Air Pollution XVI (Wit Transactions on Ecology and the Environment)

Air Pollution XVI

WIT Press publishes leading books in Science and Technology. Visit our website for the current list of titles. www.witpress.com

WITeLibrary Home of the Transactions of the Wessex Institute. Papers presented at Air Pollution XVI are archived in the WIT eLibrary in volume 116 of WIT Transactions on Ecology and the Environment (ISSN 1743-3541). The WIT eLibrary provides the international scientific community with immediate and permanent access to individual papers presented at WIT conferences. http://library.witpress.com

SIXTEENTH INTERNATIONAL CONFERENCE ON MODELLING, MONITORING AND MANAGEMENT OF AIR POLLUTION

AIR POLLUTION XVI CONFERENCE CHAIRMEN C. A. Brebbia Wessex Institute of Technology, UK J.W.S. Longhurst University of the West of England, UK

INTERNATIONAL SCIENTIFIC ADVISORY COMMITTEE J. Baldasano J.G. Bartzis C. Borrego F. Costabile J.V. de Assuncao D.M. Elsom I. Hunova J.G. Irwin G. Passerini F. Patania L. Pignato V. Popov R. San Jose K. Sawicka-Kapusta F. Schwegler E. Tiezzi C. Trozzi

ORGANISED

BY

Wessex Institute of Technology, UK

SPONSORED BY WIT Transactions on Ecology and the Environment

WIT Transactions Transactions Editor Carlos Brebbia Wessex Institute of Technology Ashurst Lodge, Ashurst Southampton SO40 7AA, UK Email: [email protected]

Editorial Board B Abersek University of Maribor, Slovenia Y N Abousleiman University of Oklahoma, USA P L Aguilar University of Extremadura, Spain K S Al Jabri Sultan Qaboos University, Oman E Alarcon Universidad Politecnica de Madrid, Spain A Aldama IMTA, Mexico C Alessandri Universita di Ferrara, Italy D Almorza Gomar University of Cadiz, Spain B Alzahabi Kettering University, USA J A C Ambrosio IDMEC, Portugal A M Amer Cairo University, Egypt S A Anagnostopoulos University of Patras, Greece M Andretta Montecatini, Italy E Angelino A.R.P.A. Lombardia, Italy H Antes Technische Universitat Braunschweig, Germany M A Atherton South Bank University, UK A G Atkins University of Reading, UK D Aubry Ecole Centrale de Paris, France H Azegami Toyohashi University of Technology, Japan A F M Azevedo University of Porto, Portugal J Baish Bucknell University, USA J M Baldasano Universitat Politecnica de Catalunya, Spain J G Bartzis Institute of Nuclear Technology, Greece A Bejan Duke University, USA

M P Bekakos Democritus University of Thrace, Greece G Belingardi Politecnico di Torino, Italy R Belmans Katholieke Universiteit Leuven, Belgium C D Bertram The University of New South Wales, Australia D E Beskos University of Patras, Greece S K Bhattacharyya Indian Institute of Technology, India E Blums Latvian Academy of Sciences, Latvia J Boarder Cartref Consulting Systems, UK B Bobee Institut National de la Recherche Scientifique, Canada H Boileau ESIGEC, France J J Bommer Imperial College London, UK M Bonnet Ecole Polytechnique, France C A Borrego University of Aveiro, Portugal A R Bretones University of Granada, Spain J A Bryant University of Exeter, UK F-G Buchholz Universitat Gesanthochschule Paderborn, Germany M B Bush The University of Western Australia, Australia F Butera Politecnico di Milano, Italy J Byrne University of Portsmouth, UK W Cantwell Liverpool University, UK D J Cartwright Bucknell University, USA P G Carydis National Technical University of Athens, Greece J J Casares Long Universidad de Santiago de Compostela, Spain, M A Celia Princeton University, USA A Chakrabarti Indian Institute of Science, India

S K Chakrabarti Offshore Structure Analysis, USA A H-D Cheng University of Mississippi, USA J Chilton University of Lincoln, UK C-L Chiu University of Pittsburgh, USA H Choi Kangnung National University, Korea A Cieslak Technical University of Lodz, Poland S Clement Transport System Centre, Australia M W Collins Brunel University, UK J J Connor Massachusetts Institute of Technology, USA M C Constantinou State University of New York at Buffalo, USA D E Cormack University of Toronto, Canada M Costantino Royal Bank of Scotland, UK D F Cutler Royal Botanic Gardens, UK W Czyczula Krakow University of Technology, Poland M da Conceicao Cunha University of Coimbra, Portugal A Davies University of Hertfordshire, UK M Davis Temple University, USA A B de Almeida Instituto Superior Tecnico, Portugal E R de Arantes e Oliveira Instituto Superior Tecnico, Portugal L De Biase University of Milan, Italy R de Borst Delft University of Technology, Netherlands G De Mey University of Ghent, Belgium A De Montis Universita di Cagliari, Italy A De Naeyer Universiteit Ghent, Belgium W P De Wilde Vrije Universiteit Brussel, Belgium L Debnath University of Texas-Pan American, USA N J Dedios Mimbela Universidad de Cordoba, Spain G Degrande Katholieke Universiteit Leuven, Belgium S del Giudice University of Udine, Italy G Deplano Universita di Cagliari, Italy I Doltsinis University of Stuttgart, Germany M Domaszewski Universite de Technologie de Belfort-Montbeliard, France

J Dominguez University of Seville, Spain K Dorow Pacific Northwest National Laboratory, USA W Dover University College London, UK C Dowlen South Bank University, UK J P du Plessis University of Stellenbosch, South Africa R Duffell University of Hertfordshire, UK A Ebel University of Cologne, Germany E E Edoutos Democritus University of Thrace, Greece G K Egan Monash University, Australia K M Elawadly Alexandria University, Egypt K-H Elmer Universitat Hannover, Germany D Elms University of Canterbury, New Zealand M E M El-Sayed Kettering University, USA D M Elsom Oxford Brookes University, UK A El-Zafrany Cranfield University, UK F Erdogan Lehigh University, USA F P Escrig University of Seville, Spain D J Evans Nottingham Trent University, UK J W Everett Rowan University, USA M Faghri University of Rhode Island, USA R A Falconer Cardiff University, UK M N Fardis University of Patras, Greece P Fedelinski Silesian Technical University, Poland H J S Fernando Arizona State University, USA S Finger Carnegie Mellon University, USA J I Frankel University of Tennessee, USA D M Fraser University of Cape Town, South Africa M J Fritzler University of Calgary, Canada U Gabbert Otto-von-Guericke Universitat Magdeburg, Germany G Gambolati Universita di Padova, Italy C J Gantes National Technical University of Athens, Greece L Gaul Universitat Stuttgart, Germany A Genco University of Palermo, Italy N Georgantzis Universitat Jaume I, Spain G S Gipson Oklahoma State University, USA P Giudici Universita di Pavia, Italy F Gomez Universidad Politecnica de Valencia, Spain

R Gomez Martin University of Granada, Spain D Goulias University of Maryland, USA K G Goulias Pennsylvania State University, USA F Grandori Politecnico di Milano, Italy W E Grant Texas A & M University, USA S Grilli University of Rhode Island, USA R H J Grimshaw, Loughborough University, UK D Gross Technische Hochschule Darmstadt, Germany R Grundmann Technische Universitat Dresden, Germany A Gualtierotti IDHEAP, Switzerland R C Gupta National University of Singapore, Singapore J M Hale University of Newcastle, UK K Hameyer Katholieke Universiteit Leuven, Belgium C Hanke Danish Technical University, Denmark K Hayami National Institute of Informatics, Japan Y Hayashi Nagoya University, Japan L Haydock Newage International Limited, UK A H Hendrickx Free University of Brussels, Belgium C Herman John Hopkins University, USA S Heslop University of Bristol, UK I Hideaki Nagoya University, Japan D A Hills University of Oxford, UK W F Huebner Southwest Research Institute, USA J A C Humphrey Bucknell University, USA M Y Hussaini Florida State University, USA W Hutchinson Edith Cowan University, Australia T H Hyde University of Nottingham, UK M Iguchi Science University of Tokyo, Japan D B Ingham University of Leeds, UK L Int Panis VITO Expertisecentrum IMS, Belgium N Ishikawa National Defence Academy, Japan J Jaafar UiTm, Malaysia W Jager Technical University of Dresden, Germany

Y Jaluria Rutgers University, USA C M Jefferson University of the West of England, UK P R Johnston Griffith University, Australia D R H Jones University of Cambridge, UK N Jones University of Liverpool, UK D Kaliampakos National Technical University of Athens, Greece N Kamiya Nagoya University, Japan D L Karabalis University of Patras, Greece M Karlsson Linkoping University, Sweden T Katayama Doshisha University, Japan K L Katsifarakis Aristotle University of Thessaloniki, Greece J T Katsikadelis National Technical University of Athens, Greece E Kausel Massachusetts Institute of Technology, USA H Kawashima The University of Tokyo, Japan B A Kazimee Washington State University, USA S Kim University of Wisconsin-Madison, USA D Kirkland Nicholas Grimshaw & Partners Ltd, UK E Kita Nagoya University, Japan A S Kobayashi University of Washington, USA T Kobayashi University of Tokyo, Japan D Koga Saga University, Japan A Konrad University of Toronto, Canada S Kotake University of Tokyo, Japan A N Kounadis National Technical University of Athens, Greece W B Kratzig Ruhr Universitat Bochum, Germany T Krauthammer Penn State University, USA C-H Lai University of Greenwich, UK M Langseth Norwegian University of Science and Technology, Norway B S Larsen Technical University of Denmark, Denmark F Lattarulo, Politecnico di Bari, Italy A Lebedev Moscow State University, Russia L J Leon University of Montreal, Canada D Lewis Mississippi State University, USA S lghobashi University of California Irvine, USA

K-C Lin University of New Brunswick, Canada A A Liolios Democritus University of Thrace, Greece S Lomov Katholieke Universiteit Leuven, Belgium J W S Longhurst University of the West of England, UK G Loo The University of Auckland, New Zealand J Lourenco Universidade do Minho, Portugal J E Luco University of California at San Diego, USA H Lui State Seismological Bureau Harbin, China C J Lumsden University of Toronto, Canada L Lundqvist Division of Transport and Location Analysis, Sweden T Lyons Murdoch University, Australia Y-W Mai University of Sydney, Australia M Majowiecki University of Bologna, Italy D Malerba Università degli Studi di Bari, Italy G Manara University of Pisa, Italy B N Mandal Indian Statistical Institute, India Ü Mander University of Tartu, Estonia H A Mang Technische Universitat Wien, Austria, G D, Manolis, Aristotle University of Thessaloniki, Greece W J Mansur COPPE/UFRJ, Brazil N Marchettini University of Siena, Italy J D M Marsh Griffith University, Australia J F Martin-Duque Universidad Complutense, Spain T Matsui Nagoya University, Japan G Mattrisch DaimlerChrysler AG, Germany F M Mazzolani University of Naples “Federico II”, Italy K McManis University of New Orleans, USA A C Mendes Universidade de Beira Interior, Portugal, R A Meric Research Institute for Basic Sciences, Turkey J Mikielewicz Polish Academy of Sciences, Poland

N Milic-Frayling Microsoft Research Ltd, UK R A W Mines University of Liverpool, UK C A Mitchell University of Sydney, Australia K Miura Kajima Corporation, Japan A Miyamoto Yamaguchi University, Japan T Miyoshi Kobe University, Japan G Molinari University of Genoa, Italy T B Moodie University of Alberta, Canada D B Murray Trinity College Dublin, Ireland G Nakhaeizadeh DaimlerChrysler AG, Germany M B Neace Mercer University, USA D Necsulescu University of Ottawa, Canada F Neumann University of Vienna, Austria S-I Nishida Saga University, Japan H Nisitani Kyushu Sangyo University, Japan B Notaros University of Massachusetts, USA P O’Donoghue University College Dublin, Ireland R O O’Neill Oak Ridge National Laboratory, USA M Ohkusu Kyushu University, Japan G Oliveto Universitá di Catania, Italy R Olsen Camp Dresser & McKee Inc., USA E Oñate Universitat Politecnica de Catalunya, Spain K Onishi Ibaraki University, Japan P H Oosthuizen Queens University, Canada E L Ortiz Imperial College London, UK E Outa Waseda University, Japan A S Papageorgiou Rensselaer Polytechnic Institute, USA J Park Seoul National University, Korea G Passerini Universita delle Marche, Italy B C Patten, University of Georgia, USA G Pelosi University of Florence, Italy G G Penelis, Aristotle University of Thessaloniki, Greece W Perrie Bedford Institute of Oceanography, Canada R Pietrabissa Politecnico di Milano, Italy H Pina Instituto Superior Tecnico, Portugal M F Platzer Naval Postgraduate School, USA D Poljak University of Split, Croatia

V Popov Wessex Institute of Technology, UK H Power University of Nottingham, UK D Prandle Proudman Oceanographic Laboratory, UK M Predeleanu University Paris VI, France M R I Purvis University of Portsmouth, UK I S Putra Institute of Technology Bandung, Indonesia Y A Pykh Russian Academy of Sciences, Russia F Rachidi EMC Group, Switzerland M Rahman Dalhousie University, Canada K R Rajagopal Texas A & M University, USA T Rang Tallinn Technical University, Estonia J Rao Case Western Reserve University, USA A M Reinhorn State University of New York at Buffalo, USA A D Rey McGill University, Canada D N Riahi University of Illinois at UrbanaChampaign, USA B Ribas Spanish National Centre for Environmental Health, Spain K Richter Graz University of Technology, Austria S Rinaldi Politecnico di Milano, Italy F Robuste Universitat Politecnica de Catalunya, Spain J Roddick Flinders University, Australia A C Rodrigues Universidade Nova de Lisboa, Portugal F Rodrigues Poly Institute of Porto, Portugal C W Roeder University of Washington, USA J M Roesset Texas A & M University, USA W Roetzel Universitaet der Bundeswehr Hamburg, Germany V Roje University of Split, Croatia R Rosset Laboratoire d’Aerologie, France J L Rubio Centro de Investigaciones sobre Desertificacion, Spain T J Rudolphi Iowa State University, USA S Russenchuck Magnet Group, Switzerland H Ryssel Fraunhofer Institut Integrierte Schaltungen, Germany S G Saad American University in Cairo, Egypt

M Saiidi University of Nevada-Reno, USA R San Jose Technical University of Madrid, Spain F J Sanchez-Sesma Instituto Mexicano del Petroleo, Mexico B Sarler Nova Gorica Polytechnic, Slovenia S A Savidis Technische Universitat Berlin, Germany A Savini Universita de Pavia, Italy G Schmid Ruhr-Universitat Bochum, Germany R Schmidt RWTH Aachen, Germany B Scholtes Universitaet of Kassel, Germany W Schreiber University of Alabama, USA A P S Selvadurai McGill University, Canada J J Sendra University of Seville, Spain J J Sharp Memorial University of Newfoundland, Canada Q Shen Massachusetts Institute of Technology, USA X Shixiong Fudan University, China G C Sih Lehigh University, USA L C Simoes University of Coimbra, Portugal A C Singhal Arizona State University, USA P Skerget University of Maribor, Slovenia J Sladek Slovak Academy of Sciences, Slovakia V Sladek Slovak Academy of Sciences, Slovakia A C M Sousa University of New Brunswick, Canada H Sozer Illinois Institute of Technology, USA D B Spalding CHAM, UK P D Spanos Rice University, USA T Speck Albert-Ludwigs-Universitaet Freiburg, Germany C C Spyrakos National Technical University of Athens, Greece I V Stangeeva St Petersburg University, Russia J Stasiek Technical University of Gdansk, Poland G E Swaters University of Alberta, Canada S Syngellakis University of Southampton, UK J Szmyd University of Mining and Metallurgy, Poland S T Tadano Hokkaido University, Japan

H Takemiya Okayama University, Japan I Takewaki Kyoto University, Japan C-L Tan Carleton University, Canada M Tanaka Shinshu University, Japan E Taniguchi Kyoto University, Japan S Tanimura Aichi University of Technology, Japan J L Tassoulas University of Texas at Austin, USA M A P Taylor University of South Australia, Australia A Terranova Politecnico di Milano, Italy E Tiezzi University of Siena, Italy A G Tijhuis Technische Universiteit Eindhoven, Netherlands T Tirabassi Institute FISBAT-CNR, Italy S Tkachenko Otto-von-GuerickeUniversity, Germany N Tosaka Nihon University, Japan T Tran-Cong University of Southern Queensland, Australia R Tremblay Ecole Polytechnique, Canada I Tsukrov University of New Hampshire, USA R Turra CINECA Interuniversity Computing Centre, Italy S G Tushinski Moscow State University, Russia J-L Uso Universitat Jaume I, Spain E Van den Bulck Katholieke Universiteit Leuven, Belgium D Van den Poel Ghent University, Belgium R van der Heijden Radboud University, Netherlands R van Duin Delft University of Technology, Netherlands

P Vas University of Aberdeen, UK W S Venturini University of Sao Paulo, Brazil R Verhoeven Ghent University, Belgium A Viguri Universitat Jaume I, Spain Y Villacampa Esteve Universidad de Alicante, Spain F F V Vincent University of Bath, UK S Walker Imperial College, UK G Walters University of Exeter, UK B Weiss University of Vienna, Austria H Westphal University of Magdeburg, Germany J R Whiteman Brunel University, UK Z-Y Yan Peking University, China S Yanniotis Agricultural University of Athens, Greece A Yeh University of Hong Kong, China J Yoon Old Dominion University, USA K Yoshizato Hiroshima University, Japan T X Yu Hong Kong University of Science & Technology, Hong Kong M Zador Technical University of Budapest, Hungary K Zakrzewski Politechnika Lodzka, Poland M Zamir University of Western Ontario, Canada R Zarnic University of Ljubljana, Slovenia G Zharkova Institute of Theoretical and Applied Mechanics, Russia N Zhong Maebashi Institute of Technology, Japan H G Zimmermann Siemens AG, Germany

Air Pollution XVI

Editors C.A. Brebbia Wessex Institute of Technology, UK J.W.S. Longhurst University of the West of England, UK

Editors C.A. Brebbia Wessex Institute of Technology, UK J.W.S. Longhurst University of the West of England, UK

Published by WIT Press Ashurst Lodge, Ashurst, Southampton, SO40 7AA, UK Tel: 44 (0) 238 029 3223; Fax: 44 (0) 238 029 2853 E-Mail: [email protected] http://www.witpress.com For USA, Canada and Mexico WIT Press 25 Bridge Street, Billerica, MA 01821, USA Tel: 978 667 5841; Fax: 978 667 7582 E-Mail: [email protected] http://www.witpress.com British Library Cataloguing-in-Publication Data A Catalogue record for this book is available from the British Library ISBN: 978-1-84564-127-6 ISSN: (print) 1746-448X ISSN: (on-line) 1734-3541 The texts of the papers in this volume were set individually by the authors or under their supervision.Only minor corrections to the text may have been carried out by the publisher. No responsibility is assumed by the Publisher, the Editors and Authors for any injury and/ or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. The Publisher does not necessarily endorse the ideas held, or views expressed by the Editors or Authors of the material contained in its publications. © WIT Press 2008 Printed in Great Britain by Athenaeum Press Ltd. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of the Publisher.

Preface

This volume contains the reviewed papers accepted for the Sixteenth International Conference on Modelling, Monitoring and Management of Air Pollution held in Skiathos, Greece in September 2008. This successful international meeting builds upon the prestigious outcomes of the 15 preceding conferences beginning with Monterrey, Mexico in 1993 and most recently in the Algarve, Portugal in 2007. These meetings have attracted outstanding contributions from leading researchers from around the world with the presented papers permanently stored in the WIT eLibrary as Transactions of the Wessex Institute (see http://library.witpress.com/) Air pollution remains one of the most challenging problems facing the international community; it is widespread and growing in importance and has clear and known impacts on health and the environment. The human need for transport, manufactured goods and services brings with it often unintended, but none the less real, impacts on the atmospheric environment at scales from the local to the global. Whilst there are good examples of regulatory successes in minimising such impacts the rate of development of the global economy bring new pressures and the willingness of governments to regulate air pollution is often balanced by concerns over the economic impact of regulation. Science is the key to identifying the nature and scale of air pollution impacts and is essential in the formulation of polices and regulatory decision-making. Continuous improvements in our knowledge of the fundamental science of air pollution and its application are necessary if we are to properly predict, assess and mitigate the air pollution implications of changes to the interlinked local, regional, national and international economic systems. Science must also be able to provide the evidence of improvements to air quality that result from implementation of the mitigation measure or the control regulation. The ability to assess and mitigate using the precautionary principle is a challenge that science must grasp and position itself to convince decision makers that uncertainty does not mean inertia. The outcomes of such activities must also be translatable into a suitable format to assist policy makers in reaching sustainable decisions and to build public acceptance and understanding of the nature and scale of the air pollution problem.

This volume brings together contributions from scientist around the world to discuss their recent work on various aspects of the air pollution phenomena. The conference series provides opportunities to foster scientific exchange, develop new collaborations amongst scientists and between scientists and policy makers or regulators, and provides a means for identifying new areas for scientific investigations. The Editors of this book wish to thank the authors for their contributions and to acknowledge the assistance of the eminent members of the International Scientific Advisory Committee with the organisation of the conference and particularly with reviewing of the submitted papers. WIT proceedings are regularly recorded and reported in abstracting services and databases including Crossref, Cambridge Scientific Abstracts, Scopus, Compendex, GeoBase, INSPEC, Thompson Index to Scientific and Technical proceedings, Scitech Book News and the Directory of Published Proceedings. The Editors Skiathos, 2008

Contents Section 1: Air pollution modelling Elevated PM10 and PM2.5 concentrations in Europe: a model experiment with MM5-CMAQ and WRF-CHEM R. San José, J. L. Pérez, J. L. Morant & R. M. González .....................................3 Analytic solutions of the diffusion-deposition equation for fluids heavier than atmospheric air F. P. Barrera, T. Brugarino, V. Piazza & L. Pignato .........................................13 Air quality forecasting in a large city P. Perez ...............................................................................................................21 Meshing effects on CFD modelling of atmospheric flow over buildings situated on ground with high terrain A. Karim, P. Nolan & A. Qubian ........................................................................29 Enhanced evaluation of a Lagrangian-particle air pollution model based on a Šaleška region field data set B. Grašič, P. Mlakar, M. Z. Božnar & G. Tinarelli ............................................39 Modelling the multi year air quality time series in Edinburgh: an application of the Hierarchical Profiling Approach H. Al-Madfai, D. G. Snelson & A. J. Geens........................................................49 Definition of PM10 emission factors from traffic: use of tracers and definition of background concentration E. Brizio & G. Genon..........................................................................................57 LNG dispersion over the sea A. Fatsis, J. Statharas, A. Panoutsopoulou & N. Vlachakis ...............................67 Modelling of air pollutants released from highway traffic in Hungary Gy. Baranka ........................................................................................................77

Dust barriers in open pit blasts. Multiphase Computational Fluid Dynamics (CFD) simulations J. T. Alvarez, I. D. Alvarez & S. T. Lougedo.......................................................85 Contribution of oil palm isoprene emissions to tropospheric ozone levels in the Distrito Metropolitano de Quito (Ecuador) R. Parra...............................................................................................................95 Some reflections on the modelling of biogenic emissions of monoterpenes in the boreal zone K. M. Granström ...............................................................................................105 An evaluation of SOA modelling in the Madrid metropolitan area M. G. Vivanco, I. Palomino, J. Plaza, B. Artíñano, M. Pujadas & B. Bessagnet ..................................................................................................115 Comparison between ozone monitoring data and modelling data, in Italy, from the perspective of health indicator assessments A. De Marco, A. Screpanti, S. Racalbuto, T. Pignatelli, G. Vialetto, F. Monforti & G. Zanini....................................................................................125 Section 2: Air quality management Dealing with air pollution in Europe C. Trozzi, R. Vaccaro & C. Leonardi................................................................137 The development and operation of the United Kingdom’s air quality management regime J. W. S. Longhurst, J. G. Irwin, T. J. Chatterton, E. T. Hayes, N. S. Leksmono & J. K. Symons ........................................................................149 Integrating local air quality and carbon management at a regional and local governance level: a case study of south west England S. T. Baldwin, M. Everard, E. T. Hayes, J. W. S. Longhurst & J. R. Merefield...............................................................................................159 Failures and successes in the implementation of an air quality management program in Mexicali, Baja California, Mexico M. Quintero-Nuñez & E. C. Nieblas-Ortiz........................................................169 Emission management system in the Russian Federation: necessity for reforming and future adaptation of the western experience A. Y. Nedre, R. A. Shatilov & A. F. Gubanov....................................................179

Outdoor air quality data analysis of Al-Mansoriah residential area (Kuwait): air quality indices results S. A. Al-Haider & S. M. Al-Salem .....................................................................189 A modelling tool for PM10 exposure assessment: an application example E. Angelino, M. P. Costa, E. Peroni & C. Sala.................................................197 Air pollution and management in the Niger Delta – emerging issues M. A. Fagbeja, T. J. Chatterton, J. W. S. Longhurst, J. O. Akinyede & J. O. Adegoke ................................................................................................207 Air quality monitoring and management for the industrialized Highveld region of South Africa G. V. Mkhatshwa ...............................................................................................217 Real time air quality forecasting systems for industrial plants and urban areas by using the MM5-CMAQ-EMIMO R. San José, J. L. Pérez, J. L. Morant & R. M. González .................................225 Section 3: Emission studies Air quality in the vicinity of a governmental school in Kuwait E. Al-Bassam, V. Popov & A. Khan ..................................................................237 Emissions of nitrogen dioxide from modern diesel vehicles G. A. Bishop & D. H. Stedman..........................................................................247 Errors in model predictions of NOx traffic emissions at road level – impacts of input data quality R. Smit ...............................................................................................................255 Air pollution from traffic, ships and industry in an Italian port G. Fava & M. Letizia Ruello.............................................................................271 Fugitive dust from agricultural land affecting air quality within the Columbia Plateau, USA B. S. Sharratt .....................................................................................................281 Emission inventory for urban transport in the rush hour: application to Seville J. Racero, M. Cristina Martín, I. Eguía & F. Guerrero ...................................291

Modeling carbon emissions from urban land conversion: gamma distribution model A. Svirejeva-Hopkins & H.-J. Schellnhuber......................................................301 Reduction of CO2 emissions by carbonation of alkaline wastewater M. Uibu, O. Velts, A. Trikkel & R. Kuusik ........................................................311 Section 4: Monitoring and measuring Characterisation of inhalable atmospheric aerosols N. A. Kgabi, J. J. Pienaar & M. Kulmala .........................................................323 Data handling of complex GC-MS signals to characterize homologous series as organic source tracers in atmospheric aerosols M. C. Pietrogrande, M. Mercuriali & D. Bacco...............................................335 Monitoring of trace organic air pollutants – a developing country perspective P. B. C. Forbes & E. R. Rohwer .......................................................................345 NIST gas standards containing volatile organic compounds in support of ambient air pollution measurements G. C. Rhoderick.................................................................................................357 A comparison of EPA and EN requirements for nitrogen oxide chemiluminescence analyzers J. Barberá, M. Doval, E. González, A. Miñana & F. J. Marzal........................367 A procedure for correcting readings in chemiluminescence nitrogen oxide analyzers due to the effect of sample pressure M. Doval, J. Barberá, E. González & F. J. Marzal ..........................................375 New measures of wind angular dispersion in three dimensions P. S. Farrugia & A. Micallef.............................................................................385 Variation of air pollution with related meteorological factors in Tripoli (case study) T. A. Sharif, A. K. El-Henshir & M. M. Treban ................................................397

Section 5: Urban air management Prediction of air pollution levels using neural networks: influence of spatial variability G. Ibarra-Berastegi, A. Elias, A. Barona, J. Sáenz, A. Ezcurra & J. Diaz de Argandona....................................................................................409 Environmental planning and management of air quality: the case of Mexicali, Baja California, Mexico E. Corona-Zambrano & R. Rojas-Caldelas......................................................419 High-resolution air quality modelling and time scale analysis of ozone and NOx in Osaka, Japan K. L. Shrestha, A. Kondo, A. Kaga & Y. Inoue .................................................429 Practical problems associated with assessing the impact of outdoor smoking on outdoor air quality: an Edinburgh study D. G. Snelson, A. J. Geens, H. Al-Madfai & D. Hillier ....................................439 Role of leaf- and rhizosphere-associated bacteria in reducing air pollution of industrial cities in Saudi Arabia M. A. Khiyami ...................................................................................................447 Section 6: Indoor air pollution Indoor concentrations of VOCs and ozone in two cities of Northern Europe during the summer period J. G. Bartzis, S. Michaelidou, D. Missia, E. Tolis, D. Saraga, E. Demetriou-Georgiou, D. Kotzias & J. M. Barero-Moreno ..........................459 Comparative study of indoor-outdoor exposure against volatile organic compounds in South and Middle America O. Herbarth, A. Müller, L. Massolo & H. Tovalin ............................................467 Sampling of respirable particle PM10 in the library at the Metropolitana University, Campus Azcapotzalco, Mexico City Y. I. Falcon, E. Martinez, A. Cuenca, C. Herrera & E. A. Zavala ...................475 Impacts of ventilation: studies on “environmental tobacco smoke” A. J. Geens, H. Al-Madfai & D. G. Snelson......................................................483 PCB contamination in indoor buildings S. J. Hellman, O. Lindroos, T. Palukka, E. Priha, T. Rantio & T. Tuhkanen...................................................................................................491

Evaluation of Indoor Air treatment by two pilot-scale biofilters packed with compost and compost-based material M. Ondarts, C. Hort, V. Platel & S. Sochard....................................................499 Section 7: Aerosols and particles The role of PM10 in air quality and exposure in urban areas C. Borrego, M. Lopes, J. Valente, O. Tchepel, A. I. Miranda & J. Ferreira .....................................................................................................511 Spatial distribution of ultrafine particles at urban scale: the road-to-ambient stage F. Costabile, B. Zani & I. Allegrini...................................................................521 Electromagnetic and informational pollution as a co-challenge to air pollution A. A. Berezin & V. V. Gridin .............................................................................533 Genotoxic and oxidative damage related to PM2.5 chemical fraction Sa. Bonetta, V. Gianotti, D. Scozia, Si. Bonetta, E. Carraro, F. Gosetti, M. Oddone & M. C. Gennaro .........................................................543 Preliminary results of aerosol optical thickness from MIVIS data C. Bassani, R. M. Cavalli & S. Pignatti............................................................551 Section 8: Air pollution effects on ecosystems Response of lichens to heavy metal and SO2 pollution in Poland – an overview K. Sawicka-Kapusta, M. Zakrzewska, G. Bydłoń & A. Sowińska .....................561 Forest ecosystem development after heavy deposition loads – case study Dübener Heide C. Fürst, M. Abiy & F. Makeschin....................................................................571 Impact of biogenic volatile organic compound emissions on ozone formation in the Kinki region, Japan A. Kondo, B. Hai, K. L. Shrestha, A. Kaga & Y. Inoue.....................................585

Section 9: Policy studies Potential contribution of local air quality management to environmental justice in England I. Gegisian, M. Grey, J. W. S. Longhurst & J. G. Irwin....................................597 Are environmental health officers and transport planners in English local authorities working together to achieve air quality objectives? A. O. Olowoporoku, E. T. Hayes, N. S. Leksmono, J. W. S. Longhurst & G. P. Parkhurst...............................................................607 A fuzzy MCDM framework for the environmental pollution potential of industries focusing on air pollution R. K. Lad, R. A. Christian & A. W. Deshpande.................................................617 A Model Municipal By-Law for regulating wood burning appliances A. Germain, F. Granger & A. Gosselin ............................................................627 Author Index ...................................................................................................637

This page intentionally left blank

Section 1 Air pollution modelling

This page intentionally left blank

Air Pollution XVI

3

Elevated PM10 and PM2.5 concentrations in Europe: a model experiment with MM5-CMAQ and WRF-CHEM R. San José1, J. L. Pérez1, J. L. Morant1 & R. M. González2 1

Environmental Software and Modelling Group, Computer Science School, Technical University of Madrid (UPM), Campus de Montegancedo, Madrid, Spain 2 Department of Meteorology and Geophysics, Faculty of Physics, Complutense University of Madrid; Ciudad Universitaria, Madrid, Spain

Abstract We have applied the MM5-CMAQ model to simulate the high concentrations in PM10 and PM2.5 during a winter episode (2003) in Central Europe. The selected period is January 15 – April 6 2003. Values of daily mean concentrations up to 75 µgm-3 are found on average of several monitoring stations in Northern Germany. This model evaluation shows that there is an increasing underestimation of primary and secondary species with increasing observed PM10. The high PM levels were observed under stagnant weather conditions that are difficult to simulate. The MM5 is the PSU/NCAR non-hydrostatic meteorological model and CMAQ is the chemical dispersion model developed by EPA (US) used in this simulation with CBM-V. The TNO emission inventory was used to simulate the PM10 and PM2.5 concentrations with the MM5-CMAQ model. The results show a substantial underestimation of the elevated values in February and March 2003. An increase on the PM2.5 emissions (five times) produces the expected results and the correlation coefficient increases slightly. The WRF/CHEM model results show an excellent performance with correct emission database. The main difference between MM5-CMAQ simulations and WRF/CHEM is the MOSAIC particle models and the “classical” MADE/SORGAM particle model used in WRF/CHEM and CMAQ respectively. MOSAIC seems to make a better job than MADE particle model for this particular episode. Keywords: emissions, PM10 and PM2.5, air quality models, air particles. WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line) doi:10.2495/AIR080011

4 Air Pollution XVI

1

Introduction

Simulations of elevated PM10 and PM2.5 concentrations have been always underestimated by modern three dimensional air quality modelling tools. This fact has focused much more attention between researchers during last years. Three dimensional air quality models have been developed during the last 15–20 years and substantial progress has occurred in this research area. These models are composed by a meteorological driver and a chemical and transport module. Examples of meteorological drivers are: MM5 (PSU.NCAR, USA) [5], RSM (NOAA, USA), ECMWF (Redding, U.K.), HIRLAM (Finnish Meteorological Institute, Finland), WRF [15] and examples of dispersion and chemical transport modules are EURAD (University of Cologne, Germany) [13], EUROS (RIVM, The Netherlands) [7], EMEP Eulerian (DNMI, Oslo, Norway), MATCH (SMHI, Norrkoping, Sweden) [2], REM3 (Free University of Berlin, Germany) [14], CHIMERE (ISPL, Paris, France) [12], NILU-CTM (NILU, Kjeller, Norway) [3], LOTOS (TNO, Apeldoorm, The Netherlands) [8], DEM (NERI, Roskilde, Denmark) [4], OPANA model [9–11] based on MEMO and MM5 mesoscale meteorological models and with the chemistry on-line solved by [6], STOCHEM (UK Met. Office, Bracknell, U.K.) [1] and CMAQ (Community Multiscale Air Quality modelling system) [16], developed by EPA (USA). In USA, CAMx Environ Inc., STEM-III (University of Iowa) and CMAQ model are the most up-to-date air quality dispersion chemical models. In this application we have used the CMAQ model (EPA, U.S.) which is one of the most complete models and includes aerosol, cloud and aerosol chemistry. In this contribution we present results from two simulations by two different models. The first air quality modelling systems is MM5-CMAQ which is a matured modelling system based on the MM5 mesoscale non-hydrostatic meteorological model and the dispersion and chemical transport module, CMAQ. The second tool is the WRF/CHEM [15] air quality modelling system, which is an on-line (one code, one system) tool to simulate air concentrations based on the WRF meteorological driver. In WRF/CHEM the chemistry transport and transformations are embedded into WRF as part of the code so that the interactions between many meteorological and climate variables and the chemistry if at hand and can be investigated. WRF/CHEM is developed by NOAA/NCAR (US) [15]. The advantage of on-line models is based on the capability to analyze all variables simultaneously and to account for all interactions (or at least, as much as possible) with a full modular approach.

2 PM10 and PM2.5 episode During the period January 15 2003 to April 5 2003 in central Europe (mainly northern part of Germany), we observe three high peaks on PM10 and PM2.5 values in several monitoring stations located in the area of North-East of Germany. The daily averages of PM10 concentrations were close to 80 µgm-3 and higher than 70 µgm-3 for PM2.5 concentrations. These values are about 4–5 times higher than those registered as “normal” values. The first peak on PM10 and PM2.5 concentrations was developed after Feb. 1 until Feb. 15. During this WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

Air Pollution XVI

5

period of time, Central Europe was under the influence of a high-pressure system coming from Russia through Poland and Souther Scandinavia. In Northern part of Germany, we found southeasterly winds and stable conditions with low winds. These meteorological conditions brought daily PM10 concentrations at about 40 µgm-3. The second peak was characterized by a sharp gradient on PM10 concentrations after Feb. 15 and until March 7. These episode reached daily PM10 concentrations up to 70 µgm-3. The meteorological conditions on March 2 (peak values) was characterized by a wind rotation composed by Southwesterly winds from Poland over the North of Germany and Northwesterly and Western winds in the Central part of Germany. Finally a third peak with values of about 65 µgm-3 on March 27 starts on March 20, ending on April 5 2003 was having a similar structure and causes to the second one.

3

Emission data

In both models, we have applied the TNO emissions [17] as area and point sources with a geographical resolution of 0.125º latitude by 0.25º longitude and covering all Europe. The emission totals by SNAP activity sectors and countries agree with the baseline scenario for the Clean Air For Europe (CAFE) program [18]. This database gives the PM10 and PM2.5 emission for the primary particle emissions. We also took from CAFE the PM splitting sub-groups, height distribution and the breakdown of the annual emissions into hourly emissions. The PM2.5 fraction of the particle emissions was split into an unspecified fraction, elemental carbon (EC) and primary organic carbon (OC). The EC fraction of the PM2.5 emissions for the different SNAP sectors were taken from [19]. For the OC fraction, the method proposed by [20] is applied as follows: an average OC/EC emission ratio of two was used for all sectors, i.e. the OC fraction were set as twice the EC fractions, except if the sum of the two fractions exceed the unity. In this case (fEC > 0.33), fOC was set as: fOC = 1 – fEC. With this prepared input, the WRF/CHEM and CMAQ took the information as it is. The hourly emissions are derived using sectordependent, monthly, daily and hourly emission factors as used in the EURODELTA (http://aqm.jrc.it/eurodelta/) exercise.

4

Observational data

Eighteen PM10 stations were selected for the comparison with the model results. Seventeen stations represent the rural background and one station represents the urban background in Berlin. All stations are located in flat or moderate hill terrain. Most of the stations are operated by the respective Federal State agencies. At four stations (Neuglobsow, Zingst, Westerland and Deuselbach, which are EMEP background stations run by the German Environmental Protection Agency, Umweltbundesamt), the observed concentrations of particulate sulphate, total nitrate (HNO3+NO3-) and total ammonia (NH3+NH4+) were available. Deuselbach, in the southwest of Germany, is located outside of the high PM10 concentration region. In addition, at the research station Melpitz [21] the concentrations of the components of secondary inorganic aerosols SO4--, NO3-, NH4+, as well as the concentrations of EC, OC and NH3 were available. WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

6 Air Pollution XVI The SO2 and NO2 concentrations at these five stations were also taken into account in the model comparison. PM2.5 observations were available at four stations: Melpitz, Waldhof, Deuselbach and Hannover. All PM10 and PM2.5 observations are based on gravimetric measurements, and the concentrations of the inorganic species in aerosol particles on ion chromotography. The chemical composition data at Melpitz result from the PM2.5 fraction, whereas the composition data from the other stations were analyzed from the PM10 particle concentrations. OC data were corrected by a factor of 1.4 to account for the nonC atoms in the particulate organic matter (OM) concentrations, which are currently not measured [22].

5

MM5-CMAQ and WRF-CHEM architectures and configurations

MM5 was set up with two domains: a mother domain with 60x60 grid cells with 90 km spatial resolution and 23 vertical layers and 61x61 grid cells with 30 km spatial resolution with 23 vertical layers. The central point is set at 50.0 N and 10.0 E. The model is run with Lambert Conformal Conical projection. The CMAQ domain is slightly smaller following the CMAQ architecture rules. We use reanalysis T62 (209 km) datasets as 6-hour boundary conditions for MM5 with 28 vertical sigma levels and nudging with meteorological observations for the mother domain. We run MM5 with two-way nesting capability. We use the Kain-Fritsch 2 cumulus parameterization scheme, the MRF PBL scheme, Schultz microphysics scheme and Noah land-surface model. In CMAQ we use clean boundary profiles for initial conditions, Yamartino advection scheme, ACM2 for vertical diffusion, EBI solver and the aqueous/cloud chemistry with CB05 chemical scheme. Since our mother domain includes significant areas outside of Europe (North of Africa), we have used EDGAR emission inventory with EMIMO 2.0 emission model approach to fill those grid cells with hourly emission data. The VOC emissions are treated by SPECIATE Version 4.0 (EPA, USA) and for the lumping of the chemical species, we have used the [24] procedure for 16 different groups. We use our BIOEMI scheme for biogenic emission modeling. The classical, Atkin, Accumulation and Coarse modes are used (MADE/SORGAM modal approach). In WRF/CHEM simulation we have used only one domain with 30 km spatial resolution similar to the MM5. We have used the Lin et al. (1983) scheme for the microphysics, Yamartino scheme for the boundary layer parameterization and [23] for the biogenic emissions. The MOSAIC sectional approach is used with 4 modes for particle modeling.

6

Model results

The comparison between daily average values (averaged over all monitoring stations) of PM10 concentrations and modeled values has been performed with several statistical tools such as: Calculated mean/Observed mean; Calculated STD/Observed STD; bias; squared correlation coefficient (R2); RMSE/Observed mean (Root Mean Squared Error); percentage within +/- 50% and number of data WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

Air Pollution XVI

7

Figure 1:

Comparison between daily average observed PM10 concentrations and model results produced by MM5-CMAQ. The model does not capture the magnitude of the PM10 peaks.

Figure 2:

Comparison between daily average observed PM10 concentrations and model results produced by WRF/CHEM. The model captures quite well the magnitude of the PM10 peaks, particularly the first one.

sets. Figure 1 shows the comparison between PM10 observed averaged daily values and the modeled values by MM5-CMAQ. The results show that MM5CMAQ underestimates about 4 times the observed peak values and particularly the WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

8 Air Pollution XVI

Figure 3:

Comparison between daily average observed PM2.5 concentrations and model results produced by MM5-CMAQ. The model does not capture the magnitude of the PM2.5 peaks.

Figure 4:

Comparison between daily average observed PM2.5 concentrations and model results produced by WRF/CHEM. The model captures quite well the magnitude of the PM10 peaks, particularly the last one.

highest one on March 2 2003. The R2 coefficient is 0.69. Figure 2 shows similar information but for the WRF/CHEM results. In this case WRF/CHEM captures quite well the magnitude of the peaks, particularly the first one. For the second and third peak, the model underestimates about 20% the peak values. The R2 coefficient is 0.61. In the case of PM2.5 Figures 3 and 4 show similar results to WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

Air Pollution XVI

9

Figure 5:

Comparison between daily average observed PM10 concentrations and model results produced by MM5-CMAQ with PM2.5 emissions multiplied by 5. The model captures quite well the magnitude of the PM10 peaks, particularly the second one.

Figure 6:

Comparison between daily average observed PM2.5 concentrations and model results produced by MM5-CMAQ with PM2.5 emissions multiplied by 5. The model captures quite well the magnitude of the PM10 peaks, particularly the third one.

WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

10 Air Pollution XVI figures 1 and 2. The R2 coefficients are 0.41 and 0.58. The squared correlation coefficient goes from 0.69 to 0.61 in the case of PM10 but increases substantially In the case of PM2.5, from 0.41 to 0.58. In WRF/CHEM both R2 coefficients (for PM10 and PM2.5) are quite close (0.58 and 0.61) but in the case of MM5-CMAQ, PM2.5 R2 coefficient is substantially lower than in the case of PM10. We performed another full experiment with MM5-CMAQ. We multiply by 5 the PM2.5 emissions provided by TNO in the whole domain. The results are shown in Figures 5 and 6. The results are surprisingly good for both species. The R2 coefficient is 0.70 and 0.48 for PM10 and PM2.5 respectively. In both cases the correlation is improved and particularly for PM2.5 although just slightly. It is difficult to explain these results but it is a fact.

7

Conclusions

We have implemented and run two different models (MM5-CMAQ and WRFCHEM) for the same episode over Northern part of Germany during the winter period of 2003 (Jan. 15-Apr. 5, 2003). WRF-CHEM made a better job than MM5-CMAQ, not only the patterns reproduce the peak values quite well but also the statistical parameters are good. The calculated mean values divided by thye observed mean value os exactly 1.0 for PM10 and WRF/CHEM on-line model. For the MM5-CMAQ this ratio is 0.28 and when we multiply the PM2.5 emissions by 5, the ratio is 1.02 which is also excellent. The bias values for WRF/CHEM, MM5-CMAQ and MM5-CMAQ (x5) are 0.09, -23.33 and 0.51 which are excellent values for WRF/CHEM and MM5-CMAQ (x5). No realistic explanation is found for the exercise related to multiply by 5 the PM2.5 emissions from TNO emission inventory. The main apparent reason why WRF/CHEM is doing much better job than normal MM5-CMAQ is the use of MOSAIC particle model based on sectional modal approach instead the “classical” approach based on MADE/SORGAM modal approach.

Acknowledgements We would like to thank Dr. Peter Builtjes for the initial guidance and suggestion for this experiment and also COST 728 project (EU) where the inter-comparison experiment was proposed.

References [1] Collins W.J., D.S. Stevenson, C.E. Johnson and R.G. Derwent, Tropospheric ozone in a global scale 3D Lagrangian model and its response to NOx emission controls, J. Atmos. Chem. 86 (1997), 223–274. [2] Derwent R., and M. Jenkin, Hydrocarbons and the long-range transport of ozone and PAN across Europe, Atmospheric Environment 8 (1991), 1661– 1678.

WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

Air Pollution XVI

11

[3] Gardner R.K., K. Adams, T. Cook, F. Deidewig, S. Ernedal, R. Falk, E. Fleuti, E. Herms, C. Johnson, M. Lecht, D. Lee, M. Leech, D. Lister, B. Masse, M. Metcalfe, P. Newton, A. Schmidt, C Vandenberg. and R. van Drimmelen, The ANCAT/EC global inventory of NOx emissions from aircraft, Atmospheric Environment 31 (1997), 1751–1766. [4] Gery M.W., G.Z. Whitten, J.P. Killus and M.C. Dodge, A photochemical kinetics mechanism for urban and regional scale computer modelling, Journal of Geophysical Research 94 (1989), D10, 12925–12956. [5] Grell, G.A., J. Dudhia and D.R. Stauffer, A description of the FifthGeneration Penn State/NCAR Mesoscale Model (MM5), NCAR/TN- 398+ STR. NCAR Technical Note, 1994. [6] Jacobson M.Z. and R.P. Turco, SMVGEAR: A sparse-matrix, vectorized GEAR code for atmospheric models, Atmospheric Environment 28(1994), 2, 273–284. [7] Langner J., R. Bergstrom and K. Pleijel, European scale modeling of sulfur, oxidized nitrogen and photochemical oxidants. Model development and evaluation for the 1994 growing season, SMHI report RMK No. 82, Swedish Met. And Hydrol. Inst., SE-601 76 Norrkoping, Sweden, (1998). [8] Roemer M., G. Boersen, P. Builtjes and P. Esser, The Budget of Ozone and Precursors over Europe Calculated with the LOTOS Model. TNO publication P96/004, Apeldoorn, The Netherlands, 1996. [9] San José R., L. Rodriguez, J. Moreno, M. Palacios, M.A. Sanz and M. Delgado, Eulerian and photochemical modelling over Madrid area in a mesoscale context, Air Pollution II, Vol I., Computer Simulation, Computational Mechanics Publications, Ed. Baldasano, Brebbia, Power and Zannetti., 1994, 209–217. [10] San José R., J. Cortés, J. Moreno, J.F. Prieto and R.M. González, Ozone modelling over a large city by using a mesoscale Eulerian model: Madrid case study, Development and Application of Computer Techniques to Environmental Studies, Computational Mechanics Publications, Ed. Zannetti and Brebbia, 1996, 309–319. [11] San José, R., J.F. Prieto, N. Castellanos and J.M. Arranz, Sensitivity study of dry deposition fluxes in ANA air quality model over Madrid mesoscale area, Measurements and Modelling in Environmental Pollution, Ed. San José and Brebbia, 1997, 119–130. [12] Schmidt H., C. Derognat, R. Vautard and M. Beekmann, A comparison of simulated and observed ozone mixing ratios for the summer 1998 in Western Europe, Atmospheric Environment 35 (2001), 6277–6297. [13] Stockwell W., F. Kirchner, M. Kuhn and S. Seefeld, A new mechanism for regional atmospheric chemistry modeling, J. Geophys. Res. 102 (1977), 25847–25879. [14] Walcek C., Minor flux adjustment near mixing ration extremes for simplified yet highly accurate monotonic calculation of tracer advection, J. Geophys. Res. 105 (2000), 9335–9348.

WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

12 Air Pollution XVI [15] Janjic, Z. I., J. P. Gerrity, Jr. and S. Nickovic, 2001: An Alternative Approach to Nonhydrostatic Modeling. Monthly Weather Review, Vol. 129, 1164–1178 [16] Byun, D.W., J. Young, G. Gipson, J. Godowitch, F. Binkowsky, S. Roselle, B. Benjey, J. Pleim, J.K.S. Ching, J. Novak, C. Coats, T. Odman, A. Hanna, K. Alapaty, R. Mathur, J. McHenry, U. Shankar, S. Fine, A. Xiu, and C. Lang. 1998. Description of the Models-3 Community Multiscale Air Quality (CMAQ) model. Proceedings of the American Meteorological Society 78th Annual Meeting Phoenix, AZ, Jan. 11-16, 264–268. [17] Visscherdijk, A. and H. Denier van der Gon, 2005. Gridded European anthropogenic emission data for NOx, SO2, NMVOC, NH3, CO, PM10, PM2.5 and CH4 for the year 2000. TNO-report B&O-AR, 2005/106. [18] Amann, M., Bertok, I., Cofala, J., Gyarfas, F., Heyes, C., Klimon, Z., 2005. Baseline Scenarios for the Clean Air for Europe (CAFE) Programme. Final Report, International Institute for Applied Systems Analysis, Schlossplatz 1, A-2361 Laxenburg, Austria. [19] Schaap, M., H. Denier van der Gon, A. Visschedijk, M. van Loon, H. ten Brink, F. Dentener, J. Putaud, B. Guillaume, C. Liousse, P. Builtjes, 2004a. Anthropogenic Black Carbon and Fine Aerosol Distribution over Europe, J. Geophys. Res., 109, D18207, doi:10.1029/2003JD004330. [20] Beekmann, M., Kerschbaumer, A., Reimer, E., Stern, R., Möller, D., 2007. PM Measurement Campaign HOVERT in the Greater Berlin area: model evaluation with chemically specified observations for a one year period. Atmos. Chem. Phys. 7, 55–68. [21] Spindler, G., K. Mueller, E. Brueggemann, T. Gnauk, H. Herrmann, 2004. Long-term size-segregated characterization of PM10, PM2.5, and PM1 at the IfT research station Melpitz downwind of Leipzig (Germany) using high and low-volume filter samplers. Atmospheric Environment 38, 5333– 5347. [22] Putaud, J., F. Raesa, R. Van Dingenen, E. Bruggemann, M. Facchini, S. Decesari, S. Fuzzi, R. Gehrig, C. Hueglin, P. Laj, G. Lorbeer, W. Maenhaut, N. Mihalopoulos, K. Mueller, X. Querol, S. Rodriguez, J. Schneider, G. Spindler, H. ten Brink, K. Torseth, A. Wiedensohler, 2004. A European aerosol phenomenology – 2: chemical characteristics of particulate matter at kerbside, urban, rural and background sites in Europe. Atmospheric Environment 38, 2579–2595. [23] Guenther et al., 1995 A. Guenther, C.N. Hewitt, D. Erickson, R. Fall, C. Geron, T. Graedel, P. Harley, L. Klinger, M. Lerdau, W.A. McKay, T. Pierce, B. Scholes, R. Steinbrecher, R. Tallamraju, J. Taylor and P. Zimmerman, A global model of natural volatile organic compound emissions, Journal of Geophysical Research 100 (1995), pp. 8873–8892. [24] Carter, W. P. L. (2007): “Development of the SAPRC-07 Chemical Mechanism and Updated Ozone Reactivity Scales,” Final report to the California Air Resources Board Contract No. 03-318. August. Available at http://www.cert.ucr.edu/~carter/SAPRC.

WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

Air Pollution XVI

13

Analytic solutions of the diffusion-deposition equation for fluids heavier than atmospheric air F. P. Barrera1 , T. Brugarino2 , V. Piazza3 & L. Pignato3 1 Dip.

di Ingegneria dei Trasporti, Universit`a di Palermo Facolt`a d’Ingegneria, 90128 Palermo, Italy 2 Dip. di Metodi e Modelli Matematici, Universit`a di Palermo Facolt`a d’Ingegneria, 90128 Palermo, Italy 3 Dip. di Ricerche Energetiche ed Ambientali Universit`a di Palermo Facolt`a d’Ingegneria, 90128 Palermo, Italy

Abstract A steady-state bi-dimensional turbulent diffusion equation was studied to find the concentration distribution of a pollutant near the ground. We have considered the air pollutant emitted from an elevated point source in the lower atmosphere in adiabatic conditions. The wind velocity and diffusion coefficient are given by power laws. We have found analytical solutions using or the Lie Group Analysis or the Method of Separation of Variables. The classical diffusion equation has been modified introducing the falling term with non-zero deposition velocity. Analytical solutions are essential to test numerical models for the great difficulty in validating with experiments. Keywords: atmospheric pollution, diffusion equation, exact solutions.

1 Introduction The classical form of the mean steady diffusion equation is valid for elementary particles of the fluid or when the foreign particles are of the same density as the fluid. If the density and dimensions are high enough to have terminal velocities vs not negligible, the distribution of the particles will be affected in various ways [1–4]. A simple approximation is to consider that the particle sinks at a rate vs and the ground acts as a permeable surface and retains all material passing through it. Using a very simple model it is possible to examine various cases. It is possible to have exact solutions of the mean steady diffusion when the turbulent diffusivity kz and the terminal velocity vd depend somehow on the height [5–7]. WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line) doi:10.2495/AIR080021

14 Air Pollution XVI

2 Mathematical model The sedimentation of the material may be allowed by introducing a convection term in the mean steady equation that becomes [8, 9]:
 ∂ ∂c ∂c = kz (z) + vs (z)c u(z) (1) ∂x ∂z ∂z where vs (z) is the deposition velocity. We assume that mean wind velocity u(z), the eddy diffusivity in z-direction kz (z) and the deposition velocity vs (z) are: u(z) = u0 zα kz (z) = k0 z

n

vs (z) = v0 z

q

(2) (3) (4)

3 Group analysis of the equation Group analysis of the (1) is transformations:  ∗  x z∗   ∗ c

performed through the one-parameter Lie group of = x + X(x, z, c) + O( 2 ) = z + Z(x, z, c) + O( 2 ) = c + C(x, z, c) + O( 2 )

(5)

where X, Z, C are the infinitesimal generators of the transformations [10–12]. Equation (1) is invariant respect to the group (5) of transformations if c∗ is the solution of eq. (1) in the star variables. In this case, the number of independent variables can be decreased. A considerable difficulty lies in the amount of the auxiliary calculations involved. We performed the calculations of the generators of the transformations group on a P.C. using the MATHEMATICA package. Since eq. (1) is linear, the infinitesimal generators of the group of invariance are of the form:   X = X(x) (6) Z = Z(x, z)   C = A(x, z)c + B(x, z) The function B(x, z) must satisfy eq. (1) and, without compromising with the generality, can be assumed equal to zero. If we normalize the parameters u0 /k0 → v0 and the variable k0 x/v0 → x, we have that X, Z and A must satisfy the following equations: nz−1 Z − z−1 αZ + X − 2Zz = 0 WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

(7)

Air Pollution XVI

15

qv0 z−2−n+q Z + nqv0 z−2−n+q Z − q 2 v0 z−2−n+q Z − nz−1 Az − v0 z−n+q Az − 2qv0 z−1−n+q Zz − Azz + z−n+α Ax = 0 −2

nz

Z + nv0 z

−1−n+q

Z − qv0 z

−1−n+q

−1

Z − 2Az − nz

(8)

Zz

− v0 z−n+q Zz + Zzz − z−n+α Zx = 0

(9)

We show now some results.

4 Similarity solutions Let us look at some similarity solutions. 4.1 α, n and q arbitrary (n − α
= 2) In this case it possible to obtain from the eq.s (7–9) the generators of group of similarity:   X = a 0 (10) Z =0   C = c1 c where a0 , and c1 are arbitrary constants. The characteristic equations are: dz dx dc = = x1 0 c1 c −

c1

x

The invariants are z and ce x1 . If we assume corresponding to the separation of variables, is

c1 x1

= −λ2 , the similarity solution,

c = e−λ x Z(z) 2

where Z(z) is solution of the following ordinary differential equation: (qv0 z1−n+q + z2−n+α λ2 )Z(z) + z(n + v0 z1−n+q )Z  (z) + z2 Z  (z) = 0 If n = 2q − α, the solution is [13] √2 2 1−q+α Z(z) = e

(v0 + v −4λ ) 0 −2+2q−2α

z

 × h1 


2 − 3q + 2α −

 qv0 v02 −4λ2

2(1 − q + α)

2 − 3q + 2α z , ; 1−q +α

 qv0 v02 −4λ2


2 − 3q + 2α − + h2 L −2(1 − q + α)

1−q+α



v02 − 4λ2 

1−q +α  1−q+α v 2 − 4λ2  1 − 2q + α z 0 , ; 1−q +α 1−q +α

WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

16 Air Pollution XVI z 1

2 6·10

-1

0

-2

22 22

4·10 c

2·10

22

0 2.5

5

10 7.5 x

Figure 1: The c(x, z), v0 = 20, q = 0.5, λ = 1, α = 1.5, h1 = 1, h2 = 0. where h1 and h2 are arbitrary constants, (−, −; ·) is the confluent hypergeometric function and L(−, −; ·) is the generalized Laguerre polynomial. The concentration is √2 2 −

c = e−λ x e 2

 × h1 

(v0 + v −4λ ) z1−q+α 0 1−q+α 2


2 − 3q + 2α −

 qv0 v02 −4λ2

2(1 − q + α)

2 − 3q + 2α z , ; 1−q +α

 qv0 v02 −4λ2


2 − 3q + 2α − + h2 L −2(1 − q + α)

1−q+α



v02 − 4λ2 

1−q +α  1−q+α v 2 − 4λ2  1 − 2q + α z 0 , ; 1−q +α 1−q +α

4.2 n = 1, q = 0 and α arbitrary We observe that in this case k0 = ku∗ where k is the Von Karman constant and u∗ is the friction velocity. The generators of group of similarity are:  a2  X = a0 + a1 x + x 2   2   1−α a1 + a 2 x (11)  Z= z + c0 z 2   1+α   C = (b0 + b1 z1+α + b1 x(1 + α)(1 + v0 + α))c where a0 , a1 , a2 , b0 , b1 and c0 are constants satisfying the conditions c0 (1 + 2v0 + α) = 0,

a2 + 2b1 (1 + α)2 = 0

Now we consider the following subcases. WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

Air Pollution XVI

4.2.1 c0 = 0, a2 = −2b1 (1 + α)2 and α = −1 The characteristic equations are: dz dc dx = (1 + α) = x z c The invariants are

and ξ = zx − 1+α ; the concentration is 1

c x

c = xf (ξ ) where f (ξ ) is solution of the following ordinary differential equation:
 ξ 1+α f  (ξ ) − ξf  (ξ ) = 0 ξ α f (ξ ) − 1 + v0 + 1+α In this case we have: f (ξ ) = e

−

z1+α x(1+α)2

 
v0 z1+α 2 + v0 + 2α ,1 + ; h1  1+α 1 + α x(1 + α)2 
z1+α 2 + v0 + 2α v0 , ; + h2 L − 1+α 1 + α x(1 + α)2

where h1 and h2 are arbitrary constants. The concentration is
 1+α  2 + v0 + 2α v0 z1+α − z 2 x(1+α) c = xe ,1 + ; h1  1+α 1 + α x(1 + α)2
 2 + v0 + 2α v0 z1+α , ; + h2 L − 1+α 1 + α x(1 + α)2 4.2.2 (1 + 2v0 + α) = 0, a0 = −1, b1 = 0 and α
= −1 The characteristic equations are: dc dz dx = 1−α = −1 c z 2 The invariants are cex and ξ = x +

α+1 2 2 ; 1+α z

the concentration is

c = e−x f (ξ ) where f (ξ ) is solution of the following ordinary differential equation: f (ξ ) − f  (ξ ) + f  (ξ ) = 0 In this case the concentration is: 1+α  1+α 

1+α  √ x √ x 2 z 2 z 2 − x2 + z1+α f (ξ ) = e + + h2 sin 3 + h1 cos 3 2 1+α 2 1+α WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

17

18 Air Pollution XVI z -2

-1

0

1

2 1.5·10

181

181

1·10c 5·10 2.5

5

180

0 7.5 x

10

Figure 2: The c(x, z), v0 = 11, α = 1, h1 = 1, h2 = 0. 4.3 n = 2, q = 1 and α = 0 The generators of group of similarity are:  a2  X = a0 + a 1 x + x 2   2    1   Z = b z + + a2 x)z log z (a  0 1   2
1  C = c2 − x(2a1 (v0 − 1)2 + a2 (2 + (v0 − 1)2 x))   8       1    − log z(2(v0 + 1)(a1 + a2 x) + a2 log z) c 8 where a0 , a1 , a2 , b0 , and c2 are arbitrary constants. If we put: a0 = 0, a1 = 1, a2 = 0, b0 = 0, c2 = 0, we have   X=x      1 Z = z log z 2      C = − 1 ((v0 − 1)2 x + (1 + v0 ) log z)c 4 The invariants are ce 4 (v0 −1) x z 2 (1+v0 ) and ξ = 1

2

1

log √ z; x

the concentration is

c = e− 4 (v0 −1) x z− 2 (1+v0 ) f (ξ ) 1

2

1

where f (ξ ) is solution of the following ordinary differential equation: f  (ξ ) +

ξ  f (ξ ) = 0 2

WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

(12)

(13)

Air Pollution XVI

c0.5 0

8 6 4

1 2

2 z

x

3 40

Figure 3: The c(x, z), α = −3, h1 = 1, h2 = 1.

4 c 3 2 1 0

8 6 4

1 2 z

2

3 4

Figure 4: The c(x, z), v0 = 3, h1 = 1, h2 = 1. the solution is
 ξ f (ξ ) = h1 + h2 erf 2 The concentration is − 14 (v0 −1)2 x − 12 (1+v0 )

c=e

z



1 log z h1 + h2 erf √ 2 x

WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

x

19

20 Air Pollution XVI

5 Conclusion We obtain analytical solutions, using Lie group analysis, for steady-state bi-dimensional turbulent diffusion equation with variable coefficients. The laws of wind speed, turbulent diffusion coefficients and terminal velocities are specified by power laws. The obtained solutions are more realistic respect to gaussian model in air pollution modeling. In future we intend, using our solutions, to solve the diffusion equation (1) for many boundary conditions.

References [1] Arya S. P., Air Pollution Meteorology and Dispersion Oxford University Press (1999). [2] Pasquill F. and Smith F. B., Atmospheric Diffusion Ellis Horwood (1983). [3] Seinfeld J. H., Atmospheric Chemistry and Physics of Air Pollution Wiley (1986). [4] Sutton O. G., Micrometeorology M.Graw Hill (1953). [5] Godson W. L., The diffusion of particulate matter from an elevated source; Archiv f¨ur Meteor. Geophys. Bioklim., Ser. A., 10, 305–327, (1958). [6] Huang C. H., On solutions of the diffusion-deposition equation for point sources in turbulent shear flow; Journal of Applied Meteorology, 250–254, (1999). [7] Rounds W., Solution of two dimensional diffusion equation; Trans. Amer. Geophys. Union, 36, 395–405, (1955). [8] Yamamoto G., Shimamuki A., Nishinomiya S., Diffusion of falling particles in diabatic atmospheres; Journal of the Meteorological Society of Japan, 48, 417–424, (1970). [9] Yeh G. T., Huang C. H., Three-dimensional air pollutant modeling in the lower atmosphere; Boundary Layer Meteorology, 9, 381–390, (1975). [10] Barrera P., Brugarino T., Group analysis and some exact solution for the thermal boundary layer; Advances in Fluid Mechanics VI, WIT Press, 52, 327– 337, (2006). [11] Barrera P., Brugarino T., Some exact solutions of two coupled non linear diffusion-convection equations; Air Pollution XIII, Modelling, Monitoring and Management of Air Pollution, Cordova, (2005). [12] Barrera P., Brugarino T., Pignato L., Solutions for a diffusion process in nonhomogeneous media; Il Nuovo Cimento B., Vol. 116 B, 8, 951–958, (2001). [13] Polyanin A. D., Zaitsev V. F. Handbook of Exact Solutions for Ordinary Differential Equations CRC Press (1995).

WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

Air Pollution XVI

21

Air quality forecasting in a large city P. Perez Departamento de Fisica, Universidad de Santiago de Chile, Chile

Abstract We describe the different air pollution statistical forecasting models that have been used in Santiago, Chile during the fall/winter period for the last ten years. Effort has been concentrated on particulate matter PM10 for which a standard of 150 µg/m3 for the 24 h average is currently established. Inputs to the models are concentrations measured at several monitoring stations distributed throughout the city and meteorological information in the region. Outputs are the expected maxima concentrations for the following day at the site of the same monitoring stations. Forecast values using neural network models are compared with the results obtained with linear models and persistence. Recently, a clustering algorithm has appeared as a potentially useful tool to detect high concentration episodes in advance. Keywords: particulate matter forecasting, neural networks, linear models.

1

Introduction

Air pollution has been a major concern in the metropolitan area of Santiago, the capital of Chile during the last 15 years. Together with Sao Paulo, Mexico City, and some Chinese cities it is considered as one of the most polluted in world. Several factors concur to create unfavorable conditions for air pollutant dispersion. The city is located in a valley that has an extension between 70 and 80 km in the north-south direction and approximately 40 km in the east-west direction. To the west we find the Andes Mountains and to the east a coastal range. Some elevations to the north and south trap the air and air pollutants in a region of poor air circulation, which is enhanced during fall and winter when strong thermal inversions prevent vertical dispersion. During this period of the year, the 24 hour moving average (24MA) of PM10 is used as an indicator of air quality.

WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line) doi:10.2495/AIR080031

22 Air Pollution XVI

Figure 1:

Location of air pollution monitoring stations in the city of Santiago, Chile. The black area represents the urban region. To the extreme right we see the Andes mountains, which spread in the north-south direction.

The main sources of PM10 in Santiago, with six million habitants, are vehicular traffic, industrial activity and heating. Although environmental policies in Santiago during recent years have implied a significant improvement in air quality, particle levels still are considerably high compared to international standards [1]. At present, the standard for the 24 hour average for PM10 in Santiago is 150 µg/m3 and the standard for the one year average is 50 µg/m3. Regulations that apply on episodes of high concentrations have to do with the definition of classes. According to the value of the maximum of the 24MA of PM10 (MO) the day is classified as class A (good) if MO < 195 µg/m3, class B (bad) if 195 µg/m3 ≤ MO < 240 µg/m3, class C (critical), if 240 µg/m3 ≤ MO ≤ 330 µg/m3 and class D if MO ≥ 330 µg/m3 .The condition of the city is given by the worst class among all official monitoring stations. On class B days, 40% of the motor vehicles without catalytic converters cannot circulate. On class C days 60% of the vehicles without catalytic converters and 20% of those with catalytic converters are not allowed to circulate. On class C days a number of industries identified as pollutant emitters are enforced to stop operation. On class D days 80% of vehicles without catalytic converters and 40% of those with catalytic converters are not allowed to circulate and more industries are required to stop. Fortunately, the last class D day in Santiago occurred in 2001. It appears very WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

Air Pollution XVI

23

convenient to have a reliable PM10 forecasting model for the city, which can be used by the authorities in order to warn the population about adverse conditions and to implement palliative actions in advance when extreme conditions are foreseen. In recent years, atmospheric particulate matter forecasting models have been proposed as an aid for air quality management in different parts of the world. Perez et al. [2] have developed neural network, linear and persistence models in order to forecast hourly values of PM2.5 several hours in advance in Santiago, Chile. Three types of neural models, a linear model and a persistence model have been reported in order to forecast the daily averages of PM2.5 in El Paso (USA) and Ciudad Juarez (Mexico) [3]. The performance of multiple linear regressions and neural network models on the forecasting of PM10 in Athens was analyzed by Chaloulakou et al. [4]. Several types of neural network models and a linear model have been used for PM10 forecasting in Helsinki [5]. A multilayer perceptron with emphasis on a novel training algorithm has been used in order to forecast the 24 hour moving average of PM10 in Shanghai [6]. G. Corani has analyzed the performance of neural networks and a linear model locally trained to forecast the daily average of PM10 in Milan [7]. Multilayer neural networks have been used for PM10 forecasting in Santiago since 2002 [8,9]. The results of most of these studies show that neural network models are more accurate than linear models for atmospheric particulate matter concentrations forecasting. A hybrid clustering algorithm (HCA) has also been proposed for forecasting tasks, and it is claimed that it can outperform neural network models [11]. This approach was used for PM10 forecasting, and the results showed a 10% improvement over neural network models [12]. They also agree that sometimes, more important than the particular method, is the appropriate choice of input variables.

2

Forecasting models

The forecasting task may be represented by the implementation of a function of the form: (1) Y = F ( x1, …, xn, z1, …, zm ) where Y is a vector with components that are the maxima of tomorrow’s 24MA at the site of the monitoring stations, x1, …, xn are past values of PM10 concentrations and z1 …, zm are measured and forecasted exogenous variables. Input variables may be selected by performing a correlation analysis with historical data. In the late nineties, restrictions associated to classes B, C, D days in Santiago were applied on the basis of persistence. This means that if on a given day concentrations reached levels within class C, for example, on the following day restrictions associated to that class were applied. This action would make sense only if the episode lasted two or more days. Since 2001, there is an official forecasting model, which consists of a set of linear equations, one for each monitoring station. The area where most of the times, the highest concentrations are observed is that covered by station O. The equation for this zone is: WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

24 Air Pollution XVI YO = 39.4 VO + 0.33 CO + 2.06 TO + 0.21 DO – 21.7 (2) where: YO: is the maximum of the 24 hour moving average of PM10 expected for the following day in µg/m3. VO: forecasted atmospheric stability for the following day, which is a discrete variable ranging from 1 to 5. CO: 24 hour average of PM10 measured at 10:00 AM of present day in station O in µg/m3. TO: temperature in ºC of the 925 hPa level measured at 12 UTC of present day at a location 80 km west of Santiago. DO: change in the last 24 h for height of the 500 hPa level measured at 12 UTC of present day at a location 80 km west of Santiago (in meters). These last two variables give important information about strength of thermal inversions expected in Santiago in the next hours. The performance of the forecasting model (worst station) may be evaluated by building a contingency table, which for year 2004 is shown in table 1. Table 1:

Contingency table for official PM10 forecasting model between April 1 and August 16, 2004. 2004

O B S E R V E D

FORECASTED A

B

C

D

TOT

%O

A

109

15

2

0

126

87

B

1

6

2

0

9

67

C

0

1

1

0

2

50

D

0

0

0

0

0

X

TOT

110

22

5

0

137

85

%F

99

27

20

X

In table 1, in columns A, B, C, D we see the number of days forecast to be in a given class against the class of the observed day, which appears in the corresponding arrow. The column %O displays the percentage of observed days by class that were forecast to be in that class. Arrow %F delivers the percentage of forecast days by class that were verified to occur. 100 - %F for each class corresponds to percentage of false forecasts. Numbers in the grey diagonal boxes are the successful forecasts by class. At the lower right corner, the overall rate of successful forecasts is registered. We observe that the performance for this year was reasonable for the identification of class B and class C days, but was poor for the large fraction of false positives on these two classes, which affects the model reliability. Starting 2003, an alternative PM10 forecasting model for Santiago was presented with the idea to increase the reliability of the instrument on which city authorities base their decisions about restrictions. It was an artificial neural WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

Air Pollution XVI

25

network model [9]. In this case, Equation (1) is a non linear function that can be schematically represented as a set of nodes connected by weights, in which an input layer contains the variables in parenthesis and the output layer contains the Y components. A hidden layer with a number of auxiliary variables was also included. The transfer function between layers was a sigmoid. Connection weights were calculated by an optimization algorithm that fitted historical data from the previous two years [12]. The inputs used in the neural model were: one hour averages of PM10 measured at 6 PM and 7 PM of the present day at each of five stations (those with highest concentrations in average), the observed difference between maximum and minimum temperature on the present day, the forecasted difference between maximum and minimum temperature on the next day and the forecasted value of an index called PMCA for the next day. This index is a discrete meteorological variable that ranges from 1 to 5 and it is a measure of atmospheric stability in the Santiago area. Table 2 shows the 2004 contingency table for the neural network PM10 forecasting model. Table 2:

Contingency table for neural PM10 forecasting model between April 1 and August 16, 2004. 2004

O B S E R V E D

FORECASTED A

B

C

D

TOT

%O

A

119

7

0

0

126

94

B

2

5

2

0

9

56

C

0

0

2

0

2

100

D

0

0

0

0

0

X

TOT

121

12

4

0

137

92

%F

98

42

50

X

We observe that the neural model is in overall more accurate than the official model (92% against 85%), it is better for identification of class C days (100% against 50%) and produces less false positives on class B and class C days. Due to change in the properties of emissions in the city, identification of high concentrations (especially class C days) has been poor with both the official model and neural model in the last two years. For year 2007, the contingency table for the neural model is shown in table 3. This result seems poor considering that the population was exposed to high concentrations of particulate matter when a class C day was verified and no restrictions were applied. It is expected that when the restrictions associated to class C days are applied, they have the effect of lowering to some extent the concentrations. A way to correct the poor performance of the neural model on class C days identification is the proposal by Sfestos and Siriopoulos [10] and Vlachogiannis and Sfestos [11] of a clustering algorithm that may be applied for WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

26 Air Pollution XVI Table 3:

Contingency table for the neural PM10 forecasting model between April 1 and September 15, 2007. 2007

O B S E R V E D

Table 4:

FORECASTED A

B

C

D

TOT

%O

A

138

1

0

0

139

99

B

14

7

0

0

21

33

C

2

3

2

0

7

29

D

0

0

0

0

0

-

TOT

154

11

2

0

167

88

%F

90

64

100

-

Contingency table for the clustering PM10 forecasting model between April 1 and September 15, 2007. 2007

O B S E R V E D

FORECASTED A

B

C

D

TOT

%O

A

120

14

1

0

135

89

B

4

13

8

0

25

52

C

0

1

6

0

7

86

D

0

0

0

0

0

-

TOT

124

28

15

0

167

83

%F

97

46

40

-

air quality forecasting. A natural adaptation of this clustering algorithm has been implemented in Santiago to solve our problem of class identification one day in advance. The algorithm works in the following manner: For a period of three year training data, we have calculated the average values of the selected input variables within the respective classes (the same variables used in the neural model) A, B, C and D (four centroids). Within every class, we constructed linear or neural networks algorithms that reproduce the values of the output variables (the maxima of 24MA for the sites of the monitoring stations on the following day). Once we have the centroid patterns for each class, we can perform a test with the following year data, by assigning a given vector to the class with centroid to the least Euclidean distance from it. After class identification, we can calculate the numerical forecasted value by using the WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

Air Pollution XVI

27

algorithm valid for that class. For an operational forecasting system, it would be desirable to generate the most accurate value for tomorrow’s class and the expected numerical value of the maximum of the PM10 concentration. The implementation of the clustering algorithm described above with 2007 data produced table 4. From this table we can verify that the clustering algorithm, having less overall accuracy compared with the neural model (83% against 88%), it has a significantly better performance in detecting class C days (86% against 29%). The false C forecasts would not be so critical considering that most of them were verified to be class B days, which also represent levels considered harmful for the people. A disadvantage of this clustering method is the discontinuity of the numerical forecasted value upon changing from one class to another.

3

Conclusion

With rather simple statistical models it is possible to generate relevant information regarding air quality for the population and authorities in a large city. We have presented several tools that have been used for air quality management in the city of Santiago, Chile and the choice of one of them over the others will depend on the goals we pursue with the forecasting. The models may be used, with the appropriate adaptations in other cities

Acknowledgement We would like to thank Fondo Nacional de Ciencia y Tecnología (FONDECYT) for support through project 1070139.

References [1] Koutrakis, P., Sax, S., Sarnat, J., Coull, B., Demokritou, P., Oyola, P., García, J., Gramsch, E., Analysis of PM10, PM2.5 and PM2.5-10 Concentrations in Santiago, Chile, from 1989 to 2001 J. Air Waste Manag Assoc 55, 342–351 (2005). [2] Perez, P., Trier, A., Reyes, J., Prediction of PM2.5 concentrations several hours in advance using neural networks in Santiago, Chile. Atmospheric Environment 34, 1189–1196 (2000). [3] Ordieres, J. B., Vergara, E. P., Capuz, R. S., Salazar, R. E. Neural network prediction model for fine particulate matter (PM2.5) on the US-Mexico border in El Paso (Texas) and Ciudad Juarez (Chihuahua). Environmental Modelling & Software 20, 547–559 (2005). [4] Chaloulakou, A., Grivas, G., Spyrellis, N. Neural Network and Multiple Regression Models for PM10 Prediction in Athens: A comparative Assessment. J. Air Waste Manag Assoc 53, 1183–1190 (2003). [5] Kukkonen, J., Partanen, L., Karppinen, A., Ruuskanen, J., Junninen, H., Kolehmainen, M., Niska, H., Dorling, S., Chatterton, T., Foxall, R., WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

28 Air Pollution XVI

[6] [7] [8] [9] [10] [11]

[12]

Cawley, G., Extensive evaluation of neural network models for the prediction of NO2 and PM10 concentrations, compared with a deterministic modeling system and measurements in central Helsinki. Atmospheric Environment 37, 4539–4550 (2003). Jiang, D., Zhang, Y., Hu, X., Zeng, Y., Tan, J., Shao, D. Progress in developing an ANN model for air pollution index forecast. Atmospheric Environment 38, 7055–7064 (2004). Corani, G. Air quality prediction in Milan: feed-forward neural networks, pruned neural networks and lazy learning. Ecological Modelling 185, 513– 529 (2005). Perez, P., Reyes, J. Prediction of maximum of 24-h average of PM10 concentrations 30 h in advance in Santiago, Chile. Atmospheric Environment 36, 4555–4561 (2002). Perez, P., Reyes, J. An integrated neural network model for PM10 forecasting. Atmospheric Environment 40, 2845–2851 (2006). Sfestos, A., Siriopoulos, C. Time series forecasting with a hybrid clustering scheme and pattern recognition. IEEE Transactions on systems, man and cybernetics, Part A 34, 399–405 (2004). Vlachogiannis, D., Sfestos, A., Time series forecasting of hourly PM10 values: model intercomparison and the development of localized linear approaches. Air Pollution XIV, edited by Longhurst, J. W. S. and Brebbia, C. A., WIT Press, 85–94 (2006). Rumelhart, D. E., Hinton, G. E., Williams, R. J., Learning Internal Representations by Error Propagation. Parallel Distributed Processing. The MIT Press, Cambridge, London, pp 318–364 (1986).

WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

Air Pollution XVI

29

Meshing effects on CFD modelling of atmospheric flow over buildings situated on ground with high terrain A. Karim, P. Nolan & A. Qubian Chemical Engineering Department, London South Bank University, UK

Abstract A Computational Fluid Dynamics (CFD) study was carried out on the wind environment and vehicle pollution dispersion from a newly built by-pass adjacent to the Whatman International site in Maidstone, UK. The site buildings are sited on ground incorporating an area of high terrain, accordingly they were modelled using Geographical Information System (GIS). The site contains a substantial number of trees of differing species which were extensively surveyed and modeled using a simple 2D momentum sink dependent on the tree Leaf Area Density (LAD). One of the important factors which has a significant effect on CFD results is the computational mesh. The purpose of this paper is to investigate the effects of using different mesh approaches for both easterly and southerly wind scenarios and the results are compared with that of field measurements taken at the site which include the CO concentration and wind velocity. The CFD results showed that the hexahedral mesh delivers a higher level of agreement with field measurements than the tetrahedral dominant mesh and this is mainly due to the higher truncation error in the tetrahedral cell type. It was also found that the tetrahedral dominant mesh can be significantly improved by applying different numerical solving techniques such as the Nodebased gradient solver. Keywords: CFD, pollution dispersion modelling, grid effect, urban environment.

1

Introduction

The Whatman International factory is sited in Maidstone, Kent / UK, the factory consists of a number of buildings which are sited on a ground characterized by WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line) doi:10.2495/AIR080041

30 Air Pollution XVI terrain forming a small valley on either side of the river Medway, whilst a new dual carriageway (A229 bypass) is sited on the upper side of the terrain. The factory manufactures special types of filters which are highly sensitive to any type of air pollutants. The main objective of the study is to investigate the pollutant dispersion and wind environment from the A229 by-pass onto the sensitive manufacturing facility of this factory. Whatman International were seeking to obtain the concentration of certain pollutants at specific sensitive locations in their factory, therefore the pollutant concentration and velocity were measured at those points in the site. The Computational Fluid Dynamics (CFD) approach was used to simulate the wind environment and to investigate the effects of CO dispersion produced by the traffic passing through the dual-carriageway on the sensitive manufacturing facility. The preprocessor GAMBIT of the commercial CFD code Fluent [4] was used to design both the geometry and mesh of this site and the Fluent solver is used to solve the Reynolds Averaged Navier Stokes (RANS) equations. However, the footprints and locations of the buildings and land terrain were obtained using GIS and those were used to build the 3D geometry for the CFD simulation, Fig. 1. From the CFD analysis point of view, one of the main points to be considered is the meshing approach which has significant effects on the prediction and accuracy of the results [1], and this is particularly true of atmospheric flow problems with sometimes very complex geometries (buildings and terrain). Not only is this true of the mesh size but also in the mesh type used (structured / unstructured, hexahedral dominant / tetrahedral dominant etc).

Figure 1:

CFD model of the Whatman Intl site and surrounding area.

This paper presents the two main meshing approaches (hexahedral and tetrahedral). The CFD results are compared with that of the field measurements of velocity and CO concentrations. Accordingly, the mesh approaches and limitations are discussed briefly in the following section, and this is followed by a discussion of the modelling technique, and the paper finishes with the results of specific simulations and their findings. WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

Air Pollution XVI

2

31

Mesh generation

Generally speaking, the domain of a CFD analysis is usually divided into a number of cells (mesh) in which the RANS equations are solved. However, there are two main types of meshing approaches, structured and unstructured. The structured mesh approach generally ensures stability in convergence, however it is impractical for the real atmospheric flow cases [2,3]. On the other hand, the unstructured mesh approach would have higher numerical errors which significantly vary depending on the type of the cells (e.g hexahedral or tetrahedral) and the degrees of the cell quality. It is worth discussing briefly the difference between the hexahedral cell shape and the tetrahedral cell shape with regards to the solution accuracy. The face value of the cells used in the Fluent software to calculate the various flow variables is dependent on the interpolation method used and that involves different schemes to calculate values from cell to cell. For example the cell face value for any given variable φ f can be determined using equation (1) (typical interpolation equation), and this is dependent on the cell centroid value φ c , cell gradient ∇φ , and the geometrical features of the cell as defined in Fig. 2. Thus if we take two near wall adjacent cells of hexahedral and tetrahedral shape shown in Fig. 2 (here the flow is assumed to be parallel to the wall), with hexahedral cells the face is on a vertical angle whilst the tetrahedral cells are inclined and this gives rise to a non-zero gradient. This very well known phenomenon results in a higher truncation error.

( )

φ f = φ c + δr (∇φ )c + δr

2

(1)

However if the flow is not aligned to the cells (not parallel), the hexahedral mesh looses its “edge” on the tetrahedral mesh scheme. For atmospheric flow problems typically the flow can be easily and naturally aligned with the hexahedral cell arrangement. Hexahedral face

φc

[δr (∇φ )c ] = 0 Figure 2:

Tetrahedral face

φf

φf φc

[δr (∇φ )c ] ≠ 0

Diagram illustrating the effect of mesh shape on the simulation accuracy.

WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

32 Air Pollution XVI For the present CFD work, two mesh schemes were used; i) The Hexahedral Core (HC) type [4] which is a tetrahedral dominant mesh ii) The second mesh type used is the wholly hexahedral type, Hexahedral-Cooper (Cooper). The results of both approaches will be discussed in order to highlight their level of agreement with field measurements. Due to the requirement of an orthogonal cell in the near wall area studies have shown improved simulations using prism shapes in the tetrahedral mesh [5]. However the present work has sought to compare between the tetrahedral mesh and hexahedral mesh primarily. For the Hex-Cooper mesh, the maximum skewness was approximately 0.9 whilst the maximum aspect ratio was 200. The Cooper mesh contained 2.7 million cells in total. The HC mesh, on the other hand, contained a maximum skewness of 0.95, however the aspect ratio is fairly low throughout the domain. The HC mesh was about 2.6 million cells.

3

Modelling technique

For the CFD modelling procedure, there are many important factors which should be carefully chosen for solving the RANS equations. Among them are the turbulence model, the domain boundary conditions and the calculation of the cell gradient values. Brief discussions on these issues are given below. For the domain boundary condition, the side flow inlet is modelled using approximated profiles suggested by Richards and Hoxey [6]. The profiles would be calculated using meteorological data taken at the nearby East Malling station. The aforementioned profiles are commonly used in literature and seem to give suitable approximations to the flow variation, they include the wind speed, turbulent kinetic energy and dissipation [7]. All other domain boundaries are specified as symmetry whilst the outlet is set as an outflow boundary [8]. A line of substantially tall trees almost entirely surrounds the Whatman site, these were modelled as 2D momentum sinks, with turbulence generation inside the canopy ignored. This was based on data from the UK Forestry Commission on Leaf area Density (LAD). For modelling turbulence the standard k- ε modified by Detering and Etling [9] is used, with standard wall functions. The vehicle pollutant from the road is released as an area emission with the turbulence generated by vehicles ignored. For the cell gradient values; the Fluent code provides three methods for calculating the gradient values of each individual cell from the surrounding cells, Green-Gauss cell-based, Green-Gauss Node-based and Least squares cell-based, for more details see [4]. However, the CFD results could be significantly improved by using the Nodebased method, this is particularly true for the arbitrary shape arrangement of the cells which are associated with the tetrahedral meshing scheme [4,10]. The Node-based method calculates the node weighted average for all surrounding cells which share the same node. This is especially critical for the tetrahedral mesh scheme because significantly more cells share any single node than the WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

Air Pollution XVI

33

hexahedral arrangement, thus the numerical errors become naturally higher. However, the iteration time would be significantly increased, and this may be more severe (double or triple) when tackling more complex problems, e.g reactions. Accordingly, in order to improve the result of the tetrahedral based mesh without seeking higher numerical schemes or more complex turbulence models the Node-based gradient method is utilized.

4 Results and discussions In order to validate the CFD models against the results of the experimental work, the two dominant wind directions of both east and south were investigated. The CFD prediction involved the use of two meshing schemes, first the Cooper scheme which is an entirely hexahedral mesh, whilst, the second one, HC is a tetrahedral dominant mesh, and the main features of those schemes are presented below. 4.1 Wake flow behind the buildings Generally speaking, for the dispersion of the CO pollutant, the wake flow behind the buildings represents one of the important fluid features which would participate in the changes of pollutant concentration locally and also in channeling and directing the flow movement downstream of the buildings. For the Cooper meshing approach, the predictions of flow movement within the wake behind the buildings were predicted with some considerable clarity and accuracy in their sizes, flow separation and flow directions. For the HC tetrahedral approach the wake formation is much less clear and the wake region is larger in size and this is typical of all the buildings in the domain. Another point worth mentioning here, is the flow reattachment point downwind of the buildings are further downstream in the HC mesh as compared to the Cooper scheme, and this is shown in Figs. 3 and 4 around a typical site building. However, the differences in results can be explained as follows; for the HC meshing approach, despite a higher number of cells filling the wake region behind the building which in fact slightly exceeded that of the Cooper scheme, but due to the random structure arrangement of the tetrahedral cells, which mean that the cell centers are in different non-equidistant locations in relation to neighboring cells, the consequences of that, are the solver will interpolate values carrying a higher numerical error which is exponentially expanded downstream, Figs 5 and 6. The hexahedral shape at the wall is known to give improved predictions due to the better resolution of turbulent energy normal to the wall [11]. Accordingly the HC mesh results showed that the flow near the ground terrain region is characterized by small fluctuations (bumps) in the velocity plot, whilst the Cooper mesh did not have these inconsistencies. WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

34 Air Pollution XVI The features discussed above had a significant effect on the CO concentration as shown in Figs. 7 and 8. The CO concentration for the HC mesh is highly irregular with sporadic high concentration points.

Road Wall and Railings

Figure 3:

Velocity contours in vertical plane (m/s)-East wind, (Cooper).

Road Wall and Railings

Figure 4:

Figure 5:

Figure 6:

Wind

Wind

Velocity contours in vertical plane (m/s)-East wind, (HC).

Mesh in vertical plane surrounding typical site building, (Cooper).

Mesh in vertical plane surrounding typical site building, (HC).

WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

Air Pollution XVI

Road

Figure 7:

Wind

CO concentration contours in vertical plane (PPM)-East wind, (Cooper). Road

Figure 8:

35

Wind

CO concentration contours in vertical plane (PPM)-East wind, (HC).

4.2 Wind velocity and Carbon Monoxide concentration The measurements of CO concentration were carried out at a location 10 m away from the road, whilst the wind velocity measurements were carried out 60m away from the road. Generally speaking, the CFD predicted results are in good agreement with those measured, see Table 1. For the CFD results, the Cooper mesh scheme produces better overall agreement with that of the measurements and the discrepancies of both CO concentration and wind velocity were 12.5% and 3% respectively for the east wind and 16% and 1% for the south wind. Given the drastic differences between the meshing schemes in predicting wake size behind buildings and the boundary layer development, accordingly, in using the HC mesh scheme, the discrepancies of both CO prediction and wind velocity were 20% and 13% respectively for the east wind, whilst it is slightly higher for the southerly wind, 32% and 26% respectively. Those high discrepancies which are usually associated with the tetrahedral meshing scheme would be significantly attributed to the numerical and truncation errors. As a result of that the distributions within the complex flow regions such as the wakes and boundary layer formation on the terrain would be significantly changed. However, the predicted velocity or CO obtained for the WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

36 Air Pollution XVI Table 1:

Wind Direction East East East South South South

Summary of the CFD simulations as compared to the field measurements.

Mesh Type Hexahedral (Cooper) Tetrahedral (HC) Tetrahedral (HC-Node based) Hexahedral (Cooper) Tetrahedral (HC) Tetrahedral (HC-Node based)

Velocity measured (m/s)

Velocity CFD (m/s)

Velocity Error (%)

CO measured (PPM)

CO CFD (PPM)

CO Error (%)

0.70

0.72

3

1.69

1.48

12.5

0.70

0.60

13

1.69

2.03

20

0.70

0.73

4

1.69

1.98

17

1.00

1.01

1

1.06

1.23

16

1.00

1.26

26

1.06

1.41

32

1.00

1.1

10

1.06

1.37

28

free-stream regions of the domain have good agreements with field measurements. The highest discrepancies of both wind velocity and CO concentrations associated with the southern wind scenario can be related to the fact that the unsettled and unsteady flow phenomena such as the wakes, flow separation and recirculation become more significant within the reference points which are used in the measurements, accordingly, the predicted CFD results would have increased numerical and truncation errors. On the other hand, for the HC tetrahedral mesh case, the CFD results were improved significantly by using the Node-based gradient method comparing to that of Cell-based method (default Fluent option), Table 1. The predicted CO and velocity at the measurement points become comparable to that obtained by the Cooper mesh scheme

5

Conclusion

Two CFD models for the Whatman international site in Kent, UK were designed and solved using fully hexahedral Cooper mesh and the HC tetrahedral dominant mesh with the same flow conditions for the easterly and southerly wind scenarios. Both models were solved with the same turbulence model and numerical schemes and with almost the same number of meshing cells. Both schemes generally predicted the same overall flow characteristics; however the HC scheme compared to the Cooper scheme produced irregular plots of CO and velocity in the wake regions and along the ground which has a varying terrain, as well as differences in the prediction of the wake sizes behind buildings. When compared to the field measurements, the HC mesh predicted lower localized velocities which led to over-prediction of the CO concentration. The HC mesh gave discrepancies of up to 16% in wind velocity and 25% in CO WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

Air Pollution XVI

37

concentration, whilst the Cooper scheme produced discrepancies of 3% in the velocity and 16% in the CO concentration. The results strongly suggested that the HC scheme, in non-complex flow regions of the domain (complex being areas of separation or recirculation regions) produced velocity and CO predictions which were quite reasonable. However the performance of the HC mesh is greatly influenced by the complex flow phenomenon which occurred for the south wind scenario and that produced much higher discrepancies, whilst the Cooper scheme was largely unaffected Furthermore it was found that the HC mesh numerical errors could be greatly reduced by running the solver using the Node-based method for calculating the cells’ gradient values instead of using the Cell-based method which is the typical method employed in most commercial codes.

Acknowledgements The authors would like to thank Dr. Olga Grant of the East Malling Research Centre, and Dr. Rona Pitman and Chris Peachey of UK Forestry Commission.

References [1] Cowan, I. R., Castro, I.P., Robins, A.G. Numerical considerations for simulations of flow and dispersion around buildings. Journal of Wind Engineering and Industrial Aerodynamics 67 & 68 pp. 535–545, 1997. [2] Huber, A., Freeman, M., Spencer, R., Schwarz. W., Bell, B., Kuehlert, K. Development and applications of CFD simulations supporting urban air quality and homeland security, AMS sixth symposium on the urban environment, Atlanta, U.S, 2006. [3] Kim, S., and Boysan, F. Application of CFD to environmental flows, Journal of Wind Engineering and Industrial Aerodynamics, 81 pp. 145– 158, 1999. [4] Fluent user guide. Fluent. Inc 2006 [5] Fothergill, C.E., Roberts, P.T., Packwood, A.R. Flow and dispersion around storage tanks. A comparison between numerical and wind tunnel simulations. Wind and structures, Vol. 5, No. 2-4 pp. 89–100, 2002 [6] Richards, P.J. and Hoxey, R. Appropriate boundary conditions for computational wind engineering models using the k-e model, Journal of Wind Engineering and Industrial Aerodynamics, 46, 47, pp. 145–153, 1993. [7] Hargreaves, D.M. and Wright, N. G. On the use of the k- ε model in commercial CFD software to model the neutral atmospheric boundary layer. Journal of Wind Engineering and Industrial Aerodynamics, doi: 10.1016/j.weia.2006.06.002, 2006 [8] Franke, J. et al Recommendations on the use of CFD in predicting pedestrian wind environment. COST Action C14 “Impact of wind and storms on city life and built environment” Working Group 2-CFD techniques, 2004 WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

38 Air Pollution XVI [9] Detering, H.W. and Etling, D. Application of the E- ε Turbulence Model to the Atmospheric Boundary Layer. Boundary Layer Meteorology 33 pp. 113–133, 1985 [10] Ferziger, J. H. and Peric, M. Computational Methods for Fluid Dynamics, Springer – Verlag: New York, 2002. [11] Fothergill, C. E., Roberts, P.T., Packwood, A.R. Flow and dispersion around storage tanks. A comparison between numerical and wind tunnel simulations. Wind and Structures, Vol. 5, 2~4 pp. 89–100, 2002.

WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

Air Pollution XVI

39

Enhanced evaluation of a Lagrangian-particle air pollution model based on a Šaleška region field data set B. Grašič, P. Mlakar, M. Z. Božnar & G. Tinarelli MEIS d.o.o., Mali Vrh pri Šmarju 78, 1293 Šmarje-Sap, Slovenia ARIANET s.r.l., Via Gilino 9, 20128 Milano, Italy

Abstract Lagrangian-particle air pollution model is required by Slovenian legislation for industrial air pollution control because it is the most efficient for small domains over complex topography. To determine its performance and efficiency for regulatory purpose a research is made. In this research a general purpose modelling system designed for local scale areas is used. The main goal of the research is to define an enhanced statistical analysis used to evaluate an air pollution model of this kind where an operational configuration of both input data and model parameters are used and a testing period with very complex dispersion conditions is used. This enhanced evaluation can help to better understand the general quality that a model can achieve in these conditions. It gives some idea on how to better evaluate and use some results that seem to be very negative simply looking to some statistical parameter. Keywords: air pollution, model evaluation, Lagrangian particle model, field data set.

1

Introduction

In accordance with the European Council Directive of 28 June 1984 on combating air pollution from industrial plants (84/360/EEC) a Slovenian Government decree on the emission of substances into the atmosphere from stationary sources of pollution came into force in April 2007. The decree defines that the performance of the air pollution model used to reconstruct air pollution

WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line) doi:10.2495/AIR080051

40 Air Pollution XVI around stationary sources must meet the requirements of complex terrain as defined in paper by Grašič et al. [1]. The Lagrangian-particle air pollution model is the one that fully satisfies all these requirements [2] among the currently available air pollution models. The main purpose of the research is an enhanced evaluation of the Lagrangianparticle air pollution model that is used for regulatory purposes over complex terrain. We are looking at its performances in severe conditions and trying to better understand and interpret some results.

2 Enhanced evaluation method Standard method to evaluate the quality and performance of an air pollution model is based on statistical analysis of measured and reconstructed data. Measured data is obtained from automatic air quality measuring stations located around source of air pollution. Reconstructed data is obtained from simulation results performed with selected air pollution model. Evaluation is performed by statistical analysis of data that is available for selected time interval T. For the selected time interval a set of data patterns must be prepared where each data pattern consist of a pair of measured and reconstructed concentration {Cm(t ), Cr (t )} . During the evaluation following three performance indices are determined: • the correlation coefficient: 1 T ∑ Cm(t ) − Cˆm Cr (t ) − Cˆr T t =0 (1) r=

(

)(

)

σ Cmσ Cr

•

•

the normalized mean square error: 1 T (Cm(t ) − Cr (t ) )2 ∑ T NMSE = t =0 Cˆ m ⋅ Cˆ r

(2)

and the fractional bias: FB = 2

Cˆ m − Cˆ r Cˆ m + Cˆ r

(3)

where Cm(t ) …measured concentration at time t, Cr (t ) …reconstructed concentration at time t, Cˆ …average concentration,

σC

…concentration standard deviation, …interval length (number of concentrations). To avoid effect of model’s inaccuracy of position and time of reconstructed concentrations an enhanced evaluation method is used. It is based on presented statistical analysis where additional reconstructed ground level concentrations around measuring station are used in comparison procedure as presented on T

WIT Transactions on Ecology and the Environment, Vol 116, © 2008 WIT Press www.witpress.com, ISSN 1743-3541 (on-line)

Air Pollution XVI

41

Figure 1. In standard evaluation procedure only one reconstructed concentration is used from the cell where station is located Cr (t ) = Cr (t , ms , ns ) . In enhanced evaluation procedure a reconstructed concentration for comparison is selected from set of reconstructed concentrations GCS(t) using best matching fuction BM. The function selects the reconstructed concentration that represents the best match according to measured concentration. Cr(t,m S,nS) Cm(t, x S,yS)

N

Cr(t,m,n) GCF(t)={Cr(t,m,n);0<m<M;0