement programs and for air quality management in general. Hoek et al. used a combination of measurement data to characterize regional and urban components of air pollution, and measures of proximity to major roads to characterize neighborhood differences related to traffic where the estimated individual exposure was a sum of regional, urban contribution and roadway contributions (Hoek et al., 2001): Exposure = Cregional + Curban + Croad

(1)

where Cregional is derived from interpolation of network monitoring, Curban is estimated from a regression relationship between air pollution and level of urbanization and Croad is assigned for those living within 100 m of a freeway or within 50 m of a major urban street. Other studies have used simple surrogates of exposure such as road proximity or traffic level on the nearest major road as surrogates of local-scale differences in pollutant concentrations (Venn et al., 2000).

4. The Border Air Quality Study In developing the exposure assessment strategy for the Border Air Quality Study, we built upon the study of Hoek et al. to enhance the assessment of traffic sources using a land use regression model (described below), adding characterization of additional sources of importance in the local airshed (woodsmoke and industrial point sources) and developing approaches to incorporate the impact of meteorology


593

and street canyons to more accurately characterize sources of variability in air pollution concentrations within the airshed and their impact on health outcomes: Exposure = Cregional + Curban + Ctraffic + Cwoodsmoke + Cpoint sources

(2)

Land Use Regression models Recent studies have measured and reported considerable spatial variability in the concentrations of traffic-related pollutants within urban areas. These “neighborhood scale” intra-urban differences tend not to be well-characterized by air quality monitoring networks. Land use regression (LUR) was first developed to address this shortcoming and has recently gained attention in the air quality management and urban planning communities. There is no standard method for conducting LUR (Figure 2), but detailed descriptions of can be found elsewhere (Jerrett et al., 2005a; Briggs et al., 1997; Henderson et al., 2007). In brief, a pollutant is measured at multiple sites specifically selected to capture the complete intra-urban range of its concentrations. Geographic attributes that might be associated with those concentrations (e.g. surrounding land use, population density, and traffic patterns) are measured around each measurement site in a Geographic Information System (GIS). Linear regression is used to correlate measured concentrations with the most predictive variables, and the resulting equation is used to estimate pollutant concentrations anywhere that all of the predictors can be measured. Concentration maps with high spatial resolution can be generated by rendering the regression model in GIS. Fig. 2 The LUR modelling procedure

Traffic sources Land use regression was initially developed in Europe to estimate individual-level exposure to traffic-related air pollutants for epidemiological studies. This need arose from (1) the infeasibility of collecting individual measurements for large populations and (2) inaccuracies inherent to crude surrogates such as distance to nearest road, or data from the nearest regulatory monitoring locations. With LUR, researchers were able to estimate individual exposures from statistical models that combined the predictive power of several surrogates based on their relationship with measured concentrations. Although interest in traffic-related health effects has favoured the development of LUR for traffic-related pollutants, the method is now being explored for other sources.

M. Brauer et al.

594

Published results are summarized in Table 1. Most of these studies were undertaken to provide exposure assessment for epidemiological research. Comparison of R2 values across study areas and pollutant types in Table 1 suggests that LUR produces consistent results regardless of location. Table 1 Summary of previous LUR studies. Investigator, Year

Henderson, 2007

Study location

Vancouver

Domain size (km2)

Mean NO2 (SD) (ppb)

R2 for NO2 (N sites)

Mean ABS (SD) (10-5 m-1)

16.2 (5.6)

0.56–0.60 (114)

0.84 (0.47)

0.39–0.41 (25)

–

–

1.28 (0.83)

0.56–0.65 (39)

2200

Larson, 2007

R2 for ABS (N sites)

Sahsuvaroglu, 2006

Hamilton

1400

16.4 (3.7)

0.76 (101)

–

–

Jerrett, 2006

Toronto

900

32.7 (10.5)

0.69 (95)

–

–

Gilbert, 2004

Montreal

1200

11.6 (3.0)

0.54 (67)

–

– 0.75 (24)

Ryan, 2007

Cincinnati

1600

–

–

0.67 (0.29)*

Ross, 2005

San Diego

2100

14.8 (5.7)

0.77 (39)

–

–

El Paso

800

20.6 (7.1)

0.81 (20)

–

–

Western Germany

3300

13.7

0.89 (40)

1.71

0.81 (40)

Netherlands Rotterdam Stockholm Munich

38000 200 150 80

15.4 (4.9) 17.5 (3.9) 10.1 (4.0) 15.2 (4.1)

0.85 (40) 0.79 (18) 0.73 (42) 0.62 (40)

1.64 (0.58) 1.79 (0.56) 1.29 (0.35) 1.84 (0.43)

0.81 (40) 0.77 (18) 0.66 (42) 0.67 (40)

Amsterdam

30

Huddersfield

300

0.63 (80) 0.61 (80) 0.72 (80)

–

–

Prague

50

Gonzales, 2005 Hochadel, 2006 Hoek, 2002 Brauer, 2003 (TRAPCA)

Briggs, 1997 (SAVIAH)

20.1–28.6 (3.4–6.7) 14.1–26.3 (5.2–7.8) 12.3–21.9 (5.7–9.9)

LUR Versus Dispersion Modeling One alternative to LUR is dispersion modeling, where emissions parameters are input into models that use physical and chemical equations to predict pollutant concentrations at individual receptors. While this is a common approach in risk assessment and air quality management evaluation, it is rarely used for epidemiological studies because dispersion models require specific inputs (traffic volume, motor vehicle fleet makeup, street configurations, industrial emissions, local meteorology, etc.) that may not be available for all areas. Even where complete input data exist, dispersion model operation requires considerable time, resources and expertise. In comparison, LUR allows flexibility in terms of inputs, resource requirements, and outputs. Land use regression models can be built on a location-by-location basis with whatever data are available. Sampling can be conducted at a flexible number of sites over a flexible period of time using a wide range of instrumentation.


595

Once data collection is complete the analyses can easily be conducted by individuals with a background in statistics and GIS. Final models can be rendered into high-resolution pollution maps. Because LUR is a stochastic approach that uses actual measurements, model estimates tend to be realistic. Dispersion models use estimated emission factors that can result in considerable disparity between model output and actual concentrations. On the other hand, dispersion models can easily be used to evaluate different emissions scenarios. Within the TRAPCA project (Brauer et al., 2003; Cyrys et al., 2007), results for LUR and dispersion models of NO2 concentrations were compared in Stockholm and Munich. In Stockholm the R2 for estimates made with the AIRVIRO1 model and measured concentrations of NO2 was 0.69, with greater correlations observed for sites located in street canyons. The LUR model had an R2 value of 0.76. The TRAPCA study concluded that AIRVIRO and LUR had similar predictive power, but the applicability of LUR in the absence of emission inventories was an attractive advantage. This finding was supported in a recent study by Cyrys et al. that compared dispersion (IMMIS net2) and LUR estimates of NO2 and PM2.5 concentrations for their study population in Munich, Germany and concluded that both methods performed equally well in estimating exposures of their study population. More recently, Briggs et al. (2006) compared LUR with a state-of-the-art dispersion model (ADMS-Urban) for NO2 and PM10 at a limited number of measurement sites (N = 18 for PM10, N = 8 for NO2) in London, England. The LUR estimates had correlations (Pearson’s coefficient, r) of 0.61 for NO2 and 0.88 for PM10 compared to the annual mean. The ADMS estimates had correlations of 0.72 and 0.81 for NO2 and PM10, respectively. These results suggest that LUR pollutant concentration estimates are of equal or better accuracy than those from dispersion models. Beyond its aforementioned flexibility, another important advantage of LUR is its applicability to specific components of particulate matter, such as elemental carbon or source-specific tracers. In contrast, sophisticated dispersion models are only available for a limited set of pollutants. In BAQS we built upon previous traffic-related pollutant LUR models in developing approaches to be used in our epidemiological analyses (Henderson et al., 2007). Briefly, passive samplers to collect NO and NO2 were deployed for two-14 day periods at 116 sites in the study area. Mean concentrations during these two periods were highly correlated with, and closely approximated annual averages from regulatory monitoring network data. Sites were selected by a locationallocation algorithm that used a crude LUR model based upon regulatory monitoring stations as a demand surface. The algorithm optimally locates samplers to maximize their ability to characterize variability in the demand surface. More samplers are located in areas predicted to have higher spatial variability. The model output was then weighted by population density to ensure adequate numbers of samplers in residential areas. In addition, PM2.5 mass was measured at a subset of 25 locations during a two-month sampling period. Integrated one-week average PM2.5 samples were collected on Teflon filters using Harvard Impactors. For a 1 http://www.indic-airviro.smhi.se/ 2 http://www.ivu-umwelt.de/e/index.html

596

M. Brauer et al.

subset of 36 sites, we measured particle absorbance (Black Carbon) using a Particle Soot Absorption Photometer in a mobile monitoring platform and adjusted these measurements for temporal variation based upon repeated measurements at a centrally-located site. For each of the 116 (and the subsets of 25 and 36) measurement sites, 55 variables were generated in a Geographic Information System (GIS) and linear regression models of NO, NO2 and Black Carbon were built with the most predictive covariates. Models were developed using both road classifications and traffic density as potential predictors. Using road classifications (to allow the model to be applied to areas without traffic density measures), for NO, the model had an R2 of 0.62 and included the number of major roads within 100 and 1,000 m radius circular buffers of the measurement sites, the number of secondary roads within a 100 m buffer, population density within a 2,500 m radius and elevation. For NO2, the model (R2 = 0.56) included the same variables as well as the amount of commercial land use within 750 m. For PM.5 the model (R2 = 0.52) included the amount of commercial and industrial land use within 300 m, the amount of residential land use within 750 m and elevation. For Black Carbon the model (R2 = 0.56) included the number of secondary roads within a 100 m buffer, distance to the nearest highway and the amount of industrial land use within 750 m. Models developed using traffic density did not show significantly higher correlations but had marginally improved agreement in evaluation analyses. The model surfaces clearly showed differences in spatial extent of primary vs secondary pollutants. For applications to epidemiological analyses, we used the models to generate smooth spatial surfaces of predicted (annual average) concentrations for the entire study area at a resolution of 10 m. The surfaces were then smoothed to remove abrupt changes and edge effects so as to more accurately reflect the measured effect of proximity to roadways (Gilbert et al., 2003). For each LUR model, the corresponding monitoring network data for each pollutant were fit with a monthly dummy variables and a covariate for linear trend (Times Series Forecasting System, SAS v 9.1). For Black Carbon, the PM2.5 trend was used as there were no corresponding regulatory monitoring network data. From these models, month-year adjustment factors were applied to each surface to estimate monthly average concentrations. Using these we then computed individual subject exposures for the same exposure windows as described above for the monitor-based approaches. Wood Smoke Residential wood smoke can be an important local source of ambient particulate matter during winter months but its distribution is often not well-characterized by regulatory monitoring networks due, in part, to the sparselylocated sources in residential areas. We developed a novel land use regression model for wood smoke based upon a mobile monitoring campaign to map the impact of wood smoke particulate matter across the BAQS study area (Larson et al., 2007). Briefly, we first identified potential hotspots for wood-burning based on property assessment data, a telephone survey on woodburning practices conducted for a local emissions inventory and topography (Su et al., 2007). A network of fixed monitors was located at potential hotspots and control sites to collect twoweek samples of PM2.5 and levoglucosan, a biomass combustion tracer compound.


597

The monitoring of levoglucosan was conducted to confirm that PM2.5 concentrations assessed by mobile monitoring and used to model spatial patterns were associated with wood smoke. On 19 cold, clear nights (9pm–1am) during the heating season we conducted mobile sampling in a vehicle equipped with a logging GPS and light-scattering nephelometer. Two routes were pre-selected to cover target areas of predicted variability in woodsmoke concentrations, traverse populated areas, and circumnavigate the fixed-location monitoring sites. These campaigns generated more than 12,000 pairs of geospatial coordinates and light-scattering coefficients (bsp) that were temporally-adjusted and merged into a single, high-resolution file for LUR analysis. To generate data for linear regression the model domain was divided into ~50 air catchments (based upon hydrological catchment basins), assuming that a given location is systematically downwind of uphill sources under stable meteorological conditions (e.g. cold, clear nights). The bsp values and predictive variables were averaged at the catchment level, and all uphill catchments within an 8km radius were assumed to contribute to the mean bsp of the downhill catchment. Variables describing the population, ethnic composition, economic status, buildings, and wood-burning appliance usage in each catchment produced an R2 value of 0.64. A similar mobile monitoring campaign was also conducted in a second geographically distinct region and comparable model results were obtained. For epidemiological analyses, the catchment area values of woodsmoke PM concentrations were assigned to tertiles with the highest tertile areas considered to be woodsmoke-exposed. Since woodsmoke is emitted seasonally, the surface was “activated” for epidemiological analyses only during the heating season. We calculated Heating Degree Days (HDD) (using 18ºC as the base value) for each of twelve two-week monitoring periods during which we collected measurements of levoglucosan and regressed the mean levoglucosan concentrations (from all six sites) against the mean HDD for the corresponding period (R2 = 0.60). From this regression, we estimated that a HDD of 12.0 was equivalent to approximately 115 ng/m3 of Levoglucosan which is above a concentration of ~100 ng/m3 at which woodsmoke-impacted areas and woodsmoke events can be differentiated from other areas and periods. Accordingly, days with HDD>12.0 were considered “woodsmoke days” and the spatial surface was applied on these days only. Point Sources The approach taken for point-source (i.e., industrial) emissions was a simple alternative to more complex dispersion/photochemical models. The metric is a proxy for residential exposure to industrial point-source emissions but no actual concentrations are estimated. We built upon the work of Yu et al. (2006) in which fixed size and shape (3 km radius and 90q angle) wedges that partially account for wind direction were used to estimate exposures from petrochemical facilities. We simplified the approach to use uniform circles, thus ignoring wind direction, to account for the hundreds of industrial point sources in our study region. The point-source exposure metric is a proximity-weighted summation of relative emissions within a given radius of each postal code (PC) centroid. The metric incurporates point source emissions and locations, postal code locations and a cut-off threshold distance. Point-source emissions and locations, used as inputs into the

M. Brauer et al.

598

CMAQ model, were obtained for CO, NOX, VOC, NH3, SO2, PM10, PM2.5, and coarse mode PM (i.e., PM10-PM2.5). If the distance between a PC and an emission source is greater than a certain value, that emission source is ignored when evaluating that PC. Given the nature of air dispersion, the variability among sources in stack height and plume rise, and the influences of complex topology and meteorology, it is not possible to identify a single value as the “correct” cut-off threshold distance. Two cut-off threshold distance values, 10 and 40 km, were separately employed to represent the approximate distance needed for a point-source plume to fully mix throughout the atmospheric mixing height. Both values (10 and 40 km) were derived from PasquillGifford curves and offer an order-of-magnitude estimate of the “impact zone” for a point source. The true size of an impact zone will vary widely over time and among sources, based on parameters such as emissions, stack height, exit velocity, meteorology, and topography. The point-source exposure metric is calculated using the following formula: Wi

¦ ^ `

j ; d ij ! x

Ej

dij

(3)

where Wi is the point-source exposure metric for postal code i, Ej is the relative emissions for point source j, and dij is the distance between postal code i and point source j. The summation is made for all point sources within distance x of the PC. To calculate relative emissions (Ej) for a point source j, emissions for each point source are first converted from a raw emission rate (tons per day) into the percentile of that source among all emitting point sources. This step is repeated for each of four pollutants (PM2.5, SOx, NOx, and VOCs). For example, a point source that does not emit SOx is assigned a percentile of zero; a source whose SOx emissions are at the 85th percentile (i.e., 85% of the SOx-emitting point sources have an emission rate that is less than this source) is assigned a SOx value of 0.85. Next, percentile scores for the four pollutants are summed to yield the relative emissions for a specific point source. The largest relative emission rate is 3.96, for a specific point source in the 99th emission percentile for all four pollutants. The lowest relative emission rate is zero, representing sources with no emissions of the four pollutants. In the Vancouver study area, each postal code has an average of 173 point sources within 10 km and 753 point sources within 40 km. Mean (st dev) values for the point-source exposure metric are 21.6 (21.8) for x = 10 km, and 41.5 (27.6) for x = 40 km. Geometric means (GSD) are 12.7 (3.5) for x = 10 km, and 30.3 (2.5) for x = 40 km. The correlation between the two exposure metrics is very high (r = 0.91). For epidemiological analyses the raw values of the point source metric were assigned to each postal code of interest as a continuous variable. Meteorology The main strength of land use regression (LUR) models is the empirical structure of the regression mapping and its relatively simple inputs/low cost (Jerrett et al., 2005a). However, the method is case- and area-specific (Briggs et al., 1997), has to date not been temporally resolved, and is aimed at long-term


599

average concentrations. Also there is little theoretical-physical basis behind its application, particularly the use of circular buffers to extract local covariates for exposure estimates. By contrast dispersion models have the potential advantage of incorporating both spatial and temporal variation of air pollution within a study area. However, dispersion models often unrealistically assume a Gaussian distribution of dispersion patterns, require extensive cross-validation with monitoring data and have relatively costly data input (Jerrett et al., 2005a). Some have addressed both of these issues, such as the use of monthly exposure fixed-sized “wedges” to identify potentially exposed areas using the monthly prevailing wind directions (Yu et al., 2006). Using wind fields, Arain et al. (2007) marginally increased NO2 concentration precision for locations downwind of major highways, while we used hydrological catchment area and uphill search algorithms rather than circular buffers for modeling residential woodsmoke PM2.5 concentrations (Larson et al., 2007). These models represent a step forward in terms of conceptualizing the physical pathways of exposure by contrast with LUR, for which a more general theoretical basis has still not been developed. To enhance the tools available for assessment of exposure to local sources, a source area analysis (SA) model (Ainslie et al., 2008) incorporating spatially and temporally refined components of ambient exposures was developed and applied to LUR covariates. This model accounts for wind speed and direction in modifying the zone of influence over receptor sites. By incorporating varying meteorological parameters while maintaining the same land use covariates, we also introduce a temporal component that allows estimation of ambient concentrations/exposures on shorter time scales. Briefly, the model uses a 3D wedge to estimate pollutant concentration. The radius (i.e., length of a wedge) is given by the distance travelled by wind in one hour and wind speed, direction, cloud cover/insolation and time of day are also used to estimate the atmospheric stability classes, wedge angle and wedge depth. Circular buffers are used for stability class A, B, E and F, and a 90o wedge for classes C and D. Wedge depth equals atmospheric mixing height and takes on a constant value at night and in the early morning for convective (A and B) classes. During the day the wedge depth of the convective classes is assumed to linearly increase reaching a peak value late in the afternoon before rapidly dropping in the early evening. For neutral (C and D) and stable (E and F) classes, the wedge depth is assumed, respectively, proportional to and square root of the wind speed. An evaluation test of the source-area and simple buffer model was performed using measurements taken at the 116 field sites used to build the LUR model (Henderson et al., 2007) and is described in detail elsewhere (Su et al., 2008). When the source area models are assigned the same radii of influence as in the LUR models, the two approaches yield similar amounts of explained variance. To investigate whether the performance of the source area model was limited by the need to interpolate windspeed and direction data to the 116 sampling sites and the aggregation of hourly data to seasonal averages, we conducted a similar comparison for the 19 regulatory monitoring sites with available meteorological data. When both hourly measurements and estimations are aggregated to seasonal averages the unmeasured hourly variability is eliminated. Accordingly the prediction powers of the source area model increases significantly and performs much better

600

M. Brauer et al.

than the LUR model using circular buffers, a result similar to the initial comparison described above. The new models have (compared to LUR) an increase of variance explained from 60% to 81% for NO prediction and from 75% to 87% for NO2. Using hourly meteorological data from a single location, reduced the source area model’s predictive power, but still showed improvement over the circular buffer approach. The better prediction powers of the source area model over LUR circular buffer model for the 19 monitoring stations demonstrates the potential strength of the new approach when we scale up temporal variations that are lost with a more typical cross-section approach in LUR. Since the LUR model’s covariates do not rely on a Gaussian assumption and do not need background concentrations, less computing power than for a dispersion model is required.

5. Conclusions Exposure assessment approaches for epidemiological studies of air pollution health impacts have progressed dramatically in recent years from traditional reliance on available regulatory ambient monitoring network data. In particular, improved approaches to characterize high resolution spatial concentration differences have been developed and applied in a number of studies. In most situations these models have focused on characterizing the impacts of traffic sources, but applications to point sources and residential wood combustion have also been developed. A limitation of existing spatial models is their emphasis on long-term average concentrations and their imperfect characterization of temporal variability in air pollution concentrations. New approaches to incorporate time-varying meteorological data into high resolution spatial concentration models are being developed but have seen only limited applications to date.

References Ainslie B, Steyn DG, Su J, Buzzelli M, Brauer M, Larson TV & Rucker M (2008) “A Source Area Model Incorporating Simplified Atmospheric Dispersion and Advection at Fine Scale for Population Air Pollutant Exposure Assessment”, Atmos Environ. 42, 2394–2404. Allen R, Larson T, Sheppard L, Wallace L, Liu LJ (2003) “Use of real-time light scattering data to estimate the contribution of infiltrated and indoor-generated particles to indoor air”, Environ Sci Technol, 37, 16, 484–3492. Arain MA, Blair R, Finkelstein N, Brook JR, Sahsuvaroglu T, Beckerman B, Zhang, L, Jerrett M (2007) “The use of wind fields in a land use regression model to predict air pollution concentrations for health exposure studies”, Atmos Environ, 41, 16, 3453–3464. Brauer M, Hoek G, van Vliet P, Meliefste K, Fischer P, Gehring U, Heinrich J, Cyrys J, Bellander T, Lewne M, Brunekreef B (2003) “Estimating long-term


601

average particulate air pollution concentrations: application of traffic indicators and geographic information systems”, Epidemiology, 14, 2, 228–239. Cyrys J, Hochadel M, Gehring U, Hoek G, Diegmann V, Brunekreef B, Heinrich J. GIS-based estimation of exposure to particulate matter and NO 2 in an urban area: stochastic versus dispersion modeling. Environ. Health Perspect. 2005 Aug; 113(8):987–92. Briggs DJ, Collins S, Elliott P, Fischer P, Kingham S, Lebret E, Pryl K, VAnReeuwijk H, Smallbone K, VanderVeen A (1997) “Mapping urban air pollution using GIS: a regression-based approach”, Int J Geogr Inform Sci, 11, 7, 699–718. Briggs D, de Hoogh K, Gulliver J (2006) “Matching the metric to need: modelling exposures to traffic-related air pollution for policy support.”, NERAM V (http://www.irr-neram.ca/about/Colloquium.html (accessed July 24, 2007). Ebelt ST, Fisher TV, Petkau AJ, Vedal S, Brauer M (2000) “Exposure of chronic obstructive pulmonary disease (COPD) patients to particles: relationship between personal exposure and ambient air concentrations”, J Air Waste Manage Assoc, 50, 174–187. Ebelt ST, Wilson WE, Brauer M (2005) “Exposure to ambient and nonambient components of particulate matter: a comparison of health effects”, Epidemiology, 16, 3, 396–405. Gilbert NL, Goldberg MS, Beckerman B, Brook JR, Jerrett M (2005) “Assessing spatial variability of ambient nitrogen dioxide in Montreal, Canada, with a landuse regression model”, J Air Waste Manage Assoc, 55, 8, 1059–1063. Gilbert NL, Woodhouse S, Stieb DM, Brook JR (2003) “Ambient nitrogen dioxide and distance from a major highway”, Sci Total Environ, 312, 1–3, 43–46. Gonzales M, Qualls C, Hudgens E, Neas L (2005) “Characterization of a spatial gradient of nitrogen dioxide across a United States-Mexico border city during winter”, Sci Total Environ, 337, 1–3, 163–173. Henderson SB, Beckerman B, Jerrett M, Brauer M (2007) “Application of land use regression to estimate long-term concentrations of traffic-related nitrogen oxides and fine particulate matter”, Environ Sci Technol, 41, 7, 2422–2428. Hochadel M, Heinrich J, Gehring U, Morgenstern V, Kuhlbusch T, Link E, Wichmann HE, Kramer U (2006) “Predicting long-term average concentrations of traffic-related air pollutants using GIS-based information”, Atmos Environ, 40, 3, 542–553. Hoek G, Fischer P, Van Den Brandt P, Goldbohm S, Brunekreef B (2001) “Estimation of long-term average exposure to outdoor air pollution for a cohort study on mortality”, J Expo Anal Environ Epidemiol, 11, 6, 459–469. Janssen NA, Hoek G, Harssema H, Brunekreef B (1999) “Personal exposure to fine particles in children correlates closely with ambient fine particles”, Arch Environ Health, 54, 299192099, 95–101. Jerrett M, Arain A, Kanaroglou P, Beckerman B, Potoglou D, Sahsuvaroglu T, Morrison J, Giovis C (2005a) “A review and evaluation of intraurban air pollution exposure models”, J Expo Anal Environ Epidemiol, 15, 2, 185–204.

602

M. Brauer et al.

Jerrett M, Burnett RT, Ma R, Pope CA, Krewski D, Newbold KB, Thurston G, Shi Y, Finkelstein N, Calle EE, Thun MJ (2005b) “Spatial analysis of air pollution and mortality in Los Angeles”, Epidemiology, 16, 6, 727–736. Koenig JQ, Mar TF, Allen RW, Jansen K, Lumley T, Sullivan JH, Trenga CA, Larson T, Liu LJ (2005) “Pulmonary effects of indoor- and outdoor-generated particles in children with asthma”, Environ Health Perspect, 113, , 499–503. Kunzli N, Medina S, Kaiser R, Quenel P, Horak F Jr, Studnicka M (2001) “Assessment of deaths attributable to air pollution: should we use risk estimates based on time series or on cohort studies?”, Am J Epidemiol, 153, 11, 1050– 1055. Larson T, Su J, Baribeau AM, Buzzelli M, Setton E, Brauer M (2007) “A spatial model of urban winter woodsmoke concentrations”, Environ Sci Technol, 41, 7, 2429–2436. Miller KA, Siscovick DS, Sheppard L, Shepherd K, Sullivan JH, Anderson GL, Kaufman JD (2007) “Long-term exposure to air pollution and incidence of cardiovascular events in women”, New Engl J Med, 356, 5, 447–458. Nieuwenhuijsen MJ (ed) (2003) Exposure Assessment in Occupational and Environmental Epidemiology, First edn, Oxford University Press, Oxford, England. Pope CA 3rd, Burnett RT, Thun MJ, Calle EE, Krewski D, Ito, K, Thurston GD (2002) “Lung cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air pollution”, JAMA, 287, 9, 1132–1141. Ross Z, English PB, Scalf R, Gunier R, Smorodinsky S, Wall S, Jerrett M (2006) “Nitrogen dioxide prediction in Southern California using land use regression modeling: potential for environmental health analyses”, J Expo Sci Environ Epidemiol, 16, 2, 106–114. Sahsuvaroglu T, Arain A, Kanaroglou P, Finkelstein N, Newbold B, Jerrett M, Beckerman B, Brook J, Finkelstein M, Gilbert NL (2006) “A land use regression model for predicting ambient concentrations of nitrogen dioxide in Hamilton, Ontario, Canada”, J Air Waste Manage Assoc, 56, 8, 1059–1069. Sarnat JA, Koutrakis P, Suh HH (2000) “Assessing the relationship between personal particulate and gaseous exposures of senior citizens living in Baltimore, MD”, J Air Waste Manage Assoc, 50, 720395124, 1184–198. Setton EM, Hystad PW, Keller CP (2005) “Opportunities for using spatial property assessment data in air pollution exposure assessments”, Int J Health Geogr, 4, 26. Strand M, Vedal S, Rodes C, Dutton SJ, Gelfand EW, Rabinovitch N (2006) “Estimating effects of ambient PM(2.5) exposure on health using PM(2.5) component measurements and regression calibration”, J Expo Sci Environ Epidemiol, 16, 1, 30–38. Su JG, Larson T, Baribeau A, Brauer M, Rensing M, Buzzelli M (2007) “Spatial modeling for air pollution monitoring network design: Example of residential woodsmoke.”, J Air Waste Manag Assoc, 57, 8, 893–900. Su JG, Brauer M, Ainslie B, Steyn D, Larson TV, Buzzelli M (2008). Comparing source area analysis with land use regression models for exposure analysis. Sci Total Environ. 390, 520–529.


603

Venn A, Lewis S, Cooper M, Hubbard R, Hill I, Boddy R, Bell M, Britton J (2000) “Local road traffic activity and the prevalence, severity, and persistence of wheeze in school children: combined cross sectional and longitudinal study”, Occup Environ Med, 57, 320277145, 152–118. Yu CL, Wang SF, Pan PC, Wu MT, Ho CK, Smith TJ, Li Y, Pothier L, Christiani DC, Kaohsiung Leukemia Research Group (2006) “Residential exposure to petrochemicals and the risk of leukemia: using geographic information system tools to estimate individual-level residential exposure”, Am J Epidemiol, 164, 3, 200–207. Zhang KM, Wexler AS, Zhu YF, Hinds WC, Sioutas C (2004) “Evolution of particle number distribution near roadways. Part II: the ‘road-to-ambient’ process”, Atmos Environ, 38, 38, 6655–6665.

Discussion S.T. Rao: Given the spatial homogeneity of sulphate levels, why do you see differences in health outcomes between within city and between cities? Since sulphate level has been decreasing due to SO2 emission reduction, do you see a trend the health outcomes? Regarding the BAQ study, do you see a commonality between pollutants in Seattle vs Vancouver and associated health outcomes? What seems to be primary driver for the health impacts? M. Brauer: There is a misunderstanding in our use of sulphate to assess air pollution exposures. Because sulphate has no known major indoor sources, we use it as an indicator of the amount of ambieny particulate matter that infiltrates indoors from outside. This allows us to more accurately assess the amount of ambient particles that people are exposed to, and when we apply this in studies of health effects we get stronger relationships compared to when we estimate exposure by just using the measured ambient concentrations. Regarding the question about the BAQ study, in general we do see similarities between our studies in Vancouver and Seattle and one common aspect is the importance of primary, combustion-source pollutants – especially those related to traffic and traffic proximity. It is however, difficult to quantitatively compare the findings since we are relaying on administrative data and there are differences in they type of data that are available. For example, in Canada we have health statistics for essentially the entire population for all encounters with the health care system whereas in Seattle much of our data are limited to those based on hospital admissions and vital statistics (births, deaths).

604

M. Brauer et al.

J. Pleim: Have land use regression models been combined with high resolution meteorology models? Would it help to include such meteorology data in the regression? M. Brauer: There have been some attempts to incorporate output from meteorological models as additional predictor variable into land use regression models – in general these have shown relatively small improvements in the model predictions, in part due to the limited spatial resolution of the meteorological models relative to the other land use regression input data and the fact that most of the land use regression models have been developed for long term (annual) estimates. Certainly in developing land use regression models for shorter time periods, incorporating meteorological information (from measurements or high resolution models) will be very helpful. P. Suppan: The presentation shows the very important link between epidemiology and air quality. Some slides show the NO distribution provided by the land use model and CMAQ. But as the results from CMAQ are far too coarse for epidemiological studies the question arises if also results from micro scale models can be introduced, or can be provided in order to improve the results. M. Brauer: Land use regression models are one approach to incorporate spatial variability in air pollution models into epidemiology. Other types of high resolution spatial models can certainly be just as useful if estimates can be developed over large study areas and can even improve over land use regression models if they are able to incorporate temporal variability in pollutant concentrations and spatial dispersion due to meteorology and/or variable emissions.

7.6 The Importance of Exposure in Addressing Current and Emerging Air Quality Issues Tim Watkins, Ron Williams, Alan Vette, Janet Burke, B.J. George and Vlad Isakov

Abstract The air quality issues that we face today and will face in the future are becoming increasingly more complex and require an improved understanding of human exposure to be effectively addressed. The objectives of this paper are (1) to discuss how concepts of human exposure and exposure science and should be applied to improve air quality management practices, and (2) to show how air quality modeling tools can be used to improve exposures estimates used for understanding associations between air quality and human health. Data from a large human exposure monitoring study is presented to demonstrate the value of exposure in understanding important air quality issues, such as health effects associated with exposure to components of particulate matter (PM), to PM of different size fractions (coarse and ultrafine), and to air pollution in near roadway environments. Various approaches for improving estimates of exposure via application of air quality modeling are discussed and results from example modeling applications are presented. These air quality modeling approaches include: the integration of regional scale eulerian air quality models with local scale gaussian dispersion models; the fusion of modeled estimates with air quality observations; the integration of air quality and human exposure modeling tools; and the use of exposure factors, such as housing ventilation, to adjust modeled estimates of ambient air quality.

Keywords Air quality, exposure, modeling, particulate mater, toxic air pollutants 1. Introduction Existing air quality standards and regulations are in place to protect public health and the environment. However, there remain uncertainties regarding whether existing standards/regulations should be adjusted to be more protective or perhaps to more effectively target air quality management activities. For particulate matter (PM) standards, there are questions regarding whether current mass based standards for PM10 and PM2.5 should be revised to address specific components or sources of PM or different PM size fractions (NRC, 2004a; U.S. EPA, 2005). There are also concerns about potential disproportionate health effects associated with air pollution “hotspots.” These “hotspot” concerns are often related to environmental C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

640

The Importance of Exposure in Addressing Current and Emerging Air Quality Issues

641

justice issues (NRC, 2004b). For these air quality management issues, understanding human exposure is critical. This paper discusses why this is the case, presents data from human exposure monitoring and modeling studies to further highlight the importance of air pollution exposure issues, and discusses how air quality models may be used to improve exposure assessment. While air pollution impacts both humans and ecological resources, this paper focuses on human health outcomes.

2. Exposure and Exposure Science Although air quality standards to protect public health are most often based on levels of a pollutant in the ambient air, people experience health impacts from the pollutants in the air they breathe, i.e., from their exposure. The United States Environmental Protection Agency defines exposure as contact (of an environmental pollutant) with the exterior of the person (U.S. EPA, 1992). The critical factors to characterize and understand human exposure to air pollution include the following. x

Spatial and temporal variability of ambient pollutant concentrations – When ambient air concentrations are relatively homogeneous in space and time, then human exposures may be more closely approximated by ambient concentrations. However, when there is significant spatial and temporal variability in ambient air concentrations, ambient levels of pollutants will more poorly represent human exposure.

x

Concentrations of ambient pollutants in microenvironments – Micro-environments are locations where people spend their time (e.g., indoors at home, indoors at work/school, in-vehicle, outdoors at home). A person’s exposure to ambient air pollution will depend greatly on concentrations of ambient pollutant in microenvironments, which in turn depend upon exposure factors such as proximity to sources, air exchange rates, penetration rates, indoor air chemistry, and indoor decay rates and removal mechanisms.

x

Human Activities – A person’s daily activities play a significant role, if not the most significant role, in characterizing human exposure. Where a person spends their time and how much time he/she spends in each location will impact that person’s exposure.

Human exposure is the critical link between ambient air concentrations and human health outcomes. The field of exposure science includes research to measure and model factors and human activities that influence magnitude, frequency, and duration of exposure to air pollutant concentrations in various microenvironments. Understanding human exposures requires an understanding of the factors that influence the spatial and temporal variability of ambient air concentrations, which in turn requires an understanding of air pollution sources, fate and transport of air pollutants, and ambient air concentrations. Therefore, the field of exposure science

642

T. Watkins et al.

also includes aspects of research related to source emission characterization, atmospheric processes, and ambient air measurement and modeling.

3. The Growing Importance of Exposure Understanding actual human exposures is critical to addressing the current and emerging air quality management issues mentioned in the introduction of this paper. Human exposure is the link between ambient concentrations and health outcomes. Many existing air quality management policies are based upon studies that associated ambient concentrations with health impacts by inferring that ambient concentrations are equivalent to actual human exposures. However, for current and emerging air quality management issues this inference may not be appropriate and understanding exposure may in fact be the critical factor to developing and implementing effective air quality management policies for these issues. The following discussion provides examples of why this is the case.

3.1. Particulate matter components, size fractions, and sources The uncertainties surrounding existing PM standards are largely based on whether specific PM characteristics, such as composition or size fraction, lead to a greater proportion of observed health impacts (NRC, 2004a; U.S. EPA, 2005). Existing standards for PM2.5 and PM10 are largely based upon epidemiological studies that found associations between ambient concentrations of PM and observed health impacts. These studies often used a central site monitor to estimate exposures to PM, which is reasonable for PM2.5 because the variability of PM2.5 across many urban areas is relatively homogeneous. However, the spatial variability of specific PM components or PM of different size fractions (e.g., ultrafine PM or coarse PM) is greater than that for PM2.5 (U.S. EPA, 2004, 2005). In addition, there are significant uncertainties regarding the microenvironmental concentrations of PM components and PM size fractions (U.S. EPA, 2004). Therefore, any epidemiological evidence or risk assessment for PM components or PM size fractions will require an improved exposure assessment due to the spatial heterogeneity of ambient concentrations. Related to the issues of PM components and size fractions is the issue of whether PM standards should be targeted at sources of PM that may be disproportionately responsible for observed health impacts. Characteristics of PM emissions vary by source, thus there is the potential that the relative toxicity of PM from different sources may also vary. Exposure will play an important role in addressing this issue as well. One approach to address this issue is to evaluate the inter-city variability of PM characteristics. The composition and characteristics of PM in different urban areas varies because the contributing sources of PM vary (McMurry et al., 2004), therefore studies that evaluate the differential exposures and health


643

impacts across multiple urban areas may provide insights into the issue regarding standards for particular PM sources.

3.2. Air pollution hotspots and environmental justice issues Recent concerns have emerged regarding whether existing regulations provide ample protection for certain subpopulations that may be vulnerable due to elevated exposures in “hotspots.” In many cases, these issues are centered around environmental justice and toxic air pollutants (NRC, 2004b). An example of such an issue is near roadway exposures and health effects. Exposure assessment will be central to addressing these issues because improved characterization of the variability of ambient concentrations, microenvironmental concentrations, and human activities will all be required to evaluate environmental policies to address potential hotspots and environmental justice issues.

4. The Detroit Exposure and Aerosol Research Study The US Environmental Protection Agency has conducted an exposure study that is generating data to confirm the importance of exposure in addressing the air quality management issues above and to provide insight for future air quality policy decisions. The Detroit Exposure and Aerosol Research Study (DEARS) was designed to describe the relationships between concentrations at a central site and residential/ personal concentrations for PM components, PM size fractions, PM from specific sources (mobile and point), and air toxics. To accomplish this, the DEARS included extensive field work conducted over three years and six seasons. The DEARS data collection sites included a central ambient air monitoring site and six exposure measurement areas (EMA) that were selected to evaluate the impact of local sources.

PM2.5

Eenzene

Ind

Ind ust r Di ial ese Tra l ffic Hi ghw Re ay gio nal

Sulfur

ust r Di ial ese Tra l ffic Hi ghw Re ay gio nal

5 4 3 2 1 0

5 4 3 2 1 0

5 4 3 2 1 0

Ind ust ria Di l ese Tra l ffic Hi ghw Re ay gio nal

Increased Variability

Fig. 1 Ratio of residential outdoor concentrations to central site concentrations for DEARS season 1

Fig. 2 Relationships between concentrations at residential outdoor sites and the central site for DEARS EMAs

T. Watkins et al.

644

mg m mg m

3

Fig. 3 Comparisons of personal, indoor, outdoor residential and ambient central site measurements for DEARS (seasons 1–4)

3

Figures 1 and 2 present data on the spatial variability of pollutants by comparing data collected at the central site to data collected at residential outdoor sites in each EMA. Figure 1 presents the ratios of concentrations at residential outdoor locations to concentrations at the central site for sulfur, PM2.5, and benzene. The ratios for sulfur, a PM component, are all very close to 1. The ratios for total PM2.5 are also very close to 1, but exhibit a bit more variability than sulfur. The ratios for sulfur and PM2.5 confirm the relative spatial homogeneity for these pollutants. However, the ratios for benzene are generally greater than 1, which shows the influence of sources, particularly in industrial and mobile source influenced EMAs. Figure 2 provides additional data to show how spatial variability differs by pollutant. For PM2.5, there is a relatively strong relationship between residential outdoor sites and the central site. For sulfur, the correlation is also very strong plus seasonal variation is observed. Again, the results for PM2.5 and sulfur are not surprising given the regional nature of these pollutants. However, there is much more variability between the central site and residential outdoor sites for 1,3 butadiene and iron. For 1,3 butadiene, the influence of local sources can be seen particularly at EMA 6, which is impacted by highway sources. For iron, the influence of local sources is even greater, particularly at EMAs 1 and 3. The DEARS also provides insights into how microenvironmental factors and human activities impact human exposures to various pollutants. Figure 3 compares measurements of various air toxics taken at the central monitoring site, at outdoor residential locations, at indoor residential locations, and personal exposure measurements. The influence of micro-environmental concentrations (including indoor sources) and human activities can be seen as total personal exposure is generally higher and exhibits more variability than indoor, outdoor, and ambient concentrations. 16 14 12 10 8 6 4 2 0 40

Benzene

m,p-Xylene

30 20 10

mg m

3

0 8 6 4 2 0 0

Acrolein

70 60 50 40 30 20 10 0 14 12 10 8 6 4 2 0 6 5 4 3 2 1 0

Toluene Personal Indoor Outdoor Ambient

o-Xylene

12 10 8 6 4 2 0 4

Ethylbenzene

1,3-Butadiene

3 2 1

1,3,5-Trimethylbenzene

0 6 5 4 3 2 1 0

4-Ethyltoluene

5. Improving Exposure Assessment with Air Quality Models Addressing the air quality management issues above will require improved human exposure assessments. Improved exposure estimates could be obtained through more spatially, temporally, and compositionally refined air quality monitoring and


645

through conducting human exposure monitoring studies. However, increased monitoring is cost prohibitive. Air quality modeling tools offer an alternative to increased monitoring. For example, air quality models can provide more spatially, temporally, and compositionally refined estimates of ambient concentrations. They also can provide scientific insights to improve the understanding and characterization of the atmospheric processes that impact spatial and temporal variability of pollutants. When air quality models are integrated with each other and with other data sources even more opportunities to improve exposure estimates arise. For example, output from air quality models can be used to fill in the spatial and temporal gaps in existing ambient monitoring data to improve the estimates of air quality and exposure estimates for health studies (Bell, 2006). Another way air quality models can be used to improve exposure estimates is by combining results from regional scale air quality models, such as CMAQ, and local scale dispersion models, such as AERMOD. Figure 4 demonstrates this concept from a modeling analysis done in New Haven, CT. The regional air quality model provides a regional background concentration upon which the influence of local stationary and mobile sources can be added. The result is a more spatially refined ambient air quality estimate that may be used to improve exposure assessments. Combining air quality modeling output with monitoring data and integrating regional and local scale air quality models provide more spatially and temporally refined estimates of ambient air quality that can be used to improve human exposure estimates. However, while these approaches address the spatial and temporal variability exposure factor, they do not address other exposure factors such as microenvironmental concentrations and human activities. Linking air quality modeling outputs with human exposure models provides an approach to address these exposure factors. Using ambient concentration inputs from air quality models, human exposure models estimate actual human exposures by modeling factors, such as pollutant penetration rates, that impact pollutant concentrations in microenvironments and then integrating human activity data including time spent in each microenvironment. Figure 5 shows results from linking air quality and human exposure models in Philadelphia, PA (Isakov et al., 2006). As shown in Figure 5, while the spatial patterns are similar, the actual human exposures to benzene are greater than ambient concentrations, most likely due to exposures experienced in high exposure microenvironments, such as in-vehicle, at gas stations, or in attached garages.

646

T. Watkins et al.

Fig. 4 Approach for combining regional scale and local scale air quality models (New Haven, CT)

Fig. 5 Results from linking air quality and human exposure models in Philadelphia, Pa (Isakov et al., 2006)

6. Summary and Conclusions To effectively address the current and emerging air quality management issues that we face today will require an improved understanding of human exposures. This paper provided examples and case studies demonstrating how improved exposure assessment will be an important tool to support environmental policy decisions on whether to revise existing or develop new air quality standards and regulations. Improved exposure assessments can also inform other air quality management activities such developing and evaluating alternative emissions control strategies and evaluating whether air quality regulations have met anticipated goals to protect human health. Air quality modeling tools offer tremendous promise for improving exposure estimates needed for air quality management activities. To date, air quality models have been used sparingly in health studies. However, as air quality modeling approaches become more sophisticated, the opportunities to use air quality models to enhance exposure assessments in health studies will grow and potentially lead to improved air quality policies. Disclaimer The views expressed in these Proceedings are those of the individual authors and do not necessarily reflect the views and policies of the United States Environmental Protections Agency (EPA). Scientists in the EPA have prepared the


647

EPA sections and those sections have been reviewed in accordance with EPA’s peer and administrative review policies and approved for presentation and publiccation.

References Bell ML (2006) The use of ambient air quality modeling to estimate individual and population exposure for human health research: a case study of ozone in the Northern Georgia Region of the United States, Environment International, 32 (5), 586–593. Isakov V, Graham S, Burke J, Ozkaynak H (2006) Linking Air Quality and Exposure Models, EM, September 26–29. McMurry P, Shepard M, Vickery J (eds) (2004) Particulate Matter Science for Policy Makers: A NARSTO Assessment. New York: Cambridge University. National Research Council (2004a) Research Priorities for Airborne Particulate Matter IV – Continuing Research Progress. Washington DC: National Academies Press. National Research Council (2004b) Air Quality Management in the United States. Washington DC: National Academies Press. U.S. EPA (1992) Guidelines for Exposure Assessment. U.S. Environmental Protection Agency, Risk Assessment Forum, Washington, DC, 600Z-92/001. U.S. EPA (2004) Air Quality Criteria for Particulate Matter (October 2004). U.S. Environmental Protection Agency, Washington, DC, EPA 600/P-99/002aF-bF. U.S. EPA (2005) Review of the National Ambient Air Quality Standards for Particulate Matter: Policy Assessment of Scientific and Technical Information, OAQPS Staff Paper (June 2005). U.S. Environmental Protection Agency, Washington, DC, EPA-452/R-05-005.

Discussion P. Builtjes: Do you see a future for bio-monitoring, by for example measuring radicals in blood? T. Watkins: Yes, I do believe that bio-monitoring holds promise for providing valuable information about human exposures. However, there are currently limited biomarkers routinely available for many air pollutants, particularly for particulate matter. Also, while bio-monitoring information is certainly useful, biomarkers indicate whether an exposure has occurred, but they do provide other important information such the source, route, duration or intensity of exposure. Therefore other tools, including modelling tools, are needed to more fully characterize exposures.

P4.4 A Construction and Evaluation of Eulerian Dynamic Core for the Air Quality and Emergency Modelling System SILAM Mikhail Sofiev, Michael Galperin and Eugene Genikhovich

Abstract The paper presents a new dynamic core of the SILAM modelling system. It is based on the original Eulerian advection algorithm combined with extended resistance scheme for vertical diffusion. Apart from the standard advantages of the Eulerian environment, it has several unique features: (i) exactly zero numerical viscosity and a possibility to utilise the sub-grid information on mass location inside a grid cell; (ii) robustness to sharp gradients of concentrations and their preservation during the transport, (iii) applicability at high Courant numbers, (iv) options for prescribing or dynamically evaluating the horizontal diffusion; (v) vertical diffusion with thick adaptive layers that utilises the sub-grid information of advection for refining the flux values.

Keywords Advection-diffusion schemes, Eulerian models, numerical modelling

1. Structure of SILAM and the New Items Historically, most of emergency-response systems are based on Lagrangian advection, often with the Monte-Carlo random-walk diffusion (Saltbones et al., 1996; Stohl et al., 1998; Sørensen et al., 2000; Sofiev et al., 2006). Systems for air quality assessments are mainly based on Eulerian approaches, such as the widely used scheme of Bott (1989) and its numerous variations. For diffusion, the K-theory and its Crank-Nicholson three-diagonal solution are the most universal. The SILAM system (Figure 1, Sofiev et al., 2006) is a flexible environment made for a wide variety of tasks, including emergency response, air quality, observation analysis, data assimilation and inverse-problem applications. It is a modular system with object-oriented code. Lagrangian and Eulerian dynamic cores utilise the same supplementary routines including meteorological, emission, and I/O servers. At present, the cores are not connected and the user has to select the type of dynamics used for the run. The advection routine (Galperin, 2000) of the new Eulerian core is connected with the adaptive vertical diffusion algorithm of Sofiev (2002), which makes use of the advection-controlled sub-grid variable (the first moment of the mass located in


699

M. Sofiev et al.

700

the grid cell) for dynamic adaptation of the vertical. The tests of the advection scheme (examples in Figure 2) show that is has zero numerical viscosity and is capable of operating at very high Courant numbers. Vertical diffusion scheme serves for the whole column and incorporates dry deposition and re-evaporation via extended resistive algorithm. Its parameterization is based on K-theory after Genikhovich et al. (2004). Acknowledgments The study was supported by TEKES-KOPRA and EU-GEMS projects, and the Network scale Atmospheric Modelling NETFAM. Control unit Dispersion interface Lagrangean pollution cloud

Eulerian data buffer

Dispersion dynamics

Emission composer

Fig. 1 Schematic picture of the main SILAM components

Area Bomb Point emission

Processing

Dispersion physics & chemistry

Interface Data storage Control flow Data flow

a

b

Fig. 2 Tests of the advection scheme: a point-type and rectangular masses in a circular vortex; 1D jet-stream with Courant number (grey) up to 9; relative concentrations are shown in black

References Bott A (1989) A positive definite advection scheme obtained by non-linear renormalization of the advection fluxes. Mon. Wea. Rev., 117, 1006–1012. Galperin MV (2000) The Approaches to Correct Computation of Airborne Pollution Advection. In: Problems of Ecological Monitoring and Ecosystem Modelling, vol. XVII, St. Petersburg, Gidrometeoizdat, pp. 54–68. Genikhovich E, Sofiev M, Gracheva I (2004) Interactions of meteorological and dispersion models at different scales. In Air Polution Modelling and its Applications XVII (eds. C Borrego, A-LNorman), Springer 2007, pp. 158–166, ISBN-10: 0-387-28255-6

A Construction and Evolution of Eluerian Dynamic Core for the Air Quality

701

Saltbones J, Foss A, Bartnicki J (1996) A real time dispersion model for severe nuclear accidents, tested in the European tracer experiment. Syst. Anal. Model. Simul. 25, 263–279. Sofiev M (2002) Extended resistance analogy for construction of the vertical diffusion scheme for dispersion models. J. of Geophys. Research-Atmosphere, 107, D12, doi:10.1029/2001JD001233. Sofiev M, Siljamo P, Valkama I, Ilvonen M, Kukkonen J (2006) A dispersion system SILAM and its evaluation against ETEX data. Atmos. Environ., 40, 674–685. Sørensen JH, Mackay DKJ, Jensen CØ and Donaldson AI (2000) An integrated model to predict the atmospheric spread of foot-and-mouth disease virus. Epidemiol. Infect. 124, 577–590. Stohl A, Hittenberger M, Wotawa G (1998) Validation of the Lagrangian particle dispersion model FLEXPART against large scale tracer experiments. Atmos. Environ. 24, 4245–4264.

P7. Air quality and human health

P7.1 A Multi-Objective Problem to Select Optimal PM10 Control Policies Claudio Carnevale, Enrico Pisoni and Marialuisa Volta

Abstract To implement efficient air quality policies Environmental Agencies require integrated systems allowing the evaluation of both the effectiveness and the cost associated to different emission reduction strategies. These tools are even more useful when considering atmospheric PM10 concentrations, a strongly nonlinear secondary pollutant. The classical approaches of cost-benefit and cost-effectiveness analysis create unique solutions, hiding possible stakeholders conflicts. In this work the formulation of a multi-objective problem to control particulate matter is proposed, defining: (a) control objectives, namely the air quality indicators and the cost functions; (b) decision variables and their constraints; (c) source-receptor models, describing the cause-effect relation between air quality indicators and decision variables. The multi-objective problem results obtained for Northern Italy are analyzed in terms of not-dominated solutions. 1. Introduction Due to nonlinearities of secondary pollutants, it is very challenging to develop sound policies considering both air quality improvements and implementation costs. In literature, to solve this control problem, multi-objective analysis (Guariso et al., 2006; Carnevale et al., 2007) has been rarely used, and only for ozone control. In this paper, a multi-objective optimization methodology for PM10 control is proposed. The nonlinear relations between decision variables and PM10 exposure index, defining the air quality objective, are described by neuro-fuzzy models, identified processing long-term simulations of GAMES multiphase modelling system performed in the CityDelta project (Cuvelier et al., 2007).

2. Methodology and Results The target of this study is to control particulate matter exposure at ground level. This issue can be attained by optimizing both air quality indicators and emission abatement costs. The emission reduction rates (decision variables) are computed by C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

715

C. Carnevale et al.

716

a two-objective mathematical programming algorithm as the ones corresponding to the most efficient strategies, with respect to both the considered objectives. The problem can be formalized as follows: min J ( T ) J ( T ) [ AQI ( E ( T )); CPI ( E ( T ))]

T 4 where E represents the precursor emissions for the reference case, T are the decision variables (namely the emission reductions) constrained to assume values in 4 , AQI(E( T )) and CPI(E( T )) are the Air Quality Index and Cost of Policy Index respectively, both depending on precursor emissions and emission reductions. Optimization methodology results suggest: (a) efficient solutions of the PM10 control problem, depicting the Pareto boundary; (b) emission reduction priorities needed to obtain a particular result; (c) costs of implementation of the different emission reduction policies. I.e. in Figure 1 the Pareto boundary (left) is shown, stressing point A (the point associated to no reductions), point B (that reduces PM10 to 32 Pg/m3 with a cost of roughly 700 Meuro), and point C (maximum PM10 reduction with maximum cost). The macrosector costs for point B and C are shown in Figure 1 (right).

Fig. 1 Pareto boundary solution of the optimisation problem (left) and emission reduction costs associated to point B and C in the Pareto Boundary, for each CORINAIR macrosector

References Carnevale C, Pisoni E, Volta M (2007) Selecting effective ozone exposure control policies solving a two-objective problem, Ecological Modelling 204, 93–103. Cuvelier C et al. (2007) CityDelta: a model intercomparison study to explore the impact of emission reductions in European cities in 2010, Atmospheric Environment, 41, 189–207. Guariso G, Pirovano G, Volta M (2004) Multi-objective analysis of ground level ozone concentration control, Journal of Environmental Management 71, 25–33.

P7.5 Air Pollution Assessment in an Alpine Valley Peter Suppan, Klaus Schäfer, Stefan Emeis, Renate Forkel, Markus Mast, Johannes Vergeiner and Esther Griesser

Abstract ALPNAP (Monitoring and Minimization of Traffic-Induced Noise and Air Pollution along Major Alpine Transport Routes) is an ongoing research project focussing on corresponding effects along several major transit routes across the European Alps. Assisting regional authorities with appropriate output, a unique cooperation of scientists within the alpine region was setup to better assess and predict the spatial and temporal distribution of air pollution and noise close to major alpine traverses. A methodology for measurement strategies and model simulations for air pollutants will be demonstrated for the Brenner traverse. Results of a field measurement campaign give detailed insights of the complexity of the atmospheric conditions and the distribution of air pollutants in the Inn valley. First results of the air quality simulations with the regional meteorology-chemistry model MCCM (MM5/chem) on a coarse resolution show the importance of a detailed emission inventory in an alpine valley. Within ALPNAP comprehensive measuring campaigns as well as detailed modelling simulation are foreseen to describe the air pollution situation in an alpine valley. In a first step a measurement campaign in the lower Inn-valley was designed to determine cross-valley air pollution and meteorological information as well as vertical profiles to determine flow regimes (valley, slope winds), mixing layer height, stability in the boundary layer and emission sources at specific locations. Covering the major part of the winter season, this data will be used for enhanced analysis as well as for the set-up and validation of corresponding models merging the wealth of different remote sensing and in-situ measurements. In a second step air quality simulations will be carried out, in order to receive detailed 3D information of the distribution of air quality parameters. In the frame work of ALPNAP, a field campaign in the Inn valley was performed between November 2005 and January 2006. The most important goals for the field campaign have been to study the spatial variation and distribution of air pollutants within a cross-section along the northern and southern slopes and to evaluate the best use of an existing slope temperature profile for mixing height and stability analyses and finally to use the field data for the set-up and validation of analysis and modelling inter-comparison studies.


723

724

P. Suppan et al.

Results The NO concentrations are clearly dominated by the traffic volume and therefore are the highest at the highway followed by the valley floor in 800 m distance to the highway. The exceedance at the valley floor as compared to the slope station is basically related to a stable layering of the valley atmosphere during nearly all the time. Outstanding high pollution episodes were found during: 20–24 December, 2005, 09–22 January and 25 January–02 February, 2006, when the detailed structure of the valley atmosphere was captured by the additional intensive measurements periods. Moreover, the data indicate that the ratio of concentrations near the highway to the background is systematically higher for NO2 (more than factor 2) than for NO. On a local scale the data allow e.g. to demonstrate the impact of nearby construction work imposing on the CO, NO and NOx measurements. The temporal variations of air pollution concentration (CO, PM10, and NO2) at the valley ground are clearly dominated by the mentioned weather conditions and emissions. As for the NO2 and CO emissions, the main source is road traffic at the highway. The NO2 and CO maxima in the morning and in the evening correspond to the local traffic maxima. In the early afternoon mixing height rises and the concentrations decrease, whereas the ozone concentrations increase. During midnight the CO concentrations show a maximum and PM10- and NO2-concentrations show a minimum. However, even detailed measurements like these cannot reflect the full temporal and spatial variability of the complex flow regimes or the horizontal and vertical distribution of chemical parameters in alpine valleys. Future numerical simulations will contribute there and the corresponding validation process will greatly benefit from the available data. Thus setting up different models in a most sophisticated way will enable for process oriented studies, model inter-comparisons and impact studies. Further information about the ongoing work can be obtained at the ALPNAP web site http://www.alpnap.org/ Acknowledgments The project ALPNAP is implemented through financial assistance from funds of the European Community Initiative Programme “Interreg III B Alpine Space”.

P4.2 Air Pollution Dispersion Modelling Arround Thermal Power Plant Trbovlje in Complex Terrain – Model Verification and Regulatory Planning Marija Zlata Božnar, Primož Mlakar, Boštjan Grašiþ and Gianni Tinarelli

Abstract The paper shows extensive verification of Lagrangian particle model Spray coupled with Minerve 3D mass consistent diagnostic wind field model in extremely complex terrain of Zasavje region in Slovenia. On-line measured emission and ambient data was used to reconstruct air pollution situation across the area for one-year time interval.

1. Introduction and Campaign Description Zasavje region is a highly industrial area located in a river canyon in central part of Slovenia. Thermal Power Plant (TPP) Trbovlje, Cement factory Trbovlje and Glass factory Hrastnik are main sources of air pollution in the area. TPP and Cement factory have just installed wet desulphurisation plants that decreased the previous level of SO2 pollution significantly. A study (Božnar et al., 2006) was done to reconstruct the current air pollution situation in the area, to quantify the expected reduction of SO2 pollution by desulphurisation plants and to model the future scenarios (new planned gas powered TPP). One year of on-line meteorological, air pollution and emission data was analysed. Across the area of interest mainly SO2 and NO2 pollution exceed the regulation limits. High pollution of the area is caused by high emissions, but it is also emphasized by local microclimatological conditions (low wind speeds, calm situations and strong thermal inversions) as the area is a highly complex terrain (canyon with steep slopes of approximately 45q, several valleys perpendicular to main canyon, hills of relative height over 1,000 m, see GoogleMaps, search keyword “Trbovlje, Slovenia”. Zasavje region therefore represents an excellent study case for the evaluation of dispersion modelling in complex terrain. In the area there exist a dense monitoring of meteorological parameters (including one SODAR profile) environmental concentrations (nine stations) and one on-line emission station on the existing TPP stack. On line measured emission and ambient data for one-year time interval from 01.07.2005 till 31.06.2006 was used to reconstruct air pollution situation across the area. C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

695

M.Z. Božnar et al.

696

2. Results and Conclusions To determine the optimal height of the new planned TPP a comparison of reconstructed ground level concentrations for stacks with different heights was performed. Extensive verification was performed to qualify the results of reconstruction. Two very good reconstructed air pollution situations are depicted on Figure 1 were measured and reconstructed values at locations Dobovec and Kovk are compared. First situation occurred before and second after the TPP desulphurisation plant installation. 1000,00 900,00 800,00

200,00

Dobovec

SO2 concentration

180,00

[µg/m3]

160,00

Kovk Measured Reconstructed

140,00

700,00 600,00

Measured

120,00

500,00

Reconstructed

100,00

400,00

80,00

300,00

60,00

200,00

40,00

100,00

20,00

0,00 12.09.2005 00:00

SO2 concentration [µg/m3]

12.09.2005 06:00

12.09.2005 12:00

12.09.2005 18:00

13.09.2005 00:00

13.09.2005 06:00

13.09.2005 12:00

0,00 12.06.2006 00:00

12.06.2006 06:00

12.06.2006 12:00

12.06.2006 18:00

13.06.2006 00:00

13.06.2006 06:00

13.06.2006 12:00

Fig. 1 Comparison of reconstructed and measured SO2 concentrations at stations Dobovec and Kovk

Good behaviour and shortcomings of Lagrangian particle model approach were identified and discussed. Extensive verification showed that foundation for good reconstruction is good database without any corrupted data measurements. Modelling of the future scenarios (operating of two desulphurisation plants and possible new TPP) will be used by the decision makers to allow or reject building of new TPP (or other objects) in the area. As the area is considered to be highly polluted, these results may also help to prepare other remediation programs for better air quality in the area.

References Božnar M, Grašiþ B, Tinarelli G (2006) Thermal power plant Trbovlje air pollution impact modelling in complex terrain., Sixth Annual Meeting of the European Meteorological Society (EMS) [and] Sixth European Conference on Applied Climatography (ECAC): Ljubljana, Slovenia, 4–8 September 2006 (EMS annual meeting abstracts, volume 3). Ljubljana: European Meteorological Society: Agencija RS zaokolje

P2.13 Analysis of Atmospheric Transport of Radioactive Debris Related to Nuclear Bomb Tests Performed at Novaya Zemlya Jørgen Saltbones, Jerzy Bartnicki, Tone Bergan, Brit Salbu, Bjørn Røsting and Hilde Haakenstad

Abstract Tropospheric transport calculations of debris from nuclear bomb tests in the atmosphere at Novaya Zemlya in October 1958 and November 1962 have been performed and compared with daily measured radioactivity in air at Norwegian monitoring sites. Our analysis of potential vorticity (PV) anomaly strongly indicate that episodic intrusion of stratospheric air into the troposphere is the most probable transport mechanism for the peaks in radioactivity measured in Norway. The first hypothesis tested was that episodic increases in radioactivity in air sampled close to the ground in Norway were caused by a direct tropospheric transport of radioactive debris from the nuclear bomb tests at Novaya Zemlya. Forward trajectory computations from Novaya Zemlya have been performed, starting at different heights in the troposphere and for these periods, just a few of the bomb tests could possibly have sent radioactive debris via a tropospheric direct route to Norway. This indicated that a direct tropospheric transport from Novaya Zemlya to the Norwegian monitoring sites was not a very likely transport mechanism (Bartnicki et al. 2004; Bergan et al., 2005). However, when calculating the age of the radioactive debris, by observing ratios between radio isotopes with different decay rates, it seems likely that radioactive debris arrived at Norwegian monitoring sites 10–40 days after the detonations; – (Bergan et al., 2005). This supports earlier observations of quite long residence times for radioactive debris in the polar stratosphere. The nuclear bomb tests in these periods had variable total yields up to several kt and can be expected to have sent radioactive debris high up into the stratosphere just after the blast (STANAG, 1994). To investigate more closely the transport mechanism responsible for peaks in the measurements, we have worked out the following screening procedure: Using ERA-40 re-analyses data from ECMWF as a start, we have performed analysis/ short prognosis of the weather situation – using the Norwegian HIRLAM–10 km model, which is operational at the Norwegian Meteorological Institute. Pattern indicating intrusion of stratospheric air into the troposphere are, e.g. low relative humidity (Rh) and high values of potential vorticity (PV). Results from the HIRLAM model for 20 October 1958 indicate just that. As shown in Figure 1, high values of PV and low values of Rh, dipping down as a tongue just west of C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

683

684

J. Saltbones et al.

Norway, just at the place and at the time of the peak in total beta activity measured in Bergen. This is a strong indication of intrusion of stratospheric air.

Fig. 1 Vertical crossection north of Bergen showing potential vorticity (blue lines), potential temperature (red lines) and relative humidity (green lines) for 20 October 1958 at 12 UTC

Acknowledgments This study was supported and partly financed by Norwegian Research Council for Science and Humanities and by Norwegian Radiation Protection Authority.

References Bartnicki J, Foss A, Saltbones J (2004) Analysis of trajectories related to nuclear bomb tests performed in Novaya Zemlya. Met. no Note 9. Norwegian Meteorological Institute. Oslo, Norway. Bergan DT, Bartnicki J, Dowdal M, Foss A, Saltbones J (2005) Analysis of trajectories related to nuclear bomb tests performed at Novaya Zemlya. In: Proceedings from The 6th International Conference on Environmental Radioactivity in the Arctic & Antarctic, 2–6 October 2005 in Nice, France (P. Strand, P. Børretzen, T. Jølle, eds). Norwegian Radiation Protection Authority, Østeraas, Norway 2005. pp. 77–86. STANAG (1994) STANAG 2103, APT-45 Vol I/II: ‘Reporting Nuclear Detonations, Biological and Chemical Attacks, and Predicting and Warnings of Associated Hazards and Hazard Areas (NATO Declassified), June 1994.

P2.2 Application of Back Trajectories Using Flextra to Identify the Origin of 137Cs Measured in the City of Barcelona Delia Arnold, Arturo Vargas, Petra Seibert and Xavier Ortega

Abstract In order to determine the possible origin of the 137Cs detected at the Energy Technologies Institute (INTE) station from the Technical University of Catalonia (Barcelona, NE Spain), a study by means of trajectory analysis has been carried out and compared with results from Physikalisch-Technische Bundesanstalt (PTB) station in Braunschweig (Germany). A relation between high detection and incoming air from highly contaminated regions is clearly visible for the PTB site but not for Barcelona. Before the Chernobyl accident, the source of 137Cs was just from past nuclear weapons tests. This 137Cs was dispersed all over the globe and the measured values were therefore similar in different locations. However, after the Chernobyl accident 137 Cs activities increased significantly in Europe and especially near the Chernobyl region. Some works (Seibert and Frank, 2004; Swanberg and Hoffert, 2001; Wershofen and Arnold, 2005) have related measured values in Germany and Scandinavia and atmospheric transport simulations, indicating that the origin of 137 Cs is found in many cases in the highly contaminated areas of Ukraine, Belarus and Russia near Chernobyl. 137 Cs is measured in weekly seven-days samples at the INTE using a highvolume air sampler. To study its possible origin, ten-day backward trajectories computed by the FLEXTRA trajectory model (Stohl et al., 1995) have been analysed for the period 2001–2004. The results are compared with those found for the PTB station at Braunschweig (Germany) for the years 1998–2003 with the same method. At Barcelona approximately the 10% of the weekly measurements give 137Cs values above the detection level. Therefore, it has been decided to consider as episodes with high 137Cs measured values those 10% with highest measured values. For these selected periods, the percentage of the total residence times below the mixing height, which is associated with measurable 137Cs upon arrival of the trajectories, has been computed in a 1º × 1º grid covering Europe. In fair agreement with Wershofen and Arnold (2005) it has been found that an increase in 137Cs concentration at the German site is related to air masses coming from the East (see Figure 1, left). High values of residence times are found near the Chernobyl region, indicating that this is a possible source for the 137Cs measured at that site. From the visual inspection of trajectories, the highly contaminated areas of the Scandinavia C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

661

662

D. Arnold et al.

appear to be a possible source. This shall be studied in the future. For Barcelona, no clear influence of 137Cs transported from Belarus and Ukraine regions is found (Figure 1, right). The main reason is that almost no trajectories passed over the contaminated areas. It appears therefore likely that the contribution of locally or regionally resuspended 137Cs makes an equal or even dominant contribution to the measured 137Cs as compared to contributions from the surroundings of Chernobyl. Thus, even a simple tool such as trajectory residence times can be used as a first approach.

Fig. 1 Residence time per grid cell (%) for the PTB (left) and the INTE (right) stations. Crosses are station locations and the circled black dot is Chernobyl. White areas have zero residence times

Acknowledgments This work is included in the Spanish Education and Science Ministry project CGL2005-04182/CLI. Authors would like to thank the PTB to allow the use of data appeared in PTB-Ra-45 report and ECMWF for data access through the Special Project MOTT. P. Seibert acknowledges the support of the European Commission through the FP6 Network of Excellence ACCENT.

References Stohl A, Wotawa G, Seibert P, Kromp-Kolb H (1995) Interpolation errors in wind fields as a function of spatial and temporal resolution and their impact on different types of kinematic trajectories. J. Appl. Meteor. 34, 2149–2165. Seibert P, Frank A (2004) Source-receptor matrix calculation with a Lagrangian particle dispersion model. Atmos. Chem. Phys. 4, 51–63. Swanberg EL, Hoffert S (2001) Using atmospheric 137Cs measurements and Hysplit to confirm Chernobyl as a source of 137Cs in Europe. 23rd Seismic Research Review: Worldwide monitoring of nuclear explosions, October 2–5. Wershofen H, Arnold D (2005) Radionuclides in the Ground-level Air in Braunschweig – Report of the PTB Trace Survey Station from 1998 to 2003. Report Radioaktivität PTB-Ra-45, ISBN 3-86509-431-7, Braunschweig.

P1.3 Assessment of the Breathability in Urban Canyons Through CFD Simulations and Its Application to Sustainable Urban Design Mário Tomé, Ricardo J. Santos, António Martins and Mário Russo

Abstract According to the European Commission´s own Impact Assessment, every year, 369,980 people die prematurely because of air pollution. CFD simulations can be used to assess the advection and turbulent diffusion of pollutants emitted by moving sources on urban canyons. Although these tools need more development and validation, they are already sufficiently robust to support the urban design process, maximising the pollutants (and other scalars such as energy) cleaning inside the canyon. On the other hand, the urban designers and planners are aware of the importance of urban air quality but do not have enough knowledge to support their planning decisions as no relevant guidelines are available. In order to evaluate the capacity that an urban canyon has to exchange pollutant in its virtual ceiling top, CFD simulations were carried-out based on both eulerian and lagrangian approaches for both stationary and dynamic urban canyons with different aspect ratios. Moreover, the breathability conditions inside the canyons are defined using classical approaches borrowed from chemical reactor theory such as pollutant residence time distribution and dimensionless concentrations maps. CFD tools can contribute to understand and help design better urban landscapes for strong cleaning of the pollutants released inside the urban canyons (Xiaomin et al., 2006). Besides many remaining questions about the difficulties of CFD to model the open atmospheric boundary layer, namely its roughness and stability (Blocken et al., 2007), there are other important issues that need more studies. We are interested to compare eulerian with lagrangian approaches for recirculating flows that arise in urban canyons of aspect ratio of 1. The 2D simulation domain consists of an infinite array of buildings of 16 × 16 m separated from each others with 16 m. The domain height is 100 m. The infinite array was modelled using the periodic boundary condition of the Fluent CFD software applied to the singular repeated domain of 32 m. A simple fortran code extrapolated this periodic solution for a six canyons domain (6 × 32 m = 192 m). In this larger domain the flow is not computed because it has previously converged in the 32 m wide periodic domain with second order accuracy. The 192 m wide domain was used to compute the eulerian dis-persion of a passive pollutant emitted by a surface with 8 m wide by 1 m height. This virtual road has an emission rate of 0.001 kg pollutant/(m2s–1). The same emission is used by a lagrangian approach in which these same area of road emits 220 discrete particles in each lagrangian time step (=0.05 s). Thus, each C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

655

M. Tomé et al.

656

numerical particle represents 1.818 × 10-6 kg of pollutant. The flow field was initialized by an exponential profile with wind velocity of 5 m s-1 at the height of 26 m (10 m + displacement = 16 m). This initial wind field does not parameterize the solution because the periodic boundary is not dependent of this initial solution. The computed wind field has a maximum horizontal velocity of 8.36 m s-1 (at the domain top) and an air mass flow rate of 720 kg s-1. The turbulence model used was the standard k-e. Due to the strong vortex the lagrangian approach with stochastic numbers revealed to be much more time consuming and also it requires more attention from the modeller as the particles can be recirculating nearly for endless. Thus, eulerian and lagrangian results can produce quite different outcomes unless the lagrangian simulation is very “well done”, which means having appropriate (small) time steps, a very long global temporal simulation time period in order to achieve “lagrangian stability”, which can be easily assessed by the particles escaping rate from the domain being equal to the number o injected particles (220) per time step. A RTD analysis, from chemical engineering, is applied in order to interpret the results of a constant flux tracer (pollutant) emitted in the second canyon. The RTD curves (Figure 2) show an average time of pollutant residence in the domain much higher than the wind passage time. The wind should approximately transport materials: to x = 64 m in approximately 4 s; to x = 128 m in approximately 12 s; to x = 160 m in approximately 16s and to x = 192 m in approximately 20 s. The RTD curves at the different proving location keep approximately the same relative distance, but the typical value is much higher: to x = 64 m E(t)max = 30 s; to x = 128 m E(t)max = 45 s; to x = 160 m in E(t)max = 50 s and to x = 192 m E(t)max = 55 s. The Eulerian approach clearly shows that the steady recirculation vortices between buildings, seen to trap the particles, are dead volumes with very small mass transfer across the virtual ceiling, even considering turbulent diffusion. Probably LES results will yield less stable vortices between buildings, and thus account more precisely for mass transfer, representing the next envisaged step in this work. Results and further discussion of these simulations are presented in the poster (available [email protected]). 100

Pollutant mass fraction (-)

7.00E-03 5.00E-03 2.00E-03 1.00E-03 5.00E-04 2.00E-04 1.00E-04 5.00E-05 3.00E-05

Y (m)

79

58

37

zoom

16

8

24

40

56

72

88

104

X (m)

120

136

152

168

184

Fig. 1 Eulerian concentration (mass fraction) field of pollutant and pathlines

Fig. 2 RTD analysis, with the system outlet boundary at different distances, from the “frozen” wind field and a step tracer emission curve

Acknowledgments R. Santos is grateful for the funding from POCI/N010/2006.

References Blocken B, Tathopoulos T, Carmeliet J (2007) CFD. Atmos. Environ. 41, 238–252.

P4.5 BOLCHEM Air Quality Model: Performance Evaluation over Italy Alberto Maurizi, Mihaela Mircea, Massimo D’Isidoro, Lina Vitali, Fabio Monforti, Gabriele Zanini and Francesco Tampieri

Abstract The modeling system BOLCHEM for air quality simulations has been run to study the evolution of tropospheric ozone over the Italian peninsula during 1999. The comparison of measured and modeled ozone time series shows that BOLCHEM predicts well the ozone concentrations. The summer cases are better simulated than the winter ones. The model configuration using SAPRC90 meets always the US-EPA criteria for the statistical indexes UPA, MNBE and MANGE.

The Italian peninsula has a very complex topography, therefore the separation of meteorology and chemistry in offline simulations can lead to a loss of potentially important information about atmospheric processes, which often have a much smaller time scale than the meteorological output frequency. Here, we show the ability of a new developed air quality model, BOLCHEM, to reproduce the observed ozone concentrations for four clear sky periods (one from winter, the others from summer season) selected based on Meteosat images of Europe. The calculated O3 concentrations were compared with measurements made at rural or semi-rural stations. The statistical measures recommended by the U.S. Environmental Protection Agency (US-EPA, 2005) for air quality model validations were also computed for hourly values of ozone concentration. BOLCHEM couples the meteorological model BOLAM (Buzzi et al., 2003) to SAPRC90 (Carter, 1990), and CB-IV (Gery et al., 1989) as alternative photochemical mechanisms. The mechanisms were chosen since they adopt different criteria for grouping the organic gases: CB-IV groups the organics according to bond type, while SAPRC90 groups them according to molecule type. Figure 1 shows that the agreement between simulated and measured ozone concentrations is good for both summer and winter. Generally, the model has difficulties in reproducing low ozone concentrations observed during the winter and during the night. A more extensive discussion of the results can be found in Mircea et al. (2007). Table 1 shows the statistical indexes recommended by US-EPA (USEPA, 2005). It can be seen that generally, UPA is lower than 35%, MNBE is lower than 15%, MANGE is lower than 30–35%.


702

BOLCHEM Air Quality Model: Performance Evaluation over Italy

703

Montelibretti 300 3

2 P g/m )

250 200

SAPRC90

150

CB4

100

obs

50 1999080717

1999080709

1999080701

1999080617

1999080609

1999080601

1999080517

1999080509

1999080501

1999070417

1999070409

1999070401

1999070317

1999070309

1999070301

1999070217

1999070209

1999070201

1999070117

1999070109

1999070101

1999060317

1999060309

1999060301

1999060217

1999060209

1999060201

1999060117

1999060109

1999060101

1999012417

1999012409

1999012401

1999012317

1999012309

1999012301

1999012217

1999012209

1999012201

1999012117

1999012109

1999012101

1999012017

1999012009

1999012001

0

Fig. 1 Simulated (SAPRC90-red line and CB-IV-pink line) and observed (green line) surface ozone concentrations at Montelibretti during one winter and three summer periods Table 1 Quantitative model performance statistics for hourly surface ozone over the whole domain during summer. MNBE and MANGE were calculated with 40 ppb cutoff to the ozone concentrations.

1–3 June

UPA SAPRC90 –20.45

CB-IV –37.15

MNBE SAPRC90 –6.98

CB-IV –19.55

MANGE SAPRC90 22.53

CB-IV 24.61

1–4 July 5–7 August

9.3 –13.96

–17.59 –28.74

–4.37 –2.37

–19.38 –16.46

25.60 22.84

27.71 23.81

The model performances are better during summer, when the photochemistry is active, than during winter. During summer, the validation exercise shows that the model configured with SAPRC90 always meet the US-EPA criteria for UPA, MNBE and MANGE, while the MNBE calculated with CB-IV ozone concentration is sometimes higher than the recommended values. Acknowledgments This work was supported by the European Commission through the Network of Excellence ACCENT and the project GEMS, and by the Italian Ministry of Environment through the Program Italy-USA Cooperation on Science and Technology of Climate Change.

References Buzzi A, D’Isidoro M, Davolio S (2003) A case-study of an orographic cyclone south of the Alps during MAP SOP, Quart. J. Roy. Met. Soc., 129, 1795–1818. Carter WPL (1990) A Detailed Mechanism for the gas-phase atmospheric reactions of organic compounds, Atmos. Environ., 24A, 481–518. Gery MW, Witten GZ, Killus JP, Dodge MC (1989) A photochemical kinetics mechanism for urban and regional scale computer modeling. J. Geophys. Res., 94 (D10), 12925–12956. Mircea M, d’’Isidoro M, Maurizi A, Vitali L, Monforti F, Zanini G, Tampieri F (2007) A comprehensive performance evaluation of the air quality model BOLCHEM over Italy, doi:10.1016/j.atmosenv.2007.10.043.

704

A. Maurizi et al.

US-EPA, U.S. Environmental Protection Agency (2005) Guidance on the use of models and other analyses in attainment demonstrations for 8-hour ozone NAAQS. EPA report EPA-454/R-05-002, 128pp.

P3. Data assimilation and air quality forecasting

P3.1 Detection of a Possible Source of Air Pollution Using a Combination of Measurements and Inverse Modelling Borivoj Rajkovic, Zoran Grsic and Mirjam Vujadinovic

Abstract Detection of a possible source of air pollution as a combination of measurements and inverse modelling based on Bayesian statistics has been proposed. The simplicity of the approach and its numerical efficiency qualifies it for the problem, especially in the operational mode. Detection of an air pollution point source, during an accidental release, requires time efficient and relatively accurate solution. These two characteristics can be fulfilled with an approach based on the Bayesian statistical method (Fuentes and Raftery, 2001; Tarantola, 2005, among others). Relative positions of the assumed source and measurement points are presented in the Figure 1, upper right panel. The first step, in finding source’s position, was to generate field of passive substance concentration by its release during 120 minutes, using a PUFF model (Grsic and Milutinovic, 2000). The grid had 301 × 301 points with spacial distance of 60 m. The time interval between two consecutive puffs was 1 minute. Values at the measurement positions where then randomly pertubatied by 5%, thus mimicking the measurement errors. The second step was to create the cluster of points, possible sources, with center positioned in the measuring point with the highest concentration observed. In the first iteration we had relatively high cluster resolution of 30 grid points. From the cluster points we calculated probability density function (pdf) of the source position, assuming that it is Gaussian (Tarantola, 2005). In the second iteration, we translated the center of a cluster, to the possition of the first pdf’s maximum. Again, we calculated pdf and its maximum and translated the center to the possition of the new maximum. Since now cluster member with maximum pdf was at the inside area of the previous cluster, we decided also to halve the spread among third cluster’s members. In the next two iterations we again translated the cluster and halved the distance between its members. The first four, consecutive, positions of the cluster are presented in Figure 1, left panel. The right lower panel shows the last cluster position and calculated source location (Ms). The error in the obtained possition, after five iterations, was 120 m (two grid points). C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

689

690

B. Rajkovic et al.

Fig. 1 The right upper panel shows relative positions of the source and observation points. The left panel presents changes of the cluster position during four iterations, while the position of the final, fifth, cluster is shown at the right lower panel, where Ms denotes modelled position of a source

Combination of measurements and Bayesian statistical approach through several iterations could result in a very accurate and yet efficient method of detecting a possible, point like, source of a pollution. Acknowledgments This work has been partialy funded by Republic of Serbia, Ministry of Science, Technology and Environment, grant no. 1197, Italian Ministry of Environment and Territories through its two projects, SINTA and ADRICOSMSTAR. The second author is partially funded through the project of the decommission of the nuclear reactor in Vinca.

References Grsic Z, Milutinovic P (2000) Air Pollution Modelling and Its Application, XIII: Automated meteorological station and the appropriate software for air pollution distribution assessment. Fuentes M, Raftery A (2001) Model evaluation and spatial interpolation by combining observations with outputs from numerical models via Bayesian Melding, Technical report no. 403, Department of Statistics, University of Washington, Seattle, WA. Tarantola A (2005) Inverse problem theory and methods for model parameter estimation, SIAM, Philadelphia.

P2.14 Development and Application of a New Model for the Atmospheric Transport and Surface Exchange of Semi-Volatile Organics Using the CMAQ Model Framework Fan Meng, Baoning Zhang, Fuquan Yang and James Sloan

Abstract PCBs and PCDD/Fs are toxic, persistent pollutants that can bioaccumulate in the food chain and become serious health hazards. Although the manufacture of materials such as PCBs has been banned in most parts of the world for some time, they are still found in significant concentrations in the environment. In view of this, it has become important to understand not only their sources, but also the mechanisms responsible for their transport in the environment. Since they are semi-volatile, it is necessary to consider their transport by atmospheric particulate matter as well as in the gas phase. We have developed a capability to simulate the atmospheric behaviour of PCBs and PCDD/Fs within the framework of the CMAQ modelling system. To describe transport on particulate matter, we have added two gas/particle partitioning models – the Junge-Pankow adsorption model and the KOA absorption model to the basic CMAQ system. We have also included gas phase chemistry of these semi-volatile organic materials as well as their atmosphere/water surface exchange processes. Using this modified model system, we have conducted simulations of the atomspheric behaviour of these materials for the years 2000 and 2002 on a domain covering most of North America. Validation studies show that both partitioning models give reasonable results when compared with available measurements of deposition rates and air concentrations. The simulations confirm that long range transport occurs by both gas phase and heterogeneous mechanisms. This causes these toxic materials to be deposited in pristine regions far from emission sources. In cases where they have entered the water table, large water bodies such as the Great Lakes can also become net sources. Polychlorinated Biphenyls (PCBs), Polychlorinated Dibenzo [p] Dioxins and Polychlorinated Dibenzo-Furans (PCDD/Fs) are toxic persistent organic pollutants (POPs) that can be transported over global scales, thereby affecting human health and the ecology. Most PCBs and PCDD/Fs are semi-volatile organic compounds that partition between gas and particle phases in the atmosphere and between the atmosphere and surface soil and water. Since the chemical transformation and removal of these POPs in the gas phase differ from that in the particle phase, the partitioning process is a key factor for simulating their fate in the atmosphere. C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

685

F. Meng et al.

686

We have added modules that simulate PCBs and PCDD/Fs to the Community Multiscale Air Quality (CMAQ) model of the U.S. EPA. This allows us to take advantage of its capability to predict real-time spatially and temporally resolved aerosol concentrations. The added components are the Junge-Pankow adsorption and KOA absorption gas/particle partitioning models to CMAQ, as well as important gas phase chemical reactions involving the PCBs and PCDD/Fs. For gas phase PCBs we also added an atmosphere/water surface exchange model. We compared the performance of the resulting model systems with existing measurements. A total of 22 PCB congeners and the 17 most toxic PCDD/F congeners are included in this model system. Gas phase PCBs concentration

Modeled(pg/m3)

Measured(pg/m3)

PCB18

PCB52 PCB101

Fig. 1 Comparison of modelled gas-phase concentration of PCBs with measurements for January 2000–July 28, 2000

Fig. 2 Modelled surface level PCB gas phase concentrations after model spinup

Figure 1 is the comparison of the predicted results for PCB18, PCB52 and PCB101 with measurements from the Integrated Atmospheric Deposition Network (IADN) sites for the year 2002. The agreement is generally good; the model is in reasonable agreement with the measurement for PCB101 and PCB 52, but overestimates PCB18, the lighter (tri-chloro) congener, which is expected to be predominantly in the gas phase The modelling domain and typical predicted gas phase PCB concentrations are shown in Figure 2. PCB emissions are from isolated continental sources and this result shows significant atmospheric transport on a continental scale. As a result, there is considerable deposition of this material into the great Lakes and related watersheds. Acknowledgments We are pleased to acknowledge the financial support of Ontario Power Generation Inc., Natural Sciences and Engineering Research Council Canada, Canadian Ortech Environmental Inc. and the Province of Ontario.

P4.3 Development of a Quasi-Real-Time Forecasting System over Tokyo Masayuki Takigawa, Masanori Niwano, Hajime Akimoto and Masaaki Takahashi

Abstract We present an evaluation of the distribution of ozone over Kanto region, calculated by using a one-way nested global/regional air quality forecasting (AQF) system. This AQF system consists of the global chemistry-transport model (CTM) part and the regional CTM part. The global CTM part is based on CHASER, and the regional CTM part is based on WRF/Chem. An experimental phase of this model system began operation in July 2006 and has been providing 15-hour forecasts of the distribution of ozone concentrations over Kanto region four times in a day. The time-evolution and horizontal distribution of chemical species calculated by this AQF system were compared to ground–based observations.

The global CTM part is based on CHASER (Sudo et al., 2002). Spectral coefficients are triangularly truncated at wavenumber 42 (T42), equivalent to a horizontal grid spacing of about 2.8. The model has 32 vertical layers. The regional CTM part is based on WRF/Chem (Grell et al., 2005). Two-domains’ calculation has been done in the regional CTM part. The outer domain covers over Japan with 15 km horizontal resolution, and the inner domain covers over Kanto with 5 km resolution. The inner and outer domain in the regional CTM have 31 vertical layers up to 100 hPa. Anthropogenic emissions except automobiles over Japan are taken from JCAP (Japan Clean Air Program) with 1 × 1 km resolution (Murano, private communication), and anthropogenic emissions from automobiles over Japan are taken from EAgrid2000 (East Asia gridded emissions inventory) with 1km × 1km resolution (Kannari et al., 2007). Surface emissions over China, North Korea, and Korea are taken from REAS with 0.5q × 0.5q resolution (Ohara et al., 2007). The lateral boundary of chemical species in the regional CTM part is taken from the global CTM part. The lateral boundary is updated every 3–hourly, and linearly interpolated for each time step. The feedback from the regional CTM part to the global CTM part is not taken into account in the present study, i.e., the one–way nesting calculation has been done between the global and regional CTMs. A 15hour forecast has been produced four times in a day with a 8–10 hour lead time since August 2006. The initial condition of meteorological field for the regional CTM part is taken from Mesoscale Model (MSM) of Japan Meteorological Agency (JMA) for each forecast, and the initial condition of chemical species is taken from the model output driven by the analysis meteorology.


697

698

M. Takigawa et al.

To evaluate the model-calculated ozone, the surface ozone mixing ratio was compared to that observed at air quality monitoring stations. In August 2006, 251 stations observed ozone within the inner domain of the regional CTM. For comparison of temporal variation, hourly observed and modelled surface ozone mixing ratios in August 2006 are shown in Figure 1. Observed ozone exceeded 100 ppbv from 3 to 6 August at Hanyuu in Saitama prefecture (3610’ 28’’ N, 13933’ 21’’ E), which is downwind from the Tokyo Metropolitan area. The maximum value in the observation was 162 ppbv at 16Z on 3 August. The model reproduced the ozone maximum on 3 August well. The maximum simulated value was 137 ppbv in the model. The model also successfully captured the decrease from 3 to 7 August, but failed to show the rapid decrease on 8 August. Three typhoons (‘Maria’, ‘Somai’, and ‘Bopha’) existed in this period, and the difficulty of predicting the meteorological field may have led to overestimation of ozone on 8 August. The observed and modelled ozone exceeded 100 ppbv on 11, 13 August, and the model overestimated ozone mixing ratio on 19 August. Modelled ozone mixing ratio was 135 ppbv, whereas the observed ozone mixing ratio was 86 ppbv. Fig. 1 Hourly observed (solid line) and modelled (dashed) surface ozone mixing ratio in August 2006 at Hanyuu in Saitama Prefecture. Units are ppbv

References Grell GA, Peckham SE, Schmitz R., McKeen SA, Frost G, Skamarock WC, Eder B (2005) Fully coupled “online” chemistry within the WRF model, Atmos. Environ., 39, 6957–6975. Kannari A, Tonooka Y, Bada T, Murano K (2007) Development of multiplespecies 1km × 1km resolution hourly basis emissions inventory for Japan, Atmos. Environ., 41, 3428–3439, doi:10.1016/j.atmosenv.2006.12.015. Ohara T, Akimoto H, Kurokawa J, Horii N, Yamaji K, Yan X, Hayasaka T (2007) Asian emission inventory for anthropogenic emission sources during the period 1980–2020, submitted to Atmos. Chem. Phys. Dis. Sudo K, Takahashi M, Kurokawa J, Akimoto H (2002) CHASER: a global chemical model of the troposphere 1. Model description, J. Geophys. Res., 107, doi:10.1029/2001JD001113.

P4.6 Evaluation of an Operational Ensemble Prediction System for Ozone Concentrations over Belgium Using the CTM Chimere Andy Delcloo and Olivier Brasseur

Abstract In the framework of operational air quality forecasts in Belgium, an ensemble prediction system based on the chemical transport model (CTM) Chimere has been implemented. The Chimere model was forced by ECMWF meteorological fields and by the EMEP emission database. The simulation domain covers Western Europe with a spatial resolution of 0.5q. The objective of these ensemble simulations consists of evaluating the impact of uncertainties from emission and meteorological data on the simulated concentrations of pollutants. Such evaluations are important in the operational context, since they contribute to reduce the risk of false alarm and inappropriate broadcast of information to the public. Indeed the forecaster has at its disposal more information to better judge to what extent a change in one or more particular input parameters can influence the modelled pollutant concentrations. A first assessment of the ensemble prediction system has been performed for ozone forecasts during summertime. Currently, the Chimere model is run considering 13 different scenarios in which some variables influencing pollutant dispersion and ozone chemistry (e.g. temperature, wind velocity, cloud cover) are perturbed. The treatment of all these simulations allows defining a confidence interval around the concentrations simulated by the reference (i.e. without change in input parameters) simulation, which contributes also to improve the accuracy of the ozone forecast. Considering physical aspects, the ensemble prediction system contributes to point out – for each specific situation – the most sensitive input parameters that influences ozone concentrations. The observations, used to validate the Chimere model are deduced from an advanced interpolation method (kriging). This “RIO-algorithm”, developed by VITO (Hooyberghs et al., 2006), provides every hour an ozone value which takes into account the representativeness of the location the user is interested in. The ozone maps created with this algorithm have a spatial resolution of 5 km2. Since a CTM is very depended on the meteorological fields, it is interesting for the forecaster to know what will be the outcome of a sudden change in these meteo fields for the next day (D + 1). Therefore we roughly changed (in a first stage) 12 parameters: Cloud cover: (( 0 or 1 ) (CLO, CL1), +50% (Cp5), –50% (Cm5));


705

A. Delcloo, O. Brasseur

706

temperature: (+2ºC (Tp2), +4ºC (Tp4), –2ºC (Tm2), –4ºC (Tm4)) and wind speed: (–1 m/s (Wm1), –2 m/s (Wm2), +1 m/s (Wp1), +2 m/s (Wp2)). As an example results are shown for latitude 51 and longitude 4.5 (Figure 1 and Table 1). To validate the results, some general statistics (bias, rmse, correlation) are calculated on this different runs for the forecasted day + 1 (D + 1): Max EPS forecast for D+1 at LAT = 51.0 and LON = 4.5 140

120

Ozone (ppb)

100

80

60

40

20 01/Jun/2006

11/Jun/2006

21/Jun/2006

02/Jul/2006

12/Jul/2006

22/Jul/2006

01/Aug/2006

Fig. 1 Boxplot of maximum daily ozone concentrations (ppb) for the time period 01/06/2006– 31/07/2006. The 10, 25, 50, 75, 90 percentiles and the mean value (black cross) are shown. The red lines corresponds to the observed values with RIO, the green line represents the standard run of Chimere Table 1 Bias, correlation and rmse for the time period 01/06/2006–31/07/2006. STD

CL0

CL1

Cm5

Cp5

Tm2

Tm4

Tp2

Tp4

Wm1

Wm2

Wp1

Wp2 -10.6

Bias

-10.2

-2.3

-27.9

-7.9

-11.4

-9.0

-9.2

-8.8

-6.7

-9.8

-9.5

-10.4

Corr

0.85

0.80

0.65

0.85

0.85

0.86

0.86

0.87

0.87

0.85

0.85

0.85

0.85

Rmse

15.6

13.9

32.8

14.4

16.6

14.5

14.8

14.4

13.6

15.4

15.3

15.8

15.8

The results show that Chimere has a high negative bias for this particular time period. When we reduce the cloudcover, the bias and rmse improves, but the correlation declines. Table 1 shows that in general the changes in wind speed and temperature are less important. Only Tp4 shows significantly better scores for bias, correlation and rmse, compared with the standard run of Chimere.

References Hooyberghs J, Mensink C, Dumont G, Fierens F (2006) Spatial interpolation of ambient ozone concentrations from sparse monitoring points in Belgium, J. Environ. Monit., 8, 1129–1135.

P2.7 Evolution of the Ozone Episodes in Northern Iberia (Cantabric and Pyrenaic regions) Under West European Atlantic Blocking Anticyclones V. Valdenebro, G. Gangoiti, A. Albizuri, L. Alonso, M. Navazo, J.A. García and M.M. Millán

Abstract How main ozone episodes registered in the Basque Country (BC), at northern Iberia (Figure 1), affect the neighbouring areas and which are de mechanisms and pathways to export pollutants to these areas is analyzed. The blocking anticyclones over the British Islands, which are behind most of the ozone episodes in the BC, are related with regional and sub-continental transport of pollutants into this region, as we have documented in previous studies (Gangoiti et al., 2002 2006). The use of a coupled high resolution RAMS-HYPACT modelling system has allowed us to find various mechanisms and pathways for the importing of pollutants into the BC (Figure 3a, b). Now we analyze how these episodes affect the neighboring areas (Cantabric and Pyrenaic regions), which are the mechanisms and pathways to export pollutants from the BC, and how the episodes dissipate. To fulfil these tasks the AirBase ozone data corresponding to the Cantabric and Pirenaic regions are analized and the previous RAMS-HYPACT simulations have been extended in time. 55N 50N

GRID #1

GRID #2

46N

1

44N

GRID #2

45N

42N

40N

GRID #3

BAY OF BISCAY 3 4 6 8 6 7

2

9

11 10

40N

35N

12W 10W 8W

30N

0

20W 15W 10W 5W

0

5E 10E 15E

6W

4W

2W

0

2E

4E

6E

200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400

HEIGHT (m)

Fig. 1 Topographic map of the three domains used by RAMS-HYPACT, with the selected sources of pollution (left), and the location of ozone monitors (right). The BC is inside grid #3

Our results show that this type of ozone episodes is recorded all along the Cantabrian coast, where they begin in a quasi simultaneous manner and last similarly. Maximum concentrations are registered earlier in the W zone (during the accumulation phase) and later in the SW of France (during the dissipation phase). Pollutants from the BC are transported towards the Atlantic Ocean and coast of Portugal following two main pathways (along the Duero Valley and along the C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

671

672

V. Valdenebro et al.

northern coast of Iberia) at the beginning of the episodes (Figure 3a); then, they can continue their transport along the coast of Portugal towards the African continent. At the final stage of the episodes (Figure 3c), there is a massive transport towards southern France and along the Ebro Valley, towards the Mediterranean Sea.

Fig. 2 Time sequences of the ozone monitors concentrations (Figure 1) during the June 2001 episode

Fig. 3 Main pathways during the episodes: (a) accumulation, (b) peak, and (c) dissipation phase

Acknowledgments The authors wish to thank the Basque Government for their supply of data and the Spanish Ministry of Science and Technology for financing.

References Gangoiti G, Alonso L, Navazo M, Albizuri A, Pérez-Landa G, Matabuena M, Valdenebro V, Maruri M, García JA, Millán MM (2002) Regional transport of pollutants over de Bay of Biscay: analysis of an ozone episode under a blocking anticyclone in west-central Europe, Atmospheric Environment, 36, 1349–1361. Gangoiti G, Albizuri A, Alonso L, Navazo M, Matabuena M, Valdenebro V, García JA, Millán MM (2006) Sub-continental transport mechanisms and pathways during two ozone episodes in northern Spain, Atmospheric Chemistry and Physics, 6, 1469–1484.

P1. Local and urban scale modelling

P1.1 Finite Volume Microscale Air-Flow Modelling Using the Immersed Boundary Method V. Fuka and J. Brechler

Abstract This contribution describes results of computation of a turbulent flow over a square cylinder by 2D large eddy simulation. Solid wall boundary conditions were described by the immersed boundary method. We choose this example as a part of validation of a CFD model we are developing for flows in geometrically complex (namely urban) areas. Square cylinder is here used as a prototype of bluff body, instead of a real 3D building, which will be computed later. For time discretisation of incompressible Navier-Stokes equations, we used the fractional step method. For spatial discretization, the finite volume method was used, but for advective fluxes, we used central high-resolution scheme of Kurganov and Tadmor (2000). The turbulence was treated in the context of ILES (implicit large eddy simulation (Drikakis, 2003)) using the Kurganov-Tadmor method. The complex geometry on Cartesian grid was described by the immersed boundary method (Kim et al., 2001). Due to the computational resources and the present state of our model we performed the calculation in 2D. Although the turbulence is a inherently a 3D phenomenon, according to Bouris and Bergeles (1999) the 2D computation can capture most of important features of the quasi-two-dimensional flow. In addition, in 2D, one can use grid with better resolution. In Figure 1 is a snapshot of the vorticity field at time t = 200. The vortex shedding is clearly visible. Strouhal number of vortex shedding was 0.12, which is slightly less than 0.13–0.14 reported by Bouris and Bergeles (1999). Averaged horizontal velocity profile at the centerline is in Figure 2. Recirculation length was 0.6, which is less than experimental value, but it is consistent with 2D calculations of Bouris and Bergeles (1999).


651

V. Fuka, J. Brechler

652

Fig. 1 Instant vorticity field at t = 200

Fig. 2 Averaged velocity profile at the centerline, solid line is the present study, other lines are other numerical results and symbols are experimental values

Acknowledgments This research has been supported by the Grant Agency of the Czech Academy of Sciences, grant no. T400300414 and by the Grant Agency of the Czech Republic, grant no. 205/06/0727.

References Bouris B, Bergeles G (1999) 2D LES of vortex shedding from a square cylinder, J. Wind Engrg. Indus. Aerodynam. 80, 31. Drikakis D (2003) Advances in turbulent flow computations using high-resolution methods, Prog. Aerospace Sci. 39, 405. Kim J, Kim D, Choi H (2001) An immersed-boundary finite-volume method for simulations of flow in complex geometries, J. Comput. Phys. 171, 132. Kurganov A, Tadmor E (2000) New high-resolution central schemes for nonlinear conservation laws and convection-diffusion equations, J. Comput. Phys. 160, 241.

P2.8 High Temporal Resolution Measurements and Numerical Simulation of Ozone Precursors in a Rural Background M. Navazo, N. Durana, L. Alonso, J. Iza, J. A, García, J.L. Ilardia, G. Gangoiti and M. De Blas

Abstract High-resolution numerical modelling results – using RAMS and HYPACT – of an ozone precursors episode, detected on a rural background area in Northern Iberia, are shown. The episode was identified by data analysis of a continuous VOC measurement system. Simulation results show that, when low temporal and spatial resolutions are used, the origin of polluted air masses affecting targets at low heights may be wrongly interpreted. A very complete database of individual non-methane hydrocarbon measurements with high temporal resolution (hourly) in a rural background atmosphere was prepared between January 2003 and December 2005. That kind of database can be used for biogenic NMHC characterization as well as for the identification of the transport and impact of anthropogenic NMHC on rural areas (Durana et al., 2006). The measurement system operated continuously in the centre of the Valderejo Natural Park in northern Iberia. Data coverage was higher than 70% for a total of 59 VOC of both anthropogenic and biogenic origin, with detection limits in the range of pptv. This database was used to describe the behaviour of these compounds, in order to identify the chemical transformations and external impacts arriving to the sampling site, highly representative of a rural background atmosphere.

Fig. 1 Concentration time series of selected aromatic VOCs between August 1 and 7, 2003

A coupled RAMS-HYPACT modelling system, applied over the target region with a temporal and spatial resolution of 1 hour, and 3 × 3 km, respectively (Gangoiti et al., 2006), was used for the duration of the episode, when high VOC levels were detected at night (Figure 1). The Toluene/Benzene ratio was characteristic of an old C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

673

M. Navazo et al.

674

air mass, which could not get to the site unless long-range transport mechanisms were involved (Figures 2 and 3). Fig. 2 Results from the meteorological and dispersion simulation (using RAMS & HYPACT models) for August 5, 2003 at 0300 UTC

Fig. 3 Corresponding N-S (left) and E-W (right) transects to Figure 2

Acknowledgments To the personnel from the Valderejo Natural Park Center, for their friendliness and logistic support; to the Basque Government for his support, and to the Spanish MEC for financing the projects MAMECOVA and TRAMA

References Gangoiti G, Albizuri A, Alonso L, Navazo M, Matabuena M, Valdenebro V, García JA, Millan MM (2006) Sub-Continental transport mechanisms and pathways during two ozone episodes in northern Spain, Atmos. Chem. Phys., 6, 1469–1484. Durana, N, Navazo M, Gómez MC, Alonso L, García JA, Ilardia JL, Gangoiti G, Iza J (2006) Long term hourly measurement of 62 nonmethane hydrocarbons in an urban area: Main results and contribution of non-traffic sources, Atmos. Environ., 40, 2860–2872.

P2.12 High Time and Space Resolution Ozone Modelling in Regional Air Quality Management of a Complex Mountain Area Using Calgrid 2.44 Carlo Trozzi, Silvio Villa and Enzo Piscitello

Abstract In this paper is reported the study of ozone pollution in Autonomous Province of Trento referred to year 2004, with a suite of meteorological (MM5 and Calmet) and photochemical (Calgrid) models in a high detailed space and time resolution over long time period. Trento province occupies a surface of 6.207 km2 (2.9% of national territory); 70% of the surface is located over 1,000 m above sea level. The resident people in Trento province is estimated as many as 450,000. Every year a number of as many as 28 million of tourists involves an additional pressure on the territory. Trento area is a very particular area (a mountain area in Alps) in which the air quality protecttion is a primary goal. CALGRID model (Yamartino et al., 1992) requires input meteorological parameters for every 1 × 1 km cell of the geographical domain. This task is accomplished by Calmet pre-processor (Scire et al., 1992), whose input was the terrain geographical features, land use, various boundary layer specific variables, hourly meteorological parameters coming from the regional network of agrometeorological stations, and upper atmosphere parameters directly modeled with the MM5 model. The non-hydrostatic MM5 model (National Center for Atmospheric Research, 2005) was set up for a coarse domain formed by cells 15 × 15 km. wide with a nest of 5 × 5 km. cells enclosing the entire Calgrid domain. The emission input for Calgrid was based on a highly detailed emission inventory following a bottom-up approach. Emissions were evaluated for 55 point sources, 416 line sources (6 highway sections and 410 extraurban road sections) and 225 area sources (municipality). Line and area sources were assigned to 1 × 1 km grid using proxy variables (mainly derived from land use maps). Emissions were disaggregated on hourly basis using surrogate variables and estimated for 150 Corinair activities. Volatile Organic Compounds emissions were speciated in SAROAD classes for input in chemical mechanism (Carter, 1990). A specific information system was used. A slightly code modified Calgrid v.2.44 model run, with case specific parameters choice and SAPRC97 chemical mechanism, produced interesting results in determining ozone concentrations, in particular for mountainous and high forested terrain, as shown in Figure 1. A subsequent study focused on medium-high populated C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

681

C. Trozzi et al.

682

places like cities in the numerous valleys included in Trento Province and legislative indexes were calculated on that area. The AOT40 index was also calculated, showing a high percentage (75% approx.) of cells exceeding this index.

190 180 170 160 150 140 130 120 110 100 90 80 70 60 50 40 30

Fig. 1 Average ozone concentration from May 1 to August 31, 2004 expressed in µg/m³ as calculated by Calgrid model

Acknowledgments The work was conducted on behalf of Autonomous Province of Trento Environmental Protection Agency.

References Carter WPL (1990) A detailed mechanism for the gas-phase atmospheric reactions of organic compounds. Atmos. Environ., 24A, 481–518. Scire JS, Insley EM, Yamartino RJ, Fernau ME (1995) A User’s Guide for the CALMET Meteorological Model. Yamartino RJ, Scire JS, Hanna SR, Carmichael GR, Chang YS (1992) The CALGRID mesoscale photochemical grid model - I. Model formulation. Atmos. Environ., 26A, 1493–1512. National Center for Atmospheric Research (2005) PSU/NCAR Mesoscale Modeling System. Tutorial Class Notes and User’s Guide, MM5 Modeling System Version 3, January 2005.

P7.3 Intake Fraction for Benzene Traffic Emissions in Helsinki Joana Soares, Ari Karppinen, Leena Kangas, Matti Jantunen and Jaakko Kukkonen

Abstract Benzene (Bz) is well known for its haema and genotoxicity and the carcinogenic effect associated with long time exposure. In urban environment, traffic is an important source for ambient air Bz concentrations. In order to quantify emission-to-intake relationships, intake fraction (iF) was defined as the integrated incremental intake of Bz released from a source (or source category) and summed over all exposed individuals during a given exposure time, per unit of emitted pollutant (Bennet et al., 2000). iF takes into account the dispersion of pollutants, locations and activity of population, and human breathing rates. The calculated iF for Bz is directly applicable to any other inert substance emitted by the traffic, e.g. CO, NOX, so the calculations also provide a ready-to-use tool for health effects studies concerning other pollutants and emission scenarios. This study calculates the spatial distribution of average benzene iF for Helsinki Metropolitan Area (HMA) using the EXPAND model (Kousa et al., 2002). The spatial Bz concentration distributions were obtained by using dispersion models: CAR-FMI (Karppinen et al., 2000) and OSPM (Berkowicz, 2000). A constant breathing rate of 1 m3/day was considered. The EXPAND results for 2000 are shown in Figure 1.

Fig. 1 Spatial distribution of total intake fraction for benzene from mobile sources in 2000

The total iF for HMA is 2.8 × 10-5 with higher values concentrated in residential/ commercial areas, ranging between 10-10 and 10-7, where people spend most of their time. The partial intake fraction (iFi) due to exposure in traffic is 3.1 × 10-6. A study conducted in California (Marshall et al., 2003), shows an iF for Bz in the same order of magnitude: 3.3 × 10-5. Moreover, a total iF of 3.7 × 10-5 was calculated for C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

719

J. Soares et al.

720

Hämeentie Street., in Helsinki, based on street canyon dispersion calculations, and activity and number of inhabitants and workers in this area (Table 1). Table 1 also shows iFi calculated for this sample of the total affected population. Table 1 Intake fraction by inhalation in Hämeentie Street (2000). Groups

Breathing rate (m3/day)

No. of people

Time of exposure (day)

iFi

Inhabitants

1

176

0.56

1.6E-05

1

113

0.37

2.0E-05

1

20,380

1.04E-04

1.1E-06 3.7E-05

Working and costumers In traffic Total iF

It is clearly shown that people travelling through the street canyon are order of magnitude less exposed than the people living and working in that area, due to the short amount of time they are spending in the street. This was also show in the HMA calculations. So if the time spent, e.g., commuting or indoors would be the same, in-traffic exposure would be much more relevant. The iF spatial variation, in particular the detailed calculations for the smallest domain, demonstrates clearly how crucial it is to have access to detailed information on traffic patterns and locations and activities of the people in order to get a reliable estimate on the real burden of pollutants on human exposure and health. Acknowledgments The financial support from CEFIC-LRI is gratefully acknowledge

References Bennet D, Margni M, McKone T, Jolliet O (2002) Intake fractions for multimedis pollutants: a tool for life cycle analysis and comparative risk assessment. Risk Anal. 22, 819–1033. Berkowicz R (2000) A simple model for urban background pollution. Environ. Monit. Assess. 65, 259–267. Karppinen A, Kukkonen J, Elolähde T, Konttinen M, Koskentalo T (2000) A modelling system for predicting urban air pollution: comparison of model predictions with the data of an urban measurement network in Helsinki. Atmos. Environ. 34, 3735–3743. Kousa A, Kukkonen J, Karppinen A, Aarnio P, Koskentalo T (2002) A model for evaluating the population exposure to ambient air pollution in an urban area. Atmos. Environ. 36, 2109–2119. Marshall JD, Teoh SK, Nazaroff WW (2005) Intake fraction of nonreactive vehicle emissions in US urban areas. Atmos Environ 39(7), 1363–1371.

P3.2 Improving Emission Inventory in Lithuania Vidmantas Ulevicius, Vytautas Vebra, Kestutis Senuta and Svetlana Bycenkiene

Abstract It was developed emission inventory and estimation system for raw data about an area and point emission sources in Lithuania: collection, emission estimation, analysis, reporting and public information. Developed emission inventtory and estimation system was used for emission estimation, sectorial and spatial analysis in EMEP grid, reports to LRTAP and NEC preparation in the year 2005. Emission inventory and estimation system’s public information subsystem exposes an emission factors’ browser, a spatial emission maps’ browser and the road transport emission calculator on the website of Institute of Physics. Developed emission inventory and estimation system is described below. Five main subsystems can be distinguished in emission inventory and estimation system according to functionality: a database management system, an emission raw data collection system, an emission estimation and analysis system, an emission data export system and emission data public information system (Figure 1). We use MySQL the database management system for emission data storage, data querying and data analysis platform; we are going to migrate to PostgreSQL database management system with PostGIS extension as more featured geographic data analysis platform. The main tasks of emission raw data collection system are: (1) import emission related data to emission inventory database from various formats; (2) provide interactive forms for data entering. We used the phpMyAdmin, Microsoft Access and MySQL Query Browser for completing these tasks. The main tasks of emission estimation and analysis system are: (1) process raw emission related data and build emission inventory; (2) perform necessary emission inventory analysis and output the results of analysis in form of tables. We have a prepared set of SQL scripts which estimate road transport emission, build emission inventory from raw emission-related data, perform key sources analysis, calculate emission in EMEP grid, perform other sectorial and spatial analysis. The road transport estimation subsystem was developed according to COPERT III methodology (Ntziachristos et al., 2000) by creating the set of SQL queries – this increased emission inventory and estimation system’s effectiveness and performance. There was also developed an emission factors’ table in emission inventory database following the national, EMEP/CORINAIR and some other methodology used by other countries emission estimation. Main task of emission data export system is export emission data stored in emission inventory database for data exchange with air quality modelling systems or other systems. The emission data export system contains data format C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

691

692

V. Ulevicius et al.

and destination specific modules. Currently we have an executive module which exports emission inventory data to the MM5-SMOKE-CMAQ air quality modelling system. The main tasks of emission data public information system are: (1) visually represent emission data on the web site available for public; (2) expose official emission inventory data browser on the web site available for public; (3) expose web-based emission calculators for personal usage on the web site available for public. The emission data public information system is built on Apache with the PHP extension platform. Currently the emission data public information system contains PHP scripts which expose an emission map’s browser, emission factors’ browser and JavaScript based road transport emission calculator on the website of Institute of Physics: http://www.fi.lt/afch.

Fig. 1 Emission inventory and estimation system

Acknowledgments This research was supported by the Agency for International Science and Technology Development Programmes in Lithuania under EUREKA WEBAIR project. The authors gratefully thank for this assistance.

References Ntziachristos L, Samaras Z, ETC/AEM (2000) COPERT III Computer programme to calculate emissions from road transport. Methodology and emission factors (Version 2.1).

P1.4 Inter-Comparison of Gaussian Plume, Street Canyon and CFD Models for Predicting Ambient Concentrations of NOx and NO2 at Urban Road Junctions Richard Hill, Peter Jenkinson and Emma Lutman

Abstract Predictions of Gaussian Plume, Street Canyon and Computational Fluid Dynamics models were compared at five road traffic junctions. CFD may not necessarily reduce uncertainties in air quality assessments, though may be useful for identifying hotspot locations. Technical guidance for the assessment of local air quality under the UK National Air Quality Strategy (NAQS) has identified that previous studies of road-trafficrelated air pollution often may not have considered the effects of junctions adequately (LAQM-TG3, 2003). The influence of road traffic junctions in urban areas is particularly significant, as these areas often have enhanced rates of emission due to traffic congestion and rates of atmospheric dispersion are often reduced, due to the effects of the surrounding buildings on wind flows (Vardoulakis et al., 2003). The combination of these two effects may result in such areas being identified as pollution hotspots through monitoring studies. Predictions of Gaussian Plume (Airviro) and Street Canyon (Aeolius) models were compared with the predictions of a Computational Fluid Dynamics (CFD) model capable of including the complex topographies that occur at urban intersections (Panache). Five case study sites were identified in order to determine the influence of different modelling techniques on the prediction of air concentrations for Local Air Quality Management. The simulations conducted in this project identified that urban buildings affect the predictions of local NOx concentrations significantly. Where streets are flanked by tall buildings these effects have long been recognised and Street Canyon models are typically used for modelling such situations. The modelling assessments for Coventry and Leicester, areas that are similar to the geometries for which Street Canyon models were developed, showed that these models provide the most suitable tools for predicting air concentrations, with similar patterns of dispersion being predicted by the CFD model. Interestingly, realistic predictions were also obtained using the Aeolius model at the Sheffield and Leeds case study sites, areas that have very different topographies to those typical of a street canyon. Model predictions from the CFD code for the Birmingham, Leeds and Sheffield sites showed that particular air concentration hotspots occurred in the wakes of C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

657

R. Hill et al.

658

large buildings and at building faces along the roadside. CFD model runtimes for these simulations and the occasional significant overprediction of concentrations, suggest that the operational use of CFD for NAQS modelling may be prohibitively time-consuming (Table 1). Moreover, CFD simulations may not necessarily reduce the uncertainties in assessments of NO2, particularly when contributions from background sources, the accuracy of emission inventories and the conversion of NOx to NO2 are considered. However, it may be appropriate to consider the types of feature shown by the CFD model to result in local pollution hotspots when developing monitoring strategies in such areas.

Fig. 1 Comparison of model predictions and measurement data (Leeds) Table 1 Statistical comparison of observed background corrected NOx concentrations with those predicted by Airviro (AI), Aeolius (AE) and Panache (CFD). Location Birmingham Coventry Leeds Leicester Sheffield

Fraction within a factor of 2 AI AE CFD 0.50 – 0.63 0.00 1.00 0.63 0.13 0.25 0.25 0.13 0.63 0.38 0.63 0.50 0.13

Normalised mean square error AI AE CFD 1.33 – 1.16 22.95 0.19 0.26 3.02 1.36 1.96 6.21 0.66 16.55 0.97 1.04 11.56

Mean bias AI 0.44 0.04 0.40 0.15 0.52

AE – 0.69 0.59 0.66 0.53

CFD 0.49 0.64 1.07 8.03 7.38

Acknowledgments The authors are grateful to Birmingham, Coventry, Leeds, Leicester and Sheffield City Councils and Defra for funding this work.

References LAQM-TG3 (2003) Part IV of the Environment Act 1995, Technical Guidance. Department for Environment, Food and Rural Affair, London. Vardoulakis S, Fisher BEA, Pericleous K, Gonzalez-Flesca N (2003) Modelling air quality in street canyons: a review. Atmospheric Environment 37(2),155–182.

P2.11 Lake Breezes in Southern Ontario: Observations, Models and Impacts on Air Quality David Flagg, Jeff Brook, David Sills, Paul Makar, Peter Taylor, Geoff Harris, Robert McLaren and Patrick King

Abstract The southwestern Ontario, Canada Border Air Quality Strategy (BAQS-Met) field campaign of summer 2007 investigates the chemical and dynamical influence of lake breeze fronts and urban environments on local and transported pollutant emissions. The presence of both local and long-range transported emissions predisposes southwestern Ontario (ON), Canada to compromised air quality (AQ). Surrounded on three sides by the Great Lakes, frequent lake-breeze fronts (LBFs) complicate this region’s air chemistry and dynamics and generate a unique challenge for air quality modelling. LBFs also contribute to local convection, initiating thunderstorms and potentially enhancing vertical transport of pollutants. The local emissions derive from multiple industrial sites in both the U.S. and Canada, including the Detroit-Windsor metropolitan area. The latter can provoke an urban heat island (UHI) circulation that interacts with LBFs, further complicating the modelling. The Border Air Quality Strategy (BAQS-Met) field study in air quality and meteorology will address the need for improved understanding of the character of this border region (U.S.–Canada). 84°

82°

83° US

43°

122

81° LAKE

LAKE ST.

111

112

LAKE

1.2.5

1.2.7

1.2.6

1.2.4

LEGEND

42° 1.2.3

121

US

Existing surface i ARWIS weather i or U.S. station OME EC supersite met i University supersite

1 2 OFig. 1 BAQS-Met domain


679

680

D. Flagg et al.

BAQS-Met is a joint study of Environment Canada (EC), the Ontario Ministry of the Environment (OME), York University, the University of Toronto (UT), University of Western Ontario (UWO) and the U.S. National Weather Service (NWS). The field study consists of a four-week intensive campaign from mid-June through mid-July 2007 and summer-long enhanced meso network (mesonet) measurements from June through August. Both campaigns concentrate on southwestern ON, adjacent to the U.S. border (see Figure 1). BAQS-Met will assess the relative importance of the chemical and dynamical influences of the Great Lakes on regional air quality. Analysis of measurements of LBF-induced vertical motion will contribute to modelling of transport forecasts, air chemistry and thunderstorm initiation. BAQS-Met will also investigate model performance of physical processes over the lakes and its contribution to regional AQ forecasting. Additional studies include modelling the Detroit-Windsor urban boundary layer to examine the role of an urban, coastal environment in tracer transport as well as a review of how threedimensional variational data assimilation (3D-Var) can improve model performance in the boundary layer. The four-week intensive campaign includes an extensive array of measurements from both fixed and mobile surface sites (over land and water) and airborne measurements. So-called ‘supersites’, stations at Bear Creek and Harrow (see Figure 1) each host measurements of gas and particles in addition to surface meteorological variables, including (not all at each site): O3, SO2, CO, NO/NOx, VOC, NH3, Real NO, NO2, NOz, PM1, PM2.5 PM10 and black carbon, with many at up to 1-minute resolution. Deployment of Environment Canada’s Canadian Regional and Urban Investigation System for Environmental Research (CRUISER), based out of Windsor, and a commercial ferry operating on Lake Erie provide mobile surface gas and particle measurements. The Rapid Acquisition SCanning Aerosol Lidar (RASCAL), stationed at Ridgetown, provides lidar measurements. Both OME and UT provide additional mobile gas and particle measurements. Meteorological measurements include a mesonet with 14 stations measuring wind speed and direction, temperature, relative humidity, passive NO2/SO2, O3 and NH3 (in addition to other sites) and, at 10 of the 14 sites, active O3 and PM2.5 measurements. These supplement existing surface stations (EC, NWS, OME), Advanced Road Weather Information System (ARWIS) stations and buoys. Tethersondes at Ridgetown and Windsor and an ozonesonde and VHF wind profiler at Harrow provide vertical profiles, complementing transects from the Canada National Re-search Council (NRC) Twin Otter aircraft, which provides 30 hours of measurement. Flight paths include multi-level transects and spirals over Lake St. Clair, Detroit-Windsor and lake shores in daytime and nocturnal boundary layers as well as single level inter-lake transects. Acknowledgments Funding provided by the Ontario Ministry of the Environment Transboundary Research Program.

P2. Regional and intercontinental modelling

P2.1 Local to Regional Dilution and Transformation Processes of the Emissions from Road Transport Dimiter Syrakov, Kostadin Ganev, Reneta Dimitrova, Angelina Todorova, Maria Prodanova and Nikolai Miloshev

Abstract The objective of the present work is to study in detail the dilution processes and chemical transformations of the generated by road transport from the local scale to the scale of the global models and on deriving some conclusions about the key parameters, which quantify the local dilution and transformation processes impact on larger scale pollution characteristics. It is expected the further development of the current work to give some clues for specification of the “effective emission indices” linking emission inventories to the emissions to be used as input in large scale models. The US EPA Models-3 system (Grell et al., 1994; Byun et al., 1998; Byun and Ching, 1999) was chosen as a modelling tool. The simulations were consecutively carried out in three nested domains. The innermost domain (D3), treated with a resolution of 10 km includes a region with very intensive road transport – the city of London and its “footprint”. The CMAQ “Integrated Process Rate Analysis” utility is used to differentiate the contribution of different dynamic and chemical processes which form the pollution characteristics in the region of interest. Two sets of emission data were used in the present study: (1) the EMEP data was used for all the countries except the UK; (2) for the UK data from the National Atmospheric Emissions Inventory, with a 1 km resolution was used. The biogenic VOC emissions were estimated, using a simple scheme recommended by Lübkert and Schöp (1989). The meteorological background input was taken from US NCEP Global Analyses data. The simulations in D3 were carried out for January and August 2002–2006 for the following emission scenarios: (1) all the emissions (detailed inventory); (2) emissions from the road transport excluded (detailed inventory); (3) all the emissions, averaged over D3; (4) emissions averaged over D3, but emissions (averaged) from the road transport excluded. The combined analysis of these scenarios will make it possible (hopefully) to clarify the role of different dynamic and chemical processes which determine the pollution from road transport pattern and time evolution. Some C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

659

660

D. Syrakov et al.

conclusions about the role of the road traffic emission inventories spatial resolution on the simulated fields and the local to regional scale interaction can also be made. The numerical experiments performed produced a huge volume of information, which have to be carefully analysed and generalized so that some final conclusions, concerning not only clarification of local scale processes of dilution and chemical transformation but also how to account for them in large scale CTMs could be made. Comprehensive survey of the output from all the numerical experiments will be possible only if some integral quantities, characterising the dilution and transformation processes within D3 domain are introduced. The conclusions that can be made at this stage of the studies are: 1. The effect of the road transport emissions is well displayed in both the concentration and process analysis fields. 2. The contributions of different processes have very complex spatial/temporal behavior and variability. 3. Even horizontally/temporally averaged process contributions may be sensitive to emission resolution. Acknowledgments The present work is supported by EC through 6FP projects ACCENT (GOCE-CT-2002-500337) and QUANTIFY (GOGE-003893), and COST Action 728.

References Byun D, Young J, Gipson G, Godowitch J, Binkowski FS, Roselle S, Benjey B, Pleim J, Ching J, Novak J, Coats C, Odman T, Hanna A., Alapaty K, Mathur R, McHenry J, Shankar U, Fine S, Xiu A, Jang C (1998) Description of the Models-3 Community Multiscale Air Quality (CMAQ) Modeling System, 10th Joint Conference on the Applications of Air Pollution Meteorology with the A&WMA, 11–16 January 1998, Phoenix, Arizona. Byun D, Ching J (1999) Science Algorithms of the EPA Models-3 Community Multiscale Air Quality (CMAQ) Modeling System. EPA Report 600/R-99/030, Washington, DC. Grell GA, Dudhia J, Stauffer DR (1994) A Description of the Fifth Generation Penn State/NCAR Mesoscale Model (MM5). NCAR Technical Note, NCAR TN-398-STR, 138. Lübkert B, Schöp W (1989) A model to calculate natural VOC emissions from forests in Europe. Report WP-89-082, IIASA, Laxenburg, Austria.

P2.6 Modelling of Atmospheric Transport of POPs at the European Scale with a 3D Dynamical Model Polair3D-POP Solen Quéguiner and Luc Musson-Genon

Abstract POLAIR3D is an Eulerian 3D atmospheric model designed to handle a wide range of applications. Thus it has been used for passive transport, for impact at European scale, for photochemistry and mercury chemistry. A new version of the model, POLAIR3D-POP, developed to study the atmospheric transport and environmental fate of persistent organic pollutant (POPs) is presented. During the atmospheric transport, POPs can be deposited and re-volatilised to the atmosphere several times before the final destination. A description of the air-surface exchange processes is included in the model to account for this multi-hop transport. The greatest families of pollutants are studied (HAP, dioxins and furans, PCB, lindane and HCB). The results from a model simulation showing the atmospheric transport for the year 2001 at the European scale are presented and evaluated against measurements from EMEP. The Persistent Organic Pollutants (POPs) are organic carbon-based chemical substances. They possess a particular combination of physical and chemical properties such that, once released into the environment, they remain intact for exceptionally long periods of time. They become widely distributed throughout the environment as a result of natural processes involving soil, water and air. They are found at higher concentrations in the food chain and are toxic to both humans and wildlife. The aim of this study is to present a 3D dynamical model, POLAIR3D-POP, describing the atmospheric transport and environmental fate of POPs at the European scale. POLAIR3D-POP is handled basic physical processes by integrating in time the Eq. (1) (Mallet et al., 2007):

wci wt div(ci .V )

div( K .ci ) F i (c) Di Ei

(1)

where i labels a chemical species, c is a vector of chemical concentrations, V is the wind vector, K is the diffusion matrix, F i combines production and loss terms of chemical reactions, Di is the deposition term (dry deposition and wet depositionscavenging), Ei stands for the emissions (surface and volumic emissions).


669

S. Quéguiner, L. Genon

670

The change in the POP concentration, ca , in the atmospheric layers with time is described in the model by the Eq. (2):

wc a

wt

1

za

Femis Fexc Oca k air ca

(2)

where za is the thickness of the layer, Femis is the emission, Fexc is the air-surface gas exchange flux, O is the wet deposition and kair is the chemical transformation rate in the air. The soil module POLAIR3D-POP has based on that of the DEHM-POP (Hansen et al., 2004) and that of the MSCE-POP (Gusev et al., 2005). We have considered six types of land use coverage (barren soil, legume/fruit, evergreen forest, deciduous forest, grassland/cereals, and water bodies). The change in the POP concentration in different underlying surfaces, cs , with time is described by the Eq. (3):

wc s

wt

1

zs

F

wet

Fexc Frun _ through k soil c s

(3)

where Frun_through is the amount of chemical running out with the excess water through the bottom soil layer (modelled just for the barren soil). Meteorological data and emissions are used as input to the model simulations. The meteorological data comes from the ECMWF and the emission input are provided by the EMEP database (EMEP, www.msceast.org). We present the model outputs for the atmospheric concentrations, the deposition fluxes and the concentrations in the different types of underlying surface for the HAPs, dioxins and furans, PCBs, lindane and HCB. Measurements from 13 stations are available and the comparison to measurements is presented for the year 2001.

References Gusev A, Mantseva E, Shatalov V, Strukov B (2005) Regional Multicompartment Model MSCE-POP. EMEP/MSC-E, Technical Report 5/2005. Hansen KM, Christensen JH, Brandt J, Frohn LM, Geels C (2004) Modelling atmospheric transport of J-hexachlorocyclohexane in the Northern Hemisphere with a 3D dynamical model: DEHM-POP, Atmos. Chem. Phys., 4, 1125–1137. Mallet V, Quélo D, Sportisse B, Ahmed de Biasi M, Debry E, Korsakissok I, Wu L, Roustan Y, Sartelet K, Tombette M, Foudhil H (2007) Technical Note: The air quality modeling system Polyphemus. Atmos. Chem. Phys., 7, 5479–5487

P2.10 Modelling the Impact of Best Available Techniques for Industrial Emissions Control in Air Quality: Setting Up Inventories and Establishing Projections R. Rodriguez, P. Maceira, J.A. Souto, J. Casares, A. Sáez and M. Costoya

Abstract Strategies for industrial emissions control mainly depend mainly on the best available techniques, BATs, their economical feasibility, and their impact over air quality. A bottom-up comprehensive methodology for industrial emissions inventories and projections is presented and applied to the estimation of emissions for different scenarios in Galicia (NW Spain). Starting from a 2001 emissions scenario, the application of different BATs was considered, and emissions projections were obtained for 2010. Finally, SMOKE (CEP, 2003) was applied to integrate the emissions inventories in air quality modelling. From the results obtained, a strong SO2 emissions reduction will be achieved, and a slight reduction of NOx emissions in the utilities sector is expected. Changes (either increments or reductions) on other pollutants emissions are feasible. An emissions inventory is a strategic tool for environmental management. In this work, a bottom-up comprehensive methodology (based in Source Classification Codes; U.S. EPA, 2004) for establishing industrial emissions inventories, is presented. The proposed methodology is applied to the estimation of industrial emissions for different scenarios in Galicia (NW Spain) using SMOKE (CEP, 2003), which is also used for a subsequent air quality evaluation. The process for establishing emissions inventories and the projections derived from the scenarios proposed is divided in three stages: (a) data acquisition and classification of emissions calculation and estimation methods; (b) structure of the emissions inventory, for multiple applications; and (c) emissions projections, based both in industrial growth rates and technological changes. This methodology was tested in the industrial emissions inventory of Galicia (NW of Spain) (Casares et al., 2005) covering 370 major industrial plants, selected for their potential atmospheric emissions from around 3,000 installations. Annual emissions inventories for 1999, 2000 and 2001 years were obtained, covering as pollutants: SO2, NOx, CO, PM10, CO2, CH4, N2O, NMVOC, PAH, benzene, Cl-HCl, F-HF, NH3, PFCs (CF4 and C2F6) and heavy metals (As, Cd, Cr, Cu, Hg, Ni, Pb, Zn, and their compounds). Finally, the inventory for 2001 was


677

R. Rodriguez et al.

678

adopted as a reference, and different emissions projections for 2010 year were obtained. Figure 1 shows results of SO2 emissions by sector for 2001 and 2010 (projection); based on existing regulations (EU Directive on large coal combustion plants), and economic growth rates expected (Mantzos et al., 2003). 0,00 0,0 4600 ,0 00 ,0 00 00 0,00 0,0 3550 ,0 00 ,0 00

kg/y ear

4 0,00 0,0 00

3 0,00 0,0 00

20 01

20 10

2 0,00 0,0 00

1 0,00 0,0 00

Oth ers

Wa ste trea teme nt

Po we r pl ants

Co gen era tio n p la nts

D iv ers e ma ch in ery

Tran sp ort e qu ip men t

M etal s

Me tal pro du cts

Lea the r

C era mi c, co nc rete …

Pl as tic s an d ru bb ers

Pe troc he mi cal in du stry

Pri nti ng

Ch emi ca l i nd us try

Fu rni ture

Pu lp an d p ap er in du stry

Tex ti l fa cto rie s

Wo od pro ce ssi ng

Min in g

Fo od pro ces si ng

0

Fig. 1 SO2: 2001 emissions and 2010 emissions projection classified in 20 industrial sectors at Galicia

A slight reduction of NOx emissions is expected. Other pollutants’ emissions change in different directions (positive and negative impacts) depending on the industrial sector considered. Acknowledgments This work was financially supported by the R&D Spanish Programme (REN2002-02988/CLI) and R&D Galician Programme (PGIDT03PXIC20901PN).

References Carolina Environmental Program (CEP) (2003) SMOKE v.2.0 User Manual. The University of North Carolina at Chapel Hill, NC, USA. Casares JJ, Rodríguez R, Maceira P, Souto JA, Ramos S, Costoya M, Sáez A (2005) Inventory, analysis and projection of industrial air pollution emissions in Galicia. University of Santiago de Compostela, Spain (in Spanish). Mantzos L, Capros P, Kouvaritakis N, Zeka-Paschou M (2003) European Energy and Transport. Trends to 2030. European Commission. U.S. Environmental Protection Agency (2004) AP-42, Fifth Edition. EPA Publications, Washington, DC, USA.

P7.6 New Approaches on Prediction of Maximum Individual Exposure from Airborne Hazardous Releases John G. Bartzis, Athanasios Sfetsos and Spyros Andronopoulos

Abstract One of the key problems in coping with deliberate or accidental atmospheric releases is the ability to reliably predict the individual exposure during the event. Due to the stochastic nature of turbulence, the instantaneous wind field at the time of the release is practically unknown. Therefore for consequence assessment and countermeasures application, it is more realistic to rely on maximum expected dosage rather than actual one. Recently Bartzis et al. (2007), have inaugurated an approach relating maximum dosage as a function of the exposure time, concentration mean and variance and the turbulence integral time scale. Such approaches broaden the capability of the prediction models such as CFD models to estimate maximum individual exposure at any time interval. In the present work a further insight is given to this methodology and an alternative correlation is proposed based on theoretical considerations. The methodology to utilize such correlation types is further justified. Recently Bartzis et al. (2007) have inaugurated an approach relating the parameter

C max 'W C

to the fluctuating intensity I and the C max 'W C

§ 'W 1 1.5 I ¨¨ © TL

'W ratio as follows: TL · ¸¸ ¹

0. 3

(1)

It is reminded that the correlation (1) has been calibrated with 1s concentration signals representing the approximation of the instantaneous concentration statistical properties. It is suggested to be applied for such cases and exposure time duration 'W t 1s . Although such restrictions do not pose any serious problem for most atomspheric applications there is a need to remove such limitations and widen the applicability of those types of approaches. The theoretical background of the present approach can be summarized in the following notations: – The instantaneous concentration pdf can be approximated by a Gamma distribution – The approximation introduced by Venkatram (1979) for time average concentration variance is nearly valid C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

725

J.G. Bartzis et al.

726

The problem closure has been obtained by assessing the FLADIS T16 field stationary data (Bartzis, 2007) and applying best fit analysis. The new obtained correlation has as follows:

C max 'W C

1 4.57 I [(1 0.37

'W 1.45 )] TL

(2)

The FLADIS T17 Experiment is similar to T16 Experiment but with different meteorological conditions (Nielsen et al., 1994). Both correlations (1) and (2) are compared against the T17 experimental data. In Figures 1 and 2 the Factor of two (FAC2) and the Index of Agreement (IA) are plotted respectively. Both methods are able to predict maximum exposure within a factor of 2, but the new model is more accurate as expected on the basis of IA. 1

Index of Agreement

Factor of 2

105

100

0.99

0.98

0.97

95 New Model

Old Model

Fig. 1 FAC2 comparison all sensors

New Model

Old Model

Fig. 2 IA comparison all sensors

Conclusions (1) An new model to estimate maximum individual dosage has been derived applicable to any exposure time based on fluctuating intensity I and the ǻĲ / TL ratio. (2) The methodology inaugurated by Bartzis et al. (2007) has been further refined taking into consideration the theoretical pdf of the instantaneous concentration.

References Bartzis JG, Sfetsos A, Andronopoulos S (2007) On the individual exposure from airborne hazardous releases. The effect of atmospheric turbulence. Journal of Hazardous Materials, accepted for publication. Venkatram A (1979) The expected deviation of observed concentrations from predicted ensemble means, Atmospheric Environment, 13, 1547–1549. Nielsen M, Bengtsson R, Jones C, Nyren K, Ott S, Ride D (1994) Design of the FLADIS field experiments with dispersion of liquified ammonia, Risø–R– 755(EN), Risø National Laboratory, Roskilde, Denmark.

P2.9 Nonlinearity in Source-Receptor Relationship for Sulfur and Nitrate in East Asia Woo-Sub Roh, Seung-Bum Kim and Tae-Young Lee

Abstract Source-receptor (S-R) relationships for sulfur and nitrate are being actively sought in East Asia where great amounts of pollutants are being emitted into the air (e.g., Park et al., 2004). Nonlinearity in the advection and chemical processes is suggested to cause problems when model is used for the derivation of S-R relationships (Bartnicki, 1999). We have examined the effects of nonlinearity in the derivation of S-R relationship for sulfur and nitrate for East Asian region using a comprehensive model. Nonlinearity problem is investigated by examining the effects of varying emission rate in a source region on the deposition of pollutant in receptor regions as in the following steps (East Asia is divided into five source/receptor regions for this study): 1. Calculation of air quality, dry and wet deposition with full emission sources. 2. Same as step (1), except that emission rate of a particular species in a particular source region is reduced to 0%, 25%, 50% and 75% of full emission. Emission reduction is considered for SOx, NOx, NH3, VOC + CO in a separate manner. 3. Calculation of the difference in the deposition amount in a receptor region between full emission and reduced emission experiments. These calculations are carried out for March and July 2002. Air quality and pollutant deposition are calculated using an Eulerian, comprehensive acid deposition model (CADM) (Lee et al., 1998). Horizontal grid size is 60 km and vertical grids are stretched with a stretch ratio of 1.15. Meteorological fields are prepared using CSU RAMS with four-dimensional data assimilation (Pielke et al., 1992) (version 4.4). One-hourly fields are produced and then supplied to CADM. The emission rates of SOx, NOx, NH3, CO, and VOCs for China, South Korea and Japan are from the Long-range Transboundary Air Pollutants in Northeast Asia (LTP) project (Park et al., 2004). Air quality and pollutant deposition are calculated using CADM for March and July 2002. Calculated monthly total deposition amounts for July are shown in Figure 1. Dry deposition of sulfur reflects the distribution of emission. But dry deposition of nitrate shows a smoother pattern than that for sulfur. Wet deposition also shows significant difference between sulfur and nitrate. Wet deposition over Japan is significant for nitrate, while it is similar to back ground level for sulfur. Total deposition amount reflects more the distribution of wet deposition, due to


675

676

W. Roh et al.

large amount of precipitation over East Asia during July. Effects of nonlinearity are being investigated.

Fig. 1 Monthly total deposition amount for July for sulfur (left panels) and nitrate (right panels). Top and bottom panels are for dry and wet deposition amounts, respectively (unit: mg/m2)

Acknowledgments This study has been supported by the Global Environment Research Center/National Institute of Environmental Research through the project “Development of next-generation model for prediction of long-range transport of air pollutants in Northeast Asia.”

References Bartnicki J (1999) Computing source-receptor matrices with the EMEP Eulerian Acid Deposition Model. EMEP/MSC-W note 5/99. Park I-S, Kim J-C, Lee D-W (eds.) (2004) Annual report for the 4th year’s Joint Research on Long-range Transboundary Air Pollutants in Northeast Asia. Secretariat of Working Group for LTP project, 2004, NIER, Korea, 392 pp. Lee T-Y, Kim S-B, Lee S-M, Park S-U, Kim D-S, Shin H-C (1998) Numerical simulation of air quality and acid deposition for episodic cases in eastern Asia. Korean. Journal of Atmospheric Sciences 1(2), 126–144.

P6. Interactions between air quality and climate change

P6.1 On the Effective Indices for Emissions from Road Transport Kostadin Ganev, Dimiter Syrakov and Zahari Zlatev

Abstract The emissions from the road traffic are in a way different from the ship and airplane emissions: (i) the road network can be pretty dense in some cells of the large scale model grid; (ii) the emissions are continuous with time; (iii) the road traffic sources are close to earth’s surface. That is why the concept of deriving effective emission indices from the interaction of an instantaneous plume with the ambient air is perhaps not so convenient in the case of road transport emissions. On the other hand, the vertical turbulent transport is a very important process near earth’s surface, which means that it is relatively easy to parameterize the vertical structure of the pollution fields and so relegate the considerations to a twodimensional problem within a layer where the emissions heterogeneity can be important for the nonlinear chemical reactions. Due to the limited volume the vertical parameterization can not be discussed in details. It will be mentioned only that it is based on the heavy particles dry deposition parameterization in the surface layer, suggested by Ganev and Yordanov (2005) and for the case of N admixtures results in the following system of equations:

wci Lci Ai Bij c j J (ci ci ) Wij c j wt

Ei ,

i 1,..., N ,

(1)

where E i ( x, y, t ) accounts for large scale pollution source, ci ( x, y, t ) and ci ( x, y, t ) are the large scale pollution content and the respective concentration averaged in the layer, ci ( x, y, t ) is the concentration above the layer, ci ( x, y, t ) (h z 0 )ci ( x, y, t ) , Ai (c1 , c 2 ,..., c N ) – the term describing sources and sinks of the i th admixture, due to chemical transformations, {Bij } and {Wij } are diagonal matrixes describing the large scale absorption by earth’s surface and gravity deposition, J is a parameter describing the pollution exchange between the near surface layer and the upper atmosphere, L is the operator describing the horizontal transport. It is assumed that the mesoscale effects on large scale characteristics are small enough, which after some additional assumptions leads to the following formulation of the small disturbances problem in a domain D :


713

K. Ganev et al.

714

wGci LGc j D ij Gc j Bij Gc j JGci 0 , D ij wt Gci (t

0) W .(GE i GBij c j ) , Gci

1 wAi h z 0 wc j

(2)

0 at D boundaries,

i 1,..., N (3)

From mass-balance considerations in Eq. (2) and after additional simplifications it ml becomes clear that the pollution outflows/inflows from cell D of the large scale ml model (the mesoscale emission disturbances) ei for a time step W can be calculated by: eiml

1 0.5W (D ij Bij ) GC jml (W ) , GC i ml

³³ Gci ( x, y,W ) dD

i 1,..., N , (4)

D ml

which actually makes the calculation of effective emissions for each time step possible. The functionals GCi ml can be calculated when the problem (2, 3) is solved for the time period >0,W @ , but there is also another way to obtain them. As the problem (2, 3) is linear, the technique of functions of influence can be applied (Marchuk, 1982; Ganev, 2004) and GCi ml can be also expressed in the form:

GCi ml W ³³ c( i ) k ( x, y,0).GE k GBkj c j dD ml *

(5)

D

ml *

The advantage is that the solutions c( i ) k of the adjoin equations in this case can be factorized – one of the multipliers accounts for the chemical transformations and the other, which accounts for the horizontal transport can be analytically obtained in an explicit form. Acknowledgments The present work is supported by EC through 6FP projects ACCENT (GOCE-CT-2002-500337) and QUANTIFY (GOGE-003893), and COST Action 728.

References Ganev K (2004) Functions of influence and air pollution models sensitivity, Compt. Rend. Acad. Bulg. Sci., 57, 10, 23–28. Ganev Ʉ, Yordanov D (2005) Parameterization of dry deposition processes in the surface layer for admixtures with gravity deposition. Int. J. Environ. Pollut., 25, 1–4, 60–70. Marchuk GI (1982) Mathematical Modeling in Environmental Problems (in Russian), Nauka, Moscow.

P5. Aerosols in the atmosphere

P5.1 Quantifying Source Contribution to Ambient Particulate Matter in Austria with Chemical Mass Balance Receptor Modeling A. Caseiro, H. Bauer, I. Marr, C. Pio, H. Puxbaum and V. Simeonov

Abstract In this work, we apply the CMB model to a set of samples collected in Vienna. Those samples were chemically characterised for a wide range of chemical species as were samples from aerosol sources. The set of samples represent periods in which the threshold value of the PM10 level was exceeded, thus providing an insight over the causes of such episodes. Particulate matter concentrations at street level have been a raising concern for many years already, in particular due to its health effects (WHO, 2006). Such situation has been the background for ambient air European legislation regarding particulate matter (Council Directive 1999/30/EC). Yet, many European cities are not in agreement (EEA, 2006). A broad understanding of the identity, the sources and their intensity, the atomspheric interactions and the sinks of particulate matter are thus required. Chemical mass balance (CMB) is based on the principle of mass conservation, so that the mass of aerosol in a given location is a linear combination of the mass emitted by each source. The use of CMB requires the knowledge of the chemical composition of the aerosol at a given location and at a given time along with the chemical composition of the different sources that contribute to it. It is then possible, using a multi-linear regression, to calculate the composition of each source to the ambient aerosol (Gordon, 1988). Vienna is the capital city of Austria with about 1.8 million inhabitants. Four sampling sites were selected for this study: Schafberg, SCH (Residential area, north-west city fringe), Rinnböckstrasse, RIN (City centre, near city-highway), Kendlerstrasse, KEN (City centre), Lobau, LOB (Background area (Danube meadows) at the south-east of the city). Chemical species accounting for a wide range of chemical classes were measured using a suite of eight analytical techniques. The 12 aerosol sources accounted for in this study are road dust, diesel exhaust, gas combustion, biomass combustion, vegetative detritus, cooking, HULIS (humic-like substances, secondary organic aerosol), ammonium sulphate, ammonium nitrate, sea salt, break wear and tyre wear.


711

712

A. Caseiro et al.

CMB is based on the assumption that the mass collected on a filter is a linear combination of the mass emitted by the different sources and that the chemical build-up of the aerosol remains constant, both during the aerosol transportation and sampling. It states that one can identify the contribution of various classes of sources by measuring the concentrations of many chemical species in ambient air samples because their composition pattern is sufficiently different. Exceedences of the 50 µg/m3 PM10 value only occurred in the cold period. From those days, a set of eleven periods was chosen to model the source contributions. The difference between the total calculated mass and the measured mass was, in average, 18% for SCH, 15% for RIN and KEN, and 13% for LOB. Secondary inorganic aerosol was a general strong source in exceedence periods. In average terms, it is by far the predominant source of PM10 in exceedence episodes, being more than two times more important than the second most contributive source in the background sites (SCH and LOB), and about one and a half time more important in the city centre sites (RIN and KEN). Road dust was the second most important source of PM10 in exceedence days. Particularly, this source was very strong for the episodes Feb1, Mar2 and Apr1. These episodes were coincident with the period subsequent to the melting of the snow, when the road gravel material, Dolomite limestone, happens to be triturated by the road traffic. Biomass burning was also a major source. The main contributor to this class is residential wood burning, which occurs mainly in the colder periods, contributing to about 15–24% of the PM10. Diesel exhaust, secondary organic aerosol and vegetative detritus were the last important sources. The former was quite constant for all the exceedence episodes and the latter less present in city centre sites. Gas combustion, cooking, salts, break wear and tyre wear were not key contributors to PM10.

References Council Directive 1999/30/EC: Limit values for sulphur dioxide, nitrogen dioxide and oxides of nitrogen, particulate matter and lead in ambient air. EEA (2006) Air pollution at street level in European cities, EEA Technical report No 1/2006, European Environment Agency, ISBN 92-9167-815-5, ISSN 17252237, © EEA, Copenhagen 2006. Gordon GE (1988) Receptor models. Environmental Science and Technology, 22, 1132–1142. Health risks of particulate matter from long-range transboundary air pollution, Joint WHO/Convention Task Force on the Health Aspects of Air Pollution, E88189 © World Health Organization Regional Office for Europe, Copenhagen 2006.

P2.5 Regional Transport of Tropospheric Ozone: A Case Study in the Northwest Coast of Iberian Peninsula Santiago Saavedra, María R. Méndez, José A. Souto, José L. Bermúdez, Manuel Vellón and Miguel Costoya

Abstract Relevant tropospheric ozone levels are frequently reached in the NW coast of the Iberian Peninsula (Galicia) during spring and summertime under high pressure conditions (Logan, 1998). In this study, the origin and associated phenomena to tropospheric ozone episodes in rural areas at that region are considered. Most of them are produced by regional ozone transport from Southern and Eastern regions of the Iberian Peninsula. In addition, analysis and simulation of a typical episode (12–22 September 2003) are presented. Tropospheric ozone episodes in rural areas of Western Europe have been reported in the past (Logan, 1998). These phenomena appear in Galicia, an Atlantic region with complex topography and strong sea influence. Therefore, a systematic analysis of ozone episodes from 2002 to 2006 was done, considering both field measurements and modelling results. For a typical episode, mesoscale modelling with PSU/NCAR MM5 (Grell et al., 1995) was applied in order to get a better understanding about the origin of O3 peaks. Setting an ozone hourly ground level concentration (glc) threshold of 150 Pg/m3, (close to first legal threshold, 180 Pg/m3) 26 episodes were identified. Then, an analysis was done considering: (a) field measurements in the region, and its surroundings; and (b) EURAD meteorological and air quality modelling (Memmesheimer et al., 2001). Analysis of regional conditions during 12–22 September 2003 typical episode shows a synoptic pattern dominated by Central-Europe anticyclone. Weak SE circulation was caused by this synoptic situation in the NW of the Iberian Peninsula, changing to southerly flux on 15th. Then, a typical summertime low pressure gradient was established at Central Peninsula: at this point, maximum ozone glc is achieved, with hourly averages above 135 Pg/m3 (reaching up to 190 Pg/m3). Period from 15 to 17 September 2003 was selected to simulate meso-E meteorology and back-trajectories using MM5. Figure 1a shows back-trajectories, covering 12-hour backwards, and showing the transport of air masses from N of Portugal, as in EURAD operational forecast. Figure 1b shows a W-E wind profile over the region, where sea breeze regime (due to the weak pressure gradient coupled with warm land temperatures) causes recirculation of air masses from the coast to inland. C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

667

668

S. Saavedra et al.

Although in other cases (Alonso et al., 2000) aged pollutants layers from the coastal O3 can be created, in this episode recirculation contributes to the entrance of O3-polluted air from the South, from upper to lower levels, increasing the effect of O3 external contribution.

Fig. 1 MM5 simulations results: (a) 12-hour back-trajectories that reach the N of Galicia on 15/September/03 at 16 UTC. Odd and even indexes show 2 and 1,000-m height, respectively; (b) simulated flows showing a convective cell produced by NW sea breeze, opposite to SE synoptic wind, on 16/September/03 at 16 UTC

Acknowledgments This work was financially supported by Endesa Generación, S.A. and the R&D Spanish Programme (CTQ2006-15481/PPQ). Air quality data were provided by Environmental Departments of As Pontes Power Plant (Endesa company), Xunta de Galicia and Junta de Castilla-León (Spain), and Ministerio do Ambiente of Portugal. EURAD operational forecasts and technical supports by MeteoGalicia and CESGA are acknowledged.

References Alonso L, Gangoiti G, Navazo M, Millán M, Mantilla E (2000) Transport of tropospheric ozone over the bay of Biscay and the eastern coast of Spain, Journal of Applied Meteorology, 39 (4), 475–486. Grell GA, Dudhia J, Stauffer DR (1995) A description of the Fifth-Generation Penn State/NCAR Mesoscale Model (MM5), NCAR/TN-398 + STR, Boulder, USA. Logan JA (1998) Trends in the vertical distribution of ozone: an analysis of ozonesonde data, Journal of Geophysical Research, 99 (D12), 25553–25586. Memmesheimer M, Jakobs HJ, Piekorz G, Ebel A, Kerschgens MJ, Friese E, Feldmann H, Geiß H (2001) Air quality modeling with the EURAD model. In proceedings of the 7th International Conference on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes, Belgirate.

P2.15 Saharan Dust over Italy: Simulations with Regional Air Quality Model BOLCHEM Mihaela Mircea, Massimo D’Isidoro, Alberto Maurizi, Francesco Tampieri, Maria Cristina Facchini, Stefano Decesari and Sandro Fuzzi

Abstract Saharan dust is transported over Mediterranean area, dust reaching often different regions of Italy. To the scope of predicting the advection of dust and its physical and chemical properties over Italy, a dust emission scheme has been implemented in the air quality model BOLCHEM, which solves simultaneously the chemical and meteorological equations. This study demonstrates the ability of BOLCHEM to predict the dust events over Italy and evaluates the impact of differrent parameterizations used in the dust production. The dust aerosols, besides of changing climate through the scattering and absorption of solar and thermal radiation, also affect the environment by fertilizing marine and terrestrial ecosystems that in turn influence the carbon cycle. Moreover, the dust particles contribute substantially to the total aerosol mass usually employed in the developing of the environmental policy regulations, therefore, a reliable forecast of dust events is mandatory over Italy, often affected by Saharan dust transport. This study describes the dust model implemented in the air quality model BOLCHEM and shows its ability to forecast dust events over Italy. The study also investigates the dependency of dust production on threshold friction velocity and number of dust size bins. The air quality model BOLCHEM (D’Isidoro et al., 2005; Mircea et al., 2007) comprises a meteorological model, an algorithm for airborne transport and diffusion of pollutants and two photochemical mechanisms. The meteorology is coupled online with the chemistry. The dust model implemented in BOLCHEM was developped by Tegen et al. (2002) and is based on the soil-derived dust emission scheme designed by Marticorena and Bergametti (1995). The horizontal and vertical dust fluxes are calculated based on the location of the preferential dust sources, soil texture, surface roughness, vegetation cover, soil moisture content and surface wind velocity. The ratio between the vertical and the horizontal dust fluxes varies with the type of soil and the size of the particle mobilized. The size distribution of the mobilized dust depends on both the surface properties (soil texture) and the surface wind speed. The threshold friction velocities used to initiate the dust emissions are computed as a function of particle size following Marticorena and Bergametti (1995), but assuming constant roughness within the model grid cells (0.001 cm). Moreover, the simulation shown here was carried out with a threshold friction velocity lowered by a factor of 0.75 since lower thresholds velocities improve model results compared to observations at global level. The comparison of the dust C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

687

688

M. Mircea et al.

event occurred on 16 July 2003, simulated by BOLCHEM and seen by the AQUA/MODIS satellite/sensor, shows that the model is able to predict well both the extent and the timing of the dust event over Italy. In both images, it can be noted that the plume of dust over the Mediterranean comes from north-west and north of Africa and goes straightforward to the center and north of Italy with only a little veil over Sicily and Messina Strait. These results substantiate that the model uses reliable surface land/soil information, meteorological conditions and transport scheme.

Fig. 1 Saharan Dust over Italy: July 16, 2003 at 13 UTC. Model simulation (left) and satellite image from Aqua/Modis (right)

Dust emissions calculated with the box dust model shows that the production of dust depends more on threshold friction velocity than on number of dust size bins. The effect of threshold friction velocity is similar for accumulation and coarse mode while the number of size bins impacts only on accumulation mode. These results will be further used in the calibration of the dust concentrations calculated by the model. Acknowledgments This work was conducted in the frame of ACCENT and GEMS EC projects, Italian MIUR project AEROCLOUDS, and were also supported by the Italian Ministry of Environment through the Program Italy-USA Cooperation on Science and Technology of Climate Change.

References D’Isidoro M, Fuzzi S, Maurizi A, Monforti F, Mircea M, Tampieri F, Zanini G, Villani MG (2005) Development and Preliminary Results of a Limited Area Atmosphere-Chemistry Model: BOLCHEM, First ACCENT Symposium, Urbino 12–16 September 2005. Marticorena B, Bergametti G (1995) Modeling the atmospheric dust cycle: 1. Design of a soil-derived dust emission scheme, J. Geophys. Res., 16415–16430. Mircea M, d’Isidoro M, Maurizi A, Vitali L, Monforti F, Zanini G, Tampieri F, (2007) A comprehensive performance evaluation of the air quality model BOLCHEM over Italy, submitted to Atmos. Environ. Tegen I, Harrison SP, Kohfeld K, Colin Prentice I, Coe M, Heinmann M (2002) Impact of vegetation and preferential source areas on global dust aerosol: results from a model study, J. Geophys. Res., 107, D21, doi:10.1029/2001JD000963.

P1.2 Simplified Models for Integrated Air Quality Management in Urban Areas B. Sivertsen, A. Dudek and C. Guerreiro

Abstract The Norwegian Institute for Air Research, NILU has been requested by the World Bank to support the Hanoi Urban Transport and Development Project (HUTDP) with the Urban Air Quality Management Subcomponent primarily executed by Hanoi’s Department of Natural Resources, Environment and Housing (DONREH). As part of the evaluation NILU has performed air quality modeling in order to assess the importance of air pollution from mobile sources in Hanoi. For this purpose, NILU has combined two models; the NILU developed air quality modelling planning system AirQUIS and the Simple Interactive Model for Better Air Quality (SIM-Air) developed by the World Bank. AirQUIS was used to simulate the computation of an emission inventory for key pollutant and estimate the impact of the sources on air quality. SIM-Air will be used to assess and evaluate health effect impact, and allowed for various policies measures, economic and technical options to be evaluated for their environmental and health impacts and cost effectiveness. This model uses simple mathematical and financial tools to evaluate different air quality management options. All the management options are linked to cost and health impacts based on percentage change in the considered option.

1. Introduction NILU has developed the AirQUIS GIS based air quality management and dissemination system to perform integrated assessment and planning for improving air quality (http://www.nilu.no/airquis/). A comprehensive management system such as AirQUIS requires large specific datasets. There is therefore a need to develop simple interactive decision support tools to assist local authorities to assist local authorities to carry out screening processes and take appropriate decisions and actions for air quality management, especially in developing countries where available data and resources are limited.


653

654

B. Sivertsen et al.

2. System Approach The concept developed by NILU can be summarised in three stages: estimation of top-down emission inventory, calculation of ambient concentrations using dispersion modelling and evaluation of health effects, management options and related costs. For estimating emissions for a specific area, NILU uses a similar integrated approach as in the Simple Interactive Model for Better Air Quality (SIM-AIR) presented by the World Bank (Shah and Saikawa, 2005). Where detailed emission inventory is not available, estimates of emissions are performed as a top down approach. In NILU’s approach, the emissions are estimated for a defined gridded domain, with user defined resolution, covering the area being studied. Input data for the modelling system include: (1) population data, (2) meteorological data, (3) emission data, (4) emission factors for source categories, (5) dose response functions, and (6) cost estimates.

3. Results and Discussion The models generated concentration distributions of NOx, PM10 and SO2 over the city of Hanoi (Sivertsen and Dudek, 2006). As a first estimate/surrogate for the total population exposure the estimated concentrations have been used together with the population distributions to estimate the person-weighted concentrations. The relative contribution from each of the vehicle categories has been estimated for NOx, PM10 and SO2. NOx exposure due to traffic emissions is caused by truck emissions (36%), motorcycles (22%), petrol driven cars (20%), diesel cars (12%) and about 8% due to buses. A similar estimate for the PM10 contributions indicated that the total population exposure in Hanoi is due in 23% to traffic sources; 15% to industrial sources and 62% to other undetermined sources (including “background”). This simple integrated decision support tool meets the requirement of a first screening of air pollution problems in defined areas and requires less computing power than existing advanced air quality management tools including complex dispersion models. Simple decision support systems of this kind may also help stakeholders air quality management planning or action plans and easier assessment of different options.

References Shah J, Saikawa E (2005) Interactive Database for Emission Analyses (IDEAHanoi) Version 1. (Developed in the East Asia Region of the World Bank.) Sivertsen B, Dudek A (2006) Support for the Review of Air Quality Management Sub-component for the Hanoi Urban Transport and Development Project. Modelling air pollution in Hanoi. Kjeller (NILU OR 83/2006)

P7.4 Source Apportionment of Particulate Matter in the U.S. and Associations with In Vitro and In Vivo Lung Inflammatory Markers Rachelle M. Duvall, Gary A. Norris, Janet M. Burke, John K. McGee, M. Ian Gilmour and Robert B. Devlin

Abstract Associations are well established between particulate matter (PM) and increased human mortality and morbidity. Fine particulate matter (particle diameter < 2.5 Pm) is most strongly linked to adverse health impacts. The toxicity of PM may depend on the PM source and composition which will vary by location. While a number of epidemiological studies have shown that certain PM sources are associated with specific health outcomes, the underlying mechanisms are still unclear. To investigate these mechanisms, continuous weekly PM2.5 samples were collected for four consecutive weeks (24 hours a day for seven days) in six cities across the U.S. as part of the Multiple Air Pollutant Study (MAPS). Sample composites were constructed for each site and particles were extracted in water. Samples were analyzed for trace metals (via Inductively Coupled Plasma – Optical Emission Spectroscopy), ions (via Ion Chromatography), and elemental carbon (via thermal methods). Sources contributing to the PM2.5 samples were identified using the EPA Chemical Mass Balance (CMB8.2) model. Both in vitro and in vivo experiments were conducted to measure a variety of toxicological outcomes. For the in vitro analysis, PM extracts were applied to cultured human lung epithelial cells and the production of different lung inflammation/injury markers (Table 1) was measured by real-time reverse transcriptase polymerase chain reaction (RT-PCR). For the in vivo analysis, particle extracts were instilled into mouse lungs at different doses (25 and 100 Pg). Indicators of lung injury and inflammation (Table 1) were measured in bronchoalveolar lavage fluid and plasma by enzyme-linked immunosorbent assay (ELISA). The relationship between the toxicological measures and PM2.5 sources was evaluated using linear regression. A few of these plots are displayed in Figure 1. For the in vitro health markers, mobile sources and secondary sulfate (from coal combustion) were related to increased IL-8 production (r2 = 0.39 and r2 = 0.79, respectively). Combustion sources and soil were associated with increases in COX2 (r2 = 0.38 and r2 = 0.48), and secondary sulfate was associated with increased HO1 (r2 = 0.51). For the in vivo health markers, wood combustion was associated with increased MIP-2 production (r2 = 0.95), whereas mobile sources were associated with increased IL1-E and TNF-D (r2 = 0.94 and r2 = 0.99, respectively). These findings confirm that PM2.5 sources are associated with specific health outcomes. C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

721

R.M. Duvall et al.

722 Table 1 In vitro and in vivo health markers analyzed. In vitro markers Interleukin-8 (IL-8) Cycolooxygenase-2 (COX-2) Heme oxygenase-1 (HO-1)

In vivo markers Macrophage inhibitory protein (MP-2) Interleukin 1-ȕ (IL1- ȕ) Tumer necrosis factor alpha (TNF-Į)

Legend: Ŷ Salt Lake City Ÿ Seattle x Phoenix ż Sterling Forest Ɣ South Bronx * Hunter College

(Sterling Forest data removed )

5.00

y = 2.77x + 0.35 R2 = 0.79

0.00 0.00

1.00

30.00

HO-1

IL-8

10.00

2.00

3.00

4.00

20.00 y = 7.58x + 0.46 R2 = 0.51

10.00 0.00 0.00

3

Secondary Sulfate (P g/m )

1.00

2.00

3.00

4.00

Secondary Sulfate (Pg/m3)

800.0

500.0

400.0

y = 329x + 255 R2 = 0.94

200.0 0.0 0.00

TNF-D

IL1- E

600.0

0.50

1.00

1.50

2.00

250.0

0.0 0.00

Mobile Sources (Pg/m 3)

y = 185x + 161 R2 = 0.99 0.50

1.00

1.50

2.00

Mobile Sources (Pg/m 3)

Fig. 1 PM sources compared to in vivo and in vitro health markers

Disclaimer Although this work was reviewed by EPA and approved for publication, it may not necessarily reflect official Agency policy.

P2.4 SPECIATE – EPA’s Database of Speciated Emission Profiles J. David Mobley, Lee L. Beck, Golam Sarwar, Adam Reff and Marc Houyoux

Abstract SPECIATE is the U.S. Environmental Protection Agency’s (EPA) repository of total organic compound (TOC) and particulate matter (PM) speciation profiles for emissions from air pollution sources. The profiles are key inputs to air quality modeling and source-receptor modeling applications. This paper addresses Version 4.0 of the SPECIATE Database. The SPECIATE Database is an important EPA product which serves as the repository for source category-specific emission speciation profiles. The profiles contain weight fractions of species of both volatile organic compounds (VOC) and Particulate Matter (PM). The weight fractions of VOC species are grouped into reactivity classes to support air quality modeling for ozone. The profiles of PM species weight fractions are specific to particle size ranges and are being used to support air quality modeling for PM and visibility. The Database has also supported air toxic assessments and is essential for source-receptor modeling applications. The Database was first computerized in 1988. Although accessibility to the Database has been sustained through the Clearing House for Inventories and Emission Factors (CHIEF) website, updates to SPECIATE have languished since the mid-1990s due to decreasing budgets. The US National Research Council in its report on Research Priorities for Airborne Particulate Matter (NRC, 2004), the Clean Air Act Advisory Committee in its report of the Air Quality Management Working Group (CAAAC, 2005), NARSTO in its Emission Inventory Assessment (NARSTO, 2005), and other groups have recommended that the Database be extensively updated and maintained in a dynamic manner. Given the importance of SPECIATE to the process of air quality management, a team was organized to undertake an update of the Database. The scope of the team’s project was to: (1) update the Database with profiles from the literature and EPA source test data sets; (2) link the new profiles to Source Classification Codes (SCCs) in the National Emissions Inventory (NEI); (3) assign any new species to reactivity classes; and (4) update the air quality models to use the new information. The final report, “SPECIATE 4.0 – Speciation Database Development Documentation (US EPA, 2006)” summarizes the development and provides guidance on use of the Database. The Database is posted on the CHIEF Website. The final version of the Database has been integrated into the Emissions Modeling Platform for subsequent research and regulatory modeling applications. The ability to speciate the emissions inventory with the new SPECIATE composite profiles will C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

665

666

J. Mobley et al.

bring substantial benefits to the fields of air quality modeling and source apportionment Results from these improved analyses will enable the development of more effective control strategies for sources of these species. Further, estimates of uncertainty in the results of air quality and receptor models will be improved by providing the most representative and up-to-date information to characterize emissions from the myriad point, area, mobile, and biogenic sources that contribute to ambient pollutant concentrations. The initiative to update SPECIATE has produced: x 2,856 PM profiles, 1,258 of which are new profiles x 1,215 gas profiles, 648 of which are new profiles x 1902 unique species, 1012 of which are new species

SPECIATE 4.0 represents a significant enhancement of the data available to characterize emissions by species and source category. Air quality modeling and source-receptor modeling applications have benefited from using these enhanced speciation profiles. Additional efforts are needed to capture new data from current testing based on data submitted via the protocol for database expansion. The user community can support the Database development by supplying electronic data with full references. Acknowledgments The authors acknowledge Ying Hsu, Randy Strait, and Frank Divita of E.H. Pechan & Associates, Inc. for their support to the SPECIATE project.

References NRC (2004) Research Priorities for Airborne Particulate Matter: IV. Continuing Research Progress, National Research Council, National Academies Press, Washington, DC. CAAAC (2005) Recommendations to the Clean Air Act Advisory Committee, Air Quality Management Working Group, January 2005. NARSTO (2005) Improving Emission Inventories for Effective Air Quality Management Across North America – A NARSTO Assessment. NARSTO-05001, September 2005. US EPA (2006) SPECIATE 4.0 – Speciation Database Development Documentation. EPA/600/R-06/161, US Environmental Protection Agency, Research Triangle Park, NC, November 2006.

P7.7 The Detroit Exposure and Aerosol Research Study Ron Williams, Alan Vette, Janet Burke, Gary Norris, Karen Wesson, Madeleine Strum, Tyler Fox, Rachelle Duvall and Timothy Watkins

Abstract The Detroit Exposure and Aerosol Research Study (DEARS) was designed to assess the impacts of local industrial and mobile sources on human exposures to air pollutants in and around Detroit, Michigan. Daily integrated measurements were made of personal exposure, and residential indoor and outdoor concentrations in six neighborhoods throughout the Detroit area. Concurrent data were collected for comparison at a central community ambient monitoring location and a regional background site. These data collected in DEARS can be used to evaluate local air quality and explore the application of air quality models to assess human exposure in an urban area.

Keywords Air quality, human exposure, modeling, particulate matter, toxic air pollutants, Detroit The Detroit Exposure and Aerosol Research Study (DEARS) will provide needed information on defining the factors that impact individual exposures to various sources of particulate matter (PM) and toxic air pollutants and will contribute to the scientific information that is needed to inform decisions on standards to protect air quality. The study included three years of field data measurements (summer 2004 through winter 2007). Approximately 120 adult participants living in detached single-family residences were randomly selected and enrolled from six neighborhood census areas. These neighborhoods were selected because they each represented a variation of potential industrial and regional source influence, housing stock, and proximity to automotive emission sources. Selected participants and/or their residences were involved in five days of summertime monitoring and five days of wintertime monitoring each year. Data collected included various PM size fractions and select pollutant concentrations such as volatile organic compounds, carbonyls, metals, and criteria pollutant gases (see Table 1). Data from the DEARS will be used to conduct analyses to improve the understanding of relationships between sources, air quality, and human exposures and to evaluate the uncertainty of using community-based monitoring as surrogates for true human exposures. Air quality modeling is an integral part of DEARS. Both the regional-scale, photochemical model, CMAQ, and the local-scale, dispersion model, AERMOD will be used. The modeling domain for CMAQ will be centered on the Detroit area and will utilize a 12 km horizontal grid resolution, while the AERMOD receptor domain will be much smaller, spanning an area of 36 by 48 km, with receptors C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

727

R. Williams et al.

728 Table 1 DEARS measurement parameters. Parameter PM2.5 (mass, elements) PMcoarse (mass, elements) EC-OC (PM2.5) EC (PM2.5) Nitrate Gases (O3, NO2, SO2) Aldehydes VOCs SVOCs PAHs Air exchange rate

Personal X – – X – X X X – – –

Indoor X X X X X – X X X X X

Outdoor X X X X X X (NO2 only) X X X X –

Ambient X X X X X X X X X X –

placed at 1 km intervals. The predicted concentrations from CMAQ and AERMOD will be combined where appropriate, using a one-atmosphere “hybrid” approach (Isakov et al., 2007). This hybrid approach allows the preservation of the granular nature of the dispersion model while properly treating the chemistry and transport offered by the photochemical model. In addition, the most recent version of CONCEPT (Consolidated Community Emissions Processing Tool) will be used to produce link-based mobile emissions for PM and air toxics to provide a more refined allocation of mobile emissions for this project and to improve the ability to analyze the local-scale impact of mobile emissions on the urban air quality. Figure 3 shows the conceptual approach for DEARS air quality modeling.

Fig. 3 Conceputal Approach for DEARS Air Quality Modeling

Disclaimer and Acknowledgments Although this work was reviewed by EPA and approved for publication, it may not necessarily reflect official Agency policy. The research was funded and conducted through contract 4D- 5653-NAEX (University of Michigan), 68-D-00-012 (RTI International), and 68-D-00-206 (Alion Corp).

References Isakov V, Irwin J, Ching J (2007) Using CMAQ for exposure modeling and the importance of sub-grid variability for exposure estimates. Journal of Applied Meteorology and Climatology, Bostan, MA, 46(9): 1354–1371, (2007).

P2.3 The Role of Sea-Salt Emissions in Air Quality Models Raúl Arasa, Maria R. Soler and Sara Ortega

Abstract In this study, we show our investigations about the effect on atmospheric aerosol concentrations caused by sea-salt emissions generated in open oceans and in surf zones. We use the CMAQ/MM5/MECA air quality modelling system, taking and do not taking into account sea-salt emissions in order to find that they produce several differences in particulate concentrations.

1. Introduction Atmospheric particles, in particular sea-salt ones, play an important role in climate and atmospheric chemistry. Several studies show the importance of sea-salt emissions coming from both open oceans and the surf zone due to their interactions with other atmospheric species. The aim of this study is to investigate the effect of both emission sources (open ocean and surf zones) on atmospheric particles concentration in Catalonia, located in the northeast part of Spain. In order to achieve this goal, we simulated a summer 2003 period using: an emission model created by the authors known as MECA, MM5 meteorological model, and CMAQ 4.5.1 photochemical model with its sea-salt module AERO4 enabled. Different simulations were performed taking and not taking into account sea-salt emissions. Preliminary results indicate that the addition of sea-salt module alters particle concentrations causing relative changes in total PM10 and PM2.5 concentrations.

2. Modelling System and Set-up PSU/NCAR mesoscale model MM5 v3.7 was used to generate meteorological fields, the inputs of the air pollution modelling system. Meteorological simulations were performed for four two-way nested domains with 27, 9 and 3 km resolutions. The coarsest domain covers Spain, part of France and part of Italy. An inner domain of 30 x 30 cells (9 km) covers Catalonia, while two other 3 km resolution domains (the smallest ones) cover two areas whose interest lie in their high pollution level measurements. Characteristics of these simulations are described in Soler et al. (2007). The chemical transport model used in this study is the U.S. EPA models-3/CMAQ v4.5.1. In order to investigate the sea-salt effect on atmospheric C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

663

664

R. Arasa et al.

particles concentration, we did two simulations: with and without the sea-salt module AERO4 enabled. MECA, developed by the authors (Ortega et al., 2006), was the emission model used. This model was applied over domains number two, three and four; while for the biggest domain the emission inventory was quantified by the top-down approach using EMEP emissions. These emissions include the most important primary air pollutants from vegetation, on-road traffic, industries, fossil fuel consumption, and domestic-commercial solvent effects.

3. Results and Discussion The simulated period, 10–14 June 2003, was characterized by an anticyclonic situation favouring the development of mesoscale winds such as diurnal and nocturnal sea breezes. AERO4 effect on modelled PM2.5 and PM10 concentrations was evaluated by calculating the concentration relative differences between enabling and disabling the AERO4 module. In PM2.5 case, results suggest that the inclusion of sea-salt emissions decrease PM2.5 concentrations during all daily periods over sea areas. This reduction reaches values up to 60%. On the other hand, in this particular case we do not detect any clear tendency during nighttime over land and shorelines. The relative changes, positive or negative, are in any case comparatively small: between –20% and 20%. During day time there is some tendency to PM2.5 concentrations increase. Eventually, the increment could be high but located over inland in small delimited points. In PM10 case, the most important effect controlling the concentration relative differences is the wind speed. Wind velocities higher than 4 ms-1 increase sea-salt concentrations, especially on coast areas where the sea breeze could be intense and the emissions coming from the surf zone are remarkable. This PM10 result was not unexpected because the parameterization used to take into account sea-salt emissions depends on particle size and increases exponentially as a wind velocity function. Acknowledgments This project was funded by the Spain Government through CGL2006-12474-C03-02 grant.

References Ortega S, Alarcón M, Soler MR, Pino D, Grasa J (2006) Cálculo de emisiones relevantes en la modelización fotoquímica mesoscalar, Medioambiente en Iberoamerica: visión desde la Física y la Química en los albores del siglo XXI 1, 171–178. Soler MR, Bravo M, Ortega S (2007) The use of meteorological and dispersion models in stratified boundary layers, Develop. Environ. Sci. 6, 199–208.

P4.7 The Use of MM5-CMAQ-EMIMO Modelling System (OPANA V4) for Air Quality Impact Assessment: Applications for Combined Cycle Power Plants and Refineries (Spain) R. San José, J.L. Pérez, J.L. Morant and R.M. González

Abstract Since 2000 the EPA Models-3 Community Multiscale Air Quality Modelling System (CMAQ) has become in one of the state-of-the-art air quality tools to perform a full analysis of the air concentrations in a determined domain in space and time. The CMAQ modelling tool incorporates several chemical mechanisms, several numerical solvers, several boundary layer parameterizations, etc. and the user has to select the best option according to their experience and knowledge to perform any air quality simulations in historical and/or forecasting mode. The MM5 meteorological mesoscale model and/or the new generation of mesoscale meteorological models based on WRF model can be used as input for CMAQ. Additionally, our laboratory has developed along the last ten years a sophisticated emission model which provides with the corresponding spatial and temporal resolution, the emission data required by CMAQ to perform the full simulations. This system – MM5-CMAQ-EMIMO (OPANA V4) – can be used to forecast air concentrations in space and time or to simulate different periods of time in the past. In order to perform an air quality impact assessment of an industrial plant such as a combined cycle power plant or a refinery, we should run our air quality modelling system for a period of time in the past according to the EU Air Quality Directives (and also US EPA Regulations for non-EU applications). The EU Directives related to the limits in air concentrations are concerned with a number of exceedances of a specific concentration along one year. There are limits for 8-hour, days and year. In this contribution we show the methodology to be used to fulfil the EU Directives for different applications – combined cycle power plants and refineries – in Spain and some results. The use of the system for one-year period to adapt the results to the actual legislation is presented.

Keywords Industrial plants, air quality, forecasting In this contribution we show the methodology used for applying the MM5-CMAQEMIMO air quality modeling system for performing air quality impact assessment according to the EU legislation. We will show the results for different applications for combined cycle power plants and incinerators in Spain. The calibration and validation process in all the studies is an essential step to carry out the fuell air C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

707

708

R.S. José et al.

quality impact assessment. Figure 1 shows one year ozone data obtained in different air quality monitoring stations in Madrid Community during 2005 and averaged and compared with ozone modeled data obtained in every grid cell where the monitoring stations are found. This plot is made with 365 days × 24 hours = 8,760 data for the full year 2005.

Fig. 1 Correlation coefficient between one year modeled ozone data (MM5-CMAQ-EMIMO) and measured data averaged over several stations in Madrid Community Air Quality Monitoring network

The correlation coefficient is 0.796 and it represents the accuracy of the results of the study. The methodology used is based on the so-called ON-OFF approach which means that a full MM5-CMAQ-EMIMO modeling system is run over three different domains with 405 × 405 km, 88 × 105 km and 24 × 24 km with 9, 3 and 1 km spatial resolution for the full year under the so-called OFF scenario which includes all biogenic and anthropogenic emissions present during 2005 in the area of the study and a full run of a so-called ON scenario which is exactly the same than OFF scenario but adding the maximum expected emissions of the proposed combined cycle power plant or incinerator. The differences ON-OFF represent the impact of the expected future industrial plant – this procedure is valid for any industrial plant not only for power plants or incinerators. The percentiles present in the EU Legislation for the different pollutants are calculated based in the full one year run. The quality of the simulations are based on the correlation coefficient and several other statistical tools currently available for this analysis.

P4. Model assesment and verification

P4.1 Tropospheric Ozone and Biogenic Emissions in the Czech Republic K. Zemankova and J. Brechler

Abstract Terms of formation, amount, spatial distribution of tropospheric ozone and the contribution of biogenic sources of VOC to ozone precursors were studied using numerical model for summer photochemical smog simulation (SMOG model). Biogenic emissions of VOC were estimated using semi-empirical model proposed by Guenther et al. (1995). North-eastern part of the Czech Republic (Hruby Jesenik area) was selected as a model domain and comparison of two different model runs with measured data from this area is presented. Only antropogenic emissions of VOC were taken into account in the first run of the model whereas biogenic emissions of VOC were added into model inputs in the second run.

Tropospheric ozone is a significant element of atmospheric pollution, especially in the summer period when photochemical reactions, which lead to ozone formation, are remarkably enhanced by intensive sunlight and high temperatures. Main ozone precursors such as nitrogen oxides (NOx) and volatile organic compounds (VOC) are released to the atmosphere not only from antropogenic sources, but they are emitted in considerable amounts from biogenic sources as well. Lagrangian puff model SMOG was used for evaluation of a difference in tropospheric ozone concentration while its formation has been simulated with and without the contribution of VOC from natural sources. Short decription of the SMOG model can be found in Section 2.1, for further details see Bednar et al. (2001). Emissions of biogenic VOC were estimated on the basis of semi-empirical model suggested by Guenther et al. (1995) and its principal ideas are described in Section 2.2. Model results from both model runs are compared with measured data in Section 3. SMOG model is a chemical transport model developed at the Department of Meteorology and Environment Protection, Charles University in Prague. It is a lagrangian puff model where a continuous plume of pollution is divided into several separate puffs preserving the original intensity of emission fluxes from individual sources. Thanks to this approximation SMOG model is able to model non-stationary situations under varying meteorological conditions. Each puff has its own trajectory according to meteorological preprocessor – model ETA. A dispersion of puffs into all three dimensions with the normal distribution as well as chemical interaction of individual puffs with each other is expected. C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

693

K. Zemankova, J. Brechler

694

Isoprene and monoterpens are considered to be the most prominent VOC released from natural sources (Simpson et al., 1995). Emission flux F (ȝg m-2 h-1) of each chemical compound is according to Guenther et al. (1995) calculated as:

F

(1)

H D J

where H is an ecosystem dependent emission factor (ȝg C m-2 h-1 at photosynthetically active radiation (PAR) flux of 1,000 ȝmol m-2 s-1 and leaf temperature of 303.15 K), D is foliar density (kg dry matter m-2) and Ȗ is a dimensionless correction factor which accounts for the influence of PAR and leaf temperature in case of isoprene and for the influence of leaf temperature only in case of monoterpene. Comparison of daily mean concentrations of tropospheric ozone calculated by the SMOG model for June 2000 (dashed line- with antropogenic emissions of VOC only, dotted line- with inclusion of biogenic emissions) with measured data (solid line) are shown below. Two model grid points were selected to correspond with monitoring stations Cervenohorske sedlo and Jesenik. Cervenohorske sedlo

Jesenik

concentration /Pg.m-3

160

220

140

120

170

100

120 80

70 60

20

40 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

days

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

days

Acknowledgments Authors would like to acknowledge Czech Hydrometeorological Institute, Prague and Ekotoxa, Opava who for providing measured values and land cover data.

References Bednar J, Brechler J, Halenka T, Kopacek J (2001) Modeling of summer photochemical smog in the Prague Region. Phys. Chem. Earth (B), 26, 129–136. Guenther A, Hewitt N, Erickson D, Fall R, Geron Ch, Graedel T, Harley P, Klinger L, Lerdau M, McKay WA, Pierce T, Scholes B, Steinbrecher R, Tallamraju R, Taylor J, Zimmerman P (1995) Global model of natural organic compound emissions. J. Geophys. Res., 100, 8873–8892. Simpson D, Guenther A, Hewit CN, Steinbrecher R (1995) Biogenic emissions in Europe. 1. Estimates and uncertainties. J. Geophys. Res.-Atmos., 100(D11), 22875–22890.

P4.8 Verification of Ship Plumes Modelling and Their Impacts on Air Quality and Climate Change in QUANTIFY EC 6FP Project Tomas Halenka, Peter Huszar and Michal Belda

Abstract The impact of emission from transportation on climate change is being quantified in EC FP6 Integrated Project QUANTIFY. In Activity 2 the analysis of the dilution and transformation of the emission from microscale at exhausts and plumes till mesoscale distribution will be provided from all modes of transportation. In this contribution the mesoscale simulations of ship emission impact on atomspheric pollution are studied with emphasis to compare the simulation with reality analyzed by means of flight measurement during the field campaign. In framework of the project the modeling studies are supposed to support the field campaign as well. The sensitivity of the impact on air quality and composition is analyzed as well with respect to ship emissions. Here the couple of non-hydrostatic model MM5 (PSU/NCAR) and Eulerian model CAMx (ENVIRON International Corporation, 2006) is used to support the measurement campaign. This couple with double nesting enables very high resolution both in meteorological conditions and chemistry in the region of interest, with outer domain of resolution 36 × 36 km, inner one with resolution 12 × 12 km covering the Channel. Meteorological fields generated by MM5 drive CAMx transport and dry/wet deposition. There are problems with the emission inventories available, emissions from EMEP 50 × 50 km database are interpolated and represent average ship emissions in the Channel, other emissions are combination of EMEP and UAEI (United Kingdom Atmospheric Emission Inventory). In our setting CB-IV chemistry mechanism is used (Gery et al., 1989). To see the impact of ships emission on the chemical composition at the surface we present here sensitivity test on outer domain simulation. In Figure1 ship corridors are well visible in ozone concentration field and the simulations clearly identify impact of ships emission on the chemical composition at the surface. The performance of the couple was tested first on pre-project campaign data (provided by H. Schlager). This pre-project campaign off-line test shows reasonable comparison between simulation and flight measurement. There are results for O3 displayed in Figure 2, where very good agreement can be seen in ship corridor low level flight, out of it the disagreement is due to the limited extent of CAMx coverage, it covers rather boundary layer processes. Another limitation is valid as the model is not working with actual emission data of individual ships but involving some average emission in model grid. Thus, resulting concentrations represent rather the average value across the grid and the comparison of ship corridor C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

709

710

T. Halenka et al.

background in the flight measurement is more appropriate. Individual peaks of NO concentration from the individual ship stacks captured by flight measurement cannot be resolved by the model (not shown), better comparison is provided when more complex chemical processes on longer time scale undergone.

Fig. 1 Pre-campaign simulation of chemical fields (ozone concentration, ppm). Left – simulation with ship emission included, right – no ship emission

Fig. 2 Comparison of pre-project campaign simulation to flight measurement for O3

Acknowledgments This work is supported in framework of EC FP6 Integrated project QUANTIFY (GOCE 003893) as well as under local support of the grant of Programme Informacni spolecnost, No. 1ET400300414 and Research Plan of MSMT under No. MSM 0021620860.

References ENVIRON International Corporation (2006) CAMx Users’ Guide, version 4.40 Gery MW, Whitten GZ, Killus JP, Dodge MC (1989) A Photochemical kinetics mechanism for urban and regional scale computer modeling. J. Geophys. Res., 94, 925–956.

P7.2 What Activity-Based Analysis and Personal Sampling Can Do for Assessments of Exposure to Air Pollutants? Doina Olaru and Jennifer Powell

Abstract This paper gives an example of the benefits of personal sampling (PS) and activity-based analysis (AA) for exposure assessment. Air quality and exposure studies traditionally assume the same exposure for people living in the same area and neglect individual mobility within the urban space and the the time spent indoors. This results in underestimation of true personal exposure. Combining PS with AA overcomes this limitation, as it tracks individuals through their daily routines and regards exposure at the contact point with pollutant. Two small longitudinal studies confirm significant differences in exposure profiles, despite similar activity spaces of the subjects. Substantial epidemiological evidence suggests that fixed monitoring stations measurements cannot assess the population exposure for several reasons: – Spatial and temporal resolution are too rough to identify peak exposures, more relevant from a health perspective (Michaels and Kleinman, 2000). – The real population exposure occurs as a result of conducting various activities in spatially sparse locations, and the fixed station measurement does not allow this detailed spatial analysis (Violante et al., 2006). – Population spends most of their time (more than 80%) indoors (Myers and Maynard, 2005). As exposure represents the bridge between air quality and human health risks in health studies (Curtis et al., 2006), it important to pinpoint places (microenvironments) with higher pollution concentration and how individuals spend their time there. The best way to estimate personal exposure is to consider the microenvironmental activity patterns and personal behaviour (Kotzias, 2005; Sørensen et al., 2003; Han and Naeher, 2006). This paper shows the value of spatial and temporal variation accounts in activity patterns and exposure concentration for exposure assessment. It reports on two GIS-based longitudinal studies conducted in Melbourne (2004 – 18 weeks winter and summer) and Perth (2006 – five weeks winter) for NO2 exposure of three female subjects, members of the same family (daughter, mother, and grandmother), living in the same house. Samplers were deployed in seven fixed locations (home – bedroom/bathroom/living, kitchen; work/school; in the car) and they were also worn by the subjects on a badge. The location of their daily activities and the routes were geocoded for spatial analysis and mapping of activity spaces and exposure C. Borrego and A.I. Miranda (eds.), Air Pollution Modeling and Its Application XIX. © Springer Science + Business Media B.V. 2008

717

718

D. Olaru, J. Powell

profiles. The methodology allowed us to gain considerable insights into the magnitude and variation of exposure, showing that assessment of exposure based on ambient concentrations is 20–30% lower than the personal exposure. This is consistent with previous findings presented by Violante et al. (2006) and Kaura et al. (2005). Indoor activities are significant contributors to exposure to NO2 and exposure concentration during walking had similar levels to those recorded in the car. The intra-individual variation of personal exposure depended on the day of the week and on the season, being higher on winter and on weekend days. The activity spaces of the subjects are much alike, as the young girl and the two women carry out activities close to their homes. However, the three female subjects are at different life-cycle stages, which influence their activity routines. This aspect, superimposed on the spatial variation in pollution found within the activity spaces, lead to variable exposure profiles. Even higher discrepancies arise when considering the intensity of activity. The two longitudinal case studies demonstrate how activity-based methodology combined with personal sampling is able to capture the spatial, temporal, and behavioural variation of exposure, identifying individuals potentially at greater risk, not only because of the concentrations of pollutants, but also due to the exposure durations and individual susceptibility to adverse effects. Acknowledgments The authors thank their colleagues Kate Boast and Paul Selleck from CSIRO – Marine and Atmospheric Research for analysing the samplers.

References Curtis L, Rea W, Smith-Willis P, Fenyves E, Pan Y (2006) Adverse health effects of outdoor air pollutants, Environment International 32, 815–830. Han X, Naeher LP (2006) A review of traffic-related air pollution exposure assessment studies in the developing world, Environment International 32, 106– 120. Kaura S, Nieuwenhuijsenb MJ, Colvile RN (2005) Pedestrian exposure to air pollution along a major road in Central London, UK, Atmospheric Environment 39, 7307–7320. Kotzias D (2005) Indoor air and human exposure assessment – needs and approaches, Experimental and Toxicologic Pathology 57, 5–7. Michaels RA, Kleinman MT (2000) Incidence and apparent health significance of brief airborne particle excursions, Aerosol Science and Technology 32, 92–105. Myers I, Maynard RL (2005) Polluted air – outdoors and indoors, Occupational Medicine 55, 432–438. Sørensen M, Autrup H, Møller P, Hertel O, Jensen SS, Vinzents P, Knudsen LE, Loft S (2003) Linking exposure to environmental pollutants with biological effects, Mutation Research 544, 255–271. Violante FS, Barbieri A, Curti S, Sanguinetti G, Graziosi F, Mattioli S (2006) Urban atmospheric pollution: personal exposure versus fixed monitoring station measurements, Chemosphere 64, 1722–1729.

Author Index A Aarnio P., 634 Ainslie B., 145, 146, 147, 148, 151, 591, 601 Akimoto H., 109, 136, 139, 699 Aksoyo÷lu S., 101 Alaviippola B., 634 Albergel A., 28 Albizuri A., 673 Alexandersen S., 200, 204 Alfarra M., 101 Alonso L., 673, 675 Alpert P., 360, 364 Andronopoulos S., 726 Anfossi D., 28, 82 Anlauf K., 163, 541 Arasa R., 665 Arnold D., 663 Astitha M., 507, 508 Astrup P., 200 Aulinger A., 298 B Baklanov A., 3, 4, 5, 6, 7, 10, 186, 201, 204, 640 Baldasano J., 54 Bartnicki J., 685 Bartzis J., 726, 727 Batchvarova E., 18, 22, 23 Bauer H., 712 Beck L., 667 Bedogni M., 387, 434 Belda M., 579, 710 Belfiore G., 28 Bergan T., 685 Bermúdez J., 669 Bewersdorff I., 298 Blas M., 675 Blot R., 516 Borrego C., 27, 191, 463 Bouchet V., 165, 436, 437, 472, 541, 542 Bowers J., 63 Božnar M., 697

Brandt J., 570 Brasseur O., 706 Brauer M., 591, 596, 597 Brechler J., 653, 695 Brook J., 163, 681 Builtjes P., 191, 280, 289, 528 Bullock O., 173 Burke J., 642, 722, 728 Buzzelli M., 591 Bycenkiene S., 693 C Carmichael G., 483, 485 Carnevale C., 428, 429, 431, 716 Carvalho A., 191, 198, 369 Casadei S., 387 Casares J., 679 Caseiro A., 712 Castelli S., 28, 81, 82 Chai T., 483, 487, 488 Chang J., 63 Chaxel E., 46, 47, 48 Chemel C., 46, 50 Cho S., 163, 170 Chollet J., 46, 47, 48 Christensen J., 570, 571 Christensen K., 200 Cimorelli A., 618, 625 Constantinescu E., 483, 491, 495 Costoya M., 669, 679 Cousineau S., 436, 541 Crevier L., 472 D D’Isidoro M., 423 Daescu D., 483 Davidson P., 227 Decesari S., 689 Delcloo A., 706 Denby B., 280, 283 Dennis R., 330, 351 Deutsch F., 254, 255, 454, 550 Devlin R., 722 Dharmavaram S., 445 Dimitrova R., 661 729

Author Index

730

D'Isidoro M., 422, 689, 703 Dore A., 127, 128, 133 Draxler R., 227, 228 Dudek A., 655, 656 Duhamel A., 436 Dumont G., 454 Durana N., 675 Duvall R., 722, 728 E Elbern H., 289, 295, 493 Elolähde T., 634 Emeis S., 724 F Facchini M., 689 Fierens F., 454 Finzi G., 428 Fisher B., 378, 488 Flagg D., 681 Forkel R., 724 Fox T., 728 Frohn L., 570, 571 Fuka V., 653 Fuzzi S., 689 G Gadgil A., 265 Galperin M., 701 Ganci F., 28 Ganev K., 661, 714 Gangoiti G., 673, 675 Ganor E., 360, 364 García J., 673, 675 Garcia V., 341 Geels C., 570 Gégo E., 341, 396, 416, 418 Genikhovich E., 182, 183, 187, 701 Genon L., 671 George B., 642 Gilliam R., 236, 237, 567 Gilliland A., 324, 325, 396, 414, 417, 418, 561 Gilmour I., 722 Godowitch J., 414, 418 Goldberg R., 607 Gonçalves M., 54 Gong S., 163, 436, 476

Gong W., 74, 163, 165, 229, 436, 437, 476, 477, 508, 541, 542 González R., 37, 708 Gracheva I., 182 Grašiþ B., 697 Griesser E., 724 Grsic Z., 691 Gryning S., 18, 19, 20, 22 Guerreiro C., 655 Guerrero P., 54 H Haakenstad H., 685 Halenka T., 579, 710 Hanna S., 63, 445 Hansen K., 562, 570 Harris G., 681 Hayden K., 541 He Z., 101 Hedegaard G., 570, 571, 572, 574, 576 Henderson S., 591, 595, 596, 597, 601 Hill R., 659 Hinneburg D., 90 Hogrefe C., 341, 396, 401, 403, 414, 417, 561, 607, 610, 614 Horálek J., 280, 281 Horii N., 136 Houyoux M., 608, 667 Hurley P., 30, 209, 211, 212, 213, 214, 215 Huszar P., 579, 710 I Ilardia J., 675 In H., 118 Irwin J., 396 Isakov V., 616, 618, 619, 642, 647, 729 Iza J., 675 J Janssen L., 254, 593 Janssen S., 454 Jantunen M., 720 Jenkinson P., 659

Author Index

731

K

M

Kaasik M., 333, 334, 532, 536, 538 Kallaur A., 472 Kallos G., 507, 508 Kaminski J., 6, 145, 149, 473 Kangas L., 720 Karppinen A., 307, 634, 636, 720 Kasibhatla P., 396 Katsafados P., 507 Kauhaniemi M., 634, 637 Keller J., 101 Kerschbaumer A., 72 Khajehnajafi S., 445 Kim S., 677 Kim Y., 118, 430, 653 King P., 681 Kinney P., 607 Kishcha P., 360, 361, 362, 364, 365 Kitada T., 244, 246 Knowlton K., 607 Kondragunta S., 227 Kordova L., 360 Korsholm U., 3, 6, 10, 11, 12 Koskentalo T., 634 Koslan K., 445 Kousa A., 634, 635, 636 Kryza M., 127 Ku J., 396 Kukkonen J., 154, 307, 333, 532, 634, 720 Kulmala M., 532, 533 Kurata G., 244 Kurokawa J., 136

Macdonald A., 541 Maceira P., 679 Makar P., 163, 165, 170, 436, 437, 476, 477, 541, 542, 681 Marin R., 324 Marr I., 712 Marshall J., 591, 720 Martilli A., 47, 145 Martins A., 657 Martins V., 191 Mast M., 724 Mathur R., 227, 232, 236, 498, 500 Matthias V., 298, 299 Maurizi A., 422, 689, 703 Mavromatidis E., 507 McConnell J., 145 McGee J., 722 McLaren R., 681 Meagher J., 227 Ménard S., 472, 541 Méndez M., 669 Meng F., 687 Mensink C., 254, 454, 550 Menut L., 369, 371, 487 Mihele C., 163 Mikkelsen T., 200, 201, 206 Millán M., 673 Miloshev N., 661 Miranda A., 191, 193 Mircea M., 422, 423, 689, 703 Mlakar P., 697 Mobley J., 218, 667 Monforti F., 422, 703 Monteiro A., 191, 194, 463, 465, 470 Moran M., 163, 170, 436, 437, 438, 440, 441, 442, 472, 541, 543 Morant J., 37, 708 Mortensen S., 200 Moussafir J., 28, 29

L Larson T., 591, 598, 601 Leaitch R., 541 Lee K., 118, 120, 229 Lee T., 677 Leithead A., 541 Li S., 163 Liggio J., 163, 170 Linkosalo T., 154, 156 Luecken D., 625, 626 Lutman E., 659 Lutz M., 72 Lynn B., 607

N Nagashima T., 136 Napelenok S., 324, 325, 327 Narayan J., 436 Navazo M., 673, 675 Neary L., 145

Author Index

732

Nethery E., 591 Nickovic S., 360, 361, 508 Nicolau J., 369 Nielsen S., 32, 200, 727 Niwano M., 109, 114, 699 Nollet V., 406 Nolte C., 561 Norris G., 722, 728 O Ohara T., 136, 137 Olaru D., 718 Ortega S., 665, 666 Ortega X., 663 Otte T., 236, 499 Özkaynak H., 616 P Pabla B., 541 Pace T., 218 Pavlovic R., 436 Pedersen T., 200 Pérez J., 37, 361, 708 Piazzola J., 516, 517 Pierce T., 218, 476, 499 Pinder R., 324, 417, 551 Pio C., 712 Pirovano G., 387 Piscitello E., 683 Pisoni E., 428, 716 Pleim J., 236, 237, 352 Porter P., 341, 396 Pouliot G., 218, 219, 498 Powell J., 718 Prank M., 333, 532 Prévôt A., 101 Princevac M., 315 Prodanova M., 661 Puxbaum H., 712 Q Qian W., 315, 581 Quante M., 298 Quéguiner S., 671 R Rajkovic B., 691 Ranta H., 154, 157

Rao S., 341, 344, 396, 414, 611 Reff A., 667 Reisin T., 81 Renner E., 90, 524 Riikonen K., 634 Rodriguez R., 679 Roh W., 677 Roselle S., 352, 498, 499 Rosenzweig C., 607 Røsting B., 685 Roy B., 218, 219, 222 Russo M., 657 Ruuskanen T., 340, 532, 534 S Saadi J., 218 Saavedra S., 669 Sáez A., 679 Salbu B., 685 Saltbones J, 685 Saltbones J., 701 Samaali M., 436, 442, 541 San José R., 37, 38, 708 Sandu A., 483, 487, 491, 495 Santos R., 657, 658 Sarwar G., 351, 352, 498, 667 Sassi M., 163, 436, 477, 541 Sauntry D., 163, 541 Schaap M., 191, 194, 195, 280, 282, 289, 290, 526, 551 Schäfer K., 724 Schayes G., 182, 183, 187 Schere K., 119, 173, 227, 325, 342, 352, 416, 499, 563, 608, 617, 626 Schröder W., 90 Segers A., 280, 289 Seibert P., 663 Senuta K., 693 Sfetsos A., 726 Sghirlanzoni G., 387, 388, 390 Siddans R., 289, 291 Siljamo P., 154, 307 Sills D., 681 Simeonov V., 712 Sistla G., 396 Sivertsen B., 655, 656 Sloan J., 101, 687 Soares J., 634, 720

Author Index

733

Sofiev M., 154, 155, 182, 186, 307, 308, 309, 313, 333, 334, 532, 534, 701 Sofyan A., 244, 245, 246, 247, 251 Sohn M., 265, 266, 267, 270, 272, 273 Soja A., 218, 221 Soler M., 665 Sørensen J., 200, 201, 203, 204, 206, 718 Souto J., 669, 679 Sreedharan P., 265, 273, 276 Stahl C., 625 Stendel M., 570, 571, 572 Stern R., 72 Steyn D., 145, 146, 147, 148, 151, 152, 591 Strapp J., 541 Strawbridge K., 163, 170 Stroud C., 163, 170, 436, 477, 541 Strum M., 728 Su J., 591, 598, 601 Sugata S., 136 Suppan P., 724 Sutton M., 127, 130 Sykes I., 64, 445, 447 Syrakov D., 661, 714 Szykman J., 218

Tjemkes S., 289 Todorova A., 661 Tomé M., 657 Tong D., 625, 629 Trozzi C., 683

T

W

Takahashi M., 109, 699 Takigawa M., 109, 699 Talbot D., 472 Tampieri F., 422, 689, 703 Tang S., 127, 128, 130 Tang Y., 483, 485 Tanimoto H., 136, 137 Taylor P., 398, 681 Tchepel O., 463 Tedeshi G., 516 Terrenoire E., 406, 407 Teshiba M., 109 Theobald M., 127, 134 Timmermans R., 289, 291 Tinarelli G., 28, 29, 81, 82, 697

Watkins T., 642, 728 Wayland R., 227 Wesson K., 728 White J., 63 Wiens B., 163 Williams R., 39, 528, 642, 728 Witlox H., 445, 447, 451 Wolke R., 90, 524 Wong D., 236

U Ulevicius V., 693 Uno I., 136, 137, 139 V Valdenebro V., 673 Vankerkom J., 254 Vargas A., 663 Vautard R., 289, 295, 369, 370, 389, 407 Vebra V., 693 Vellón M., 669 Venkatram A., 315, 316 Vergeiner J., 724 Vette A., 642, 728 Vieno M., 127, 128 Villa S., 683 Vitali L., 422, 703 Vogel B., 6, 351, 353, 354 Volta M., 422, 428, 716 Vujadinovic M., 691

Y Yamaji K., 136, 138, 141 Yang R., 687 Young J., 236

734

Z Zanini G., 422, 423, 703 Zanoni A., 387, 390 Zemankova K., 695 Zhan T., 315 Zhang B., 687

Author Index

Zhang J., 102, 163, 445, 476, 541, 545, 594 Zheng Q., 436 Zlatev Z., 714

Subject Index A AERMOD, 251, 253, 355, 360, 656, 658, 663, 687, 769 aerosol composition, 114, 137, 139 aerosol concentration, 155, 160, 268, 371, 372, 556, 560, 563 aerosol concentrations, 139 aerosol feedbacks, 39, 46 air quality, 40, 42, 44, 49 Air Quality Forecast, 264, 270, 512 AirQUIS, 696 analytical solution, 218, 220 AOD, 327, 335 AOT, 154–160, 330 Asian dust, 52, 155 atmospheric transport model, 163, 169, 170, 578 B BEIS3, 519, 539, 544, 608, 648 C

CMAQ, 51, 73, 80, 90, 91, 154, 155, 172, 187, 209, 265, 336, 362, 363, 424, 477, 537, 538, 539, 540, 601, 639, 646, 663, 666, 687, 703, 729, 736, 749, 769 coastal area, 176, 288, 452, 556 Complex Geometries, 695 Complex Terrain, 739 CORINAIR, 92, 292, 471, 567, 570, 735, 758 critical load, 421 D Data assimilation, 301, 318, 325, 336, 345, 353, 362, 371, 387, 527, 733 Dispersion Modelling, 71, 88, 676, 712, 739 DMS, 44 DREAM, 400 E

CAFE, 291, 294, 297, 328, 470 Calgrid, 108, 110, 725 Calmet, 725 CAMx, 81, 137–138, 143, 427, 428, 430, 433–435, 461, 547, 548, 550 CBL, 249 CFD model, 64, 70, 73, 81, 694, 701 chemical transport model, 42, 110 Chemical Transport Model, 39, 46 Chernobyl accident, 47, 705 Chimere,449, 747, 748 Chlorine Chemistry, 491 City-Delta, 334 Clean Air Act, 256, 648, 651, 654, 709, 710 Climate Change, 51, 466, 554, 601, 608, 732, 746

elemental carbon, 108, 112, 470, 477, 637, 763 EMEP, 81, 92, 128, 163, 232, 292, 325, 339, 371, 372, 410, 422, 435, 447, 471, 475, 566, 570, 611, 622, 703, 708, 713, 720, 735, 747, 751 emission reduction, 74, 80, 92, 114, 138, 189, 428, 468, 471, 473, 615, 645, 668, 673, 757, 758 ensemble Kalman filter, 318, 327, 523, 537 EURAD, 711 F Fluent, 699 frame, 118, 123, 427, 468, 470, 474, 732, 765 735

Subject Index

736

G

O

Great Lakes, 216, 276, 729 grid resolution, 88, 97, 167, 265, 273, 275, 372, 413, 620, 666, 769

one-way nesting, 74, 219, 231 organic carbon, 137, 173, 232, 269, 422, 461, 713 Ozone concentrations, 90, 379, 465, 612

H heterogeneous chemistry, 46, 231, 477, 516, 589 HIRLAM, 39, 191, 238, 346, 372, 374, 727 I Iberian Peninsula, 91, 236, 404, 711 Inverse modelling, 369 L Lagrangian Particle Model, 64 Long-Term Simulation, 172 low wind speed, 59, 64, 141 M Mediterranean region, 400 mercury deposition, 209 mercury emissions, 536 microscale urban flow, 124 mineral dust, 402, 407, 554 Minerve, 739 mixing height, 63, 640, 641, 705, 765, 766 MM5, 73, 99, 137, 155, 229, 231, 272, 283, 289, 337, 343, 392, 397, 429, 458, 505, 513, 602, 607, 618, 621, 647, 660, 666, 672, 704, 711, 725, 736, 765 MODIS/TERRA, 154 multi-objective analysis, 757 MUSCAT, 43, 52, 126, 564

P PAH, 337, 721 particle formation, 128, 131, 293, 550, 572 particulate sulfate, 525, 547 personal exposure, 632, 643, 686, 759, 769 photochemical simulation, 379 plume dispersion, 123, 132 point sources, 168, 200, 248, 292, 373, 515, 539, 582, 621, 634, 639, 640, 648, 725 PREV’AIR, 410 public health, 380, 631, 633, 647, 651, 656, 682 R RAMS, 117, 173, 179, 557, 621, 715 receptor modelling, 435 regional climate scenarios, 602 remote sensing data, 39 S Saharan Dust, 132, 400, 731 sensitivity analysis, 162, 369, 594, 615 SILAM, 191, 218, 345, 371, 572 SKIRON, 408, 548, 550 SORGAM, 110 SPM, 281 surface roughness, 171

N

T

nested grid, 225, 659 nitrate concentration, 594

TKE, 120, 146

Subject Index

U urban area, 55, 76, 83, 124, 199, 394, 427, 666, 674, 681, 718, 762, 769 urban flow, 124 urban heat island, 82, 723

737

Air Pollution Modeling and its Application XIX (NATO Science for Peace and Security Series C: Environmental Security)

Air pollution modeling and its application XIX

Energy and Environmental Challenges to Security (NATO Science for Peace and Security Series C: Environmental Security)

Environmental Security and Public Safety (NATO Science for Peace and Security Series C: Environmental Security)

Mini-Micro Fuel Cells: Fundamentals and Applications (NATO Science for Peace and Security Series C: Environmental Security) (NATO Science for Peace and Security Series C: Environmental Security)

Geographic Uncertainty in Environmental Security (NATO Science for Peace and Security Series C: Environmental Security)

Nanotechnology - Toxicological Issues and Environmental Safety (NATO Science for Peace and Security Series C: Environmental Security)

Energy Options Impact on Regional Security (NATO Science for Peace and Security Series C: Environmental Security)

Advanced Combustion and Aerothermal Technologies: Environmental Protection and Pollution Reductions (NATO Science for Peace and Security Series C: Environmental Security)

Nanomaterials: Risks and Benefits (NATO Science for Peace and Security Series C: Environmental Security)

Nanomaterials: Risks and Benefits (NATO Science for Peace and Security Series C: Environmental Security)

Virus Diseases and Crop Biosecurity (NATO Science for Peace and Security Series C: Environmental Security)

Environmental Security and Environmental Management: The Role of Risk Assessment (NATO Science for Peace and Security Series C: Environmental Security)

Complexity and Security (Nato Science for Peace and Series) (Nato Science for Peace and Series)

Green Chemical Reactions (NATO Science for Peace and Security Series C: Environmental Security)

Wastewater Reuse - Risk Assessment, Decision-Making and Environmental Security (NATO Science for Peace and Security Series C: Environmental Security)

Harbour Protection Through Data Fusion Technologies (NATO Science for Peace and Security Series C: Environmental Security)

Remote Sensing of the Atmosphere for Environmental Security (NATO Science for Peace and Security Series C: Environmental Security)

Management of Intentional and Accidental Water Pollution (NATO Science for Peace and Security Series C: Environmental Security)

Air Pollution Modeling and its Application XIV (Air Pollution Modeling and Its Application)

Exposure and Risk Assessment of Chemical Pollution - Contemporary Methodology (NATO Science for Peace and Security Series C: Environmental Security)