The Eruption of Soufriere Hills Volcano, Montserrat, From 1995 to 1999
Dedicated to Peter Francis
Geological Society Publications Society Book Editors A. J. FLEET (CHIEF EDITOR) P. DOYLE F. J. GREGORY J. S. GRIFFITHS A. J. HARTLEY R. E. HOLDSWORTH A. C. MORTON N. S. ROBINS M. S. STOKER J. P. TURNER
Society books reviewing procedures The Society makes every effort to ensure that the scientific and production quality of its books matches that of its journals. Since 1997. all book proposals have been refereed by specialist reviewers as well as by the Society's Books Editorial Committee. If the referees identify weaknesses in the proposal, these must be addressed before the proposal is accepted. Once the book is accepted, the Society has a team of Book Editors (listed above) who ensure that the volume editors follow strict guidelines on refereeing and quality control. We insist that individual papers can only be accepted after satisfactory review by two independent referees. The questions on the review forms are similar to those for Journal of the Geological Society. The referees' forms and comments must be available to the Society's Book Editors on request. Although many of the books result from meetings, the editors are expected to commission papers that were not presented at the meeting to ensure that the book provides a balanced coverage of the subject. Being accepted for presentation at the meeting does not guarantee inclusion in the book. Geological Society Special Publications and some other books series are included in the ISI Index of Scientific Book Contents, but they do not have an impact factor, the latter being applicable only to journals. More information about submitting a proposal and producing a Society Publication can be found on the Society's web site: www.geolsoc.org.uk.
It is recommended that reference to all or part of this book should be made in one of the following ways: DRUITT, T. H. & KOKELAAR, B. P. (eds) 2002. The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society. London, Memoirs, 21. LUCKETT, R., BAPTIE, B. & NEUBERG, J. 2002. The relationship between degassing and rockfall signals at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 595-602.
GEOLOGICAL SOCIETY MEMOIRS NO. 21
The Eruption of Soufriere Hills Volcano, Montserrat From 1995 to 1999
EDITED BY
T. H. DRUITT Universite Blaise Pascal, Clermont-Ferrand, France and
B. P. KOKELAAR University of Liverpool, UK
2002 Published by The Geological Society London
THE GEOLOGICAL SOCIETY The Geological Society of London (GSL) was founded in 1807. It is the oldest national geological society in the world and the largest in Europe. It w;as incorporated under Royal Charter in 1825 and is Registered Charity 210161. The Society is the UK national learned and professional society for geology with a worldwide Fellowship (FGS) of 9000. The Society has the power to confer Chartered status on suitably qualified Fellows, and about 2000 of the Fellowship carry the title (CGeol). Chartered Geologists may also obtain the equivalent European title, European Geologist (EurGeol). One fifth of the Society's fellowship resides outside the UK. To find out more about the Society. log on to www.geolsoc.org.uk. The Geological Society Publishing House (Bath, UK) produces the Society's international journals and books. and acts as European distributor for selected publications of the American Association of Petroleum Geologists (AAPG), the American Geological Institute (AGI). the Indonesian Petroleum Association£ (IPA), the Geological Society of America (GSA), the Society for Sedimentary Geology (SEPM) and the Geologists' Association (GA). Joint marketing agree ments ensure that GSL Fellows may purchase these societies' publications at a discount. The Society's online bookshop (accessible from www.geolsoc.org.uk) offers secure book purchasing with your credit or debit card. To find out about joining the Society and benefiting from substantial discounts on publications of GSL and other societies worldwide, consult www.geolsoc.org.uk, or contact the Fellowship Department at: The Geological Society. Burlington House. Piccadilly. London W1J OBG: Tel. —44 (0)20 7434 9944; Fax +44 (0)20 7439 8975; E-mail: enquiries@ geolsoc.org.uk. For information about the Society's meetings, consult Events on www.geolsoc.org.uk. To find out more about the Society's Corporate Affiliates Scheme, write to
[email protected].
Published by The Geological Society from: The Geological Society Publishing House Unit 7 Brassmill Enterprise Centre Brassmill Lane Bath BA1 3JN, UK (Orders'. Tel. +44 (0)1225 445046 Fax +44 (0)1225 442836) Online bookshop: http://bookshop.geolsoc.org.uk The publishers make no representation, express or implied, with regard to the accuracy of the information contained in this book and cannot accept any legal responsibility for any errors or omissions that may be made. @ The Geological Society 2002. All rights reserved. No reproduction, copy or transmission of this publication may be made without written permission. No paragraph of this publication may be reproduced, copied or transmitted save with the provisions of the Copyright Licensing Agency, 90 Tottenham Court Road, London W1P 9HE. Users registered with the Copyright Clearance Center, 27 Congress Street, Salem, MA 01970, USA: the item-fee code for this publication is 0435-4052/02/S 15.00. British LibraryCCataolologuingin Publication Data A catalogue record for this book is available from the British Library. ISBN 1-86239-098-3 ISSN: 0435-4052 Typeset by Aarontype Ltd, Easton, Bristol, UK Printed by Cambrian Press, Aberystwyth, UK.
Distributors USA AAPG Bookstore PO Box 979 Tulsa OK 74101-0979 USA Orders: Tel. +1 918 584-2555 Fax +1 918 560-2652 E-mail bookshop@ aapg.org Affiliated East-West Press PVT Ltd G-l/16 Ansari Road. Daryaganj. New Delhi 110002 India Orders: Tel. +91 11 327-9113 Fax +91 11 326-0538 E-mail affiliat@ nda.vsnl.net.in Japan Kanda Book Trading Co. Cityhouse Tama 204 Tsurumaki 1-3-10 Tama-shi Tokyo 206-0034 Japan Orders: Tel. +81 (0)423 57-7650 Fax +81 (0)423 57-7651
Contents
Foreword Preface Acknowledgements In Memorium Peter Francis Background and overview of the eruption Setting, chronology and consequences of the eruption of Soufriere Hills Volcano, Montserrat (1995-1999): KOKELAAR, B. P. The eruption of Soufriere Hills Volcano, Montserrat (1995-1998): overview of scientific results: SPARKS, R. S. J. & YOUNG, S. R. The Montserrat Volcano Observatory: its evolution, organization, role and activities: ASPINALL, W. P., LOUGHLIN, S. C, MICHAEL, F. V., MILLER, A. D., NORTON, G. E., ROWLEY, K. C., SPARKS, R. S. J. & YOUNG, S. R. The volcanic evolution of Montserrat using 40Ar/39Ar geochronology: HARFORD, C. L., PRINGLE, M. S., SPARKS, R. S. J. & YOUNG, S. R. Volcanic processes, products and hazards Growth patterns and emplacement of the andesitic lava dome at Soufriere Hills Volcano, Montserrat: WATTS, R. B., HERD, R. A., SPARKS, R. S. J. & YOUNG, S. R. Dynamics of magma ascent and lava extrusion at Soufriere Hills Volcano, Montserrat: MELNIK, O. & SPARKS, R. S. J. Mechanisms of lava dome instability and generation of rockfalls and pyroclastic flows at Soufriere Hills Volcano, Montserrat: CALDER, E. S., LUCKETT, R., SPARKS, R. S. J. & VOIGHT, B. Pyroclastic flows and surges generated by the 25 June 1997 dome collapse, Soufriere Hills Volcano, Montserrat: LOUGHLIN, S. C., CALDER, E. S., CLARKE, A., COLE, P. D., LUCKETT, R., MANGAN, M. T., PYLE, D. M., SPARKS, R. S. J., VOIGHT, B. & WATTS, R. B. Eyewitness accounts of the 25 June 1997 pyroclastic flows and surges at Soufriere Hills Volcano, Montserrat, and implications for disaster mitigation: LOUGHLIN, S. C., BAXTER, P. J., ASPINALL, W. P., DARROUX, B., HARFORD, C. L. & MILLER, A. D. Deposits from dome-collapse and fountain-collapse pyroclastic flows at Soufriere Hills Volcano, Montserrat: COLE, P. D., CALDER, E. S., SPARKS, R. S. J., CLARKE, A. B., DRUITT, T. H., YOUNG, S. R., HERD, R. A., HARFORD, C. L. & NORTON, G. E. Small-volume, highly mobile pyroclastic flows formed by rapid sedimentation from pyroclastic surges at Soufriere Hills Volcano, Montserrat: an important volcanic hazard: DRUITT, T. H., CALDER, E. S., COLE, P. D., HOBLITT, R. P., LOUGHLIN, S. C., NORTON, G. E., RITCHIE, L. J., SPARKS, R. S. J. & VOIGHT, B. Episodes of cyclic Vulcanian explosive activity with fountain collapse at Soufriere Hills Volcano, Montserrat: DRUITT, T. H., YOUNG, S. R., BAPTIE, B., BONADONNA, C., CALDER, E. S., CLARKE, A .B., COLE, P. D., HARFORD, C .L., HERD, R. A., LUCKETT, R., RYAN, G. & VOIGHT, B. Modelling of conduit flow dynamics during explosive activity at Soufriere Hills Volcano, Montserrat: MELNIK, O. & SPARKS, R. S. J. Computational modelling of the transient dynamics of the August 1997 Vulcanian explosions at Soufriere Hills Volcano, Montserrat: influence of initial conduit conditions on near-vent pyroclastic dispersal: CLARKE, A. B., NERI, A., VOIGHT, B., MACEDONIO, G. & DRUITT, T. H. Hazard implications of small-scale edifice instability and sector collapse: a case history from Soufriere Hills Volcano, Montserrat: YOUNG, S. R., VOIGHT, B., BARCLAY, J., HERD, R. A., KOMOROWSKI, J.-C, MILLER, A. D., SPARKS, R. S. J. & STEWART, R. C. The 26 December (Boxing Day) 1997 sector collapse and debris avalanche at Soufriere Hills Volcano, Montserrat: VOIGHT, B., KOMOROWSKI, J.-C., NORTON, G. E., BELOUSOV, A. B., BELOUSOVA, M., BOUDON, G., FRANCIS, P. W., FRANZ, W., HEINRICH, P., SPARKS, R. S. J. & YOUNG, S. R. Generation of a debris avalanche and violent pyroclastic density current on 26 December (Boxing Day) 1997 at Soufriere Hills Volcano, Montserrat: SPARKS, R. S. J., BARCLAY, J., CALDER, E. S., HERD, R. A., KOMOROWSKI, J.-C., LUCKETT, R., NORTON, G. E., RITCHIE, L. J., VOIGHT, B. & WOODS, A. W. Sedimentology of deposits from the pyroclastic density current of 26 December 1997 at Soufriere Hills Volcano, Montserrat: RITCHIE, L. J., COLE, P. D. & SPARKS, R. S. J. The explosive decompression of a pressurised volcanic dome: the 26 December 1997 collapse and explosion of Soufriere Hills Volcano, Montserrat: WOODS, A. W., SPARKS, R. S. J., RITCHIE, L. J., BATEY, J., GLADSTONE, C. & BURSIK, M. I. Pyroclastic flow and explosive activity of the lava dome of Soufriere Hills volcano, Montserrat, during a period of no magma extrusion (March 1998-November 1999): NORTON, G. E., WATTS, R. B., VOIGHT, B., MATTIOLI, G. S., HERD, R. A., YOUNG, S. R., DEVINE, J. D., ASPINALL, W. P., BONADONNA, C., BAPTIE, B. J., EDMONDS, M., HARFORD, C. L., JOLLY, A. D., LOUGHLIN, S. C., LUCKETT, R. & SPARKS, R. S. J. Tephra fallout in the eruption of Soufriere Hills Volcano, Montserrat: BONADONNA, C., MAYBERRY, G. C., CALDER, E. S., SPARKS, R. S. J., CHOUX, C., JACKSON, P., LEJEUNE, A. M., LOUGHLIN, S. C., NORTON, G. E., ROSE, W.I., RYAN, G. & YOUNG, S. R. Numerical modelling of tephra fallout associated with dome collapses and Vulcanian explosions: application to hazard assessment on Montserrat: BONADONNA, C., MACEDONIO, G. & SPARKS, R. S. J.
vii ix xi xiii
1 45 71 93
115 153 173 191 211 231 263 281 307 319 349 363 409 435 457 467
483 517
vi
CONTENTS
Dynamics of volcanic and meteorological clouds produced on 26 December (Boxing Day) 1997 at Soufriere Hills Volcano. Montserrat: MAYBERRY, G. C, ROSE, W. I. & BLUTH, G. J. S. Monitoring of airborne particulate matter during the eruption of Soufriere Hills Volcano, Montserrat: MOORE. K.R.. DUFFELL. H.. NICHOLL, A. & SEARL. A. Geophysical and gas studies Seismicity, gas emission and deformation from 18 July to 25 September 1995 during the initial phreatic phase of the eruption of Soufriere Hills Volcano, Montserrat: GARDNER, C. A. & WHITE. R. A. Spaceborne radar measurements of the eruption of Soufriere Hills Volcano, Montserrat: WADGE, G., SCHEUCHL, B. & STEVENS. N. F. The relationship between degassing and rockfall signals at Soufriere Hills Volcano, Montserrat: LUCKETT. R.. BAPTIE. B. & NEUBERG, J. A model of the seismic wavefield in gas-charged magma: application to Soufriere Hills Volcano. Montserrat: NEUBERG. J. & O'GORMAN, C. Observations of low-frequency earthquakes and volcanic tremor at Soufriere Hills Volcano. Montserrat: BAPTIE. B.. LUCKETT. R. & NEUBERG, J. Variation in HC1/SO2 gas ratios observed by Fourier transform spectroscopy at Soufriere Hills Volcano. Montserrat: OPPENHEIMER, C., EDMONDS, M., FRANCIS, P. & BURTON, M. Index
539 557
567 583 595 603 611 621 640
Foreword Volcanoes are the most violent surface expression of the Earth's internal energy. Only impacts of large extra-terrestrial bodies can match the explosive release and devastation of the largest volcanoes. Indeed for some of the most dramatic events the Earth has seen - the large terrestrial extinctions of animal life - the jury is still out as to whether they were brought about by meteoritic impact or by wide-scale effects of volcanic activity. Volcanoes have it too when it comes to sustained visual impact. Earthquakes, tsunamis and avalanches all cause massive devastation, but it is accomplished in the blink of an eye, and floods rise with a progressive and depressing inevitability. Volcanoes are simply the most spectacular of the destructive natural hazards to life on Earth. To those who are far enough away to view them in safety, volcanoes can offer a truly awe-inspiring pyrotechnic display of the Earth's innate power - a natural, spectacular son et lumiere. For this reason from time immemorial they have exerted a siren-like attraction for geologists, photographers, filmmakers and many others. And, like the sirens of ancient fable, they have lured to their death all too many of those who dared to get too close. Indeed volcanoes inspired such awe in the ancient world that their own mythology sprang up about them. Cyclops, the one-eyed giant who all-unprovoked threw rocks great distances to kill shepherds tending their flocks, we know today as Mount Etna. The giant was also able to cause springs to flow where he struck the ground - it is not uncommon for groundwater flows to be disrupted during volcanic episodes. Like its neighbouring islands in the Caribbean, Montserrat exists solely because of volcanic activity. It is a volcanic island formed by the progressive accumulation of layers of lava and debris erupted on the sea floor. It is a small Caribbean island, which, along with a host of others, is located close to where the slowly moving crust of the Atlantic takes leave of the surface and plunges down into the Earth's interior. All life forms, humans included, are so eager to find new habitats that as soon as a volcano has been inactive for a hundred years or so, and sometimes sooner, it is colonized by a flora and fauna. Montserrat had long been inactive and, besides being well situated for fishing and tourism, and a little agriculture, it supported a resident population of over ten thousand. The art of volcanic prediction is still too poorly developed to be very useful and when in 1995 the volcano showed signs of renewed activity the population simply hoped that it would soon die down. It had been quiescent for about 350 years. Understandably, people who in their own lifetime have known only a gently steaming mountain are not inclined to believe that things are about to change. But change they did. Devastating pyroclastic flows overwhelmed the southern half of the island with its villages and smallholdings. In the north, the infrastructure of life was disrupted and part buried by settling ash. The people had no experience of active volcanoes and could not imagine how rapidly the behaviour of the volcano could change and how unpredictable it was. The inescapable speed and heat of flows of incandescent ash were beyond their comprehension. Although they were warned, many were reluctant to abandon their homes on official advice and chose to take the risk. Sadly some paid the ultimate price. And when this happened in spite of their best efforts, some on the ground had to live with the nagging doubt as to whether, had they tried just one more time, they could have persuaded the farmers to leave. In the world league of volcanic eruptions the ongoing Montserrat eruption does not rate very high. What was unusual, indeed unique, about Montserrat was the combination of two special circumstances. First, because of the risk to life and the presence of an indigenous population with no escape other than to leave the
island, resources were available to monitor the volcano that would simply not have been there for a purely scientific study. Whether those resources were really adequate is another matter. Second, there has been no recent opportunity anywhere in the world to study an oceanic island arc volcano during eruption. Different kinds of recent volcanic activity have been studied elsewhere in Iceland, in Hawaii, in Washington State, in southern Italy and in Japan, but Montserrat offered a unique opportunity to study this kind of oceanic eruption with modern techniques. This combination of circumstances has made it possible to document the behaviour of the volcano in considerable detail and to do so with the collaboration of geologists from a wide range of organizations and from many countries. Intriguingly, there is one important feature of the Montserrat eruption that is little known elsewhere. The fine ash that is common in many eruptions and which buried buildings on Montserrat to depths of three metres or more is very unusual: it contains, and in places largely comprises, very fine particles of silica in an unusual crystalline form - minute particles of the mineral cristobalite. These are uncomfortably similar in their characteristics to other fine particles that damage the respiratory system, and were regarded as potentially a significant health hazard on the island. Occurrences such as this present governments with major moral dilemmas. Montserrat is a British dependency many thousands of kilometres from the UK and therefore difficult, not to say very expensive, to support. In the early years of the renewed eruption, the infrastructure and much of the farming land was destroyed. Resettlement was offered to the inhabitants and eventually the majority of them accepted the offer. But what are our obligations to the others? The economy of the island is now extraordinarily tenuous and life there can continue only if external support is maintained. It looks as if the northern part of the island will be relatively safe from the direct products of eruption for the foreseeable future. But it will be decades before soils develop and before agriculture can be re-established in the south. The eruption is still in progress and has now lasted for more than six years, longer than in virtually all similar lava-dome eruptions around the world. It seems that Soufriere Hill Volcano is evolving into a persistently active state that could continue for decades. And so, alongside the scientific investigations, a complex human drama was playing as well, and geoscientists, for whom volcanology had been their somewhat esoteric and rather academic specialization, suddenly found themselves at the frontline of its practical application where life at times had much in common with a war-zone. There are other lessons to be learned as well. Volcanology has been very much a minority discipline within UK Earth Sciences. It has been kept alive by the efforts of a few outstanding and energetic individuals of real academic distinction. No value for money or relevance criteria would have suggested that volcanology was worthy of more than peripheral support. But without this infrastructure of knowledge and experience the UK would have had no indigenous capability to cope with the events on Montserrat. Many of the results of the scientific studies at Montserrat are presented in this volume. They represent an unparalleled suite of detailed observations that will add significantly to the understanding of volcanic hazards. In due course, they will lead to a better understanding of how volcanoes work and to a better ability to predict their behaviour. I am proud that the Geological Society is the publisher of this volume. Ron Oxburgh President of the Geological Society of London
Preface The andesitic dome-building eruption of Soufriere Hills Volcano has wreaked havoc on the small Caribbean island of Montserrat. About half of this 'Emerald Isle' has been rendered barren and uninhabitable, almost two-thirds of the original population has left, and 19 lives have been lost, all as a direct result of the volcanic activity. Many Montserratians have suffered multiple evacuations and displacements from their homes, and the economy has been severely affected by the loss of infrastructure and farmland, and by the adverse impact on tourism. The centre of the former capital, Plymouth, today lies partially buried under metres of volcanic debris, and several villages have been swept away by catastrophic flows of incandescent ash. As this book goes to press, the eruption continues and as yet shows no signs of abating. The still-populated northern half of the island is to some extent recovering and the people are learning to live with the volcano. Nevertheless, fine airborne ash from intermittent major events continues to penetrate into many homes and to pose a health hazard, while boulder-laden floods following the torrential rains of all-too-frequent hurricanes impede recovery of property. This Memoir presents results of monitoring and associated research over a five-year period, from the onset of the eruption in July 1995 until November 1999. Scientists on active volcanoes need to balance their essential activities in monitoring, assessing hazards, advising local and national authorities, and informing the public, with programmes of basic research into causes, mechanisms and consequences of the volcanic activity under scrutiny. Mitigation of volcanic hazards is most effective when there exists good understanding of the physical and chemical processes controlling the system. Monitoring and research on Montserrat evolved from mainly remote surveillance by the Seismic Research Unit of the University of the West Indies in Trinidad, before the eruption, to establishment of a multinational and multidisciplinary team at the Montserrat Volcano Observatory (MVO). The observatory, which moved its location three times as the eruption slowly escalated between July 1995 and September 1997, developed enhanced capabilities over the period as funding and technical assistance became available. The slow escalation of the crisis gave the scientists time to build effective monitoring infrastructures and team management before the most devastating phases of the eruption in 1997. Monitoring included use of short-period and broadband seismometers to detect and locate all types of earthquakes, Global Positioning Satellite (GPS) and Electronic Distance Measurement networks to detect upheaval of the volcano, measurements by Correlation Spectrometer and Fourier Transform Infra-red spectroscopy to monitor gas exhalations, and photogrammetry and GPS-based methods to determine magma extrusion rates and volumes. Petrological and geochemical studies of magmatic products were carried out as the eruption progressed. Monitoring was initially aided by the close proximity to the volcano of the MVO, which until September 1997 benefited from line-of-sight observations. Analysis and cross-correlation of emerging multi-parameter data sets, together with the development and application of physical models of magma ascent, degassing and extrusion, led to greatly increased understanding of the system dynamics and origin of the various signals being monitored as the eruption unfolded. This in turn influenced subsequent monitoring strategies. The large numbers of visiting scientists, who undertook diverse studies at the MVO, greatly enhanced the effectiveness and innovation of the work of the core teams. Many diverse phenomena related to the ascent and extrusion of andesitic magma have now been studied at close quarters. These include phreatic explosions, dome collapses with formation of pyroclastic flows and surges, dispersal of ash plumes, vertically directed explosions with fountain collapse, sector collapse followed by debris avalanche, lateral blast and violent pyroclastic density current, and a remarkable period of disintegration of a lava dome with little or no accompanying magma extrusion or precursory seismic activity. Cyclic activity, registered in seismic and deformation data, in compositions and fluxes of magmatic gases, and in rates of magma
extrusion, has been attributed to interactive physical processes in the conduit and extruding lava dome. Cycles with timescales ranging from several weeks to several hours are now understood in terms of the ascent, degassing, rheological stiffening and pressurization of crystal-rich andesitic magma, with complex feedback effects and multiple regimes of behaviour. The eruption has permitted detailed documentation and better understanding of processes and associated signals that lead to lava dome instability and to explosive decompression, and of the physical behaviour and characteristics of associated pyroclastic currents. It has also provided the opportunity for development of different methods in hazards assessment and zonation, including formal elicitations of international scientific expertise and statistical treatments of eruption-scenario models and associated uncertainties. The eruption and its consequences constitute an important case of volcanic crisis management and of the interactions between scientists, authorities and populace on a small island, with significant lessons for the future. The decision was taken to publish as much as possible of the material pertaining to the 1995-1999 eruptive period in one book. Previous notable benchmark eruptions, such as of Mount St Helens in 1980 and Mount Pinatubo in 1991, have been similarly followed by scientific monographs that are the main repositories of data and interpretations concerning those eruptions. The books constitute an invaluable resource for those researchers, volcano observatories and government agencies concerned with volcanic risk minimization. Two series of short reports on the Montserrat eruption up to August 1997 were published during 1998 in Geophysical Research Letters. This Memoir contains 30 papers, many of which address the chronology, dynamics, products and associated hazards of the eruption. It also includes papers specifically on the associated geophysics and geochemistry, although geophysical aspects in particular constitute significant parts of many of the other papers. Four introductory papers provide overviews of the eruption chronology and consequences, of the scientific results, of the evolution, organization, role and activities of the Montserrat Volcano Observatory, and of the volcanic development of Montserrat through time. A large photographic record of the 1995-1999 eruptive period is included. Using author consensus, we have adopted certain conventions for clarity and consistency throughout the Memoir. We consider the recent events to represent a single eruption of Soufriere Hills Volcano, which, following six years of increased seismicity, started on 18 July 1995 and continues at the time of publication. We use the convention that there have been two lava domes to date during the eruption. One grew intermittently from mid-November 1995 to midMarch 1998, undergoing multiple collapses as it did so. This was followed by a twenty-month interval during which sectors of the first dome collapsed with little or no associated extrusion and commonly without seismic precursors. The second dome started growing in midNovember 1999. This Memoir concerns only that period of the eruption up to the resumption of magma discharge in mid-November 1999. Individual lava extrusions during dome growth are referred to as shear lobes, or, for brevity, lobes. Lobes are referred to by the date of their first appearance (e.g. 17 July 1996 Lobe). We have also tried, for the benefit of non-specialist readers, to simplify the terminology used in the Memoir, particularly that concerning pyroclastic density currents. In most cases during the 1995-1999 period it was possible to distinguish fairly clearly, using conventional criteria, between pyroclastic density currents predominantly of high-concentration and those predominantly of lowconcentration, and also between their respective deposits. We have retained the terms pyroclastic flow and pyroclastic surge respectively for these phenomena. We distinguish genetically between domecollapse, fountain-collapse and surge-derived pyroclastic flows. The descriptive terms block-and-ash flow and pumice-and-ash flow are also used respectively for the former two phenomena. Only in one case, the catastrophic collapse of 26 December 1997, have we retained the general term pyroclastic density current, because significant vertical and lateral gradients in particle size and
PREFACE concentration are inferred to have been present in the moving current so that neither pyroclastic flow nor pyroclastic surge seems appropriate. Monitoring the eruption from 1995 to 1999 was the work of more than 120 people, including Montserratian scientific, technical and clerical staff along with scientists and technicians from the British Geological Survey, the Seismic Research Unit on Trinidad, the US Geological Survey Volcanic Crisis Assistance Team, and universities in the UK and other countries including the USA, Puerto Rico and France. Many were graduate students in volcanology or young, MVO-trained staff and much of the collection and analysis of data from the eruption was accomplished by these young people, who also participated in the daily jobs of hazards assessment and public education. Although tragic in so many ways, the eruption has served as a training ground for a large cohort of young volcanologists who are now well prepared to deal with future volcanic crises around the globe. It is a tribute to all of the people involved
that, after more than five years of volcanic activity and with constant changing of observatory personnel, the vast amount of information collected is in such good order that it could be compiled and analysed for the papers in this Memoir. A hundred years ago. on 8 May 1902, a cloud of hot gas and pyroclastic debris from a lava dome on Montagne Pelee swept over the town of St Pierre on the French Caribbean island of Martinique and claimed about 28000 lives. Many of the processes that drove that eruption probably resembled those at Soufriere Hills. This Memoir registers the enormous advances made since that time, both in our understanding of lava-dome eruptions and in the sophistication of the tools we use to monitor them. Much, however, remains to be learned. Nature will not reveal many secrets in one go. Tim Druitt & Peter Kokelaar Editors
Acknowledgements The editors and authors of this book wish to pay their respect to the people of Montserrat, who have suffered greatly during the present crisis. They have borne their hardship and, for many, the loss of homes, family and friends, with courage, good nature and dignity. The contributions of all staff members of the Montserrat Volcano Observatory (MVO) over the 1995-1999 period are acknowledged. The local staff in particular formed the backbone of the team, often under conditions that were stressful for them, their families and their communities. They are: Venus Bass, Owen Butler, Levar Cabey, Sharon Charles, Thomas Christopher, Paulette Cooper, Billy Darroux, Deneese Fenton, Franklyn Greenaway, Linda Halloran, David Lea, Sunny Lea, Chelston Lee, Leroy Luke, Pops Morris, Karney Osborne, Joel Osborne, Alwyn Ponteen, Graham Ryan, Patch Silcott, George Skerrit, Tappy Syers, Bill Thorn, Bill Tonge, Jackie Weekes, Daisy Weekes, Dave Williams and Pyiko Williams. We also list the overseas scientists and technicians that rotated through the MVO during the period, while apologizing to any inadvertently forgotten: Stephane Acounis, Godfrey Almorales, Christian Antenor, Sayyadul Arafin, Willy Aspinall, Wilkie Balgobin, Brian Baptie, Jenni Barclay, Peter Baxter, Costanza Bonadonna, Eliza Calder, Robert Carsley, Tom Casadevall, Caroline Choux, Amanda Clarke, Paul Cole, Mark Davies, Peter Day, Pierre Delmelle, Joe Devine, Laurance Donnelly, Sarah Dornan, Tim Druitt, Hayley Duffell, Peter Dunkley, Neil Dyer, Marie Edmonds, John Ewert, Glenn Ford, Peter Francis, Davie Galloway, Cynthia Gardner, Gilbert Hammouya, Chloe Harford, Richard Herd, Rick Hoblitt, Claire Horwell, Paul Jackson, Mike James, Art Jolly, Chris Kilburn, Pete Kokelaar, Jean-Christophe Komorowski, Joan Latchman, Anne-Marie Lejeune, Andy Lockhart, Sue Loughlin, Richard Luckett, Lloyd Lynch, Adam Maciejewski, Maggie Mangan, Glen Mattioli, Gari Mayberry, Bill McGuire, Angus Miller, Dan Miller, Kate Moore, Mick Murphy, Tom Murray, Jurgen Neuberg, Gill Norton, Clive Oppenheimer, Ouchi Osuji, Dave Petrie, Luchman Pollard, John Power, Dave Pyle, Chandrapath Ramsingh, Tony Reedman, Lucy Ritchie, Ritchie Robertson, Geoff Robson, Lizzette Rodriguez, Keith Rowley, Willie Scott, Desmond Seupersad, John Shepherd, Bennett Simpson, Alan Smith, Stefan Soil, Steve Sparks, Mark Stasiuk, Nicki Stevens, Rod Stewart, Dai Stewart, John Stix, Sharon Teebenny, Glenn Thompson, John Tomblin, Jane Toothill,
Patrick Tuchais, Terry Turbitt, Jean-Pierre Viode, Barry Voight, Geoff Wadge, Colin Walker, Matthew Watson, Robert Watts, Randall White, and Simon Young. It has been a privilege for MVO scientists to have worked, often under hazardous conditions, with helicopter pilots of exceptional skill, in particular Jim McMahon, Alex Grouchy, Laurance Linskey and Barry Lashley. The MVO was financed mainly by the UK Government Department for International Development (DFID), formerly the Overseas Development Administration (ODA), and by the Government of Montserrat. Research grants and studentships were provided by the UK Natural Environment Research Council (NERC), the US National Science Foundation (NSF) and the French Centre National de la Recherche Scientifique (CNRS). The work at the MVO during 1995-1999 was supported on Montserrat by two Governors, HE Frank Savage and HE Tony Abbott, by three Chief Ministers, Reuben Meade, Bertrand Osborne and David Brandt, and by Frankie Michael and Horatio Tuitt of the Emergency Department. The observatory also benefited from the help of Frank Hooper and Chris Burgess, the police force, the Royal Montserrat Defence Force, Radio ZJB, the Government Information Service, the Aid Management Office of DFID, the Governor's Office, and ministers and officers of the Government of Montserrat. The production of this Memoir was skilfully overseen by Angharad Hills. Funding for colour printing was generously provided by the Geological Society of London. The following people kindly took the time to provide reviews of the articles, and helped us ensure high standards of science and presentation: Willy Aspinall, Charlie Bacon, Peter Baxter, Mike Branney, Ray Cas, Bernard Chouet, Brian Dade, Mike Dungan, Jon Fink, Armin Freundt, JeanLuc Froger, Jennie Gilbert, Lori Glaze, Richard Hiscott, Rick Hoblitt, Jean-Christophe Komorowski, Tak Koyaguchi, Steve McNutt, Pete Mouginis-Mark, Setsuya Nakada, Augusto Neri, Chris Newhall, Harry Pinkerton, John Power, Richie Robertson, Bill Rose, Mauro Rosi, Dave Rothery, Steve Self, Steve Sparks, John Stix, Brad Sturtevant, Don Swanson, Greg Valentine, Jim Vallance, Ben van Wyk de Vries, Barry Voight, Richard Waitt, Randy White, Stan Williams, Lionel Wilson, Andy Woods and Simon Young.
In Memorium PETER WILLIAM FRANCIS, 1944-1999 Professor of Volcanology, The Open University
Like many people, I first encountered Peter through his book Volcanoes. It was listed amongst about thirty titles that I was advised by my college to digest before going to university (I confess it was the only one of them I read). It accompanied me for several months on a trip to Indonesia - my dishevelled copy is annotated with comments on the characteristics of Javanese volcanoes that resembled those Peter described and illustrated. I had no idea that, five years later, I would begin my doctoral studies at the Open University under Peter's supervision. Numerous colleagues have related very similar stories of how Peter's books inspired them to follow careers in volcanology. Peter was born in 1944 in Mufulira, Zambia, where his father worked as a pharmacist. After his family resettled in England, he attended Bournemouth Grammar School and Reading School, and then studied at Imperial College, London, where he earned a first class honours degree in geology in 1966. While at Imperial, Peter went to a talk at the college's Expeditions Society given by a filmmaker, Tony Morrison, on the subject of a 'ghost' mining town high in the southern Bolivian Andes. The lecture fired the imagination of a small group, which put together an expedition to the region. Peter joined as Equipment Officer, and took responsibility for the medical kit. They set off for South America by sea in July 1966, taking a Landrover with them, and before long Peter was enjoying his first taste of remote volcanoes. They produced a weighty report on return, which included a contribution from Peter that reveals youthful manifestations of his well known sense of humour. In a section captioned 'Unnecessary Items' he admitted that: 'an overestimate on the part of the Equipment Officer led to rather too much toilet paper being carried. In this case, better too much than too little'. Under the heading 'Items we should have had', and following a list that includes pressure lights and a water pump, he reported that: 'certain members of the party acquired an obsession for looking at themselves in a looking-glass. Future expeditions are recommended to carry a full-length mirror to satisfy this desire'. Peter remained at Imperial for his doctoral research, which he completed in 1969. Supervised by Janet Watson, he wrote his thesis on the structural geology of the island of Barra in the Outer Hebrides of Scotland. Peter's adventurous, but sometimes accidentprone, character is typified by a story he used to tell of solo climbing in his field area (presumably in pursuit of some lofty rock samples). Negotiating a tricky cliff section, he lost his footing, but dropped short of the ground thanks to a rope securely attached to an overhang. He then swung helplessly and hopelessly for several hours before being able to reattach himself to the rock face. As an old friend of Peter's recalled, another of his tricks in this period was to coast his Landrover down the 260 metre descent into Castlebay without touching the brakes. He also took the opportunity to dive, retrieving items from Second World War wrecks, and enjoyed taking ammunition apart, so that he could dry the cordite and throw handfuls on to the stove for a bit of light relief while waiting out the autumn storms. Peter's flair for organizing fieldwork in remote and sometimes hostile environments was put to use on a further expedition to the Central Andes in 1970, this time to north Chile. He contributed his expertise in structural geology to a group that included George Walker. He was now truly engrossed by volcanology, and the research led to his early paper on the San Pablo and San Pedro volcanoes and their pyroclastic deposits (Francis et al. 1974). Recruited by Ian Gass to the Department of Earth Sciences at the Open University in 1971, Peter initiated geochemical studies of Andean volcanic rocks (chiefly in collaboration with Richard Thorpe) and made further contributions to the understanding of
ignimbrites. Always excited (if sometimes frustrated) by new technology, he was quick to recognize the potential of the first Landsat multispectral satellite images for mapping such vast and arid terrains as the Central Andes. By this means, he was the first to observe that Cerro Galan, in remote northwest Argentina, is the core of a resurgent caldera (Francis et al. 1983). He followed this discovery up by leading an Anglo-Argentinian expedition to the volcano in 1982, which included Steve Sparks. Steve later recalled driving across the fascinating and breathtaking altiplano landscape on the way up to the base camp, whilst listening to Peter's tape of Bach's St Matthews Passion, as a 'high point' in his life (anyone who telephoned Peter at home regularly will have spotted his enthusiasm for baroque music). The project received military support, and was the last joint operation between the British and Argentine armies before the Falkland Islands/Las Malvinas conflict. His name was subsequently splashed over the Argentine tabloid papers as the man behind 'COMPLOT, a fictional campaign of British espionage. But coming under suspicion for spying was nothing new to Peter. During fieldwork in Peru, he once spent three nights in police custody after being apprehended photographing some intriguing ash dunes of the 1600 Huaynatputina fall deposit, a little too close to a military base. He had with him a set of US aeronautical charts, the only maps available of the region, and for fear of what might happen if these were discovered, tore them to shreds which he slipped through the cracks in the floorboards. He was later moved to another station for detention, where he assisted the police by picking the lock for the main door, which they were unable to open. Flying above the freshly emplaced debris avalanche and blast deposits of the 1980 Mount St Helens eruption, the new landscape around the volcano reminded Peter of the hummocky geomorphology visible on satellite images of many Andean volcanoes. This led him to make important contributions to the understanding of catastrophic sector collapses and debris avalanche deposits. This included the first description (Francis et al. 1987) of what is arguably the best-preserved debris avalanche deposit in the world at Volcan Socompa in northern Chile. At this time, he also engaged in archival research on the tsunami generated by the mighty 1883 eruption of Krakatau (Francis 1985). In the 1980s, taking leave from the Open University, Peter held an appointment as Senior Visiting Scientist at the Lunar and Planetary Institute in Houston, Texas, where he studied Martian volcanism (e.g. Francis & Wood 1982; Francis & Wadge 1983), and was subsequently a Visiting Professor in the Planetary Geosciences Division of the University of Hawaii at Manoa until 1991. This led to several productive collaborations with US-based scientists. Peter became involved in the volcanology team for NASA's current Earth System Enterprise spaceborne mission. It was while in Washington DC for a NASA team meeting that he met Mary George, a senior British civil servant then seconded to the British Embassy, whom he married in London in June 1991 (while Pinatubo erupted!). By the mid-1980s, a new generation of Landsat satellites had arrived, providing higher spatial resolution and extending into the short wavelength infrared region. While surveying Andean volcanic summits for signs of recent activity using these new image data, Peter made the chance discovery of an infrared 'glow' within the summit crater of Lascar volcano in northern Chile in 1985 (Francis & Rothery 1987). Until then, this volcano had not been recognized as active but, one year later, its lava dome exploded, showering the distant town of Salta in Argentina with ash. The modern era of eruption detection and early warning from satellites had arrived.
OBITUARY
Fig. 2. Peter would throw on the brakes while driving around Montserrat if he saw decent-looking mango windfalls on the road. His Montserratian colleagues considered him eccentric for this, but he enjoyed supplementing the limited island diet of roti and goat stew with fresh fruit in this wav.
Fig. 1. Peter in the operations rooms at MVO, Mongo Hill. Peter was usually distinctively dressed for the job, though unusually in this scene he has not put his sandals over dark socks. In the early 1990s, Peter alighted on the volcanological potential of developments in infrared spectroscopy that had led to the manufacture of portable and robust equipment suitable for field use. Peter obtained a commercial Fourier transform spectrometer and initiated the most complete investigations carried out with this technique to date. These included field measurements in Italy, Nicaragua and Montserrat (e.g. Francis et aL 1998). Peter led a major ECfunded project to study gas and aerosol emissions from Mount Etna, which was completed shortly before his death. This brought together colleagues from Italy, France and the UK in a memorably convivial campaign that combined in-plume sampling, ground-based, airborne and satellite remote sensing studies, and outstanding dining on funghi porcini dishes in Nicolosi's fine restaurants. Peter was particularly enchanted by the simple beauty of rural Sicily, and was seldom happier than when telephoning Mary from the hilltop village of Novara di Sicilia, high in the Peloritan mountains, and commanding a magical view of the Eolian islands on a clear day. Peter carried out three tours of duty as Deputy Chief Scientist at Montserrat Volcano Observatory (MVO), and attended several scientific assessment meetings (Figs 1 and 2). He was on-island during the extraordinary period of cyclic activity in late 1997, and the 3 July 1998 dome collapse, and co-wrote the corresponding scientific reports (Young et al. 1997; Antenor et al. 1998). His contributions at MVO spanned research, surveillance and hazard management, focusing on many aspects including the crater wall stability, lahars, gas monitoring, and delineation of hazard zones. Many people came to respect him through the positive, considered and equitable way in which he engaged in scientific debate on the crisis. His decision-making was always informed by the immense breadth and depth of his expertise and reading, his practical knowledge of the vagaries and traits of volcanoes, and his irreproachable humanitarian instincts. He was renowned for his modesty, fairmindedness and level-headed judgement, even during times when egos obstructed the path of science, and when contentious issues raised the hackles of others. His contribution was described by one MVO scientist as the "true and calm voice of reason'.
Perhaps unsurprisingly, he had a love-hate relationship with Montserrat. He found the close proximity to an ongoing eruption utterly absorbing and deeply educational. I remember him remarking how he had never appreciated just how much of a filthy mess a volcano can make when it dumps ash on settled areas. He was by turns fascinated and frustrated by the interwoven scientific, cultural and political complexities to crisis management (Francis 1996; Aspinall et al. 1998). In scientific mode, the torpid pace of island life would get to him; off-duty, he would happily unwind on the verandah at Mongo Hill with a gin and tonic, or go for a world-class sunset swim at Little Bay. Peter w7as instrumental in setting up the long-term gas surveillance programme for MVO. Motivated by the need to complement geophysical and geodetic monitoring data, and his career-long interest in remote sensing techniques, he successfully applied for emergency funds from the UK Natural Environment Research Council in 1996 to purchase a Correlation Spectrometer for use by the observatory. Several scientific papers and PhD thesis chapters arose from the measurements of SO2 flux it obtained (Young et al. 1998; Watson et al. 2000; Oppenheimer et al. 1998c). He also initiated a more experimental programme of remote surveillance of Soufriere Hill's gas composition, using open-path Fourier transform spectroscopy (Fig. 3). The dataset revealed the ratios of sulphur to chlorine in the summit gas plume. It is now the longest running of its kind for any eruption, and several publications have arisen from the work (e.g. Oppenheimer et al. 1998a. b). This work informed Peter's broader perspective on degassing during dome-building eruptions, which he elaborated on at a conference on andesite volcanoes, convened by the Royal Society of London, shortly before his death (Francis et al. 2000). Peter was a co-organizer of the meeting. It was the last time many of us were to enjoy his company. Peter had a tremendous impact on the teaching of Earth sciences at the Open University. Always an innovator, he launched the OU's first Earth system science course. and became Director of Teaching for the Earth Sciences Department in 1996. In 1998. he was awarded the title of Professor of Volcanology in recognition of his research and teaching accomplishments. With a rare gift for writing (inspired by the clarity of Lawrence's prose in Seven Pillars of Wisdom) and a drive to communicate science to the wider audience, Peter was a great popularizer of volcanology and planetary science through books and magazine articles. His early Volcanoes book, published by Penguin in the 1970s (Francis 1976). was the only popular text on volcanoes at the time, and was avidly read by everyone who got involved at Mount St Helens. It turned many people on to a subject that had previously been considered
OBITUARY
Fig. 3. Peter measuring volcanic gases in Plymouth with the infrared spectrometer in 1996. just old rocks. He later revised it for Oxford University Press, whilst on leave in Hawaii (Francis 1993). Peter earned the respect and admiration of colleagues all over the world, and those of us who became his friends deeply miss his intellectual agility, generous spirit, wise counsel, and whimsical, usually provocative, sense of humour. He was an inspirational supervisor who ceaselessly nurtured the intellectual and career development of his research students and postdoctoral assistants. In his spare time, Peter was an amateur pilot, an accomplished sailor, an avid bookworm, and a keen walker and jogger. He loved aphorisms and mottos. One of his favourites was 'carpe diem'. While there was so much he still wanted to do, and so much more he would have contributed to volcanology, no-one could deny that he lived life to the full, along the way inspiring thousands of people with his books, lectures, and Open University courses, and, above all, enriching and changing the lives of Mary, family and friends. Another proverb that Peter used to delight in pointing out was 'man that is born of woman hath but a short time to live'. Perhaps he knew something. Clive Oppenheimer With special thanks and acknowledgements to C. Foster, T. Morrison, S. Self, D. Rothery, S. Sparks, G. Norton, R. Stewart, P. Cole, J. Barclay, B. Voight, W. Aspinall, G. Woo, P. Baxter, K. Rowley, and M. Francis. References ANTENOR, C., BONADONNA, C., FRANCIS, P., LUCKETT, R., ROBERTSON, R., NORTON, G., ROWLEY, K., WALKER, C., WATTS, R. & MVO STAFF. 1998. The events of July 3, 1998. MVO Special Report 7.
ASPINALL, W., FRANCIS, P., LYNCH, L., ROBERTSON, R., ROWLEY, K., SPARKS, S. & YOUNG, S. 1998. Scientists at the sharp end in a disaster zone. Nature, 393, 728. FRANCIS, P. 1976. Volcanoes. Penguin, Middlesex. FRANCIS, P.W. 1985. The origin of the 1883 Krakatau tsunamis. Journal of Volcanology and Geothermal Research, 25, 349-363. FRANCIS, P. 1993. Volcanoes: a Planetary Perspective. Oxford. FRANCIS, P. 1996. Volcanoes - Dangers hang over dependency. Nature, 383, 28. FRANCIS, P. W. & ROTHERY, D. A. 1987. Using the Landsat Thematic Mapper to detect and monitor active volcanos: an example from Lascar volcano, northern Chile. Geology, 15, 614-617. FRANCIS, P. W. & WADGE, G. 1983. The Olympus Mons aureole - formation by gravitational spreading. Journal of Geophysical Research, 88, 8333-8344. FRANCIS, P. W. & WOOD, C. A. 1982. Absence of silicic volcanism on Mars - implications for crustal composition and volatile abundance. Journal of Geophysical Research, 87, 9881-9889. FRANCIS, P. W., ROOBOL, M. J., WALKER, G. P. L., COBBOLD, P. R. & COWARD, M. 1974. The San Pedro and San Pablo volcanoes of northern Chile and their hot avalanche deposits. Geologische Rundschau, 63, 357-388. FRANCIS, P. W., O'CALLAGHAN, L., KRETZSCHMAR, G. A., Thorpe, R. S., SPARKS, R. S. J., PAGE, R. N., DEBARRIO, R. E., GILLOU, G. & GONZALES, O. E. 1983. The Cerro Galan ignimbrite. Nature, 301, 51-53. FRANCIS, P. W., GARDEWEG, M., RAMIREZ, C. F. & ROTHERY, D. A. 1987. Catastrophic debris avalanche deposit of Socompa volcano, northern Chile. Geology, 13, 600-603. FRANCIS, P., BURTON, M. & OPPENHEIMER, C. 1998. Remote measurements of volcanic gas compositions by solar FTIR spectroscopy. Nature, 396, 567-570. FRANCIS, P., HORROCKS, L. & OPPENHEIMER, C. 2000. Monitoring gases from andesite volcanoes, Philosophical Transactions of the Royal Society, 358, 1567-1584. OPPENHEIMER, C., FRANCIS, P. & MACIEJEWSKA, A. 1998#. Volcanic gas measurements by helicopter-borne fourier transform spectroscopy. International Journal of Remote Sensing, 19, 373-379. OPPENHEIMER, C., FRANCIS, P. & MACIEJEWSKI, A. 19986. Spectroscopic observation of HC1 degassing from Soufriere Hills volcano, Montserrat. Geophysical Research Letters, 25, 3689-3692. OPPENHEIMER, C., FRANCIS, P. & STIX, J. 1998c. Depletion rates of SO2 in tropospheric volcanic plumes. Geophsyical Research Letters, 25, 2671-2674. WATSON, I. M., OPPENHEIMER, C., VOIGHT, B, FRANCIS, P. W., CLARKE, A., STIX, J., MILLER, A., PYLE, D. M., BURTON, M. R., YOUNG, S. R., NORTON, G., LOUGHLIN, S., DARROUX, B. & MVO STAFF. 2000. The relationship between degassing and deformation at Soufriere Hills volcano, Montserrat. Journal of Volcanology and Geothermal Research, 98, 117-126. YOUNG, S., FRANCIS, P. W., BARCLAY, J., CASADEVALL, T. J., GARDNER, C. A., DARROUX, B., DAVIES, M. A., DELMELLE, P., NORTON, G. E., MACIEJEWSKI, A. J. H., OPPENHEIMER, C., STIX, J. & WATSON, I. M. 1998. Monitoring SO2 emission at the Soufriere Hills volcano: implications for changes in eruptive conditions. Geophysical Research Letters, 25, 3681-3684. YOUNG, S. R., COLE, P. D., CALDER, E. S., BAPTIE, B. A., BONADONNA, C., FRANCIS, P. W., HERD, R. A., JACKSON, P., LOUGHLIN, S. C., LUCKETT, R., NORTON, G. E., ROWLEY, K., SPARKS, R. S. J., WATTS, R. & MVO STAFF. 1997. Dome collapse and Vulcanian explosive activity, September to October 1997. MVO Special Report 5.
Setting, chronology and consequences of the eruption of Soufriere Hills Volcano, Montserrat (1995-1999) B. P. KOKELAAR Earth Sciences Department, University of Liverpool, Liverpool, L69 3BX, UK (e-mail:
[email protected])
Abstract: The eruption on Montserrat during 1995-1999 was the most destructive in the Caribbean volcanic arc since that of Mont Pelee (Martinique) in 1902. It began on 18 July 1995 at the site of the most recent previous activity, on the flank of a c. 350-year-old lava dome within a sector-collapse scar. Phreatic explosivity occurred for 18 weeks before the onset of extrusion of an andesitic lava dome. Dome collapses produced pyroclastic flows that initially were confined by the sector-collapse scar. After 60 weeks of unsteadily accelerating dome growth and one episode of sub-Plinian explosivity, the dome eventually overtopped the confining scar. During 1997 almost two-thirds of the island was devastated following major dome collapses, two episodes of Vulcanian explosivity with fountain-collapse pyroclastic flows, and a flank failure with associated debris avalanche and explosive disruption of the lava dome. Nineteen people were killed directly by the volcanic activity and several were injured. From March 1998 until November 1999 there was a pause in magma ascent accompanied by reduced seismic activity, substantial degradation of the dome, and considerable degassing with venting of ash. The slow progress and long duration of the volcanic escalation, coupled with the small size of the island and the vulnerability of homes, key installations and infrastructure, resulted in a style of emergency management that was dominantly reactive. In order to minimize the disruption to life for those remaining on the island, following large-scale evacuations, scientists at the Montserrat Volcano Observatory had to anticipate hazards and their potential extents of impact with considerable precision. Based on frequent hazards assessments, a series of risk management zone maps was issued by administrative authorities to control access as the eruption escalated. These were used in conjunction with an alert-level system. The unpreparedness of the Montserrat authorities and the responsible UK government departments resulted in hardship, ill feeling and at times acrimony as the situation deteriorated and needs for aid mounted. Losses and stress could have been less if an existing hazards assessment had registered with appropriate authorities before the eruption.
The eruption of Soufriere Hills Volcano during 1995-1999 devastated the small Caribbean island of Montserrat, which is an Overseas Territory of the UK. Approximately 60% of the island, including the most densely populated districts, was designated unsafe for human habitation. Of the original population of c. 10 500, 92% suffered evacuations and many families were relocated two or three times. At the climax of the crisis, in 1997, almost 1600 people were accommodated in basic temporary shelters, and by early 1998 roughly 70% of the population had left the island. Most of the administrative, commercial and industrial facilities were destroyed or rendered inaccessible, as were the airport, harbour and prime agricultural land. Also lost was much of the verdant paradise that attracted tourists and numerous residential migrants from the North American winter, all of whom contributed significantly to Montserrat's economy. More than two-thirds of businesses were closed by October 1998. Insurance companies curtailed or withdrew cover as the eruption escalated in August 1997, which was just before most of the losses were incurred. Consequently, the local financial institution concerned with mortgages and savings collapsed. The Montserrat economy, only recently in budgetary surplus, was plunged back into dependency on UK financial aid. Unofficial insurance industry sources estimate that total losses could be as much as £1 billion if real estate is not recovered. Whereas health problems were less than in many other natural disasters with catastrophic onset (e.g. monsoon floods), the protracted emergency led to considerable psychological distress and related health problems for many Montserratians (Clay et al. 1999). Perhaps more challenging still for the future population will be the linking of disaster recovery with sustainable development (see Possekel 1999). The eruption and associated hazards escalated only slowly, step by step from 1995 through to 1997, and from the outset many inhabitants indicated a strong preference to remain on the island. Understandably, the Government of Montserrat wished to preserve life as near to normal as possible and to avoid jeopardizing the longterm viability of the island community. The UK Government policy was that people would be supported to remain on the island as long as there was a viable safe area. Given this scenario, a reactive strategy for emergency management was inevitable. The management strategy adopted was to react to changing levels of risk as they were identified, rather than immediate and complete evacuation to
an entirely safe area. Consequently, considerable importance was placed on scientific monitoring, hazard anticipation, risk assessment and communication of risks to officials and the public. There were no contingency plans. Many actions taken by both the UK Government and the Government of Montserrat were driven stepwise by events in the volcanic escalation. Initially, because there was no clear understanding of how the eruption might develop, much of the on-island emergency management involved solutions for the short term. Similarly, UK Government departments attempted initially to deal with the crisis using normal institutional arrangements. However, as the eruption escalated it became clear that some aspects of the handling of the emergency were unsatisfactory and that longer-term solutions were required. In 1997 the House of Commons Select Committee on International Development recommended an independent evaluation of the UK Government's response to the Montserrat volcanic emergency (IDC 1997). The terms of reference requested identification of key findings and lessons learnt. The present author was asked to evaluate the scientific monitoring and risk assessment, and to lay out the course of the volcanic developments against which the emergency management response could be charted. Aspects of this paper originate in work for the evaluation study, which was published recently (Clay et al. 1999). This paper is mainly to provide a factual record that sets the scene for the more analytical scientific contributions that follow. Plates 1-20 provide a pictorial narrative, principally concerning the characteristic styles of eruption and their effects. Some aspects of the handling of the emergency are analysed to distinguish problems and their possible solutions. Facets of the history of Montserrat are given (mainly tabulated), because they are relevant to understanding the plight of the people and their perceptions regarding both the handling of the emergency and the aid provided by the UK Government (see also Fergus 1994; Pattullo 2000). The judgements in this paper are those of the author. However, the author has benefited from full co-operation of the scientists involved, with generous provision of information and guidance. Although not explicitly woven into the narrative, it should be noted that, in addition to the core scientific team from the Caribbean and the UK, scientists from France, Puerto Rico and the USA made valuable and considerable contributions in the crisis management.
DRUITT, T. H. & KOKELAAR, B. P. (eds) 2002. The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 1-43. 0435-4052/02/S15 © The Geological Society of London 2002.
B. P. KOKELAAR Setting of the volcanic crisis Geological setting Montserrat is in the northern part of the Lesser Antilles volcanic island arc (Fig. 1). The arc results from westward subduction of Atlantic oceanic lithosphere beneath the Caribbean plate. From Martinique southwards to Grenada the arc comprises a closely spaced double chain of volcanoes, the eastern elements of which date back to the Eocene. From Martinique northwards an extinct eastern volcanic chain, of Eocene to mid-Oligocene age, diverges via Marie Galante to Sombrero, while Montserrat lies in the western active chain that extends to seamounts up to 100 km NW of Saba. The northern arc volcanoes are founded on a Cretaceous oceanic island arc (Bouysse & Guennoc 1983; Wadge 1986). Earthquake studies indicate that the subducted oceanic slab is segmented into three main parts with differing dips and slip vectors (Wadge & Shepherd 1984). Montserrat is above the northern segment, overlying crust that is no more than 30 km thick, an asthenospheric wedge that extends to 130km depth, and a Benioff zone that dips westwards at 50-60°. Montserrat is the top of a compound volcanic
edifice that extends from 1 km above sea level, at Soufriere Hills Volcano, to 700-900 m below sea level, where the basal diameter is c. 25-30 km. Building of a volcano here probably initiated in the Miocene (c. 9 Ma; Briden et al. 1979). when the axis of the northern part of the arc migrated westwards. However, the oldest exposed rocks are Pliocene (c. 2.6 Ma) and most are Pleistocene or younger in age (see Harford et al. 2002). Recent convergence of the Atlantic and Caribbean plates has been quite slow, at 20-40 mm a !. and magma productivity has consequently been low ( 3-5 km 3 M a - 1 km-1 of arc). especially in northern and southern parts of the arc (see Wadge 1984; M acdonald et al. 2000). Basalts of the northern volcanic chain are predominantly of low-K or medium-K type (low-K tholeiite and low-K calcalkaline; Rea 1982). with compositional trends through to andesite and dacite mainly controlled by polybaric crystal fractionation with limited magma hybridization. Montserrat is predominantly composed of porphyritic andesites. with basaltic rocks represented in volumetrically minor outcrops (South Soufriere Hills; Rea 1974) and common mafic inclusions in the more evolved rocks. The Soufriere Hills andesites erupted during 1995-1998 clearly implicate basalt in their petrogenesis and ascent. Experimental and petrological
Fig. 1. Location of Montserrat in the Lesser Antilles volcanic island arc (modified from Wadge 1986).
SETTING AND CHRONOLOGY OF THE ERUPTION
Fig. 2. Map showing Montserrat as it was before the eruption, which initiated in July 1995 through Castle Peak (lava dome). studies show that basalt at 1050°C invaded and mingled with hydrous andesitic crystal mush (c. 60-65% crystalline with interstitial melt containing c.4-5wt% H2O) at c. 830-860°C, heated it, and then erupted with it as inclusions (Barclay et al. 1998; Devine
et al. 1998; Murphy et al. 1998, 2000). These studies, together with analyses of seismic and deformation signals of conduit processes (Aspinall et al. 1998; Mattioli et al. 1998; Voight et al. 1999), are consistent with tapping of a long-lived reservoir >6 km below the
Fig. 3. Maps of (a) prehistoric fans of pyroclastic and lahar deposits of Soufriere Hills Volcano (modified from Roobol & Smith 1998). showing these as preferred sites for homes, key installations and infrastructure, and (b) the extent of areas devastated by pyroclastic flows during 1995-1999 (after Cole et al. 2002).
SETTING AND CHRONOLOGY OF THE ERUPTION vent. The form of the reservoir, however, is not well defined and the controls and duration of its replenishment are only just beginning to be understood (see Murphy et ai 2000). Cyclicity of seismic and magmatic activity detected on scales of about six to seven weeks can be related to processes in the conduit and magma chamber (Denlinger & Hoblitt 1999; Melnik & Sparks 1999, 2002; Voight et al. 1999; Wylie et al. 1999). The recurrence interval of approximately 30 years for volcanoseismic crises at Montserrat (Table 1) seemingly reflects the frequency of substantial perturbation of the magmatic system by influx of basalt from greater depths. The understanding of this volcanic plumbing system behaviour is an exciting prospect for the future, with particular significance for possible quantitative forecasting of eruptions. Physiography Montserrat is a small island, approximately 16.5km north to south and 10km east to west (c. 100km 2 ; Fig. 2, Plates 1,2). Its topography is dominated by four main volcanic massifs, each with many valleys and ridges radiating towards and truncated at a coastline predominantly of steep cliffs. The massifs, from north to south, Silver Hill (403m), Centre Hills (740m), Soufriere Hills (preemption 914m at Chances Peak) and South Soufriere Hills (756m), each represent composite eruptive centres, mainly of andesitic lavas, although deep erosion has modified most original volcanic landforms. The less substantial St George's Hill, Garibaldi Hill and Roche's Bluff mainly comprise volcaniclastic deposits. New representative 40 Ar/ 39 Ar age determinations are presented by Harford et al. (2002). Soufriere Hills Volcano, which is the youngest centre, retained little-modified primary features in its sector-collapse scar (English's Crater) and Castle Peak lava dome within (see Fig. 2, Plate 2B), although these are now substantially obliterated. Thermal waters found widely on Montserrat, including the numerous hot springs and fumaroles (soufrieres) on the flanks of Soufriere Hills Volcano (main ones labelled 1-4 in Fig. 2), reflect a sustained deep supply of both magmatic heat and volatiles to overlying aquifers (Chiodini et al. 1996; Hammouya et al. 1998). The gentler slopes flanking the volcanic massifs in the southern two-thirds of Montserrat are composed of volcaniclastic deposits from Soufriere Hills Volcano and were the sites favoured for habitation and infrastructure (Figs 2, 3a, Plates 1, 2A). The capital town, Plymouth, and its environs on the west coast, the airport on the east coast, and numerous communities in between on the northern flanks of the volcano, were all built on incised prehistoric fans primarily of pyroclastic flow and lahar deposits from the volcano (Plate 1A, B; see Roobol & Smith 1998). Montserrat's climate is maritime subtropical. In the period from 1992 to 1997, winds towards the west tended to prevail at low levels (1-5 km altitude) and high levels (20-30 km), and towards the east at intermediate levels (8-18 km), with standard deviations for directions over the 30km of altitude of 30-162 (Bonadonna et al. 2002). During 1997, ash plumes from dome-collapse pyroclastic flows and Vulcanian explosions commonly ascended to 15 km, and tephra was dispersed by intermediate-level winds, at different times, towards the north, NW, NE, south, SW and SE. Average rainfall ranges from 1 ma" 1 near sea level to >2.5ma - 1 in the hills (Possekel 1999). Torrential rain is associated with hurricanes that all too frequently track northwestwards through the eastern Caribbean. Before the 1995 eruption the vegetation in the hills was mainly secondary forest (little indigenous forest remaining), whereas on the less steep slopes and volcaniclastic fans there was mainly bush or cultivated land (Plates 1, 2, 3A). It was Montserrat's originally lush and exotic vegetation that earned it the epithet 'Emerald Isle of the Caribbean', recalling the verdant homeland of the early Irish colonists. Deforestation and inappropriate land use, however, coupled with the torrential rain, left many slopes eroded and susceptible to landslides. Montserrat's small size and predominantly rugged terrain severely constrained on-island options for volcanic risk mitigation. The location of most human activity and infrastructure in highly vulnerable areas maximized the impact of the eruption (Fig. 3b)
5
A brief history of Montserrat leading to the eruption during 1995-1999 In 1998 Montserrat became a UK Overseas Territory, having previously been a UK Dependent Territory. Although tragically the eruption had just rendered Montserrat once more dependent on UK financial aid, the change of title constituted one further advance from a history of some 300 years of British colonial status, commercial exploitation and slavery (see Fergus 1994). Basic features of social justice were secured only quite recently. Power invested in a Montserratian Chief Minister with a ministerial government dates from 1961, although issues concerning national security and international relations remain the business of the Governor of Montserrat, who is answerable to the UK Government. The duality of governance of Montserrat caused some problems in the management of the volcanic crisis, as described by Aspinall et al. (2002). Table 1 gives key developments in Montserrat's emergence into a free and economically viable small-island nation. Alongside this are charted the volcanic activity and volcanoseismic surveillance, both on Montserrat and on other Caribbean islands, as they bear on Montserrat's preparedness for the Soufriere Hills eruption. Prior to the devastation inflicted by Hurricane Hugo in 1989, Montserratians had acquired good standards of accommodation, education and health-care services, and, with UK aid following the storm, they were on the verge of almost complete recovery when disaster struck again. Chronology, nature and nomenclature of the volcanic crisis The 1995 eruption of Soufriere Hills Volcano involved a slow, incremental escalation of volcanic activity and associated hazards, after several years of precursory seismic activity. With the small size of the island, and with the population located mainly on the flanks of the active volcano (Fig. 3a, Plates 1, 2A), the slow escalation caused several distinct problems in emergency management. These are outlined both in the following narrative chronology of events and in succeeding sections. Key developments of the eruption, emergency responses and effects on the population of Montserrat are listed in Table 2 (see also Plates). Precursors Increased seismicity in the vicinity of Montserrat was initially detected in April 1989. Eighteen low- to moderate-intensity swarms of volcanotectonic earthquakes close to Soufriere Hills Volcano were registered at intervals from January 1992 and particularly from mid- to late 1994, before the first phreatic explosion on 18 July 1995 (Ambeh & Lynch 1996; Aspinall et al. 1998; Table 1, Fig. 4). Hot springs and fumaroles (soufrieres) on the volcano flanks showed little change prior to the eruption, although in March 1995 Galway's Soufriere showed pronounced magmatic signatures in 3 He/ 4 He and 13C, like those subsequently measured in gas from the andesitic lava dome (Hammouya et al. 1998). The unrest was on schedule according to the previously recognized c. 30 year cyclicity of volcanoseismic crises at Montserrat (Table 1; Wadge & Isaacs 1988), but none of the detected precursors was a clear, unambiguous indicator of an imminent eruption. Styles of eruption and pyroclastic fallout Following an opening phreatic phase, most of the eruption from 1995 to 1998 involved slow, unsteady ascent and extrusion (0.511 m3 S-1) of andesitic magma of high to extremely high viscosity (c. 106 to >10 I4 Pas), which formed a composite lava dome comprising numerous shear lobes (Sparks et al. 1998; Voight et al. 1999; Sparks & Young 2002; Watts et al. 2002). However, magma also erupted explosively from the conduit on several occasions, including two protracted intervals. On the first occasion (17-18 September 1996), during sub-Plinian explosive activity, fragmentation in the
B. P. KOKELAAR
6
Table 1. Historical development of Montserrat and the region leading to the 1995 eruption Occurrence
Volcanic activity
c. 3950 BP
English's Crater forms (Roobol & Smith 1998).
Sector collapse.
c. 3000 BP 1493
South American Amerindians first settle on island. Columbus sails along west coast of island (apparently deserted), and names it Santa Maria de Monserrate after an abbey in mountains near Barcelona (Spain) where a similar rugged profile occurs.
1500s to early 1600s; likely 1620s
Castle Peak dome forms in English's Crater with several pyroclastic layers deposited (Young et al. 1996, 1998; S.R. Young pers. comm. 2000).
1632
Irish Catholics first colonize as religious refugees from nearby St Kitts. shortly afterwards joined by Irish Catholic dissidents from Virginia.
Mid- 1600s
English colonial control established. Arrival of exiles from Ireland (deported by Cromwell) and transportees (criminals). African slaves imported, mainly to work in sugar plantations. Population comprises Anglo-Irish plantation owners, poorer Irish servants and increasing numbers of slaves. Frequent raids by French and Caribs. French capture island: restored in Treaty of Versailles.
Date
Late 1600s- 1700s 1782-1783 Early 1800s 1834
1838-
1866
1897-1898
1902 1933-1937
1936 1951 1959 1961
1966-1967
1967
1971-1972
1976 1979 1981 1987 1988 April 1989
Slave population exceeds 6500. Abolition of slavery by Act of UK Parliament. Full emancipation of slaves. Former slave population continues to struggle to establish independent peasant culture, being compromised by land-lease arrangements (share-cropping) and political control by whites. UK tightens control by establishing Crown Colony rule. Governor (appointed in UK) heads Legislative Council with six members appointed by him. New hot springs and fumaroles (Gages Lower Soufriere) initiated on volcano flank. Eruption at St Vincent kills c. 1500 persons. Eruption at Martinique (Mont Pelee) kills c. 29 500. Perret makes observations 1934-1937 (Perret 1939); Royal Society expedition in 1936 consequent upon petition following destructive earthquakes in 1935 and concern about possible major eruption (MacGregor 1938). Gages Upper Soufriere reactivated.
Andesitic lava and ash eruption(s).
Volcanoseismic crisis.
Volcanoseismic crisis.
New constitution includes four elected members of Legislative Council. Universal adult suffrage introduced. Share-cropping ended. Political power wrested from white merchant-planter class. Ministerial government established and led by first Chief Minister. Governor retains responsibility for national security, civil service and foreign relations. Galway's and Tar River Soufrieres become more active. Movement of magma from > 10km depth inferred (Shepherd et al. 1971). Affiliation with West Indian Federation rejected, effectively reaffirming colonial status; Montserrat becomes a UK Dependent Territory. Sale of 600 acres of prime land to North Americans and Europeans initiates relative economic boom.
Volcanoseismic crisis.
Eruption at St Vincent. On neighbouring Guadeloupe, phreatic explosions at La Soufriere lead to evacuation of 72000 persons (Fiske 1984); estimated cost c. £200 million. Eruption at St Vincent. Montserrat no longer in need of UK budgetary aid. Wadge & Isaacs (1987) report on hazards due to Soufriere Hills Volcano submitted to sponsors, noting Plymouth to be vulnerable. Lesser Antilles Volcanic Assessment Seminar hosted by Seismic Research Unit (Trinidad) and attended by Montserrat government representative(s). Wadge & Isaacs" (1988) findings published in an international journal. Seismic Research Unit prompted to deploy second and third seismic stations at Montserrat.
Seismic activity escalates above background. continued
7
SETTING AND CHRONOLOGY OF THE ERUPTION Table 1. (continued] Date
Occurrence
17 September 1989
Hurricane Hugo totally destroys 20% of homes, with 50% severely damaged; nearly 25% of population homeless. Plymouth devastated; total damage estimated at £150 million. 1 1 persons killed. £3 million from UK as emergency aid. Average wind speeds c. 240 km h^1 (c. 67ms~') with gusts over 300 km h" 1 . All three seismic stations destroyed. £16.8 million capital aid programme approved in UK. New government headquarters, library and hospital, all built in Plymouth. Seismic stations restored by Seismic Research Unit.
1991
January 1992 July 1993
Mid- 1994
End Nov.-Dec. 1994
Volcanic activity
Earthquakes occur in distinct swarms. Hypocentres located up to 1015km depth (Robertson et al 2000). Governor assists Montserrat Government in initiating upgrade of disaster preparedness. Three additional seismic stations established. Direct links of two stations to Seismic Research Unit, via Antigua, restored. Hypocentres up to 10-1 5 km depth. Head of Seismic Research Unit gives public interviews to reassure population concerning the earthquakes.
Early 1995
Data from six seismic stations telemetered to Emergency Operations Centre in Plymouth, with possible events forwarded to Seismic Research Unit. National Disaster Action Plan (manual) delivered, with virtually no reference to volcanoes. Potato crops fail on volcano flank.
18 July 1995
Early mid-afternoon: jet-engine-like roaring noises, sulphurous smell and ash fallout. Population of Montserrat c. 10500.
Start of volcanoseismic crisis.
Increasing seismicity.
Intense swarm of felt earthquakes.
Soufriere Hills Volcano erupts.
Table 2. Progress of volcanic activity with related emergency responses and effects on the population of Montserrat Date
Volcanic activity
Response
18 July 1995
Onset of eruption.
Emergency Operations Centre activated in Plymouth. On-island population c. 10500.
28 July 1995
Military contingency evacuation plans completed both for removal to north and off-island. Long Ground temporarily evacuated. First major evacuation of 6000 from southern and eastern areas, which lasted for 2 weeks.
21-22 August 1995
Major phreatic explosions.
Mid- to late November 1995 1-2 December 1995
Onset of lava-dome growth.
Long Ground and White's Yard evacuated.
Onset of minor pyroclastic flows.
Second major evacuation of 6000, which lasted for 1 month.
January 1996 21 March 1996 31 March 1996 3-4 April 1996
Onset of major pyroclastic flows.
April 1996
Civilian contingency evacuation plans (Operation Exodus) initiated. Government of Montserrat confirms acceptance of budgetary aid conditions. On-island population c. 9000 Governor declares state of public emergency. Plymouth and southern areas evacuated finally. Population in temporary shelters 1366 (gradually declined until August 1997). Voluntary Evacuation Scheme gives evacuees leave to remain in UK for 2 years. Risk management zone map introduced (Fig. 5a). On-island population c. 7500. £25 million aid for 2 years agreed.
May 1996 August 1996
Pyroclastic flows reach the sea.
September 1996
First major magmatic explosion; ballistic Revised risk management zone map issued, dated October (Fig. 5b). blocks >1 m diameter wreck Long Ground. First Galway's Wall crisis. Revised risk management zone maps issued dated November and December (Fig. 5c, d). Red Alert requiring further evacuations (19 December) largely ignored. Dome material overtops Galway's Wall On-island population c. 6000. Revised risk management zone map for first time. issued (Fig. 5e).
End November into December 1996
February 1997
continued
B. P. KOKELAAR
8
Table 2. (continued) Date
Volcanic activity
Response
May 1997
Dome growth switches to north and escalates.
Early to mid-June 1997
Increasing dome-collapse and pyroclastic flow activity in northern drainages.
25 June 1997
Pyroclastic flows kill 19 persons and injure 8. Surge-derived pyroclastic flow unexpectedly reaches vicinity of Cork Hill.
Population in temporary shelters 775: some residents still refuse to evacuate high-risk zones. Revised risk management zone map issued 6 June (Fig. 5f ): increased risk at airport is explicit. Emergency jetty handed over to Government of Montserrat. Cork Hill evacuated; airport and Plymouth port closed. Search and rescue initiated.
27 June 1997 4 July 1997
Revised risk management zone map issued, requiring further evacuation of western areas (Fig. 5g). Large pyroclastic flows frequent and encroaching Plymouth.
Revised risk management zone map issued (Fig. 5h). abandoning microzonation and designating all hazardous areas as Exclusion Zone.
First series of (13) cyclic repetitive Vulcanian explosions and associated radially directed fountain-collapse pyroclastic flows.
1 160 persons in temporary shelters. £6.5 million emergency housing scheme announced to accommodate 1000 in north of island. Evacuation of areas just north of Belham River. 1598 persons in temporary shelters. Formal assessment by MVO presented to UK Government. UK Foreign Secretary establishes inter-departmental Montserrat Action Group with Ministerial and Cabinet Office monitoring. Assisted Passage (to UK) Scheme announced, to aid relocation.
July 1997 August 1997
September 1997
22 September-21 October 1997
Second series of (75) cyclic repetitive Vulcanian explosions
November 1997 2-5 December 1997 26 December 1997
Revised risk management map issued places Salem and Old Towne in Exclusion Zone (Fig. 5i). Chief Minister visits London securing commitments to aid development of northern Montserrat. Frequent ashfall causes extreme nuisance: many remaining islanders decide to leave. On-island population 3338 Scientists meet in Antigua to produce formal assessment for UK Government: validated on 19 December by Chief Scientific Adviser.
Galway's Wall sector collapse and violent pyroclastic density current.
Early 1998 February 1998
Heavy ashfall in Central Zone.
March 1998
Cessation of magma ascent.
20-21 April 1998 21 May 1998
On-island population 3000. Recommended evacuation of Central Zone generally not heeded. UK Foreign Secretary visits Montserrat. Robbery of bank vault in Plymouth constitutes most significant opportunistic crime of the emergency. UK Government spend on aid related to the volcanic emergency totals c, £56 million Scientists meet in UK to produce formal assessment for UK Government. Evacuees allowed to settle indefinitely in UK.
11 June 1998
UK Government commitment of £75 million over 3 years to 2001. and indicative £25 million for 2001-2002.
14-16 July 1998
Scientists meet on Montserrat: formal assessment confirms lower risk levels.
30 September 1998 October 1998
Revised risk management map issued (Fig. 5j). Phased reoccupation of areas north of Belham River allowed. 427 people still housed in shelters. Montserrat Sustainable Development Plan published. Inquest verdict on deaths of June 1997 published (11 January); it criticizes both UK and Montserrat Governments. Montserrat Country Policy Plan agreed. Revised risk map issued allows Daytime Entry north of Plymouth (Fig. 5k). Assisted Return Passage Scheme (from UK) begins. On-island population recovered to c. 4500.
November 1998 January 1999 12 April 1999 1 May 1999 Mid-1999 November 1999
Lava dome substantially degraded by collapses; ash-venting frequent.
Renewed dome growth.
SETTING AND CHRONOLOGY OF THE ERUPTION
Fig. 4. Hypocentres of earthquakes that occurred in early to middle stages of the Montserrat volcanic emergency (after Aspinall et al. 1998). conduit may have descended to depths close to the magma reservoir (Robertson et al. 1998). By March 1998 a total cumulative volume of 0.3 km3 of magma had been erupted (Watts et al. 2002). From March 1998 until November 1999 there was a pause in magma extrusion,
9
during which time the dome became substantially reduced by collapses and ultimately divided by a deep chasm (Norton el al. 2002). The renewed extrusion from November 1999, which is not dealt with in this Memoir, is considered as forming a second dome (Watts et al. 2002; Sparks et al 2002). Thus the adopted convention is that there is one eruption, ongoing at the time of writing (April 2001), and only one dome formed until November 1999. The growth and partial collapse behaviour of this first dome, however, was extremely varied and on several occasions involved rebuilding from the mouth of the volcanic conduit. The eruption during 1995-1999 involved five main styles of volcanic activity. Each characterized a distinctive phase in the eruptive history, but overlapped with other styles. Phreatic explosions. These were produced by sudden and/or sustained jet-like releases primarily of heated groundwater. They blasted out mainly old volcanic rock, forming small craters, and characterized the opening phase of the eruption (see Plates 2B, 3). A powerful explosion on 28 July was associated with the opening of a new vent and another, on 21 August, produced a slow-moving, cold, dilute pyroclastic density current that precipitated the first evacuation of Plymouth (Table 2). Technically, if the explosions produced ash that included fragments of new (juvenile) magma, the explosions would be referred to as phreatomagmatic. Although juvenile fragments may have been included (e.g. Boudon et al. 1998), this was not clearly demonstrated. The term "phreatic phase' applies to this early activity as it registers the significant involvement of groundwater, irrespective of whether explosions ejected any juvenile material. The heat source is inferred to have been newly arisen magma and associated released volatiles (Gardner & White 2002). Lava-dome growth with dome-collapse pyroclastic flows. Extrusion of andesitic lava, at a rate mainly in the range 0.5-11 m3 s"1 (Sparks et al. 1998; Sparks & Young 2002), formed a steep-sided composite dome comprising numerous shear lobes and spines (e.g. Plates 4, 7, 17). Partial dome collapse due to gravitational instability commonly produced, in genetic terms, dome-collapse pyroclastic flows (Plates 5B, 6A, 8). Watts et al. (2002) document the dome growth, while Calder et al. (1999, 2002) tabulate volumes of deposits formed from dome-collapse pyroclastic flows, their runout parameters and the dates of occurrence. Cole et al. (2002) present sedimentological analyses of the pyroclastic deposits and their parent flows. To derive approximate dense-rock equivalent (DRE) volumes of andesite from deposit or collapse-scar volumes, Calder et al. (2002) utilize bulk densities of 2 x 10 3 kgm~ 3 for deposits, 2.2 x 10 3 kg m~3 for the dome and associated carapace breccias and 2.6 x 103 kgm~ 3 for the andesite. In this paper, the volume of material given as having collapsed is that of the partially fragmental dome, with a bulk density of 2.2 x 103 kgm~ 3 . Minor collapses that formed the talus apron around the dome and had runouts of 2 peaked and then diminished (Robertson et al. 2000; Gardner & White 2002). On 21 August at 08:02 local time (LT), on what became known as Ash Monday, a particularly energetic phreatic explosion produced an ash plume that reached c. 3 km altitude, as well as a slow-moving, cold, dilute pyroclastic surge that enveloped Plymouth for approximately 15 minutes. Car headlights were virtually useless in the darkness there. The surge cloud appeared menacing (more threatening than the actuality) and many inhabitants became frightened and spontaneously evacuated the town. As there could be no guarantee that conditions would not rapidly become dangerous, an official evacuation was ordered. Areas south of Richmond Hill, including Plymouth (Fig. 2), were evacuated on 22 August (c. 5000 people), and, with explosions continuing, the entire area south of the Belham River was evacuated on 23 August (a further c. 1000 people). On 27 August a new vent opened on the north flank of Castle Peak dome as
14
B. P. KOKELAAR
explosions continued. There was controversy between the scientific teams regarding the safety of a return of the population to the south, but partial reoccupation north of Plymouth was permitted on 4 September in order to allow evacuees to gain shelter and prepare for Hurricane Luis. Many had been accommodated only in tents. Hurricane Luis struck on 5 September, with storm noise disrupting much of the ongoing seismic recording. Remaining evacuated areas, including Plymouth, were reoccupied on 7 September or shortly afterwards, while there were more phreatic explosions. The US Geological Survey's Volcanic Crisis Assistance Team departed on 10 September. On 25 September, a new, small (c.4 x 10 4 m 3 ) mound-like mass of lava with a central spine was confirmed to be growing on the SW side of Castle Peak dome (see Plate 3C; Robertson et al. 2000, fig. 3). This growth was thought to reflect either upheaval of old rock due to the shallow emplacement of new magma, or possibly the emergence of a cooled part of a new plug. Throughout the period from the initial explosions until 25 September, there was no clear evidence within the seismic activity of ascent of a substantial amount of new magma, and, similarly, ground deformation monitors in position at that time detected none. However, the flux of SO2 and the variety of seismic signals at shallow levels are taken to indicate that magma may already have migrated to a shallower level before and/or soon after the first phreatic explosion (Gardner & White 2002). This being the case, the upheaval of late September represented renewed ascent of magma. Between 25 September and 30 November (65 days), the volcano was mostly obscured by low atmospheric cloud and there was disagreement concerning the risk attributable to the presence of new magma. Scientists from the UK and USA, mindful of the rapid onset of dangerous activity at St Vincent (1979), Mount St Helens (1980) and Pinatubo (1991), were concerned that the risk on the volcano flanks might be considerable and they recommended evacuation of Long Ground, near to the open side of English's Crater (Fig. 2). Concern was also expressed that the probable existence and implications of the occurrence of new magma had not yet been communicated to Montserratians. On the other hand, the leader of the scientific team (from SRU) adopted a less precautionary stance, which was supported by the Chief Minister. This position was probably influenced by the considerable economic impacts of the evacuation of 72000 people from the flanks of La Soufriere on Guadeloupe in 1976, which was deemed, by some, to have been unnecessary (see Fiske 1984). The elderly and infirm were re-evacuated from southern Montserrat on 5 October. The airport was temporarily closed owing to ash fallout on 23 October, Plymouth was inundated with ash on 31 October, 4 November and 9 November, and southern villages were similarly affected in early November, all due to continued phreatic explosivity. The initiation of continuous lava-dome growth, on 15 November, was marked by remotely monitored deformation in the area of English's Crater and by intense hybrid earthquake activity at depths 1 m) in pyroclastic flow deposits (Grunewald et ai 2000). The friction marks occur as polished and striated slickensided surfaces on a thin layer (5-10 mm) of cataclastic rock (Fig. 16a). Friction marks occur on all sides of blocks, commonly in several directions. Petrographic studies show
64
R. S. J. SPARKS & S. R. YOUNG
that the friction marks are similar to tectonic mylonites and that in the largest marks frictional melting took place to form pseudotachylites (Fig. 16b). Grunewald et al (2000) calculated that the imposition of the weight of a large block (1-10 m) on a small area of a few square centimetres with relative sliding velocities of a few metres per second is sufficient to cause confining stress conditions that are comparable to those experienced on large tectonic faults and to cause frictional melting at over 1400CC. A major implication of the formation of friction marks is that Bingham or sliding-block models of pyroclastic flows cannot be realistic. The flows are best perceived as rapid agitated granular flows in which blocks jostle, rotate and occasionally slide past one another with strong frictional contacts (Drake 1990; Druitt 1998). The magmatic explosive activity during the eruption has provided a superb opportunity to study Vulcanian explosions, which were recorded in detail by video, photography and seismicity (Druitt el al. 2002b). The explosive activity always followed a major collapse of the dome and thus was triggered by rapid unloading of the conduit. As discussed previously and in Voight el al. (1999), the two the series of Vulcanian explosions in 1997 involved repetitive and cyclic explosions over extended periods that can be linked to ground deformation and seismic cycles. Typical Vulcanian explosions lasted a few tens of seconds and were followed by vigorous ash-venting lasting from 30 minutes to an hour. Eruption columns ranged from 3 to 15km high and all but one of the Vulcanian explosions involved fountain collapse, generating pumice-and-ash flows, which were emplaced along all the main drainages (Fig. 7a). The explosions were pulse-like and involved ejection of ballistics up to 1.7km from the dome. The origin of the cyclic and pulse-like character of the explosions has been considered by Melnik & Sparks (2002b), who found that pulse-like behaviour is predicted if the magma fragments at some critical overpressure. Pumice clasts from the explosions are typically angular and platy (see photographs in Druitt et al. 2002b) and are similar to the shapes of clasts produced in laboratory experiments by Adilbirov & Dingwell (1996), in which natural dome rocks and pumices, containing a high-pressure gas, were abruptly decompressed. This similarity supports the idea of brittle fragmentation of a vesicular magma as the cause of the explosions. Melnik & Sparks (20026) found that the characteristics of the Vulcanian explosions can be described by models of the explosive expansion of already vesiculated magma in the upper conduit with no further mass transfer during the event. The eruption has also provided the opportunity to compare models of interaction with the atmosphere and fountain collapse during explosive eruptions, as discussed further by Clarke et al. (2002). The eruption has given the opportunity to investigate the contrasted features of tephra fall deposits generated from plumes above block-and-ash flows related to dome collapses and from magmatic explosive eruptions. The dispersal of tephra is dominated by the local wind conditions in the Caribbean, with low-level (up to 4-6 km) winds blowing to the west and high-level winds to the east. As expected, the tephra deposits from the lofting ash plumes above the block-and-ash flows are very fine-grained and aggregation processes, often with formation of accretionary lapilli, are dominant (Bonadonna et al. 2002). In contrast, the Vulcanian tephra fall deposits are usually strongly bimodal as a consequence of having two sources of tephra, namely relatively coarse-grained lapilli and ash derived from the high vertical columns above the vent, and finegrained plumes of low height generated above the pumice flow (Bonadonna et al. 2002). The abundant cristobalite discovered in the finest respirable components of the ash fall deposits associated with block-and-ash flows has posed a possible health threat to the islanders (Baxter et al. 1999). The cristobalite forms by vapourphase crystallization and devitrification in the dome and is concentrated in the fine elutriated ash by selective crushing of the groundmass in the pyroclastic flows. Detailed geochemical studies (Horwell et al. 2001) show that the physical fractionation is a consequence of the fine-grained groundmass being significantly weaker than the phenocrysts. Baxter et al. (1999) found that the dome rock contained 3-6% cristobalite, but the respirable ash (0.4km 3 ). Thus the basalt magma can be expected to continue to cool and crystallize and supply further gases during residual volcanic activities. The interpretations on the role of the mafic magma in some of the eruptive phenomena discussed above must remain speculative. Clearly there is a major challenge for volcanology: to monitor and understand these deeper processes, which are quite likely to be fundamental to the eruptive process.
Concluding remarks At the close of the twentieth century substantial progress has been achieved in understanding volcanoes. Significant advances have been made as a consequence of technical developments, enhanced funding of science in general and an increasingly quantitative approach. An important factor in progress has been the detailed study of particular eruptions by large multidisciplinary teams using sophisticated techniques. Such concentrations of effort and expertise on single eruptions have been the hallmark of volcanology in the last decades of the twentieth century with investigations of Kilauea, Etna, Piton de la Fournaise on Reunion, and Krafla being examples for basaltic volcanoes and Mount St Helens, Mount Pinatubo, Mount Redoubt and Mount Unzen being examples for more silicic volcanoes. Each investigation has contributed to identifying new phenomena, compiling substantial databases of geophysical, geological and geochemical information, and stimulating development of theoretical work and new ideas. Such case studies have added to the fundamental understanding that ultimately underpins the practical objective of mitigating the effects of volcanic eruptions. The investigations at Soufriere Hills Volcano have benefited from these previous studies. This paper has outlined some of the particular contributions of the Soufriere Hills eruption to the reservoir of knowledge and scientific understanding. Despite the advances in volcanology, eruptions like that at Soufriere Hills Volcano highlight that there is much to do. In assessing the scientific work on Montserrat, Sir Robert May (Chief Scientific Adviser of the UK Government in 1995-2000) was generally complimentary about the work of the MVO, but remarked that the understanding of volcanoes was still largely based on natural history rather than on quantitative science. While we do not completely accept Sir Robert's analysis, his remarks highlight a number of issues that will occupy volcanologists in the next century. Despite the importance of natural history, it is remarkable how few really well-documented examples of eruptions actually exist to develop robust empirical or natural laws of how volcanoes behave. During the Soufriere Hills eruption, the MVO team found that they had remarkably little past experience and information to draw on. It soon became apparent that the only compelling analogue for the eruption, for which there was a significant amount of high-quality documentation and modern data, was the dacitic dome eruption of Mount Unzen (1991-1995). We also found that much of the information that we needed to know about dome eruptions was not in a readily accessible form, with only the Smithsonian database and the pioneering paper of Newhall & Melson (1983) providing some systematic information on dome-forming eruptions. There were several stratigraphic studies of dome complexes and their products, including Wadge & Isaacs (1988) for Montserrat. The information in such studies is necessarily partial and incomplete due to the vagaries of geological preservation. This information needed to construct a natural history of dome-forming eruptions is actually very limited. Our science might be compared with meteorology, where scientists are faced with phenomena of equal complexity, such as hurricanes.
Meteorologists, however, have a good understanding of the natural laws of hurricanes, because there are several hurricanes per year to study. At the moment there are very few cases of well-documented dome eruptions. Sir Robert was perhaps a little unfair in his implication that volcanology has not advanced into a quantitative science. Certainly there have been substantial developments in volcano geophysics and modelling of volcanic processes, which have been aided by the huge increases in computer power. Mathematicians, engineers and physicists have involved themselves in the science and geologists have tended to embrace these more quantitative approaches. Nevertheless Sir Robert's perspective has some truth to it. To a large extent we do not yet understand the data that come from monitoring active volcanoes very well; at least not well enough to be able to make confident forecasts. A few examples are worth giving. The SO? flux has varied greatly during the Soufriere Hills eruption (Young et al. 1998a), but was a decrease in flux good or bad news? A flux decline might mean gas was being stored and pressure was building prior to explosive activity (bad news) or it might mean that magma ascent was declining (good news). In the period of residual volcanic activity from March 1998 to November 1999 there were sporadic volcanotectonic earthquake swarms and small-scale but still impressive explosions and ash-venting (Norton et al. 2002). Prior to the resumption of dome growth in November 1999 the scientists involved in Montserrat struggled with interpreting this residual volcanic activity. Was this activity the sign of new attempts of the magma to rise again or the sign of settling of a disturbed system? The volcano rather than the scientists answered the question. The advances made at Soufriere Hills suggest that the challenges are even more formidable than we had thought. For Soufriere Hills one major advance has been to develop a strong case that the main parameters that are monitored (seismicity, ground deformation and gas fluxes) are governed largely by shallow-level processes in the conduit. The signals seem largely to be telling us about the processes of magma ascent, degassing and rheological stiffening with consequent pressure fluctuations in the upper parts of the conduit (Voight et al. 1999; Denlinger & Hoblitt 1999; Wylie et al. 1999; Melnik & Sparks 1999, 2002a). While this may be a significant advance, the basic driving forces for the eruption are much deeper. It is disconcerting to be faced with the notion that monitoring data may contain little information on what is going on in the magma chamber and deeper crust. We have little to go on with regard to deeper processes. Petrology and geochemistry can provide important constraints and help unravel complex processes, and they have provided important inputs to the long-term assessments of the volcano during the Soufriere Hills crisis. However, such studies are too long and complex to make substantial contributions to real-time monitoring, forecasting, and on-the-spot judgements. Seismic tomography is both expensive and timeconsuming and its resolution insufficient to indicate any more than that there are anomalous volumes of melted rock beneath volcanoes, a fact that most petrologists would no doubt argue is selfevident! There are therefore formidable challenges in development of instruments, techniques and understanding that will allow us to see through the dominant shallow processes to decipher the magma chamber and source processes. Advances in modelling work are improving understanding, but also emphasizing just how far we have still to go in prediction and forecasting. Published models of volcanic flows are predominantly steady-state and deterministic. Only recently have models been developed that start to explore strong non-linear behaviour and unsteady flows. The implications of these new models are not entirely encouraging. Models that explore gas loss and transitions between extrusive and explosive activity show that systems can be extremely sensitive to very small changes in conditions, and can show multiple steady solutions for exactly the same conditions (Jaupart & Allegre 1991; Woods & Koyaguchi 1994; Woods 1995; Melnik & Sparks 2002a). Thus some forms of volcanism may prove at a certain level of detail inherently unpredictable. Switches can occur between one eruptive state and another with little warning.
ERUPTION OF SOUFRIERE HILLS VOLCANO: OVERVIEW and the issue of whether some changes in eruptive state can be forecast or predicted may become a major focus in research. Unsteadiness is a major feature of much volcanic activity and models will need to develop an understanding of unsteadiness. One example of such a study concerns the explosive activity on Montserrat (Melnik & Sparks 2002b). It seems probable that volcanic systems, as anticipated by Shaw (1988), can exhibit chaotic behaviour. Inspection of the mathematical description of many of the processes that control volcanic flows shows that they are often highly non-linear with mathematical structures that make chaotic behaviour likely. This topic of mathematical modelling applied to volcanology has yet to be developed in earnest. Thus advances in modelling concepts are casting some doubt on determinacy as a valid concept. Chaos, however, does not necessarily mean that accurate forecasting is an unattainable objective. Hurricanes are part of a chaotic system (the atmosphere), but the combination of theoretical advances and observations on many examples allow hurricanes to be tracked, courses anticipated in probabilistic terms, and energetics and dynamics understood. The word 'prediction' may become outmoded in the new century, with the concepts of forecasting and anticipation becoming more appropriate to the nature of volcanoes. Another issue concerns the funding of volcano science. To a considerable extent funding in volcanology revolves around crises. Rather large amounts of money become available once a crisis starts, particularly if a wealthy developed country is involved in some way. A conservative estimate of the cost of the scientific work so far on the Soufriere Hills Volcano is US$ 9 million, far more than the UK has spent on academic volcanological research over the past decade. Funding then declines if the activity declines and there is a struggle to support even the most basic monitoring capability on dormant but dangerous volcanoes. There are still remarkably few volcanoes with good baseline monitoring and detailed stratigraphic studies combined with precision geochronology and quantitative hazards investigations. With the exception of funding quanta acquired during a crisis, funding levels make volcanology a poor relation of other areas of the science. It is not quite clear why this should be. Volcanoes excite the public's interest just as much as black holes and exploration of the planets and arguably more than particle physics. Yet the Earth's volcanoes largely remain poorly monitored and few of the 40 to 50 eruptions per year are documented in much detail. We seem unable to make our presence felt as a big science. However, with an expanding world population and some megacities adjacent to dangerous volcanoes it is likely that this coming century will have eruptions that will seriously threaten large numbers of people. Montserrat is a microcosm of the problems that will arise when volcanoes like Vesuvius and Popocatepetl have major eruptions. The authors acknowledge the work of our many colleagues at the Montserrat Volcano Observatory, some of which is summarized in this overview. R.S.J.S. acknowledges a NERC Research Professorship. Thanks to G. Norton, R. Stebbles, P. Cole and M. Murphy for help with some of the figures. P. Kokelaar and R. Hoblitt provided helpful reviews. References ADILBIROV, M. & DINGWELL, D. B. 1996. Magma fragmentation by rapid decompression. Nature, 380, 146-148. AMBEH, W. B., LYNCH, L. L., CHEN, J. & ROBERTSON, R. E. A. 1998. Contemporary seismicity of Montserrat, West Indies (abstract). In: An, W., PAUL, A. & YOUNG ON, V. (eds) Transactions of the 3rd Geological Conference of the Geological Society of Trinidad and Tobago and the 14th Caribbean Geological Conference, Vol. 1. ANDERSON, T. 1908. Report on the eruptions of the Soufriere in St. Vincent in 1902, and on a visit to Montagne Pelee in Martinique. The changes in the district and the subsequent history of the volcanoes. Philosophical Transactions of the Royal Society, Series A 208(ii), 275-332. ANDERSON, T. & FLETT, J. S. 1903. Report on the eruptions of the Soufriere St Vincent, and a visit to Montagne Pelee in Martinique. Part 1. Philosophical Transactions of the Royal Society, Series A 200, 353-553.
67
ASPINALL, W. P., MILLER, A. D., LYNCH, L. L. ETAL. 1998. Soufriere Hills eruption, Montserrat 1995-1997: volcanic earthquake locations and fault plane solutions. Geophysical Research Letters, 25, 3397-3400. ASPINALL, W. P., LOUGHLIN, S. C., MICHAEL, F. V. ET AL. 2002. The Montserrat Volcano Observatory: its evolution, organization, role and activities. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 71-91. BARCLAY, J., CARROLL, M. R., RUTHERFORD, M. J., MURPHY, M. D., DEVINE, J. D., GARDNER, J. & SPARKS, R. S. J. 1998. Experimental phase equilibria: constraints on pre-eruptive storage conditions of the Soufriere Hills magma. Geophysical Research Letters, 25, 3437-3440. BAXTER, P. J., BONADONNA, C., DUPREE, R. ET AL. 1999. Cristobalite in volcanic ash of the Soufriere Hills Volcano, Montserrat: Hazards implications. Science, 283, 1142-1145. BONADONNA, C., MAYBERRY, G. C., CALDER, E. S. ET AL. 2002. Tephra fallout in the 1995-1998 eruption of Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 483-516. CALDER, E. S., COLE, P. D., DADE, W. B. ET AL. 1999. Mobility of pyroclastic flows and surges at the Soufriere Hills Volcano, Montserrat. Geophysical Research Letters, 26, 537-540. CALDER, E. S., LUCKETT, R., SPARKS, R. S. J. & VOIGHT, B. 2002. Mechanisms of lava dome instability and generation of rockfalls and pyroclastic flows at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 713-190. CHOUET, B. A. 1996. Long period volcano seismicity: its source and use in eruption forecasting. Nature, 380, 309-316. CHRISTIANSEN, R. L. & PETERSON, R. L. 1981. The 1980 Mount St. Helens eruption: Chronology of the 1980 eruptive activity. In: LIPMAN, P. W. & MULLINEAUX, D. R. (eds) The 1980 Eruptions of Mount St. Helens, Washington. US Geological Survey, Professional Papers, 1250, 17-30. CLARKE, A. B., NERI, A., VOIGHT, B., MACEDONIO, G. & DRUITT, T. H. 2002. Computational modelling of the transient dynamics of August 1997 Vulcanian explosions at Soufriere Hills Volcano, Montserrat: influence of initial conduit conditions on near-vent pyroclastic dispersal. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 319-348. CLYNNE, M. A. 1999. A complex magma mixing origin for rocks erupted in 1915, Lassen Peak, California. Journal of Petrology, 40, 105-132. COLE, P. C., CALDER, E. S., DRUITT, T. H., HOBLITT, R., ROBERTSON, R. E. A., SPARKS, R. S. J. & YOUNG, S. R. 1998. Pyroclastic flows generated by the eruption of the Soufriere Hills Volcano, Montserrat. Geophysical Research Letters, 25, 3425-3428. COLE, P. D., CALDER, E. S., SPARKS, R. S. J. ETAL. 2002. Deposits from dome-collapse and fountain-collapse pyroclastic flows at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 231-262. CORTES, G. P. & RAIGOSA, A. 1997. A synthesis of the recent activity of Galeras volcano, Colombia:seven years of continuous surveillance 1989-1995. Journal of Volcanology and Geothermal Research, 77, 101-114. COUCH, S., SPARKS, R. S. J. & CARROLL, M. R. 2001. Mineral disequilibria in lavas explained by convective self-mixing in open magma chambers. Nature, 411, 1037-1039. DADE, W. B. & HUPPERT, H. E. 1999. Long run-out rockfalls. Geology, 26, 803-806. DENLINGER, R. P. & HOBLITT, R. P. 1999. Cyclic behavior of silicic volcanoes. Geology, 27, 459-462. DEVINE, J. D., MURPHY, M. D., RUTHERFORD, M. R. ETAL. 1998. Petrologic evidence for pressure-temperature conditions and recent reheating of andesite magma at Soufriere Hills Volcano, Montserrat, West Indies. Geophysical Research Letters, 25, 3669-3672. DRAKE, T. G. 1990. Structural feature in granular flows. Journal of Geophysical Research, 95, 8681-8696. DRUITT, T. H. 1998. Pyroclastic density currents. In: GILBERT, J. S. & SPARKS, R. S. J. (eds) The Physics of Explosive Volcanic Eruptions. Geological Society, London, Special Publication, 145, 145-182.
68
R. S. J. SPARKS & S. R. YOUNG
DRUITT, T. H., CALDER, E. S., COLE, P. D. ET AL. 20020. Small-volume, highly mobile pyroclastic flows formed by rapid sedimentation from pyroclastic surges at Soufriere Hills Volcano, Montserrat: an important volcanic hazard. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 263-279. DRUITT, T. H., YOUNG, S. R., BAPTIE, B. ET AL. 20026. Episodes of cyclic Vulcanian explosive activity with fountain collapse at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 281-306. EDMONDS, M., PYLE, D. M. & OPPENHEIMER, C. 2001. A model for degassing at the Soufriere Hills Volcano, Montserrat, West Indies, based on geochemical data. Earth and Planetary Science Letters, 186, 159-173. ElCHELBERGER, J. C., CARRIGAN, C. R., WESTRICH, H. R. & PRICE, R. H. 1986. Non-explosive silicic volcanism. Nature, 323, 598-602.
FINK, J. H. & GRIFFITHS, R. W. 1990. Radial spreading of viscous gravity currents with solidifying crust. Journal of Fluid Mechanics, 221, 485-510. FINK, J. H. & KIEFFER, S. W. 1993. Estimates of pyroclastic flow velocities from explosive decompression of lava domes. Nature, 363, 612-615. FOLCH, A. & MARTI, J. 1998. The generation of overpressure in felsic magma chambers by replenishment. Earth and Planetary Science Letters, 163, 301-314. GIL CRUZ, F. & CHOUET, B. A. 1997. Long period events, the most characteristic seismicity accompanying the emplacement and extrusion of a lava dome in Galeras volcano, Colombia in 1991. Journal of Volcanology and Geothermal Research, 77, 121-158. GRIFFITHS, R. W. 2000. The dynamics of lava flows. Annual Review of Fluid Mechanics, 32, 477-518. GRUNEWALD, U., SPARKS, R. S. J., KEARNS, S. & KOMOROWSKI, J. C. 2000. Friction marks on blocks from pyroclastic flows at the Soufriere Hills Volcano, Montserrat: implications for flow mechanisms. Geology, 28, 827-830. HARFORD, C. L. & SPARKS, R. S. J. 2001. Recent remobilisation of shallowlevel igneous rock on Montserrat revealed by hydrogen isotope composition of amphiboles. Earth and Planetary Science Letters, 185, 285-297. HARFORD, C., PRINGLE, M. S., SPARKS, R. S. J. & YOUNG, S. R. 2002. The volcanic evolution of Montserrat using 40 Ar/ 39 Ar geochronology. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 93-113. HATTORI, K. 1993. High-sulfur magma. A product of fluid discharge from underlying mafic magma: evidence from Mount Pinatubo, Phillipines. Geology, 21, 1083-1086. HAYASHI, J. N. & SELF, S. 1992. A comparison of pyroclastic flow and debris avalanche mobility. Journal of Geophysical Research, 97, 9063-9071. HEIKEN, G., & EICHELBERGER, J. C. 1980. Eruptions at Chaos Crags, Lassen Volcanic National Park, California. Journal of Volcanology and Geothermal Research, 7, 443-481. HILDRETH, W. 1981. Gradients in silicic magma chambers: implications for lithospheric magmatism. Journal of Geophysical Research, 86, 10153-10192. HOBLITT, R. P., MILLER, C. D. & VALLANCE, J. W. 1981. Origin and stratigraphy of the deposit produced by the May 18 directed blast. In: LIPMAN, P. W. & MULLINEAUX, D. R. (eds) The 1980 Eruptions of Mount St Helens, Washington. US Geological Survey, Professional Papers, 1250, 401-419. HORWELL, C. J., BRANA, L. P., SPARKS, R. S. J., MURPHY, M. D. & HARDS, V. 2001. A geochemical investigation of fragmentation and physical fractionation in pyroclastic flows from the Soufriere Hills Volcano. Montserrat. Journal of Volcanology and Geothermal Research, 109, 247-262. HUPPERT, H. E. & SPARKS, R. S. J. 1988. The generation of granite by intrusion of basalt into the continental crust. Journal of Petrology, 29, 599-624. JACKSON, P., SHEPHERD, J. S., ROBERTSON, R. E. A. & SKERRIT, G. 1998. Ground deformation studies at the Soufriere Hills Volcano, Montserrat, 1 Electronic distance meter studies. Geophysical Research Letters, 25, 3409-3413. JAUPART, C. 1998. Gas loss from magmas through conduit walls during eruption. In: GILBERT, J. S. & SPARKS, R. S. J. (eds) The Physics of Explosive Volcanic Eruptions. Geological Society, London, Special Publication, 145, 73-90.
JAUPART, C. & ALLEGRE, C. 1991. Gas content, eruption rate and instabilities of eruption in silicic volcanoes. Earth and Planetary Science Letters, 102, 413-429. KOKELAAR, B. P. 2002. Setting, chronology and consequences of the eruption of Soufriere Hills Volcano, Montserrat (1995-1999). In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society. London. Memoirs, 21, 1-43. LACROIX, A. 1904. La Montague Pelee and ses eruptions. Masson, Paris. LOUGHLIN, S. C., CALDER, E. S., CLARKE. A. B. ET AL. 2002. Pyroclastic flows and surges generated by the 25 June 1997 dome collapse, Soufriere Hills Volcano, Montserrat. In: DRUITT. T. H. & KOKELAAR. B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society. London. Memoirs. 21, 191-209! LUCKETT, R., BAPTIE, B. & NEUBERG. J. 2002. The relationship between degassing and rockfall signals at Soufriere Hills Volcano. Montserrat. In: DRUITT, T. H. & KOKELAAR. B. P. (eds) The eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 595-602. MACGREGOR, A. G. 1938. The Royal Society expedition to MONTSERRAT. B.W.I. The volcanic history and petrology of Montserrat with observations on Mt Pele in Martinique. Philosophical Transactions of the Royal Society, B229, 1-90. MATTHEWS, S. J., JONES, A. P. & BRISTOW. C. S. 1992. A simple magmamixing model for sulfur behaviour in calcalkaline volcanic rocks: mineralogical evidence from Mount Pinatubo 1991 eruption. Journal of the Geological Society of London, 149, 863-866. MATTHEWS. S. J, GARDEWEG. M. C. & SPARKS. R. S. J. 1997. The 1984 to 1996 cyclic activity of Lascar Volcano. Northern Chile: cycles of dome growth, dome subsidence, degassing and explosive eruption. Bulletin of Volcanology, 59, 72-82. MATTHEWS, S. J., SPARKS. R. S. J. & GARDEWEG. M. C. 1999. The Piedras Grandes-Soncor eruptions. Lascar Volcano. Chile; Evolution of a zoned magma chamber in the Central Andean upper crust. Journal of Petrology, 40. 1891-1919. MATTIOLI, G. S., DIXON, T. H., FARINA. F.. HOWELL. E. S., JANSMA, P. E. & SMITH, A. L. 1998. GPS meausrements of surface deformation around Soufriere Hills Volcano. Montserrat from October 1995 to July 1996. Geophysical Research Letters, 25. 3417-3420. MATTIOLI, G. S., RODRIGUEZ, L. SMITH. A. L.. JANSMA. P. E. & HERD. R. A. 2000. Changes in magmatic flux recorded by campaign and continuous GPS geodesy of Soufriere Hills. Montserrat. BWI (abstract). Geosciences 2000. Geological Society of London. MELNIK, O. & SPARKS, R. S. J. 1999. Nonlinear dynamics of lava dome extrusion. Nature, 402, 37-41. MELNIK, O. & SPARKS, R. S. J. 2002#. Dynamics of magma ascent and lava extrusion at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR. B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London. Memoirs, 21, 153-171. MELNIK, O. & SPARKS, R. S. J. 20026. Modelling of conduit flow dynamics during explosive activity at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society. London, Memoirs, 21. 307-317. MILLER, A. D., STEWART, R. C., WHITE, R. A. ET AL. 1998. Seismicity associated with dome growth and collapse at the Soufriere Hills Volcano, Montserrat. Geophysical Research Letters, 25. 3401-3405. MILLER, T. P., CHERTKOFF, D. G., EICHELBERGER. J. C. & COOMBS. M. L. 1999. Mount Dutton Volcano, Alsaka: Aleutian Arc analogue to Unzen Volcano, Japan. Journal of Volcanology and Geothermal Research, 89. 255-302. MOGI, K. 1958. Relations between the eruptions of various volcanoes and the deformations of the ground surfaces around them. Bulletin of Earthquake Research Institute, 36, 99-134. MURPHY, M. D., SPARKS, R. S. J., BARCLAY, J. ET AL. 1998. The role of magma mixing in triggering the current eruption of the Soufriere Hills Volcano. Montserrat. Geophysical Research Letters, 25. 3433-3436. MURPHY, M. D., SPARKS, R. S. J., BARCLAY. J.. CARROLL, M. R. & BREWER, T. S. 2000. Remobilisation of andesite magma by intrusion of mafic magma at the Soufriere Hills Volcano. Montserrat. West Indies. Journal of Petrology, 41. 21-42.
ERUPTION OF SOUFRIERE HILLS VOLCANO: OVERVIEW NAKADA, S. & MOTOMURA, Y. 1999. Petrology of the 1991-1995 eruption at Unzen: effusion pulsation and groundmass crystallization. Journal of Volcanology and Geothermal Research, 89, 173-196. NAKADA, S., SHIMIZU, H. & OHTA, K. 1999. Overview of the 1990-1995 eruption at Unzen Volcano. Journal of Volcanology and Geothermal Research, 89, 1-22. NEUBERG, J. 2000. Characteristics and causes of shallow seismicity in andesite volcanoes. Philosophical Transactions of the Royal Society Series A 358, 1533-1546. NEUBERG, J., BAPTIE, B., LUCKETT, R. & STEWART, R. C. 1998. Results from the broadband seismic network on Montserrat. Geophysical Research Letters, 25, 3661-3665. NEWHALL, C. G. & MELSON, W. G. 1983. Explosive activity associated with the growth of volcanic domes. Journal of Volcanology and Geothermal Research, 17, 111-131. NORTON, G. E., WATTS, R. B., VOIGT, B ETAL. 2002. Pyroclastic flow and explosive activity at Soufriere Hills Volcano, Montserrat, during a period of virtually no magma extrusion (March 1998 to November 1999). In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 467-481. PERRET, F. A. 1939. The volcano-seismic crisis at Montserrat 1933-37. Publications of the Carnegie Institute, Washington, 512, 1-76. POWELL, C. F. 1938. The Royal Society expedition to Montserrat, B.W.I. Final Report. Philosophical Transactions of the Royal Society, Series A 237, 1-34. REA, W. J. 1974. The volcanic geology of Montserrat. Journal of the Geological Society, 130, 341-366. RITCHIE, L. J., COLE, P. D, & SPARKS, R. S. J. 2002. Sedimentology of deposits from the pyroclastic density current of 26 December 1997 at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 435-456. ROBERTSON, R., COLE, P., SPARKS, R. S. J. ET AL. 1998. The explosive eruption of Soufriere Hills Volcano, Montserrat, West Indies, 17 September, 1996. Geophysical Research Letters, 25, 3429-3432. ROBERTSON, R. E. A., ASPINALL, W. P., HERD, R. A., NORTON, G. E., SPARKS, R. S. J. & YOUNG, S. R. 2000. The 1995-98 eruption of the Soufriere Hills volcano, Montserrat. Philosophical Transactions of the Royal Society, 358, 1619-1637. ROOBOL, M. J. & SMITH, A. L. 1998. Pyroclastic stratigraphy of the Soufriere Hills Volcano, Montserrat: implications for the present eruption. Geophysical Research Letters, 25, 3393-3396. ROSE, W. I. 1973. Pattern and mechanism of volcanic activity in Santiaguito volcanic dome. Bulletin Volcanogique, 37, 73-94. SATO, H., FUJII, T. & NAKADA, S. 1992. Crumbling dacite dome lava and generation of pyroclastic flows at Unzen Volcano. Nature, 360, 664-666. SHAW, H. R. 1988. Mathematical attractor theory and plutonic-volcanic episodicity. In: Cni-Yu KING & SCARPA, R. (eds) Modelling of Volcanic Processes. Vieweg, Berlin, 161-206. SHEPHERD, J. B., TOMBLIN, J. F. & Woo, D. 1971. Volcano-seismic crisis in Montserrat, West Indies. Bulletin Volcanologique, 35, 143-163. SHEPHERD, J. B., ASPINALL, W. P., ROWLEY, K. C., PEREIRA, J., SIGURDSSON, H., FISKE, R. S. & TOMBLIN, J. F. 1979. The eruption of the Soufriere Volcano, St. Vincent, April-June 1979. Nature, 282, 24-28. SHEPHERD, J. B., HERD, R. A., JACKSON, P. & WATTS, R. B. 1998. Ground deformation studies at the Soufriere Hills Volcano, Montserrat, 2, Rapid static GPS measurements, June 1996 to June 1997. Geophysical Research Letters, 25, 3413-3416. SPARKS, R. S. J. 1976. Grain size variations in ignimbrites and implications for the transport of pyroclastic flows. Sedimentology, 23, 147-188. SPARKS, R. S. J. 1997. Causes and consequences of pressurization in lava dome eruptions. Earth and Planetary Science Letters, 150, 177-189. SPARKS, R. S. J., SIGURDSSON, H. & WILSON, L. 1977. Magma mixing: mechanism of triggering explosive acid eruptions. Nature, 267, 315-318. SPARKS, R. S. J., YOUNG, S. R., BARCLAY, J. ET AL. 1998. Magma production and growth of the lava dome of the Soufriere Hills volcano, Montserrat: November 1995 to December 1997. Geophysical Research Letters, 25, 3421-3424.
69
SPARKS, R. S. J., MURPHY, M. D., LEJEUNE, A. M., WATTS, R. B., BARCLAY, J. & YOUNG, S. R. 2000. Control on the emplacement of the andesite lava dome of the Soufriere Hills volcano, Montserrat, by degassing-induced crystallization. Terra Nova, 12, 14-20. SPARKS, R. S. J., BARCLAY, J., CALDER, E. S. ETAL. (2002) Generation of a debris avalanche and violent pyroclastic density current on 26 December (Boxing Day) 1997 at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 409-434. STIX, J., TORRES, C. R., NARVAEZ, M. L., CORTES, J. G. P., RAIGOSA, J. A., GOMEZ, M. D. & CASTONGUEY, R. 1997. A model of vulcanian eruptions at Galeras volcano, Columbia. Journal of Volcanology and Geothermal Research, 77, 285-304. TAYLOR, G. A. M. 1958. The 1951 eruption of Mount Lamington, Papua. Bureau of Mineral Resources of Australia: Geology & Geophysical Bulletin, 38, 1-117. VOIGHT, B., HOBLITT, R. P., CLARKE, A. B., LOCKHART, A. B., MlLLER, A. D.,
LYNCH, L. L. & MCMAHON, J. 1998. Remarkable cyclic ground deformation monitored in real time on Montserrat and its use in eruption forecasting. Geophysical Research Letters, 25, 3405-3408. VOIGHT, B., SPARKS, R. S. J., MILLER, A. D. ET AL. 1999. Magma flow instability and cyclic activity at Soufriere Hills Voclano, Montserrat, British West Indies. Science, 283, 1138-1142. WADGE, G. & ISAACS, M. C. 1988. Mapping the volcanic hazards from Soufriere Hills Volcano, Montserrat, West Indies using an image processor. Journal of the Geological Society, 145, 541-551. WALLACE, P. J. 2001. Volcanic SO2 emissions and the abundance and distribution of exsolved gas in magma bodies. Journal of Volcanology and Geothermal Research, 108, 85-106. WALLACE, P. J. & GERLACH, T. M. 1994. Magmatic vapour source for sulfur dioxide released during volcanic eruptions: evidence from Mount Pinatubo. Science, 265, 497-499. WATTS, R. B., HERD, R. A., SPARKS, R. S. J. & YOUNG, S. R. 2002. Growth patterns and emplacement of the andesite lava dome at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 115-152. WHITE, R. A., MILLER, A. D., LYNCH, L. & POWER, J. 1998. Observations of seismic hybrid events at Soufriere Hills Volcano, Montserrat: July 1995 to September 1996. Geophysical Research Letters, 25, 3657-3660. WOLFE, E. W. & HOBLITT, R. P. 1996. Overview of the eruptions. In: NEWHALL, C. G. & PUNONGBAYAN, R. S. (eds) Fire and Mud: Eruptions and lahars of Mount Pinatubo, Philippines. University of Washington Press, 3-20. WOODS, A. W. 1995. The dynamics of explosive volcanic eruptions. Reviews of Geophysics, 33, 495-530. WOODS, A. W. & KOYAGUCHI, T. 1994. Transitions between explosive and effusive volcanic eruptions. Nature, 370, 641-644. WOODS, A. W., SPARKS, R. S. J., RITCHIE, L. J., BATEY, J., GLADSTONE, C. & BURSIK, M. I. 2002. The explosive decompression of a pressurized volcanic dome: the 26 December 1997 collapse and explosion of Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoir, 21, 457-466. WYLIE, J. J., VOIGHT, B. & WHITEHEAD, J. A. 1999. Instability of magma flow from volatile-dependent viscosity. Science, 285, 1883-1885. YOUNG, S. R., SPARKS, R. S. J., ROBERTSON, R. E. A., LYNCH, L. L. & ASPINALL, W. P. 1997. Eruption of Soufriere Hills in Montserrat continues. Transactions of the American Geophysical Union, 78, 401-405. YOUNG, S. R., FRANCIS, P. W., BARCLAY, J. ETAL. 19980. Monitoring SO2 emission at the Soufriere Hills volcano: implications for changes in eruptive conditions. Geophysical Research Letters, 25, 3681-3684. YOUNG, S. R., SPARKS, R. S. J., ROBERTSON, R., LYNCH, L., MILLER, A. D., SHEPHERD, J. B. & ASPINALL, W. A. 19986. Overview of the Soufriere Hills Volcano and the eruption. Geophysical Research Letters, 25, 3389-3392.
The Montserrat Volcano Observatory: its evolution, organization, role and activities W. P. ASPINALL1 2, S. C. LOUGHLIN1'3, F. V. MICHAEL4, A. D. MILLER1'5, G. E. NORTON1 6, K. C. ROWLEY1'7, R. S. J. SPARKS 18 , S. R. YOUNG1 1 Montserrat Volcano Observatory, Mongo Hill, Montserrat, West Indies 2 Aspinall & Associates, 5 Woodside Close, Beaconsfield, Bucks, HP9 1JQ, UK (e-mail:
[email protected]) 3 British Geological Survey, Murchison House, West Mains Road, Edinburgh, EH9 3LA, UK 4 Emergency Department, Government of Montserrat, St John's, Montserrat, West Indies 5 GEOWALKS, 24 Argyle Place, Edinburgh, EH9 1JJ, UK 6 British Geological Survey, Keyworth, Nottingham, NG12 5GG, UK 1 LAN DAT A Ltd, Trinidad & Tobago 8 Department of Earth Sciences, University of Bristol, Bristol, BS8 1RJ, UK
Abstract: The Montserrat Volcano Observatory (MVO) is a statutory body of the Government of Montserrat and is the organization responsible for volcano monitoring operations on the island. It was formed shortly after the first phreatic explosions from Soufriere Hills Volcano occurred on 18 July 1995, and evolved from a hastily created, interim entity to a fully established volcano monitoring operation. Participating scientific teams have been drawn mainly from the Seismic Research Unit of the University of the West Indies, the US Geological Survey, the British Geological Survey and universities from various countries including the USA, UK, France and Puerto Rico. Despite its hurried inception, the MVO has been able to provide timely, high quality hazard advice to the civil authorities and has maintained an exceptional documentary record of all scientific aspects of the eruption. Its public education and information efforts have been extensive and there have been unusually high levels of interaction between scientists and the civil authorities, and between scientists and the public, both within Montserrat and outside in the wider world. The experience of setting up and running the MVO, under difficult and stressful conditions, has exemplified the advantages of teamwork and flexibility within monitoring operations and the benefits of openness and clarity in public interactions. Novel techniques have been applied to the appraisal of hazards and advances in scientific understanding have proved invaluable for risk assessment and management.
The first historic eruption of Soufriere Hills Volcano, on the island of Montserrat (Fig. 1), began on 18 July 1995 with phreatic explosions from within the central prehistoric collapse scar (English's Crater) of this dome-complex volcano. This activity lasted until November 1995, and was followed by an extended period of dome growth that ended abruptly in early March 1998 (Young et al. 1998&). There then followed a 20-month interval in which no new magma was extruded and there was only mild activity involving ashventing, explosions, generally low levels of seismicity, and several collapses of the dome (Norton et al. 2002). In mid-November 1999, however, a new lava dome began to grow in the big chasm that had been cut into the remnants of the 1995-1998 dome by collapses in 1998 and 1999. This new dome increased steadily in size until it collapsed almost entirely on 20 March 2000, in a sequence of flows to the east, down the Tar River valley. After this, dome growth recommenced with increased vigour and has continued into 2002. Magma production rates have largely determined eruptive style, with high rates being associated with major dome collapses and explosions. This eruption, protracted over six years, has caused tremendous disruption to the lives of all residents on the island and has necessitated the continued presence on the island of a full-time scientific team and a dedicated monitoring facility. The initial slow rate of dome growth and gradual escalation of the crisis were major factors in the development of the monitoring effort, bearing in particular on early prognostications on the course of the volcanic activity. The setting, major event chronology and most important consequences of the eruption of Soufriere Hills Volcano from 1995 to 1999 are described by Kokelaar (2002). For the past six years, the Montserrat Volcano Observatory (MVO) has been the organization responsible for continuous monitoring of the volcano. The formation of a smoothly functioning scientific team during a crisis, with staff drawn from various nationalities and organizations, is inherently difficult. For the establishment of a full-scale observatory operation, the stressful situation in Montserrat, with a complex and unsatisfactory overall management structure and funding deficiencies, combined to create some significant problems. Notwithstanding these difficulties, however, a great deal of valuable scientific work has been done by the MVO, particularly in the documentation
of the eruption. There has been rapid publication of accounts of activity and scientific results in various media (e.g. Ahmad 1996; Montserrat Volcano Observatory Team 1997; Young et al. 1997; Ambeh et al. 1998). A collection of papers focusing on the first two years of the eruption was published in a two-part special section of Geophysical Research Letters (Aspinall et al. 1998a; Young et al. 1998c), and overviews of the scientific advances appear in Robertson et al. (2000) and in Sparks & Young (2002). There have been extensive efforts in public education and the MVO has become a primary source of information for the local administrative authorities and the public, as well as a major attraction for visitors to the island. More importantly, the continued operation of a volcano-monitoring facility on Montserrat is essential to the present and future occupation of the island, and recent administrative changes to the status of the MVO have placed its managerial and financial structure on a stable basis. There are many valuable lessons to be learned from the experience of setting up and running the MVO; important insights and issues of a detailed scientific nature are treated in the companion papers in this volume and elsewhere (see references). This paper summarizes some of the unique features of the new institution and outlines essential elements in its evolution, structure and organization. However, myriad other factors and influences shaped events and outcomes in Montserrat during the eruption from 1995 onwards, and these cannot all be covered in one paper. A book by Pattullo (2000) and a comprehensive report by Clay et al. (1999) each document many other aspects of the broader picture.
Administrative setting The inception and evolution of the MVO took place against a background of complex administrative arrangements, which characterized Montserrat as a British Dependent Territory (now a United Kingdom Overseas Territory). The administration of the island involves both a Governor, who represents the Queen and the UK government, and a locally elected Government, led by a Chief
DRUITT, T. H. & KOKELAAR, B. P. (eds) 2002. The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 71-91. 0435-4052/02/S15 © The Geological Society of London 2002.
71
72
W. P. ASPINALL ETAL.
Fig. 1. Map of Montserrat showing the location of the volcano, main place names used in the text and the migration of the MVO to the north of the island during the period 1995-1999.
THE MONTSERRAT VOLCANO OBSERVATORY
Minister. There is limited self-government, with a legislature and civil service, but the UK has responsibility for defence, internal security and foreign relations. For self-governance, there is an Executive Council consisting of the Governor, the Chief Minister, the Attorney-General, the Financial Secretary and three more ministers, and an eleven-member Legislative Council, of which seven are elected by universal adult suffrage (from age 18), two are nominated and two are appointed ex-officio. The UK government department with administrative responsibility for Montserrat is the Foreign and Commonwealth Office (FCO). However, this department has virtually no financial resources with which to respond to a natural disaster and the actual financial and logistic responsibility for Montserrat in an emergency devolved to the Overseas Development Administration (ODA, now called the Department for International Development - DFID). Another complication in relation to the Foreign Office is that Montserrat affairs have been dealt with internally both by the West Indies and Atlantic Department in London and by the Dependent Territories Regional Secretariat based in Barbados. The intricate relationships between all these ministries, departments and organizations, and their respective roles in the crisis, were not always clear and, as a result, were confusing to many of the scientists responding to the emergency. When activity escalated dramatically on Montserrat in July 1995, an operational base was immediately established on the island by the Seismic Research Unit (SRU) of the University of the West Indies (UWI) and, from this foundation, the Montserrat Volcano Observatory evolved into a fully functional local organization staffed by a multinational, multi-institutional team, utilizing a variety of geophysical, geological and geochemical techniques to monitor the volcano. Funding for these activities was provided mainly by the Government of Montserrat and by the UK government. The initial absence of suitably trained local staff and the longevity of the eruption necessitated a major turnover of staff at the MVO: over 100 scientists and technical staff have worked there during the past six years. At the height of the crisis, a staff complement of up to eight UK scientists (generally appointed through the British Geological Survey), three SRU scientists and technical staff could be mustered, eventually supplemented by seven Montserratian technical and administrative staff. In addition, staff and scientists from the US Geological Survey (USGS) and from universities in the UK, the USA, France and Puerto Rico have been involved in MVO operations from time to time. Generous support has been received also from the Institut de Physique du Globe de Paris, through staff from its Volcano Observatories in Guadeloupe and Martinique. From the resident Montserrat population itself, many volunteers have provided invaluable assistance with observations, routine support work and night-duty shifts at the Observatory. Evolution of the Montserrat Volcano Observatory The genesis of a permanent volcano observatory on Montserrat can be traced back to the periods of increased seismic activity that began in Montserrat in the early 1990s, first recognized by the SRU from their regional monitoring activities. This activity prompted SRU to strengthen the seismic network on the island. Within the eastern Caribbean, SRU is the regional agency for earthquake and volcano surveillance and this group monitored activity on Montserrat for several decades with permanently installed seismographs and regular field visits. Indeed, the history of the Unit's active involvement in Montserrat goes back to, and beyond, the last significant volcanoseismic crisis there, in 1966 (Shepherd et al. 1971). Their experience of handling volcanic crises in the region includes the eruptions of the Soufriere of St Vincent, in 1971-1972 (Aspinall et al. 1972) and 1979 (Shepherd et al. 1979), as well as numerous other volcanic-seismic swarms and related studies. Between January 1992 and July 1995, the seismic stations on Montserrat and the surrounding islands recorded more than 15 episodic swarms of small earthquakes. These originated at depths
73
of 10-15 km and had epicentres defining a NE trend off the east coast of Montserrat (Ambeh et al. 1998). As the swarms recurred, the SRU upgraded and augmented its seismic network on Montserrat, made field measurements of dry tilt, and observed the volcano to watch for changes in fumarolic activity. They were assisted by scientists from the Guadeloupe Volcano Observatory, who undertook annual sampling visits to the main fumaroles and thermal spring areas for physicochemical monitoring (Hammouya et al. 1998). These latter efforts, however, did not reveal any significant changes in fumarole activity or chemistry indicative of an impending eruption.
Establishing monitoring operations on Montserrat Immediately following the first phreatic explosions on 18 July 1995, the SRU established an operational base on Montserrat to provide scientific advice on the state of the volcano directly to local authorities. Acting on the advice of SRU, the Government of Montserrat then invited scientists from the USGS Volcano Disaster Assistance Program (VDAP) and the Guadeloupe Volcano Observatory to join the SRU team on-island to assist with their monitoring activities. The SRU and USGS scientists, together with one UK scientist and two student volunteers recruited in early 1995 from the local secondary school, formed the embryonic observatory team. A temporary facility was established near the government headquarters in Plymouth in July 1995 and was first called the Soufriere Hills Volcano Observatory. The physical location of the Observatory moved (Fig. 1) from Plymouth to the Vue Pointe Hotel, on the northern shore of Old Road Bay in August 1995, and then to a rented villa in Old Towne in October 1995 (Fig. 2), at which time the name changed to the Montserrat Volcano Observatory. The USGS VDAP team spent six weeks on Montserrat, from late July 1995, and reinforced the pre-existing seismic network with the establishment of additional short-period seismograph stations. In addition, three electronic tiltmeters were installed and a programme of SO2 monitoring was started, using a correlation spectrometer (COSPEC). An automated seismic data acquisition and processing system was also installed, which eventually replaced the pre-existing Soufriere system software that had been the basis of the data acquisition for the SRU network. At these early stages of the crisis (i.e. from August to November 1995), the Foreign and Commonwealth Office commissioned short visits to Montserrat by individual UK consultants with experience of volcanic activity in the Caribbean. These scientists functioned essentially as advisors to the Governor and to the UK government, but also assisted the MVO with routine monitoring work. In October 1995, the first of a series of short-term contracts from the ODA supported the direct involvement of staff from the British Geological Survey (BGS), to work with the incumbent monitoring team. Later fixed-term contracts with the BGS supported staff from that organization, UK university scientists and students, and other specialists from time to time (the number of UK personnel at the MVO increased from two in January 1996 to seven in July 1996). The local complement of staff at the observatory also increased in October 1995 with the secondment of several Montserratian civil servants from other government departments. The day-to-day running of the MVO was managed by a Chief Scientist, who was responsible for co-ordinating the scientific work and for reporting to the Government of Montserrat and to the Governor. During the first year of its operation, the Head of SRU fulfilled this role, with various other members of SRU acting on his behalf when he was off-island. By this time, the need was being recognized for a sizeable scientific, technical and administrative staff to monitor the volcano 24 hours a day. In an extended undertaking, personnel were drawn from various institutions: most came from the SRU and from the UK. UK scientists were assembled from the British Geological Survey and from the universities with strong research expertise in volcanology (principally, the universities of Bristol, Lancaster, the Open University and Cambridge). The USGS, the Institut de Physique du Globe de Paris, the
74
W. P. ASPINALL ETAL.
Fig. 2. Team members observing a small Vulcanian eruption cloud on 3 July 1997, from the pool-deck of Eiffel House, Old Towne, the location of MVO (South) from October 1995 to September 1997 (photo t W.P. Aspinall).
Universite Blaise Pascal, Penn State University, Brown University and the University of Puerto Rico provided significant assistance. However, it was initially far from clear who was responsible for what amongst all these different institutions and individuals. At times, relationships became very strained over priorities for scientific work and responsibilities for giving advice.
Early challenges and difficulties In part at least, the initial tensions over scientific obligations and priorities can be traced to the very difficult question of what level of response was appropriate at the time, given the uncertainty surrounding the future course of the crisis. Only two decades earlier, in 1976, a volcanic crisis on the neighbouring island of Guadeloupe, with many volcanological similarities to conditions in Montserrat in 1995, had failed to develop into a significant, lifethreatening eruption after a major evacuation of the population had been ordered. The circumstances surrounding the decision to instigate the evacuation became the focus of intense debate among volcanologists (see Tazieff 1980 (and preceding contributions in the same journal referenced therein); Fiske 1984), and the management of the emergency was the subject of a wide-ranging international commission of inquiry, sponsored by the French government. One major factor that was considered in analyses of the Guadeloupe episode was the enormous modern economic cost of an unfulfilled
warning of a dangerous eruption, notwithstanding the haunting political spectre of an earlier Caribbean volcanic catastrophe, the 1902 eruption of Mount Pelee in Martinique that had claimed nearly 30000 lives. Against this backdrop, it is understandable that in 1995 some in the eastern Caribbean, particularly in Montserrat itself, were reluctant to initiate urgent precautionary measures until the extent of the threat could be better ascertained. They would have been very conscious of the potential for a devastating impact on the economy of their small island, which had only just been reestablished after the ravages of Hurricane Hugo, in 1989. On the other hand, the USGS VDAP scientists, some of whom had witnessed the catastrophic consequences of the eruptions of Mount St. Helens and Pinatubo, took a more pessimistic view of the outlook and advised accordingly. In these circumstances, the UK government had to be mindful of its ultimate responsibility for the welfare of the people of Montserrat. As the higher political directorate, and one steeped in recent public disasters and health scares, UK politicians and civil servants were probably more 'risk averse' than their counterparts in the Montserrat government. They were certainly in the position to assert their judgement and authority more forcefully. This dichotomy in political attitudes generated dissonances within the scientific group. Contention developed as differences emerged as to the perceived level of risk and, with it, other difficulties arose, relating to attitude, personality and management styles. These were exacerbated in the first weeks of the crisis by the disparities in funding and resourcing that were initially available to. or deployed by, the various groups making up the team. It was not conducive to team spirit and collaboration that the regional scientists were obliged to make use of separate, cheaper accommodation, away from others, and had to endure restrictive transport arrangements and so on; this provoked a 'them-and-us' atmosphere and reduced opportunities for the informal scientific dialogue (and banter) that is so invaluable in crisis circumstances. A simple instance of a contentious issue was the accumulating cost of frequent and extended international phone calls from the nascent observatory facility. This was taken for granted as a necessity by many of the visiting scientists, but represented a major item of expense for the host government. It is unfortunate that, at this early stage, it was not possible to forge a unified approach to these issues within the whole MVO team. A large measure of concurrence existed, and the UK involvement by now included several former staff members of SRU, all familiar with the constraints faced by that organization. Thus, while the regional team from SRU were initially disinclined to widen the scope of monitoring (and scientific advice) much beyond that which could be sustained by or through the local government, some other scientists were less restrained about pressing forward with recommendations for strengthening the scientific efforts. This, in effect, implied seeking direct financial support from the UK government. As things stood then, it is perhaps not surprising that differing perceptions of the balance to be struck between economic imperative and acceptable risk overlapped into another area where substantial difficulties always arise in a volcanic crisis: the articulation of scientific uncertainty in the forecasting of future activity. Where the public is exposed to a given hazard, natural or otherwise, the lack of sufficient knowledge to quantify the attendant risks with useful accuracy impinges upon a government's ability to comprehend, and then justify to itself and the public, the extent to which steps should be taken to eliminate or reduce those risks. In the first few weeks of the Montserrat crisis there was perhaps, at times, some unwarranted scientific dogmatism about what might or might not happen at the volcano, especially in terms of the eruption turning magmatic and explosive. The confounding effect of these diverging, categorical scientific stances was then compounded for a short while by an overall diminution in communication between scientists and the civil authorities. The result was a dip in confidence in the MVO team and. with it, some loss of public credibility; this was not fully restored until later, when a consensual approach was achieved.
THE MONTSERRAT VOLCANO OBSERVATORY
Widening the scientific involvement Thus a situation developed, as the crisis started to drag on and deepen slowly, in which there was increasing pressure from the UK government in particular for the wider involvement in their deliberations of more scientists and specialists. This reflected both the natural instinct of politicians to canvass as many opinions as possible, and the anticipated strictures of impending new UK government internal policy guidelines. The latter enjoined politicians and civil servants to organize and engage wide-ranging scientific consultation in any science-related decision-making process, the relevant precepts emerging from high-profile health risk crises, such as BSE (as formulated in documents such as Office of Science and Technology 1997, 1998, for example; see also The Royal Society 1999; Curnow 1999). This pressure, to extend the spectrum of scientific feedback in respect of Montserrat, and the need to develop a team capable of functioning 24 hours per day, resulted in an increasing presence of UK personnel at the MVO and this, in turn, put additional strain on SRU's efforts to manage all the scientific elements of the crisis. In short, SRU was under-resourced for sustaining what was turning out to be a protracted and growing full-scale crisis management role on one island. As an institution it still had many other regional commitments to fulfil, but initially showed some reluctance to seek additional technical and scientific assistance once the VDAP team had left. Moreover, during the first two years of the eruption, ODA/ DFID, as the main funding agency, had virtually no direct contact with the SRU concerning either monitoring in Montserrat or the needs of the MVO. There was only very limited involvement of SRU/UWI in this process at any level, although they were the regional organization which, up to that time, had held full responsibility for volcano monitoring in Montserrat. The UK government (ODA/DFID) placed all contracts for scientific monitoring work in Montserrat through BGS as their principal source of geological advice. This one-sided arrangement served to sour inter-institutional relationships, at both administrative and scientific levels, and matters were not helped by changes in SRU's status within the University of the West Indies. In 1996, the SRU, having been a selfadministered regional research unit since its inception, was subsumed into the bigger physics department, where teaching and theoretical science are the mainstays. Thus, at an administrative level within the university, there was a shift to a directorate with other concerns, and with little experience of the demands and ramifications of mounting a major crisis response. Apparently, there was limited appreciation of the opportunity that such a prominent eruption afforded for reinforcing the SRU group's work by international exposure and external funding. The changes impinged on the deployment in the field of individual SRU scientists and technical staff, and, together with some residual internal difficulties that were legacies from the start of the eruption, the group's former responsiveness, resilience and flexibility were significantly curtailed. With all these internal and external factors taken together, it is not surprising that SRU, a small team labouring under severe constraints in difficult conditions, started to lose some momentum in its lead role at the MVO. Fundamentally, however, the difficulties faced by all the MVO scientific personnel, jointly and as members of separate institutions, stemmed mainly from the political duality that existed in a territory with the simultaneous involvement of two sets of governments.
Changes in the MVO management Recognizing the unsatisfactory nature of the arrangements at the MVO, an 'audit' of the situation was commissioned by the Governor in January 1996. Its main purpose was to consider medium-term needs (i.e. the following three years) and the longer-term development of the MVO and associated scientific programmes. The Audit Report, published immediately following the audit on-island, concluded that the eruption could be long-lived and that there was a
75
need to establish and staff a well equipped and effective monitoring facility. The report recommended the installation of new seismological equipment, the training of local staff and the construction of a purpose-built permanent observatory building. It also recommended the establishment of a management board and greater coordination in handling the UK contribution to the monitoring effort. While the Audit Report was duly completed in January 1996 and its recommendations apparently accepted, most of these (save for the procurement and installation of a new state-of-the-art seismic network) proved to have an inordinately long gestation period, or were not implemented at all. During the course of 1996, there was a shift in the role and responsibilities of the main agencies involved in, and responsible for, the management of the MVO. At the start of the eruption, the MVO was administered locally through the Development Unit within the Chief Minister's Office. In August 1996, responsibility for the MVO along with all other aspects of emergency operations was transferred to the Governor's Office. An administrative manager was employed to assist with the co-ordination of non-scientific aspects of the MVO work and to manage local staff. In September 1996, a decision was taken that the MVO should be jointly managed by the BGS and SRU, with the position of Chief Scientist alternating between staff drawn from the SRU and the UK. Each Chief Scientist took on a tour of duty that lasted between four and six weeks, typically requiring a commitment of 14 or more hours per day, seven days a week. Added to this were frequent disturbed nights when the volcano was very active. The rotation arrangement was intended to reduce the stress associated with taking responsibility for all scientific endeavours at an erupting volcano and to ensure that the individuals concerned functioned at optimal efficiency in the circumstances, without suffering undue fatigue that might affect judgement. The changes of Chief Scientist caused some inconsistency in management, unavoidable shifts in emphasis and difficulties in developing long-term strategic policies for the MVO. However, these shortcomings were ameliorated by posting to the team selected scientists who undertook longer stints of duty as Deputy Chief Scientist. These individuals took on a significant proportion of the routine administrative work of the observatory and provided much-needed continuity in management, while still acting fully as members of the scientific team. As the crisis developed, the direct load of responsibility on the Chief Scientist was also relieved by posting nominated Senior Scientists on missions to the Observatory. These were all experienced volcanologists who could assist with the tasks of interpreting observations and data for day-to-day hazard assessment and risk mitigation. With the establishment of a Chief Scientists' Committee in early 1997, an attempt was made to develop policies that were more consistent. This committee met three times during 1997 and 1998 and, in between, there were frequent contacts with the Chief Scientist in post via the Internet, so that assessments and views coming from the MVO had the understanding and support of all the scientists who filled the Chief Scientist role at other times. Before this happened, however, the BGS had submitted in November 1996 a proposal to DFID for rationalizing the management of all scientific input to the MVO that was funded directly by the UK government. The project was to cover a two-year period beginning April 1997 and provided for the overseas input to the MVO to be managed by the BGS, through whom all UK funding to the MVO would be channelled. For a variety of administrative reasons, the project was not fully approved until mid-1997, and then there were further delays in the signing of a contract between UWI and the UK Natural Environment Research Council (NERC) for the involvement of the SRU. This was not finalized until early 1998. The very slow speed of completion led to further difficulties between the two principal institutions involved in monitoring the volcano. With the eventual inception of the two-year contract, the formal reporting line to the UK government then became one that went from BGS to DFID. This involved meetings with DFID staff and weekly reports from the Observatory to DFID on the state of the volcano. DFID also received scientific feedback from the Foreign Office through the informal input of scientific advice or indirectly
76
W. P. ASPINALL ET AL.
via the Governor's Office. These multiple reporting channels and complex inter-relationships continued to lead to occasional misunderstandings and confusion during the crisis. The on-island reporting mechanism - bi-weekly briefings by the Chief Scientist to local administrators (see below) - remained the most important direct interface to the civil authorities and was largely unaffected by the overarching administrative changes.
Funding arrangements By late 1997, funding for the MVO was thus derived jointly from the Government of Montserrat (local and recurrent expenses) and from the UK government through DFID (foreign scientific support and major expenditure on monitoring equipment). The Government of Montserrat supported the local staff, MVO's local budget and the direct participation of the SRU during the period July 1995 to April 1997. Various short-term contracts between DFID, the BGS and individual consultants supported the UK and other nonSRU overseas scientists until April 1997. The new two-year DFID contract, which began (back-dated) in April 1997, provided the means for continued involvement of all overseas scientists at the MVO, including those by now being seconded from the SRU. Initially, the authorities had been slow to appreciate the significance of the longevity of the eruption and the crisis, and to understand the implications of this for scientific monitoring and its funding at an active volcano. This had led to a number of administrative responses of a short-term nature and the early development of the MVO as if it were in a conventional civil service setting, which in turn allowed little flexibility in spending. The commitment by DFID to a two-year period of stable funding for the monitoring activities improved matters, but the scientific team was still unable to respond as rapidly to breakdown of equipment or a need to acquire new equipment as they would have liked. This was because the BGS, as manager of a publicly funded project, was also bound by restrictive procedures that allowed little room for local flexibility (Clay el al 1999).
MVO management board Formation of a Management Board for the MVO, which had been recommended by the January 1996 Audit Report, was finally completed in October 1996. However, the first meeting of the Board did not occur until 7 March 1997. At that meeting, the MVO was
declared an entity of the Government of Montserrat; a mission statement was approved and decisions taken on the management structure, staffing policy, funding and long-term development of the MVO. As was the case with the Audit Report, few, if any, of the Board's recommendations were fully implemented. The Board itself had no direct role in the management of the MVO and did not devise any clear mechanisms for ensuring that decisions were implemented. There was no clear line of authority to link the Board to the MVO operations. Furthermore, the changes in administrative and funding arrangements that were put in place during late 1996 to mid-1997 were paralleled by the shift, noted above, in the relative roles of the SRU and BGS within the MVO: BGS now co-ordinated and directed the involvement of all nonSRU overseas scientists (who were in the majority), and took over effective management of nearly every aspect of scientific work at the MVO.
Relocation, another review of MVO needs, and further management changes Outside the Observatory, the escalating volcanic activity in the middle and later parts of 1997 caused another relocation of the MVO base after nearly two years in Old Towne (Fig. 1). to its present location at Mongo Hill (Fig. 3). The need for a permanent building for MVO had by then been well established. However, recurring problems regarding management and financing, and the longevity of the eruption itself, pointed to the need for a wide-ranging governmental review to ensure the continuance of the MVO beyond the eruption. This led to a Joint Comprehensive Review (JCR) of the MVO, which was commissioned by DFID in February 1998. The main purpose of this was to define the objectives for, and outputs required from, the MVO and to formulate plans on how this could be achieved cost-effectively in the short, medium and longer term. The review was intended to formulate the basis for future funding and management of the MVO. initially for the next threeyear period. The final report was delivered in mid-1998 and contained recommendations on all aspects of the MVO's operations. The JCR recommended that the MVO should become a statutory body of the Government of Montserrat, managed by a twotiered board system. A Management Board would meet annually to determine policy while an Operational Board (with wider representation) would meet quarterly and be responsible for implementation of policy and accountability. A Director with an appropriate
Fig. 3. The Mongo Hill building that became MVO (North) in September 1997. The site is on the northern side of the Centre Hills, near St John's, looking north with no view of the volcano. Windows are shuttered in anticipation of an imminent hurricane (photo G. E. Norton r BGS NERC).
THE MONTSERRAT VOLCANO OBSERVATORY
scientific background was to be recruited as soon as possible and arrangements made for continued provision of specialist advice, whenever needed. It was recommended that all funding to the MVO should be arranged through a single channel. Provision of a forward observation post, which would give the MVO team a clear view of the volcano, was seen as a critical operational issue. On 15 July 1998, the second meeting of the Management Board of the MVO was convened. This meeting had become somewhat overdue and attempted to deal with several important issues. However, most of the discussion centred on the findings of the JCR, while several on-going matters were left outstanding. Despite this, a number of key decisions were taken at this meeting. The recommendation of the JCR to establish the MVO as a statutory body was endorsed; the composition of both boards was discussed and agreed upon. It was decided that an Interim Director, with responsibility for overall management of the MVO, would be appointed as soon as possible. In addition, separate subcommittees were established to make specific recommendations on the future rationalization of the two seismological networks operated by the MVO and on the recruitment of a permanent director, respectively. Unlike previous occasions, follow-up action was taken on all the main decisions of the Board. An Interim Director of the MVO was appointed (who held post from October 1998 to June 1999), members of the Operational Board were selected and a draft bill was formulated for the incorporation of the MVO as a statutory body of the Government of Montserrat. This bill subsequently passed into law in August 1999. The appointment of the Interim Director allowed administrative action to be taken forward in a number of areas. The Operational Board held several meetings and made recommendations for consideration and endorsement by the Main Board on the establishment of a permanent observatory building, monitoring strategy, funding and a number of policy issues. As a result, the MVO now has a full-time Director who currently supervises seven local technical and administrative personnel and one contracted seismologist. They are assisted as needed in the short-term by one or more professional geoscientists, called down through the BGS/DFID arrangements. Funding is provided by the UK government through budgetary aid to the Government of Montserrat and includes provision for the construction of a permanent observatory. The wider social, political and economic contexts of the Montserrat eruption crisis, within which the evolution, vicissitudes and frequent restructuring of the MVO and its management are but a small part, have been exhaustively evaluated by Clay et al. (1999), on behalf of DFID.
77
Fig. 4. Map showing the positions of key stations of the MVO analogue short-period network and the original installed locations of the five broadband and three single-component seismometers that comprise the MVO digital broadband network. (Some instruments of both networks were co-located at the same sites.) The kernel of the short-period network was in place before the eruption started in July 1995, and the combined networks operated together from October 1996. There have been subsequent modifications due to damage from volcanic activity. Also shown are the locations of some temporary short-period analogue stations that were deployed during the crisis.
Monitoring Dedicated monitoring of Soufriere Hills Volcano remains the main objective of the MVO. This has involved routine application of various established techniques for acquiring sufficient data to provide reasoned scientific advice to civilian authorities, but with a strategy allowing flexibility so that additional techniques can be introduced if different hazards emerge as specific threats to the population. Routine monitoring entails three main activities: visual observation of the volcano, operation of the seismological network and ground deformation measurements (Figs 4 and 5). Studies of dome morphology and volume, and deposit mapping are on-going. Gas monitoring is done through direct and remote sampling. Environmental monitoring, in the form of groundwater, rainwater and ash analyses and atmospheric geochemistry measurements, has been undertaken since 1995. Other techniques used include: petrological and geochemical analysis of the erupted products, gravity surveys and rock-strength measurements. Video and still photographs are taken whenever activity dictates and conditions permit. A summary of the main areas of activity at the MVO is given in Table 1; in respect of scientific results, an overview can be found in Sparks & Young (2002). Visual observations have been made from a helicopter and from various observation points around the volcano since eruptive activ-
ity began in July 1995. These observations have been vital to assessing the state of the volcano and have been supplemented with basic photography using small- and medium-format still cameras and video photography. The data collected have allowed the detailed and accurate documentation of the evolution of volcanic activity, and of changes in dome morphology and dome volume. Two seismic monitoring networks have been used by the MVO. A short-period network using analogue telemetry has been in place since July 1995, including some stations that had been installed by the SRU before the crisis. The network consists mainly of vertical 1 Hz instruments, although two three-component instruments have been deployed. Throughout the eruption at least five and usually eight or more stations have been continuously transmitting data by VHP radio telemetry or telephone lines to a PC-SEIS dataacquisition system (Lee 1989) installed at the MVO by the USGS VDAP team. Events are inspected routinely, classified by type (Miller et al. 1998) and locations computed using the HYPO71PC program (Lee & Valdes 1989). The seismic signals are monitored 24 hours a day at the Observatory. Signals from four of the stations are written to paper drum recorders to give a real-time view of the seismic activity. A network of five three-component Guralp CMG40T broadband seismometers was installed in October 1996 and operated in conjunction with the short-period network (Fig. 4).
Fig. 5. (a) Network of stations used for precise levelling (dry tilt) monuments and electronic tiltmeter installations, (b) Electronic distance meter (EDM) networks and measurement sites on and around Soufriere Hills Volcano. Occupation of dry tilt stations was discontinued in 1996. and eruption damage to electronic tiltmeters and EDM targets reduced these networks to a single EDM line by 1998.
THE MONTSERRAT VOLCANO OBSERVATORY Table 1. Summary of main areas ofMVO activity, October 1998 to April 2000 Seismology Processing of earthquake and seismic signals from SP and BB networks Analysis and archiving of seismic data Manning of the operations room Ground deformation GPS measurements (once every one to two months) Electronic distance measurements (depending on visibility) Electronic tilt (continuous, telemetered) Theodolite measurements and photographs for dome volume calculations (depending on visibility) Geology Deposits sampling (as activity permits) Field mapping (as activity permits) Genera! Visual observations (daily) Airfall ash and water sampling (as activity dictates) Gas monitoring with SO2 tubes (every two weeks) and COSPEC (approx. three times per week) Mudflow monitoring (as required) Temperature measurements at pyroclastic flow deposits (every two months) Fumarole and hot springs sampling (two or three times/year, as conditions permit) Video/photos of volcano (as conditions permit) Special projects Hazard mapping Risk assessments and updates Dedicated monitoring for authorized activities in Exclusion Zone Preparation of public education/outreach material Electronics Maintenance and repair of electronic and field equipment Maintenance and operation of computer network Video archive Outreach Briefing of government officials (twice weekly and as activity dictates) MVO reports (daily, weekly, monthly, special) Press interviews Weekly radio interviews Public lectures; guided tours of the MVO (as necessary; officially once to three times per month on weekdays) Posters at the MVO and elsewhere Videos, and television and radio programmes on the volcano and volcanism Web pages Collaboration with other scientists and groups
79
stations located on the western flank of the volcano (Fig. 5a, b). Concern over the stability of Gages wall, overlooking Plymouth, led to the introduction of a Leica TCI 100 Total Station. This was used with single or triple prism targets to monitor possible movement on Gages Wall and Castle Peak dome. Six target sites were initially established in late September 1995. The network was radically modified between November 1995 and January 1996 when further instrument and target sites were added (Jackson et al. 1998). Destruction of reflectors during the progressive encroachments of the volcanic activity onto the surrounding flanks undermined the effectiveness of this network, which, by mid-1998, was abandoned when the single remaining reflector on the northern flank was lost. A new network was established in and around Long Ground in February 1999 to monitor possible movement on the eastern side of the volcano and the line from the side of Chances Peak to Amersham was reoccupied in June 1999 after a two-year period of inaccessibility (Fig. 5b). The ground deformation programme was reinforced from April 1996 by measurements using the portable Leica System 300, dualfrequency, differential global positioning system (GPS; Fig. 6). This allowed resolution of changes in line lengths between points on the volcano into vertical and horizontal components. The equipment was used in the rapid static mode and data processed using the Leica SKI software system (Shepherd et al. 1998). In addition to the MVO GPS system, the University of Puerto Rico (UPR) has maintained a GPS network at Soufriere Hills Volcano since July 1995 (Mattioli et al. 1998). At present, this latter network consists of three permanent GPS stations, which continuously transmit data to the MVO and to the UPR via the Internet. The MVO established a permanent GPS station at Harris in April 1998. Closer collaboration with the UPR GPS programme from February 1998 allowed the use of data collected by the UPR network in processing data by MVO GPS stations. During early 1998 and again in March 1999, the MVO GPS network was upgraded with new receivers and another permanent station established at South Soufriere Hills. A further MVO permanent GPS station was installed at Spring Estate in June 1999. Details on the early crisis period results, obtained from the GPS programme at Soufriere Hills Volcano, can be found in Shepherd et al. (1998) and Mattioli et al. (1998). Electronic tiltmeters have been used intermittently since July 1995 when three stations were initially deployed at Spring Estate, Amersham and Long Ground (Fig. 5a). The stations used high-gain tilt sensors with bubble-type biaxial platform tiltmeters and digital telemetry designed and built by the USGS Cascades Volcano Observatory (Murray et al. 1996). Two instruments were moved to Chances Peak in December 1996 because of increased concern over
Management Work programme and staffing etc. Helicopter operations Acquisition, repair and maintenance of monitoring equipment Scientific and other professional visitors Housing/accommodation Collaboration with National Trust and Tourist Board on development of visitors centre Specific operational tasks can change from time to time, but it is anticipated that most of these MVO responsibilities will remain for the foreseeable future.
Three single-component Integra LA100/F seismometers were also included in the network. These data are transmitted as 24-bit digital signals by UHF radio telemetry to the Observatory where it is acquired by a SEISAN data-acquisition system. Further details of these networks along with the data collected during the first two years of operation can be found in Aspinall et al. (1998b), Miller et al. (1998) and Neuberg et al. (1998). Ground deformation monitoring began with the use of a Wild NA2 precision level and 3-metre invar levelling rods at three dry tilt
Fig. 6. MVO field team, Rick Herd and Tappy Syers, having uncovered a marker pin beneath thick new surge deposits, are setting up the Tar River GPS station for overnight measurement, January 1999. The ruins of the Tar River Estate House are in the background. The view is looking NE, out to sea from the volcano (photo © R. E. A. Robertson).
80
W. P. ASPINALL ET AL.
the possibility of sector collapse, and others were placed at Hermitage and Gages Mountain in 1998 (Fig. 5a). However, apart from the stations on Chances Peak, there have been no perceptible variations or trends in the data collected from these tiltmeters (Voight et al. 1998). The Chances Peak stations were important for discerning cyclic patterns of conduit and dome pressurization and these variations were used, together with seismicity changes, to provide accurate short-term forecasts of activity during June to August 1997 (Voight et al. 1998, 1999). Estimates of dome volume have been made since January 1996 using compass and Abney level surveys, supplemented by photographs and theodolite measurements (Watts et al. 2002). Repeated observations and photos from fixed positions allowed estimates of dome volume to be made from trigonometry and triangulation, and from simple geometric measurements between successive photographic images (Sparks et al. 1998). In August 1996, a GPS kinematic survey method using mobile equipment and laser rangefinding binoculars was developed and used to determine dome volume. During late 1998, a computer program for such GPS data, based on an approach similar to the photographic method, was developed and used for estimating dome volumes (Herd et al. 1998). Measurements of the volume, distribution and character of pyroclastic deposits have also been vital for quantifying and characterizing the evolution of the eruption. Regular calculations of deposit volumes using the laser range-finding binoculars and kinematic GPS method, together with isopach mapping of ash-fall thickness (Bonadonna et al. 2002), have provided information on the total volume of magma extruded (Sparks et al. 1998). Changes in topography of pathways for pyroclastic flows were monitored giving direct input to hazard assessments, such as the numerical modelling of runouts (Wadge et al. 1998). Detailed mapping, sampling and temperature measurements of the new deposits (Cole et aL 1998, 2002; Calder et al. 1999; Komorowski et al. 2000; Druitt et al. 2002; Loughlin et al. 2002b) provided critical data for assessing the likely runouts and areal extents of impact of different phenomena. Coupled with studies of the geology of the Soufriere Hills complex (Roobol & Smith 1998; Harford et al. 2002), this resulted in more accurate assessments of the likely range of future behaviour of the volcano. Petrological work (Devine et al. 1998(7,6; Barclay et al. 1998; Murphy et al. 1998, 2000) has given indications as to the state of the magma chamber and the ascent rate of the magma. Gas monitoring has involved various techniques throughout the course of the eruption. The environmental effects of sulphurous gases have been monitored by filter packs (Allen 1996), SO2 diffusion tubes and measurements of acidity levels in rainfall and surface water (Norton 1997). Remote sampling of gas concentrations has been undertaken using a Mini-COSPEC correlation spectrometer (Young et al. 1998a) and open-path Fourier transform infra-red spectroscopy (Oppenheimer et al. 1998; Edmonds et al. 2001); these measurements helped elucidate cyclic emission patterns and deeper processes. Direct sampling of dome gases was done on one occasion and sampling of the associated hydrothermal system (Hammouya et al. 1998; Boudon et al. 1998) was carried out routinely during the early stages of the eruption until sample sites became too dangerous or covered by volcanic deposits. All these diverse scientific monitoring activities have been directly geared to providing information for hazard assessments and crisis management.
Crisis management The crisis management of the Montserrat volcanic eruption has been characterized by extremely close and frequent interactions of the scientific team with the civilian authorities and the public. In addition, a custom Montserrat Alert System was drafted that was specific to the conditions on the island. This has been supplemented by volcanic hazard mapping, vulnerability assessment, risk maps, formal and informal reports, scientific briefings, public meetings, news-media interviews and articles.
During the first few months of the eruption, a framework was set up to facilitate interaction among scientists and also between the scientists and the administrative authorities. An existing volcanic hazard and risk assessment, prepared originally by Wadge & Isaacs (1987, 1988), was initially adopted and then adjusted, as the nature of the eruption became clearer. (Although this work had been in existence for several years, its findings were not utilised in good time for mitigation prior to the start of the eruption; see Kokelaar 2002). The MVO subsequently produced a series of hazard maps for the authorities and worked closely with civilian authorities to produce risk maps. These showed geographical zones indicating the level of danger locally and also whether occupation was advised or allowed. The maps were published in the local newspaper and widely circulated as hand-outs. The conclusions of specially convened scientific meetings of the MVO team provided the basis for an accompanying consensus opinion on the appropriate state of activity to apply. At government briefings, this advice was normally presented by the Chief Scientist on behalf of the team.
Government briefings The MVO has provided frequent, regular briefings to the local authorities on the state of the volcano ever since the onset of eruptive activity at Soufriere Hills Volcano in July 1995. As a result, key officials have become thoroughly familiar with both the extent of scientific understanding of the volcano and the implications of any scientific uncertainty. These briefings were, and remain, integral to decision making for official response and mitigation measures. From July to September 1995, the Chief Scientist, often accompanied by another senior team member, provided daily briefings to the crisis management teams. As the pattern of events became more certain and dome growth began, the briefings became more structured and were given at least weekly to the crisis management team. This consisted of the Governor and the Chief Minister, senior civil servants and aid-agency personnel. In addition, the Governor and Chief Minister attended at least one of two internal scientific meetings held weekly at the MVO to discuss the state of the volcano. This continued until the middle of 1997. when a system of bi-weekly briefings was initiated, which remains in effect to the present. The first of these is usually given on a Monday to members of the Volcanic Executive Group, the highest decision-making body involved in disaster management on the island, comprising the Governor, the Chief Minister, senior ministers of government, the Chief of Police, the Chief Scientist and the Head of the Emergency Table 2. The Montserrat Alert Scheme, December 1995 Volcanic activity
Alert stage
Background seismicity with no new surface manifestation of volcanic activity.
0 (White)
Low-level local seismic activity, ground deformation and mild phreatic activity.
1 (Yellow)
Dome building in progress, periodic collapses generating rockfalls and occasional pyroclastic flows. Moderate level of seismic activity with no sudden changes.
2
(Amber)
Change in style of activity anticipated within a few days. Pyroclastic flows common with associated light ash fall. High level of seismic activity.
3
(Orange)
Major dome collapse under way, with large pyroclastic flows and heavy ash fall. Explosive event possible if the activity continues.
4
(Red)
Ongoing large explosive eruption with heavy ash fall.
5
(Purple)
The Alert Scheme was used as a guide when providing advice to the authorities about activity at the volcano. Although no specific actions are listed for the administrators, contingency plans draw ? n up by the Government's Emergency Department were based on the identified alert levels.
THE MONTSERRAT VOLCANO
Department. The second briefing is given on a Friday to a larger group called the Volcanic Management Support Group, which consists of senior civil servants and key sector agencies. The briefings are usually given by the Chief Scientist (now the Director) or Deputy Chief Scientist and consist of a synopsis of activity for the week along with an up-to-date assessment of what may be expected from activity during the ensuing week. Advice is also given on various aspects of crisis management, from the scientific perspective. As noted earlier, one of the peculiarities of Montserrat is the existence of a complex civil administration, with two major ministries of the UK government, the Governor and the local government all involved in the response to the crisis. Inevitably these arrangements at times led to a lack of clarity about where, and in what form, essential scientific information should be provided. The drawbacks of such a situation were exemplified early in the eruption by the appointment of independent consultants to the FCO and DFID at the same time as there was an on-going reporting line in Montserrat, originating from the embryonic MVO. This caused the SRU scientists in Montserrat to feel that their role was being undermined and their advice countered by separate, independent opinion being expressed in the UK. While their concern was soon acknowledged and resolved within the scientific team, the issue of the garbling or misinterpretation of the content of urgent scientific advice by uninitiated intermediaries remained a source of problems, even when only one channel of reporting was meant to be active. Thus, the problem of how to communicate uniform and coherent scientific advice to several interested arms of two governments, in parallel, in two separate countries in different time zones, has
OBSERVATORY
81
been a significant challenge throughout the eruption. It is a complication that may arise in other countries where there are both federal and local strands of government, for instance. The introduction of six-monthly risk assessments from August 1997 onwards, formally prepared by a group of senior scientists, improved the situation. These reports are delivered to all interested parties simultaneously and the principle has been established that the Chief Scientist (more recently, the Director) should be the channel for all immediate scientific advice. The problems that were encountered earlier on in Montserrat serve to emphasize how important it is for scientists to establish clear lines of communication in a volcanic crisis, although this can prove far from straightforward if duality of political accountability exists.
Alert Scheme From the onset of the eruption, an Alert Scheme has served as a useful tool for guiding civil response actions to varying levels of volcanic activity. The scheme has undergone several iterations. The first system used was a generic one patterned on that from the Office of the United Nations Disaster Relief Co-ordinator (UNDRO) handbook on Volcanic Emergency Management (Table 2). It was not specifically tailored to the peculiarities of the environment on Montserrat and proved inadequate to guide all necessary mitigation efforts as the eruption slowly evolved. The explosive activity that occurred in mid-September 1996 was a significant departure from the extrusive style of eruption that had
Table 3. The Montserrat Alert Scheme, March 1997 Volcanic activity
Alert stage
Background seismicity with no new surface manifestation of volcanic activity.
0 (White)
All zones occupied. Review and update emergency plans on an ongoing basis.
Low-level local seismic activity, ground deformation and mild phreatic activity.
1 (Yellow)
Maintain readiness of key personnel, systems and procedures. Keep stock of critical supplies Local evacuations may be necessary in Zone A. Zones B and C on standby.
Dome building in progress, periodic collapses generating rockfalls and occasional pyroclastic flows. Moderate level of seismic activity with no sudden changes.
2 (Amber)
Zone A: No access Zone B: Access limited to short visits by residents, officials and approved visitors with rapid means of exit Zone C: Daytime-only visits by residents, approved commercial activities and agriculture Zone D: Day- and night-time occupation by residents; high level of alert maintained Zones E, F: Full occupation by residents with national contingency plan for evacuation in readiness Zone G: Full occupation
Change in style of activity anticipated within a few days. Pyroclastic flows common with associated light ashfall. High level of seismic activity.
Major dome collapse under way, with large pyroclastic flows and heavy ash fall. Explosive event possible if the activity continues.
Ongoing large explosive eruption with heavy ashfalls.
(Orange)
4 (Red)
(Purple)
Actions by administrators and general public*
Zones A, B: No access Zone C: Access limited to short visits by residents and workers with means of rapid exit Zone D: Daytime occupation for essential services and agriculture, residents allowed access in daytime. Essential services operate with standby transport and evacuation plans in place Zone E: Full occupation with high level of alert maintained. Schools operate with standby transport Zone F: Full occupation by residents with contingency plan for evacuation. Warn of ashfalls in Zones E and F Zone G: Full occupation Montserrat Standing Operating Procedures for Red Alert in place All schools closed as required. People with special needs removed from Zones E and F Zones A, B, C, D: No access; rapid evacuation of all remaining persons Zone E: Rapid evacuation; warn of potential for gravel, pumice and ashfall Zone F: Warn of potential for gravel, pumice and ashfall Zone G: Full occupation Zones A, B, C, D, E: No access, rapid evacuation of all remaining persons Zone F: Initiate evacuation; warn of potential for gravel, pumice and ashfall Zone G: Warn of potential for ash hazards
* Recommended actions to minimize significant casualties. Zones as defined on Volcanic Risk Management Map. This Alert Scheme was drafted with closer collaboration between the disaster managers and the MVO and, as such, incorporated specific actions to be taken by administrators and the general public. The Scheme had accompanying risk maps (e.g. Fig. 7) that illustrated the zones mentioned. These maps, and a detailed listing of the specific villages within each zone, were distributed to the public by the civil authorities.
W. P. ASPINALL ET AL.
occurred up to that date. Consequently, it was necessary to revise the Alert Scheme to incorporate new conditions for potential explosion hazards, with corresponding changes concerning vulnerability and response. This resulted in the first draft of an Alert Scheme specifically for Montserrat (Table 3), which provided a detailed guide for actions to be taken by civil authorities within defined risk zones around the volcano at specified alert levels determined by the MVO team (Fig. 7). This, in turn, was revised a number of times as activity escalated. During early December 1996, when the Galway's segment of English's Crater appeared to be under severe stress, it emerged that differences in interpretation of the emergency procedures existed between scientists at the MVO, the civil administrators and the public. These procedures came under closer scrutiny again in mid-December 1996 when, upon raising the alert level in response to increased activity at the volcano, only a small percentage of the residents in a particular community that was deemed vulnerable heeded the official notice to evacuate. The increased activity that triggered the raising of the level on this occasion had all but died down by the time the public announcement for evacuation was made. Once volcanic activity had encroached on most of the built-up areas and vital services around the volcano in the tragic events of 25 June 1997, the Alert Scheme could no longer be retained as a microzonation management tool. The approach was then simplified to a system whereby the entire island was divided into a Northern Zone, a Central Zone and an Unsafe or Exclusion Zone, with severe restrictions being placed on entry into the Exclusion
Moving from Zones G to A represents an increasing risk, based on an evaluation of the volcanic hazard. The status of each zone is dependent on the Alert Level. Fig. 7. Volcanic Risk Management Map from 6 June 1997, drafted by the MVO in consultation with Montserrat government officials and disaster managers. This and the maps in Figures 8 and 9 were based on hazard maps produced for the main types of volcanic events that were expected from Soufriere Hills Volcano and were used along with population distribution and infrastructure maps to delineate risk boundaries (map redrawn here for clarity).
Zone (Fig. 8). This system was considered applicable for as long as the dome-building eruption continued and, with minor administrative changes, has remained in effect ever since. Small modifications to the boundary line of the Exclusion Zone were made as activity at the volcano decreased during 1998 (Fig. 9). However, scientific consensus on what activity level was appropriate at any particular time was seldom easy to achieve and this stimulated the development by the MVO team of structured decision procedures for the purpose.
Expert elicit at ion As an aid to discharging the responsibility to provide good and timely advice to the civilian decision makers in the form of an agreed scientific position, a formalized procedure for eliciting expert judgement was adopted by the MVO in August 1995. The need became clear at this early stage in the crisis when the nature and magnitude of future volcanic activity was most uncertain and time for repeated and protracted scientific debates amongst a large group was just not available. Arriving, by committee, at an agreed position on what advice to give the authorities against a deadline each day, took an increasing amount of the scientists' time, just when they urgently needed to press ahead with observations and measurements. This became frustrating both for them and for the authorities, who preferred to have a rapid and definitive answer provided as punctually as possible. The formalized procedure that was introduced is based on the 'classical model' for structured expert judgement (Cooke 1991), following the suggestion first made at the International Symposium on Large Explosive Eruptions in Rome in 1993 to consider using this approach for producing a collective scientific opinion in a volcanic crisis (Aspinall & Woo 1994). The method performs weighted combinations of expert judgements, where weights are determined by 'calibration* and 'information' performance on questions for which the true values are known, or become known a posteriori. However, in application, the procedure had to be adapted to the needs of real-time crisis management. In open meetings, where some preferred to hold their counsel, others were more forceful in giving their opinions so. from the outset, it was considered better to focus the procedure on quantifying the range of conservatism ("informativeness') of each individual's views, rather than rely too heavily on a hurried and questionable calibration score. (The problems of calibrating the expertise of a group of volcanologists are non-trivial at the best of times, let alone with an eruption going on outside the window!). By emphasizing individual informativeness in this way, there was an implicit assumption that all members of the team had roughly equal expertise, a reasonable supposition for the initial scientific group assembled in Montserrat. In fact, when the concepts of this approach were being introduced to the administrators and scientists they were novel to most, and several pressed for the scheme to be administered in such a way that no single participant was ever zero-weighted: all views would be used with some weight in the decision process. However, in order to bring some element of calibration into play, a preliminary set of five suitable seed questions was hastily prepared. While this set was aimed primarily at measuring the individual's informativeness factor, it was also used to provide a basic measure of how well each person might make quick judgements on issues related to safety and hazard mitigation in an emergency. The latter element of the exercise was undertaken, with general agreement, as an exploratory trial of the method in application in a live crisis. As the emergency progressed, three limitations to the elicitation procedure emerged. First, having once used the initial set of 'seed questions' for calibration, for uniformity these had to be repeated for scientists subsequently joining or replacing others in the team. Over the course of three years, a total of more than 60 individual scientists and technical specialists participated in the different elicitations (although there were generally only between five and 20 present at any one time). To everyone's credit, the answers to the
83
THE MONTSERRAT VOLCANO OBSERVATORY
N
Fig. 8. Volcanic Risk Management Map for Montserrat as it stood in September 1997. Microzonation was discontinued after 4 July 1997, when volcanic activity became more severe.
Exclusion Zone Central Zone
No admittance except for scientific monitoring and National Security Matters Residential area only, all resident on heightened state of alert. All resident to have rapid means of exit 24 hours per day Hard hat area all residents to have hard hats and dust masks.
Northern Zone
Area with significantly lower risk, suitable for residential and commercial occupation Limit of high direct volcanic hazard
seed questions were not leaked to new arrivals at the Observatory. Secondly, an experienced technical facilitator was not always available in Montserrat to supervise the elicitations and thirdly, there was a fluctuating mandate to undertake formal elicitations, as Chief Scientists rotated, or as conditions and levels of anxiety varied. There were divided opinions amongst the scientific group as regards the utility and soundness of the method, based as it is on concepts of subjective probability. These were new to many, and, with an experienced technical facilitator not always available at the MVO, the technique was not used continuously. However, the majority of scientists supported the approach, particularly for issues requiring frequent but tricky, marginal decisions, for which it proved a very successful aid. After being discontinued in late 1995, the methodology was reintroduced following the 17 September 1996 explosion. It was then used more widely in the decision-making process, being integrated into the Alert Scheme and used on a weekly basis to assist the Chief Scientist to determine the appropriate level of alert. The method continued in operation in this context until the Alert Scheme was simplified in July 1997. Its use for assessing the appropriate alert level gave continuity to the decision-making
process and provided a traceable record through time of the views of the scientific team (see, for example, Fig. 10). Further details on the theory and potential application of this technique to volcanic crises are given in Aspinall & Woo (1994) and an account of its introduction and use at the MVO is contained in Aspinall & Cooke (1998). The same structured elicitation procedure was also employed by the scientific team to progress the strategic hazard and risk assessment reviews that were commissioned later on in the crisis.
Risk and hazard assessments With the change in eruptive style that occurred in 1997, a more formalized approach to hazard and risk assessment was invoked. The first of what became regular reviews of the state of the volcano and its activity was undertaken in December 1997 and involved most of the senior scientists who had worked at the MVO. These assessments provided a considered and comprehensive review of the most recent activity and, to meet the requirements of the civil
84
W. P. ASPINALL ETAL.
proper consistency to the outcomes. The structured elicitation approach to this problem is one more example of how creative hazard management techniques have been developed within the MVO during the crisis. All the risk assessments have been updated at intervals of approximately six months (or as circumstances demanded). They have depicted most likely outlooks for future activity which consistently, as time proceeded, were not overturned by more extreme events or major surprises in the behaviour of the volcano. These assessments thus form a valid basis for addressing the question of 'acceptable risk'.
Acceptable risk
Bottom of Belham River valley and all areas south, across Centre Hills to Pelican Ghaut Limit of high direct volcanic hazard
Fig. 9. Volcanic Risk Management Map of Montserrat, revised in September 1998. The boundaries of this map remained unchanged until April 1999 when some minor administrative changes were implemented by the civil authorities in consultation with the MVO.
authorities, arrived at a prognosis for the immediate following period (six months) and the more distant future (five years or more). Various possible eruptive scenarios for this volcano, which these reviews developed on the basis of available evidence, were usually summarized in the form of event-probability trees. This provides a compact format for presentation of such information to decision makers. The formalized elicitation methodology, used with the Alert Scheme, has also been used extensively to provide consensus probability value distributions for input onto the branches of the event trees (see Fig. 11), in conjunction with traditional decisionconferencing and numerical modelling. In turn, these probability distributions serve as inputs to Monte Carlo models for simulating population casualty numbers, which form an integral part of the volcanic risk assessments that the MVO undertakes for guiding civil mitigation decisions. As the eruption progressed and its behavioural patterns changed, the event trees became increasingly more specific in terms of threats to particular areas and, as population zone occupancies changed, the casualty impact simulations also grew in complexity and detail. Figure 12 shows examples of successive F-n exposure curves for Montserrat (where the F-n curve is a risk assessment concept representing the exceedance probability of frequency of a cumulative number of casualties). In constructing and running these models from one assessment to the next, it was essential that the varied opinions from a diverse and changing group of experts could be incorporated in a coherent manner in order to maintain a rational balance to the input assumptions and a
In Montserrat, it was acknowledged by all parties that the administrative authorities held responsibility for specifying the level of acceptable risk throughout the volcanic crisis. From the onset of the eruption, their stated policy was to set the tolerability threshold at a low level, and to maintain it there. Whenever the risk to an area near the volcano was deemed to be approaching life-threatening levels, evacuation of such areas was undertaken. Having accepted accountability for determining the threshold of risk acceptability, the government also assumed the responsibility for providing relief assistance to the population that was evacuated from threatened areas. Bans were placed on re-entry until the danger was judged to have subsided and. in order to apply them, police checkpoints were established at key entry points. The purpose of these checkpoints was to limit access to authorized essential personnel only, thereby restricting the number of entries to a manageable level, in accordance with emergency and riskmanagement plans and capability. However, because of a number of social and political factors, these exclusions were not totally enforceable. Furthermore, for some people living and working near the boundary of an exclusion zone, it was inevitable that the small, residual risk involved there (about which they were advised) was not an effective restraint. By making unauthorized entry to the Exclusion Zone, many individuals in Montserrat have been able to set their own level of acceptable risk. The actual threshold that such individuals assigned to themselves varied throughout the crisis. At the start of the crisis, public knowledge of the volcano was not as informed as later, and uncertainty about the future was greatest. Many persons, having heard accounts of the 1902 Martinique disaster, accepted only a low risk and heeded evacuation recommentions. As the eruption progressed, and it was felt that their knowledge of the hazards and experience of this particular volcano increased, personal risk thresholds among some of the population appeared to change and many individuals seemed willing to accept higher levels than previously. The majority, however, were prepared to be guided by the advice and information provided by the MVO's outreach efforts.
Outreach Throughout the full duration of this persisting crisis, the MVO has taken many positive steps to help inform all the public of Montserrat on the hazards posed by their volcano (see Table 4 for a summary). They have undertaken a series of educational activities, guided by principles of openness about what is known and frankness about what is uncertain or unpredictable.
Policy A commitment to an open policy as regards information has been maintained by the MVO throughout the crisis and, as far as is practicable, a public open-door policy within the Observatory itself has been the norm (at whichever location it happened to be).
THE MONTSERRAT VOLCANO OBSERVATORY
85
Fig. 10. Time-line chart of repeated appraisals of the volcanic Alert Level using expert judgement elicitations in a weighted voting scheme, together with a record of the Alert Level in force. This representation of scientific opinion, in which the current level setting is assessed against thresholds to move up or down, provided decision makers with both the central (optimal) result of the poll and a measure of the spread of views, thus allowing them to select what level of conservatism to accept. More often than not, the distributions of scientific opinion were skewed, with longer tails to the more cautious end of the range.
This policy attitude evolved directly from public concerns about the openness of the MVO during the early stages of the crisis, when, crucially, there were differences of scientific opinion about whether the volcano would erupt. Then, it was perceived by some that the scientists and civil authorities were jointly withholding information and bad news from the public; later, others felt it was possible that some scientists were working to a hidden political agenda. (As a UK territory, there was no freedom of information act enshrined in law in Montserrat.) To minimize these concerns, the civil authorities endorsed the MVO open information policy and actively supported the scientists in their public outreach.
Outreach activities For five years, the MVO team has engaged in an extensive programme of public information and communication, in partnership with civil officials, welfare and health professionals, schools, churches and the media. Many public education exercises were co-ordinated by the Government Information Service and the Government of Montserrat Emergency Department in response to specific, identified requirements. All these efforts have resulted in a highly informed population. However, it would be unrealistic to pretend that even this intensive programme has resolved all related problems: the longlived character of the eruption and the protracted nature of the crisis on a small island have caused particular difficulties.
During July-September 1995, daily reports and ad hoc scientific briefings were given by the Chief Scientist on local radio and television. From early 1996 to mid-1998, morning and evening reports (which were circulated to key government agencies and other interested organizations by fax) were read out on the local radio station. During periods of heightened activity, additional lunch-time radio updates were, and still are, presented live by MVO staff. These enable topical issues to be explained and ensure that the latest volcanic situation is promptly understood by the public. From early in 1997, one or more of the MVO senior scientists (and usually the Chief Scientist) has been interviewed on the radio, usually once a week on a Friday, for half an hour or more. Special radio call-in programmes, allowing the public to interact directly with the scientists and local technical staff, have been aired frequently. These are effective in clarifying issues of concern to the individual caller and important topics of consequence for the wider public. From time to time, individual MVO staff have given public lectures on more specialized aspects of volcanology. These are always broadcast live on national radio and are invariably followed by intense, but enlightened, questioning from the audience. Articles about the volcano have been contributed regularly to the main local newspaper. In addition, the Montserrat National Trust and the MVO published jointly a monthly magazine called SeismiCity News, which provided further explanation of the science. This publication contained sections aimed specifically at children, because the latter were believed to be influential opinion-formers
W. P. ASPINALL ETAL.
ACTUAL COURSE OF ERUPTION
Fig. 11. A summary event probability tree for early stages of the Montserrat eruption, showing elicited probability values attached to branches depicting different potential eruptive scenarios. Such trees prompt the scientists to identify all possible scenarios and help communicate their forecasts to public officials, summarizing on various timescales the hierarchy of hazards against which mitigation measures can be planned. At the MVO. the structured expert elicitation methodology was used to update probabilities in a consistent manner; associated uncertainty distributions were, in turn, used as inputs to Monte Carlo risk simulations of population impact (see Fig. 12).
within families. The magazine was started in early 1997, following a period in December 1996 when escalating volcanic activity, rapid and frequent changes in alert level, and general public unease raised concerns about the MVO's credibility (see below). SeismiCity News was sponsored by DFID and circulated to the public free of cost. Other public information bulletins and information sheets, circulated freely to the public, have been prepared in collaboration with the Government of Montserrat's Emergency Department. Any organization or group on Montserrat that wished for further information or advice from the MVO scientists has been able to obtain it. Informal talks and tours of the Observatory are given to groups and individuals on request. Scientists make visits to schools, workers' groups, religious groups and commercial enterprises whenever possible or appropriate. For instance, from April 1996 to June 1997, when Plymouth was evacuated but the essential services still maintained their operations there, a great deal of effort was invested in making certain that the workers involved were fully aware of the risks that they faced. During visits made to these workplaces, talks were given and lengthy question periods allowed. When some residents of Spanish Point on the northeastern flank of the volcano refused to move away from danger, MVO scientists visited them in their homes with social workers and presented information about the risks that they were running. Bulletin boards were placed in central locations in the main occupied town of Salem and at the airport, with weekly updates to the displays being prepared by two school students, helped by Observatory staff.
Negative feedback All these public education and information efforts have been fully supported by the Government of Montserrat, the UK government and other agencies, and there have been strong, positive comments on their benefits from all sections of the Montserratian community (including those abroad). Of course, there have also been times
when negative comments were received from the public. Some were made on occasions when the public became aware of internal difficulties within the MVO and these usually concerned the independence, impartiality, cultural sensitivity, or otherwise, of the contributions being made by some members of the team (because of weight of numbers and the political backdrop, a lot of these criticisms tended to be aimed at UK scientists). More generally, however, adverse comments were forthcoming during periods when there appeared to be differences in perception between the MVO and the public regarding the level of hazards posed by the volcano and what mitigation steps were most appropriate. At times when strong eruptive activity waned and visible activity was low for weeks on end. many of the public quickly felt that the volcano was doing nothing at all and posed little or no threat, despite the fact they were informed that the monitoring instruments continued to indicate that the eruption had not ended. To make matters worse, these periods were commonly followed by episodes when the volcano suddenly escalated to highly dangerous activity, often in only a matter of hours. Keeping the population sufficiently alert to the dangers during such lulls in activity was a continuing, major difficulty for the scientists and, on some occasions, led to a few vocal people expressing public disagreement with MVO advice as to the level of risk involved. Examples included the scientists' recognition in early July 1997 that the style of eruption had changed to a more explosive form that might threaten areas further afield, and later, in 1998. that the key bridge over the Belham River was likely to be submerged by lahars. Both warnings were greeted by disbelief and scorn by a few small sections of the community. The general situation was not helped, however, by the fact that, in some instances, although the scientific team had correctly identified the incipient dangers, these did not always materialize promptly. In the case of the major 26 December (Boxing Day) 1997 collapse and pyroclastic surge (Sparks ei al. 2002) that devastated the southwest part of the island, scientific anxiety had been elevated for over a vear before the event.
87
THE MONTSERRAT VOLCANO OBSERVATORY
0.0001
1
10
100
1000
No. casualties N Fig. 12. Typical cumulative F-n risk curves from three risk assessments for Montserrat. The curves show estimated exceedance probabilities for a given number of casualties occurring in the next six months, based on Monte Carlo risk simulations combining identified volcanic hazard scenarios with population numbers in given areas. Such estimates are useful to civil authorities concerned with the tolerability of risk and may be compared with societal risks from other natural disasters, such as earthquakes or hurricanes. The examples shown here indicate how the perceived risk levels dropped between April 1998 and February 1999 while the volcano was in residual eruptive activity mode.
At other times, the credibility of the scientists seemed to be questioned in the public mind because specific instances of heightened activity were not 'predicted'. One case was the magmatic explosion of 17 September 1996. Even though the general likelihood of such an event taking place at some time had been repeatedly expressed, and despite the fact that the scientists had invariably indicated that specific temporal predictions were not possible, when the explosion did occur there was fairly widespread censure of the MVO for a perceived 'failure to perform'. Nevertheless, throughout the crisis, the majority of the general public were very supportive of the work of the Observatory scientists and appreciative of their efforts. Overall, confidence in the scientists and their assessments generally improved as events that had been forecast did, indeed, come to pass and as the team got better at both anticipating the behaviour of the volcano and communicating their insights and understanding. Perhaps one shortcoming of the public education programme that has been pursued by the MVO has been the absence of a mechanism for gauging the effectiveness of the outreach. This has been due mainly to the lack, until September 1999, of a suitably trained permanent member of staff to co-ordinate this particular activity. Apart from the daily scientific reports, many of the methods used to disseminate information to the public have not been sustained throughout the eruption. There have been one-off responses to particular concerns within the scientific team regarding public awareness and safety, and there have been instances that were the result of the (short-term) presence at the MVO of personnel who had specific interests in public education. Greater support for a wider range of outreach activities became possible with the recruitment to the MVO of a full-time Information Officer. However, there was one occasion when comments were deliberately solicited by the MVO from the public. This was organized immediately following the tragic events of 25 June 1997 and consisted of a series of interviews with people who were in the Exclusion Zone at the time, and with those involved in rescuing survivors (Loughlin et al. 2002a). While the information gathered was intended to provide documentary evidence on details of the volcanic episode itself, the responses also allowed the MVO to investigate how well people felt they had been informed overall on the dangers associated with the eruption. The results suggest that, predominantly, MVO's efforts at public education had been
Table 4. Summary of main methods used by the MVO to disseminate information to the public, August 1995 to April 2000 Method
Start date
End date
MVO reports (daily - morning) MVO reports (daily - evening) MVO reports (weekly) Scientific reports (weekly) Scientific reports (monthly) Scientific reports (quarterly) Special volcanic activity reports Lunch-time updates Friday evening radio interview Radio updates at times of high activity SeismiCity News Government Information Service 'Scientific Explanation' Montserrat Reporter special articles Public posters in Salem and at airport Targeted meetings (e.g. to present risk map revisions) Public lectures (occasional) Visits to schools (occasional) Visits by schools to MVO (occasional) Visits to MVO by various sectors of the population (occasional) Video nights (occasional) Local access TV (occasional) Antigua Broadcasting Service broadcasts (occasional) Off-island visits/talks to Montserratian communities abroad
August 1995 August 1995 April 1999 November 1995 June 1998 May 1999 September 1996 February 1997 February 1997 December 1996 January 1997 1997 1996 December 1996 July 1995 August 1995 November 1995 November 1995 July 1995 November 1995 November 1995 November 1995 November 1995
April 1998 March 1999 Continuing April 1998 April 1999 Continuing Continuing November 1997 Continuing Continuing March 1999 Continuing Continuing June 1997 Continuing Continuing Continuing Continuing Continuing Continuing Continuing Continuing Continuing
W. P. ASPINALL ET AL.
effective in informing the public of the risks. They also showed that individuals' decisions regarding entry into high-risk areas on the volcano were very complex and could not be attributed to any single causative factor. Some people in Montserrat appeared willing to take high risks if the perceived return or benefit to them or the community was considered of sufficient importance (e.g. to continue farming), or if the alternatives were repugnant (e.g. living in shelters). These interpretations, and other contributory factors, were to a large measure confirmed by the findings of the jury in the Coroner's Inquest into the fatalities of 25 June 1997, in the verdict handed down in January 1999.
The Internet One important and, for the region, novel aspect of the MVO's outreach has been the use of the Internet to communicate with the wider community of scientists, with non-resident Montserratians who were otherwise unable to obtain up-to-date information on the state of the volcano and with volcano enthusiasts generally. This was achieved through co-operation with the Michigan Technological University, where a Web page containing MVO reports and other information was maintained for most of the eruption. However, this new medium also showed how relationships between scientists, and between scientists and politicians, can be easily jeopardized by such communications if care is not taken. One group unwisely put more information and interpretation into a report on their homepage than the MVO wished to issue at the time. In addition, insensitive phrasing was used to express some of the group's views. This infelicitous 'report' was openly available to the surfing public and, of course, provoked further claims of information suppression on the island. At the same time, the content of the report also succeeded in irritating or offending some scientific colleagues. In September 1999, the MVO was able to establish its own Web page (). MVO scientists also used the Internet to solicit advice and maintain contact with professional colleagues throughout the world. This enabled them to access a wider variety and greater depth of scientific thought than would have been available otherwise. As one example, in October 1996 there were fears that a sector collapse could generate a large volcanic landslide and associated tsunami. Contact with numerical modelling specialists in France and the UK via the Internet enabled sophisticated calculations to be done, the results of which (Heinrich et ai 1999) allayed concerns as to the potential size of any tsunami generated in this way. Thus, in principle and in practice, the Internet allows an observatory to consult very rapidly with the world's leading experts. For the group of Chief Scientists who were responsible for the MVO during the first three years of the crisis, communication by e-mail was also used effectively to keep each of them informed of developments on Montserrat between individual rotations in post. In the wider context of scientific advancement, the MVO has encouraged collaboration with external scientists who wish to undertake research at Soufriere Hills Volcano. Attempts have been made to ensure that such collaboration is done within a defined framework and following established protocols. A set of guidelines was drafted in 1997 to serve as operational procedures for researchers working at the MVO and for those wishing to collaborate with the MVO, Due to a number of problems, mainly relating to the evolving management structure at the MVO, described above, these protocols have not always been strictly adhered to or consistently implemented. However, most scientists who have worked on Montserrat have respected customary standards of scientific etiquette while on the island. Summing up A new institution, the Montserrat Volcano Observatory, has been established on Montserrat to monitor the Soufriere Hills volcano.
While the MVO owes its existence to the eruption that started in July 1995, it was not created from a standing start: it is founded on the Seismic Research Unit's long experience and knowledge of earthquakes and volcanic crises in the islands of the eastern Caribbean, and on the strength of UK volcanology, supported by the scientific project management experience of BGS. The nature of the present crisis in Montserrat, which changed from one that was initially expected to be short-lived, with high levels of public sensitization-to-threat, to one that became unusually protracted, has presented difficulties in sustaining both public anxiety and official caution. As a result, fresh new challenges were continually arising for the scientific team, both in respect of the behaviour of the volcano itself and in terms of necessary interactions with administrators and the public. Developing new capabilities for monitoring and instituting innovative mitigation measures within a besieged community, with a frequently changing cast of participants, has been at the heart of the effort. Many valuable lessons have been learned about the processes of interfacing with decision makers, key actors and stakeholders, and the public, some of which are being transferred through regional disaster preparedness organizations to other islands. For long periods in the Montserrat crisis, the micromanagement of hazards on a small island with an erupting volcano and a population intent on remaining, pushed the society and its politicians to use to the limit such guidance as the scientists could provide: the implications of this experience, and how it might redefine the scientific role in future volcanic crises, are worthy of more detailed appraisal. It is to the credit of all involved that the performance of the MVO in terms of the quality and timeliness of its scientific advice was as good as it turned out to be. The establishment of a local entity with responsibility for monitoring a single volcano, operating at arm's-length from the regional agency, represents a departure from the conventional way of tackling the problem. Individual observatory buildings and dedicated local teams have been set up before in the region, e.g. on St Vincent and on the French islands of Martinique and Guadeloupe, but these all operate under the aegis of bigger organizations. The exceptional circumstances in Montserrat (dual political hierarchy; protracted, slowly escalating eruption) were the major factors that contributed to the setting up of the MVO as a separate body. Responding to this particular crisis required an extended range of skills and staffing capacity, and thus demanded more resources than the SRU could readily provide. This raises the fundamental question (which will probably recur in future at other small volcanic island states in the region) as to whether an alternative scheme for resourcing the monitoring effort through the SRU could have been implemented. It is too facile to argue that there were traditional colonial political pressures making sure that management control remained in the hands of the home donor country, and that this is the basic explanation for the way things developed as they did. In the earliest stages of the crisis, attempts were made to configure a viable way forward for SRU as lead team in Montserrat, but a combination of singular personalities, unique institutional conditions and characteristics, and many other internal and external factors, were sufficient to divert the evolution of the MVO into other channels, at least for the time being. The SRU experience in Montserrat also underlines yet again the cliched but difficult challenge that volcanologists in such regions need to be able to convince their governments to make long-term investments in hazard mitigation measures and related resources so that good quality, cost-effective solutions are available when disaster strikes. In the present case, the solution to the issue of staffing a monitoring team for dealing with an ongoing crisis was found by combining, on a rotating basis, a corps of young scientists and PhD students with a kernel of experienced volcanologists from various institutions. While this has probably been beneficial overall and has been invaluable as a training regime for the young scientists, it does depart from the way volcanic eruptions have been handled in the region in the past, and at many volcanoes elsewhere in the world. This course of action has not been without drawbacks: for the Chief
THE MONTSERRAT VOLCANO OBSERVATORY Scientist or Director, additional people also mean the managing of extra overheads and resources (vehicles, radios, hot-suits, etc.), and increased anxiety about safety of personnel. The ambitions, goals and experience of some of those involved may be more oriented to research and data collection for that purpose, than to contributing to hazard assessment and mitigation. For students, the circumstances of a major volcanic crisis are not always ideal for fostering their best interests, either academically or in personal terms. Relationships can become very intense and strained working at close quarters in such conditions, and there is seldom the opportunity to relax and regroup that exists in the usual university environment. Furthermore, it is debatable to what extent any student should be immersed in the decision-making process of a live crisis that might result in a bad catastrophe, such as the fatalities of 25 June 1997: some individuals in those circumstances may be inclined to feel, quite unduly, that they share a significant responsibility for the outcome, as a member of the team. That having been said, the Soufriere Hills eruption crisis has clearly been a major stimulus for many young scientists, including, most importantly, several young Montserratians. In recent discussions, the two governments and the various agencies involved in Montserrat have been seeking to map out the long-term direction of the MVO and its eventual integration into regional monitoring. However, the resurgence of dome growth in November 1999 may delay these initiatives and immediate attention is being focused on drawing up a monitoring strategy for coping with yet another major episode of persistent eruptive activity. While this is the current preoccupation, it is foreseen that, in due course, the legacy of resources, knowledge and experience gained at the MVO during this eruption will pass locally into the custodianship of trained Montserratians and will enhance the future management of volcanic crises through a strengthened regional agency. In the meantime, with the eruption ongoing, the new institution has the opportunity to play an important role in advancing volcano monitoring and volcanology, while continuing to fulfil its responsibilities to the people and authorities on Montserrat. The people of Montserrat, so badly affected by the disaster in their island, have given wholehearted support to the MVO and their encouragement and expressions of confidence have been greatly appreciated by all the scientists involved. The MVO has benefited from the hard work and commitment of local staff and individuals who selflessly devoted their time and energies when families were under threat and homes were being lost to the volcano, and from the contributions of scientists and technical staff from the Seismic Research Unit of the University of the West Indies, in particular L. Lynch and R. Robertson. Funding for the MVO has been provided by the Government of Montserrat and by the UK government through the Department for International Development. For relevant authors, this paper is published by permission of the Director, British Geological Survey (NERC) and some images are reproduced by permission of the British Geological Survey NERC - all rights reserved [IPR/7-60]. References AHMAD, R. (ed.) 1996. Abstracts from The Second Caribbean Conference on Natural Hazards and Disasters: Science, Hazards And Hazard Management, 9-12 October 1996, Kingston, Jamaica. Publication No. 1 Unit for Disaster Studies, Department of Geography and Geology, University of the West Indies, Jamaica, 26-41. ALLEN, A. G. 1996. The influence of local volcanic emissions on the atmospheric environment of Montserrat, West Indies (Abstract). Applied Geoscience 96, Warwick. AMBEH, W. B., LYNCH, L. L., CHEN, J. & ROBERTSON, R. E. A. 1998. Contemporary seismicity of Montserrat, West Indies (abstract). In: ALI, W., A. PAUL, A. & YOUNG ON, V. (eds) Transactions of the 3rd Geological Conference of the Geological Society of Trinidad and Tobago and the 14th Caribbean Geological Conference, Port of Spain, Trinidad, Vol. 1, 1-2. ASPINALL, W. P. & COOKE, R. M. 1998. Expert judgement and the Montserrat Volcano eruption. In: MOSLEH, A. BARI, R. A. (eds) Proceedings of the 4th International Conference on Probabilistic Safety Assessment and Management PSAM4, 13-18 September 1998, New York City, USA, Vol. 3, 2113-2118.
89
ASPINALL, W. P. & Woo, G. 1994. An impartial decision-making procedure using expert judgement to assess volcanic hazards. Atti dei Convegni Lincei, Rome, 112, 211-220. ASPINALL, W. P., SIGURDSSON, H. & SHEPHERD, J. B. 1972. Eruption of Soufriere Volcano on St. Vincent Island 1971-1972. Science, 181, 117-124. ASPINALL, W. P., LYNCH, L. L., ROBERTSON, R. E. A., ROWLEY, K. C, SPARKS, R. S. J., VOIGHT, B. & YOUNG, S. R. 1998a. The Soufriere Hills Eruption, Montserrat, British West Indies: Introduction to Special Section, Part 1. Geophysical Research Letters, 25, 3387. ASPINALL, W. P., MILLER, A. D., LYNCH, L. L., LATCHMAN, J. L., STEWART, R. C., WHITE, R. A. & POWER, J. A. 1998b. Soufriere Hills eruption, Montserrat 1995-1997: volcanic earthquake locations and fault plane solutions. Geophysical Research Letters, 25, 3397-3400. BARCLAY, J., RUTHERFORD, M. J., CARROLL, M. R., MURPHY, M. D., DEVINE, J. D., GARDNER, J. & SPARKS, R. S. J. 1998. Experimental phase equilibria constraints on preemptive storage conditions of the Soufriere Hills magma. Geophysical Research Letters, 25, 3437-3650. BONADONNA, C., MAYBERRY, G. C., CALDER, E. S. ET AL. 2002. Tephra
fallout in the eruption of Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 483-516. BOUDON, G., VlLLEMANT, B., KOMOROWSKI, J-C, ILDEFONSE, P. & SEMET,
M. P. 1998. The hydrothermal system at Soufriere Hills Volcano, Montserrat (West Indies): characterisation and role in the ongoing eruption. Geophysical Research Letters, 25, 3693-3696. CALDER, E. S., COLE, P. D., DADE, W. B. ET AL. 1999. Mobility of pyroclastic flows and surges at the Soufriere Hills Volcano, Montserrat. Geophysical Research Letters, 26, 537-540. CLAY, E., BARROW, C., BENSON, C., DEMPSTER, J., KOKELAAR, P., PILLAI, N. & SEAMAN, J. 1999. An Evaluation of HMG's Response to the Montserrat Volcanic Emergency. Department for International Development Evaluation, London, Report EV635. COLE, P. D., CALDER, E. S., DRUITT, T. H., HOBLITT, R., ROBERTSON, R., SPARKS, R. S. J. & YOUNG, S. R. 1998. Pyroclastic flows generated by gravitational instability of the 1996-1997 lava dome of Soufriere Hills. Geophysical Research Letters, 25, 3425-3428. COLE, P. D., CALDER, E. S., SPARKS, R. S. J. ET AL. 2002. Deposits from dome-collapse and fountain-collapse pyroclastic flows at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoir, 21, 231-262. COOKE, R. M. 1991. Experts in Uncertainty. Oxford University Press, Oxford. CURNOW, R. N. 1999. Unfathomable nature and Government policy (the address of the President of the Royal Statistical Society). The Statistician, Journal of the Royal Statistical Society, Series D 48(4), 463-476. DEVINE, J. D., MURPHY, M. D., RUTHERFORD, M. J. ET AL. 1998a. Petrologic evidence for pre-eruptive pressure-temperature conditions, and recent reheating, of andesitic magma erupting at the Soufriere Hills Volcano, Montserrat, West Indies. Geophysical Research Letters, 25, 3669-3672. DEVINE, J. D., RUTHERFORD, M. J. & GARDNER, J. E. 1998b. Petrologic determination of ascent rates for the 1995-1997 Soufriere Hills Volcano andesitic magma. Geophysical Research Letters, 25, 3673-3676. DRUITT, T. H., CALDER, E. S., COLE, P. D. ET AL. 2002. Small-volume, highly mobile pyroclastic flows formed by rapid sedimentation from pyroclastic surges at Soufriere Hills Volcano, Montserrat: an important volcanic hazard. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 263-279. EDMONDS, M., PYLE, D. M. & OPPEHEIMER, C. 2001. A model for degassing at the Soufriere Hills volcano, Montserrat, West Indies, based on geochemical data. Earth and Planetary Science Letters, 186, 159-173. FISKE, R. S. 1984. Volcanologists, journalists, and the public: a tale of two crises in the Eastern Caribbean. In: Explosive Volcanism: Inception, Evolution and Hazards. National Research Council (US) Geophysics Study, National Academy Press, Washington DC, 170-176. HAMMOUYA, G., ALLARD, P., JEAN-BAPTISTE, P., PARELLO, F., SEMET, M. P. & YOUNG, S. R. 1998. Pre- and syn-eruptive geochemistry of volcanic gases from Soufriere Hills of Montserrat, West Indies. Geophysical Research Letters, 25, 3685-3688. HARFORD, C. L., PRINGLE, M. S., SPARKS, R. S. J. & YOUNG, S. R. 2002. The volcanic evolution of Montserrat using 40 Ar/ 39 Ar geochronology.
90
W. P. ASPINALL ET AL.
In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 93-113. HEINRICH, P., GUIBORG, S., MANGENEY, A. & ROCHE, R. 1999. Numerical modeling of a landslide-generated tsunami following a potential explosion of the Montserrat volcano. Physics and Chemistry of the Earth (A), 24(2), 163-168. HERD, R., CALDER, E., BARCLAY, J. ET AL. 1998. Methods used to survey the lava dome and associated products at the Soufriere Hills Volcano, Montserrat. Montserrat Volcano Observatory Open File Report 98/16. JACKSON, P., SHEPHERD, J. B., ROBERTSON, R. E. A. & SKERRITT, G. 1998. Ground deformation studies at Soufriere Hills volcano, Montserrat I: Electronic distance meter studies. Geophysical Research Letters, 25, 3409-3412. KOKELAAR, B. P. 2002. Setting, chronology and consequences of the eruption of Soufriere Hills Volcano, Montserrat (1995-1999). In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 1-43. KOMOROWSKI, J-C., NORTON, G., BOUDON, G. ET AL. 2000. Field-relationships and texture of the striking debris-avalanche deposit from the 1997 flank-collapse of Soufriere Hills Volcano, Montserrat (abstract). Geoscience 2000. LEE, W. H. K. 1989. PC-SEIS: a toolbox for seismic data acquisition, processing, and analysis. In: LEE, W. H. K. (ed.) IASPEI Software Library, Vol. I. LEE, W. H. K. & VALDES, C. M. 1989. HYPO71PC. In: LEE, W. H. K. (ed.) IASPEI Software Library, Vol. I, 203-236. LOUGHLIN, S. C., BAXTER, P. J., ASPINALL, W. P., DARROUX, B., HARFORD, C. L. & MILLER, A. D. 2002«. Eyewitness accounts of the 25 June 1997 pyroclastic flows and surges at Soufriere Hills Volcano, Montserrat, and implications for disaster mitigation. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 211-230. LOUGHLIN, S. C., CALDER, E. S., CLARKE, A. B. ET AL. 2002/7. Pyroclastic flows and surges generated by the 25 June 1997 dome collapse, Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 191-209. MATTIOLI, G. S., DIXON, T. H., FARINA, F., HOWELL, E. S., JANSMA, P. E. & SMITH, A. L. 1998. GPS measurement of surface deformation around Soufriere Hills Volcano, Montserrat from October 1995 to July 1996. Geophysical Research Letters, 25, 3417-3420. MILLER, A. D., STEWART, R. C., WHITE, R. A. ET AL. 1998. Seismicity associated with dome growth and collapse at the Soufriere Hills Volcano, Montserrat. Geophysical Research Letters, 25, 3401-3404. MONTSERRAT VOLCANO OBSERVATORY TEAM 1997. The Ongoing Eruption in Montserrat. Science, 276, 371-372. MURPHY, M. D., SPARKS, R. S. J., BARCLAY, J. ET AL. 1998. The role of magma mixing in triggering the current eruption at the Soufriere Hills Volcano, Montserrat, West Indies. Geophysical Research Letters, 25, 3433-3436. MURPHY, M. D., SPARKS, R. S. J., BARCLAY, J., CARROLL, M. J. & BREWER, T. S. 2000. Remobilization of andesite magma by intrusion of mafic magma at the Soufriere Hills volcano, Montserrat, West Indies. Journal of Petrology, 41, 21-42. MURRAY, T. L., EWERT, J. W., LOCKHART, A. B. & LAHUSEN, R. G. 1996. The integrated mobile volcano-monitoring system used by the Volcano Disaster Assistance Program (VDAP). In: SCARPA, R. & TILLING, R. I. (eds) Monitoring and Mitigation of Volcano Hazards, Springer, Berlin, 315-362. NEUBERG, J., BAPTIE, B., LUCKETT, R. & STEWART, R. 1998. Results from the broadband seismic network on Montserrat. Geophysical Research Letters, 25, 3661-3664. NORTON, G. E. 1997. Summary of the environmental effects of volcano emissions, Soufriere Hills volcano, Montserrat. Montserrat Volcano Obervatory Open File Report 97/15. NORTON, G. E., WATTS, R. B., VOIGHT, B. ET AL. 2002. Pyroclastic flow and explosive activity at Soufriere Hills Volcano, Montserrat, during a period of virtually no magma extrusion (March 1998 to November 1999). In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 467-481.
OFFICE OF SCIENCE AND TECHNOLOGY 1997. The Use of Scientific Advice in Policy Making. Department of Trade and Industry, London. OFFICE OF SCIENCE AND TECHNOLOGY 1998. The Use of Scientific Advice in Policy Making: Implementation of the Guidelines. Department of Trade and Industry, London. OPPENHEIMER, C., FRANCIS, P. & MACIEJEWSKI, A. J. H. 1998. Spectroscopic observation of HC1 degassing from Soufriere Hills Volcano, Montserrat. Geophysical Research Letters, 25, 3681-3684. PATTULLO, P. 2000. Fire from the Mountain. Constable, London. ROBERTSON, R. E. A., ASPINALL, W. P., HERD, R. A.. NORTON. G. E., SPARKS, R. S. J. & YOUNG, S. R. 2000. The 1995-1998 eruption of the Soufriere Hills volcano, Montserrat, WI. Philosophical Transactions of the Royal Society, London, Series A 358(1770), 1619-1637. ROOBOL, M. J. & SMITH, A. L. 1998. Pyroclastic stratigraphy of the Soufriere Hills Volcano, Montserrat: Implications for the present eruption. Geophysical Research Letters. 25. 3393-3396. SHEPHERD, J. B., TOMBLIN, J. F. Woo, D. A. 1971. Volcano-seismic crisis in Montserrat. West Indies 1966-67. Bulletin Volcanologique. 35. 143-163. SHEPHERD, J. B., ASPINALL, W. P.. ROWLEY. K. C., PEREIRA, J.. SIGURDSSON, H., FISKE, R. S. & TOMBLIN, J. F. 1979. The eruption of Soufriere Volcano, St. Vincent April-June 1979. Nature, 282, 24-28. SHEPHERD, J. B., HERD, R. A.. JACKSON.. P. & WATTS, R. 1998. Ground deformation measurements at the Soufriere Hills Volcano, Montserrat II: Rapid static GPS measurements. June 1996-June 1997. Geophysical Research Letters, 25. 3413-3416. SPARKS, R. S. J. & YOUNG. S. R. 2002. The eruption of Soufriere Hills Volcano, Montserrat (1995-1999): overview of scientific results. In: DRUITT, T. H. & KOKELAAR. B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society. London. Memoirs, 21, 45-69. SPARKS. R. S. J., YOUNG. S. R., BARCLAY. J. ET AL. 1998. Magma production and growth of the lava dome of the Soufriere Hills Volcano. Montserrat, West Indies: November 1995 to December 1997. Geophysical Research Letters, 25, 3421-3424. SPARKS, R. S. J., BARCLAY, J., CALDER. E. S. ET AL. 2002 Generation of a debris avalanche and violent pyroclastic density current on 26 December (Boxing Day) 1997 at Soufriere Hills Volcano. Montserrat. In: DRUITT, T. H. & KOKELAAR. B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21. 409-434. TAZIEFF, H. 1980. Letter to the Editor. Journal Volcanology and Geothermal Research, 8, 3-6. THE ROYAL SOCIETY 1999. Science and Society: a response to the inquiry by the House of Lords Science and Technology Select Committee. Statement 12 99, London. VOIGHT, B., HOBLITT, R. P., CLARKE. A. B., LOCKHART, A. B.. MILLER. A. D.. LYNCH, L. & McMAHON. J. 1998. Remarkable cyclic ground deformation monitored in real-time on Montserrat, and its use in eruption forecasting. Geophysical Research Letters, 25. 3405-3408. VOIGHT, B.. SPARKS, R. S. J., MILLER. A. D. ET AL. 1999. Magma flowinstability and cyclic activity at Soufriere Hills Volcano. Montserrat. B.W.I. Science. 283. 1138-1142. WADGE, G. & ISAACS, M. C. 1987. Volcanic Hazards from Soufriere Hills Volcano, Montserrat, West Indies. Report to the Government of Montserrat and the Pan Caribbean Disaster Preparedness and Prevention Project. WADGE, G. & ISAACS. M. C. 1988. Mapping the volcanic hazards from the Soufriere Hills Volcano, Montserrat. West Indies using an image processor. Journal of the Geological Society, London, 145. 541-551. WADGE, G., JACKSON, P., BOWER, S. M., WOODS, A. W. & CALDER. E. 1998. Computer simulations of pyroclastic flows from dome collapse. Geophysical Research Letters. 25. 3677-3680. WATTS, R. B., HERD, R. A., SPARKS. R. S. J. & YOUNG, S. R. 2002. Growth patterns and emplacement of the andesitic lava dome at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR. B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs. 21. 115-121. YOUNG, S., SPARKS, S., ROBERTSON. R., LYNCH, L. & ASPINALL, W. 1997. Eruption of Soufriere Hills Volcano in Montserrat Continues. EOS, Transactions, American Geophysical Union, 78(38), 404. 408-409. YOUNG, S. R., FRANCIS, P. W., BARCLAY, J. ET AL. 19980. Monitoring SO2 emission at the Soufriere Hills volcano: Implications for changes in eruptive conditions. Geophysical Research Letters. 25. 3681-3684.
THE MONTSERRAT VOLCANO OBSERVATORY YOUNG, S. R., SPARKS, R. S., ASPINALL, W. P., LYNCH, L., MILLER, A. D., ROBERTSON, R. E. A. & SHEPHERD, J. B. 1998b. Overview of the eruption of Soufriere Hills volcano, Montserrat, 18 July 1995 to December 1997. Geophysical Research Letters, 25, 3389-3392.
91
YOUNG, S. R., VOIGHT, B., ROWLEY, K., SPARKS, R. S. J., ROBERTSON, R. E. A., LYNCH, L. L. & ASPINALL, W. P. 1998c. The Soufriere Hills Eruption, Montserrat, British West Indies: Introduction to Special Section, Part 2. Geophysical Research Letters, 25, 3387.
The volcanic evolution of Montserrat using 40Ar/39Ar geochronology C. L. HARFORD1, M. S. PRINGLE2, R. S. J. SPARKS1 & S. R. YOUNG3 1
Department of Earth Sciences, Bristol University, Bristol BS8 1RJ, UK (e-mail:
[email protected]) 2 Scottish Universities Research and Reactor Centre, East Kilbride G75 OQF, UK Montserrat Volcano Observatory, Mongo Hill, Montserrat, West Indies
Abstract: 40 Ar/ 39 Ar dating has facilitated a substantial reinterpretation of the volcanic evolution of Montserrat. Three volcanic centres with non-overlapping volcanic activity are identified: Silver Hills (c. 2600 to c. 1200ka); Centre Hills (at least c. 950 to c. 550 ka); South Soufriere Hills-Soufriere Hills (at least c. 170ka to present). The geochronological data show that old xenocrysts are common in the porphyritic andesite, implying that reliable ages are best obtained by dating the groundmass. Soufriere Hills evolved from early eruptions dominated by two-pyroxene andesite to eruptions of hypersthene-hornblende andesite at c. 1 lOka. Between the two varieties of andesite there was an interlude of mafic volcanism at c. 130ka to form South Soufriere Hills. There is evidence of tectonic uplift of early products of the complex along with older submarine volcanic rocks. Consideration of stratigraphy and age data indicates that only a proportion of the dome-forming eruptions are recorded as domes in the geological record. Older products are removed from the subaerial edifice by sector-collapse events. The timeaveraged eruption rate of the South Soufriere Hills-Soufriere Hills centre is estimated at 0.005 m3 s"1 (c. 0.15 km3 ka" 1 ) (dense rock equivalent). The ongoing eruption is very similar in style to previous activity at Soufriere Hills, and future activity is likely to pose similar hazards. Soufriere Hills have been characterized by alternations of periods of enhanced activity and periods of dormancy, both lasting of the order of 104 years. During periods of elevated activity several major dome-forming eruptions are separated by quiescent interludes lasting less than c. 103 years. The ongoing eruption may mark the onset of a fourth period of enhanced volcanic activity at Soufriere Hills.
Geochronology is important for understanding the evolution of young volcanic systems. High-quality age data can help constrain trends of spatial, geochemical and petrological evolution of volcanism. Such data also allow estimation of rates of volcanic processes and magmatic evolution, including eruption rates and fluctuations thereof. Work at Crater Lake, Oregon (Bacon & Lanphere 1990), Mount Adams, Washington (Hildreth & Lanphere 1994), Tatara-San Pedro complex, Chile (Singer et ai 1997) and Santorini Volcano, Greece (Druitt et al 1999) illustrates how a detailed and high-precision geochronological framework can enhance studies of volcanic systems. This paper presents a study of the evolution of volcanism on the island of Montserrat, West Indies, using 40Ar/ 39 Ar geochronology. Methods have been developed only relatively recently to provide precise age information over the age range typical of many volcanoes (tens to hundreds of thousands of years). Radiocarbon dating is only applicable to deposits younger than c. 45 ka, due to the short half-life of 14C. In addition, the radiocarbon timescale is precisely calibrated only back to 21 ka, and is susceptible to contamination problems, particularly in older samples. Furthermore, only deposits containing carbon may be dated, ruling out lava flows, domes and some pyroclastic deposits. K-Ar geochronology, in particular the 40 Ar/ 39 Ar variant, has been the most successfully applied radiometric technique on volcanic rocks less than several hundred thousand years old. Here 40 Ar/ 39 Ar geochronology is applied to a relatively difficult problem, a comparatively low-K volcanic system that is susceptible to alteration in a tropical environment. The island of Montserrat is located in the northern part of the Lesser Antilles island arc, and is over 16km long (north-south) and 10km wide (east-west). Montserrat is built almost exclusively of volcanic rock. The latest manifestation of the island's extrusive history has been the growth of an andesitic lava dome from 1995 to present at Soufriere Hills Volcano, with associated dome-collapse pyroclastic flows and explosions (Robertson et al 2000; Young et al 1998). The island's past volcanic activity is of importance for interpreting this ongoing and any future activity. Previous investigations on Montserrat have suggested that the island has evolved over the last 4 million years with at least six volcanic centres (Fig. 1). MacGregor (1938) carried out the first geological and petrological work on Montserrat. Rea (1970, 1974) developed MacGregor's work and produced a geological map. Two radiocarbon dates and four conventional K-Ar dates (three of the latter were later published in Briden et al. 1979) facilitated Rea's
(1970, 1974) reinterpretation of the geological evolution of the island (Fig. 1). However, the results in Briden et al. (1979) on other islands of the Lesser Antilles have since been found to be inaccurate (Carlut et al. 2000). A further eight conventional K-Ar dates were published by Le Gall et al. (1983), although problems were acknowledged for some of the dates obtained. The precision of all the existing K-Ar ages is poor (Icr = 7 to 41%). The work presented here confirms that significant uncertainties, often even greater than the reported analytical uncertainties, are associated with conventional K-Ar geochronological studies. In order to aid volcanic hazard investigations Baker (1985) and Wadge & Isaacs (1988) studied, and obtained radiocarbon dates for, the youngest deposits of Soufriere Hills. Roobol & Smith (1998) carried out detailed work on the pyroclastic stratigraphy of Soufriere Hills and obtained further radiocarbon ages. The radiocarbon ages cover only the uppermost of the three major units of the pyroclastic stratigraphy and extend back only as far as 31 560 0 years BP. Roobol & Smith (1998) found no carbon to date in the lower stratigraphic units of Soufriere Hills. The geology of Montserrat Montserrat comprises three major massifs: Silver Hills in the north, Centre Hills in the centre, and South Soufriere Hills-Soufriere Hills in the south (Figs 1 and 2). In addition, Garibaldi Hill and St George's Hill form two smaller, isolated topographic highs. The three major massifs are geomorphologically distinct in terms of maturity of erosion (Fig. 2). This prompted Davis (1926) to deduce correctly their relative ages from oldest to youngest as Silver Hills, Centre Hills, South Soufriere Hills-Soufriere Hills. The island's interior, with the exception of areas affected by the ongoing eruption, is densely vegetated. Exposures are thus largely limited to coastal cliffs, roadcuts, and occasional steep interior cliffs. Andesitic deposits produced by dome-forming eruptions dominate the island, although South Soufriere Hills are predominantly basaltic to basaltic-andesite in composition. Major products on Montserrat are remnants of andesitic domes, andesitic breccias representing dome talus, block-and-ash-flow deposits originating from dome collapse, lahar deposits, debris-avalanche deposits, and usually thin and subordinate tephra-fall deposits. There are areas of hydrothermally altered rocks, including active fumarole fields (soufrieres) related to Soufriere Hills Volcano (Fig. 2).
DRUITT, T. H. & KOKELAAR, B. P. (eds) 2002. The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 93-113. 0435-4052/02/S15 © The Geological Society of London 2002.
93
94
C. L. HARFORD ET AL.
Fig. 1. Shaded relief maps, derived from a digital elevation model (DEM), showing evolution of volcanism according to different authors. This work shows that volcanism has migrated from north to south, in contrast to previous interpretations involving oscillation of volcanism. Montserraf s three major massifs show a marked decrease in erosional maturity from Silver Hills to Centre Hills to South Soufriere Hills-Soufriere Hills.
Silver Hills Silver Hills are deeply eroded. They comprise massive, predominantly andesitic lava, interpreted as parts of eroded lava domes, and sequences of volcaniclastic beds, which once would have formed fan deposits around the lava domes. Extensive areas are hydrothermally altered, for example at Yellow Hole. Debrisavalanche deposits are prominent, for example at Little Bay.
Centre Hills Centre Hills are also significantly eroded. High coastal cliffs surround the hills. The cliffs are significantly higher on the west coast (1 km) pyroclastic flows were generated. There is evidence that most, but not all, large to major dome collapses at Montserrat were related to pulses in extrusion rate. Dome extrusion rates were 4-6 m3 s-1 in the two or three days before the 29 July 1996 collapse and most other large to major collapses were preceded by one to five days of enhanced dome-growth rates (Table 1: e.g. December-January 1997 collapses; 25 June 1997; 3 August 1997). This is established by volumetric data for some cases (Figs 7 and 10), and interpreted indirectly for others; for example, by changes in tilt or by the onset of intense swarms of hybrid earthquakes. For much of the eruption, the occurrence of hybrid earthquakes appears to have been directly associated with the onset of instability of the lava dome. The tilt data identify significant pressure pulses spaced five to seven weeks apart, with the best examples beginning 22 June 1997 (just prior to the 25 June collapse) and 31 July 1997 (Voight et al. 1999). The timing of several other collapses (e.g. 21 September, 4 November and 26 December 1997) also fit this pattern, although tilt data were then unavailable. The tilt cycles have been interpreted in terms of shallow pressurization (5 m were particularly prevalent in the Harris-Bethel area where slope was reduced and the flow slowed down. Hydrothermally altered blocks in the block-and-ash
flow deposits were probably entrained from the substrate (see Cole et al. 2002), but may have included blocks of dome rock altered by fumarolic activity. Slightly vesicular andesite blocks were present, particularly in the thin margins of the coastal block-and-ash flow deposits. Low-density debris, such as tree trunks and domestic gas canisters, were also abundant at flow margins. Grain-size analyses of three typical block-and-ash flow deposit samples show the poor sorting ( > 2.5) and the coarse nature of these deposits (median diameter < 1 ), compared to the other deposits (Fig. 16). Further accounts of block-and-ash flow deposits, including 25 June deposits, are given by Cole et al. (2002).
Pyroclastic surge deposits Extensive pyroclastic surge deposits (Fig. 1) defined a broad fanshape emanating northwards from the dome. Fine-grained surge deposits were also observed in the upper part of Tyre's Ghaut, a few hundred metres west of Mosquito Ghaut. These presumably were deposited by a surge cloud that expanded rapidly almost from source and surmounted an intervening ridge. Site 1 (Figs 1 and 17) shows two units of broadly equal thickness each comprising massive, fines-rich ash and lapilli. It is believed that each unit represents the surge deposit from flow pulses 2 and 3. A very thin layer of ash at the base may have been deposited from the poorly developed surge and/or lofting ash clouds associated with flow pulse 1. Alternatively, the ash layer may have been deposited from earlier pyroclastic flows, for example the flow on 17 June. The foot of Windy Hill (site 2; Figs 1 and 17) was affected only by a surge associated with block-and-ash flow pulse 3 and two layers were identified. The deposits varied in thickness, filling minor
204
S. C. L O U G H L I N ET AL.
PYROCLASTIC FLOWS OF 25 JUNE 1997
205
On the coastal plain, distinctive fine-grained deposits occurred at the distal margins of the block-and-ash flow lobes. They were typically up to 0.5m thick, thinning gradually towards the margins, and they lacked blocks or other coarse debris. They did not extend farther than 100 m beyond the flow deposits. It could be argued that these distal deposits were simply from fine-grained block-and-ash flows, the inevitable result of slow transport of fine-grained material, but the well sorted nature of the deposits suggests that they were surge-related.
Surge-derived pyroclastic flow deposits
Fig. 16. Median grain size (Md ) versus sorting ( =( 84- 16)/2) for the 25 June 1997 pyroclastic deposits. Tie lines link samples from the same location. Filled squares = block-and-ash flow deposits. Circles = fines-poor basal layer of the surge deposits. Open squares = fines-rich upper layer of the surge deposits. Triangles = surge-derived pyroclastic flow deposits.
depressions (e.g. furrows and hollows tens of centimetres deep) in the pre-existing topography. Sections here and at site 3 (Figs 1 and 17) typically showed a lower fines-poor layer of friable medium to coarse ash and lapilli that thickened into depressions and thinned onto highs. In ploughed fields this lower layer thinned and pinched out onto the ridges between furrows oriented approximately perpendicular to the flow direction. This layer could locally be subdivided further into a lower, finer-grained part rich in sheared soil clasts and vegetation fragments, and an upper, coarser part, rich in small fragments of dome rock (rare clasts up to 10cm). The upper layer was a brown, fines-rich ash layer of almost constant thickness that mantled the topography. Gas-escape structures were common within the upper layer (pipes of fines-poor ash and lapilli) and were usually rooted on fragments of carbonized organic debris or coarse clasts. Plumes of smoke and gases could be seen rising from small vents in the surge deposits in the days after the 25 June event, and were probably caused by burning vegetation. The deposits were typical of pyroclastic surge deposits formed following other dome collapses during this eruption (Cole et al. 2002). The upper fines-rich layer usually mantled the ground and normal grading of coarse clasts occurred locally (e.g. at site 3; Figs 1 and 17). Fines-poor samples had a median diameter (Md ) of -0.5 to 3 whereas fines-rich surge deposits had a Md of 2.5 to 4 (Fig. 16). Blocks 6
>3.5
N/C
Tar River valley
Deposit volumes are non-DRE and include the associated pyroclastic surge deposits. Duration measured from seismic records. N/C, not calculated.
Mosquito Ghaut was inundated in a similar fashion as small rockfalls first overtopped English's Crater above the valley on 13 June 1997. These were followed by progressively longer flows (1.5km runout) over the next few days. On 17 June a block-and-ash flow travelled 4 km down Mosquito Ghaut (Fig. 4a). Eight days later, on 25 June, a flow with three block-and-ash flow pulses travelled down Mosquito Ghaut to within 50m of the sea, 6.8km from the dome (Fig. 4b) (Loughlin et al 2002a,b). The first minor rockfalls occurred down Fort Ghaut in June 1996. A year later, dome growth concentrated on the western side formed rockfalls in Fort Ghaut on 14 June 1997, and by 16 June pyroclastic flows had reached Gages Lower Soufriere, 2km west of the dome. Dome-collapse flows on 31 June 1997 had runouts of 3.5km and reached the eastern margins of Plymouth for the first time (Fig. 1). A sustained dome collapse on 3 August 1997 formed pyroclastic flows that travelled 4 km to the west, impacting large parts of Plymouth (Fig. 1). This was followed, between 4 and 12 August, by a series of 13 Vulcanian explosions (Druitt et al. 2002b), each involving generation of fountain-collapse pumice-andash flows (Fig. 5). Dome growth after 12 August filled the crater formed by these explosions and produced numerous block-and-ash flows onto Farrell's Plain (around Farrell's Yard; Fig. 1) to the north. On 21 September 1997, large-volume block-and-ash flows travelled down Tuitt's Ghaut impacting the airport 6km to the NE (Fig. 6). This was followed by a second series of 75 Vulcanian explosions from 22 September to 21 October. All but one of the explosions generated pumiceous pyroclastic flows by fountain collapse. The flows travelled simultaneously down several valleys, including Tuitt's Ghaut, Tar River valley, Fort Ghaut, White River valley and White's Ghaut. A map of pyroclastic flow deposits formed by a typical Vulcanian explosion on 18 October 1997 is shown in Figure 7. The runouts of fountain-collapse flows were generally between 3 and 6 km. Rapid dome growth at the end of October 1997 once again filled the Vulcanian explosion crater. Growth switched to the southern margin above Galway's Wall at the end of October. Extensive dome collapses occurred on 4 and 6 November 1997, forming
block-and-ash flows in the White River valley and extending the fan on the southern coast (Fig. 3d). In November and December 1997, dome growth was once again focused on the southern side and on 26 December a sector collapse involving both the lava dome and the flanks of the old volcanic edifice formed a debris avalanche and associated high-velocity pyroclastic density current (Ritchie et al. 2002; Sparks et al. 2002; Voight et al 2002). Dome growth ceased 2.5 months later, in March 1998. Block-and-ash flows occurred sporadically following cessation of dome growth (Norton et al. 2002). These formed by both dome collapse and explosions. Flows formed during the post-domegrowth period were quite varied in both volume and nature. One of the largest dome collapses took place on 3 July 1998 and produced block-and-ash flows that moved down the Tar River valley and reached the sea, as well as pyroclastic surges that spilled northeastwards 300 m out of the valley where they impacted the village of Long Ground (Fig. 1). Collapses on 12 November 1998 down Fort Ghaut were characterized by relatively fine-grained, valleyconfined pyroclastic flows with few blocks >2m in size. Pyroclastic surges developed only within the initial 1 km, and blocks in the distal regions, 5km from the dome, were typically 3m to locally 1 m above the upper surface of the surrounding deposit.
Lithology of block-and-ash flow deposits In the available sections through them, block-and-ash flow deposits were generally massive and ungraded, although grading was observed in some exposures (Figs 10 and 11). At one locality in the 21 September 1997 block-and-ash flow deposit, the lower
half of the 3 m thick deposit was reverse graded and the upper half was normally graded. This resulted in a distinctive coarse central zone (Figs lOa and 11c), and similar grading was observed in other sheet-like deposits from large dome collapses (Fig. 11b; e.g. 20 Jan 1997). Gas-escape structures were abundant within the deposits (Fig. 11b & c). The block-and-ash flow deposits were poorly sorted ( o > 2), containing fragments that ranged from fine ash to dense blocks up to 15m in size. Grain-size analyses of the matrix fraction (3 cm) for block-and-ash flow deposits formed by (a. b) dome-collapse and (c-f ) pumice-and-ash flow deposits.
from narrow, channelized and lobate to thin sheets. Many deposits were valley-confined, with flat upper surfaces and abrupt frontal terminations. The largest-volume block-and-ash flow deposits formed thin sheets with diffuse tapering margins, and had a pattern of narrow ridges and furrows on their upper surface oriented approximately in the flow direction. Visual observations of block-and-ash flows, rounding of the edges and corners of blocks, and slickensided surfaces of large blocks (Grunewald et al. 2000) indicate that many of the blocks in the flows were tumbling. Particle interactions were thus important during large parts of the runout. The larger dome collapses formed block-and-ash flow deposits of quite variable thickness, and many deposits were massive in section. The local reverse-to-normal grading of some block-and-ash flow deposits might have related to amalgamated block-and-ash flow deposits formed during a single collapse. Low-density accidental material was concentrated at the margins of block-and-ash flow deposits.
Pyroclastic flows associated with Vulcanian explosions Observations Pyroclastic flows resulted from fountain collapse associated with 13 Vulcanian explosions between 4 and 12 August 1997 and 74 Vulcanian explosions between 22 September and 21 October 1997 (Table 1). Each explosion lasted a few tens of seconds and the pyroclastic flows generally formed from a single collapse pulse (Druitt et al. 2002/7). Collapse occurred from 300 to 650m above the crater rim at 950m a.s.l. In each case, an initial surge cloud travelled at initial velocities of 6 0 m s - ' near the collapsing fountain and rapidly decelerated before lofting (Druitt et al. 2002b). Following this, dense pumice-and-ash flows emerged from beneath the decelerating cloud, where they were channelled along the valleys and then spread out onto unconfined plains at about 1 0 m s - 1 (Figs 7 and 14a). Ash plumes that developed above the pumice-andash flows were weak in comparison (Fig. 14b) to those formed above dome-collapse flows of the same runout length.
Pyroclastic flows formed by Vulcanian explosions in August 1997 formed thin veneers of coarse pumice and lithics around the upper regions of the valleys draining the volcano (Fig. 5). Finegrained pyroclastic surge deposits were extensive in the first 2 km around the site of fountain collapse. At distances greater than 2km. pyroclastic surge deposits were poorly developed or absent (Figs 5 and 7).
Effects
of pumice-and-ash flows
Pumice-and-ash flows showed similar effects to block-and-ash flows where, in the final 0.5 km of runout, some trees and telegraph poles that were engulfed remained standing (Fig. 14c). Such a feature was observed in a number of locations, e.g. Fort Ghaut. Spanish Point and White River valley. Pumice-and-ash flows did not burn through the trunks of buried trees, although small branches were partly charred on the outer surfaces and other flammable material, e.g. paper and foam mattresses, was also partially charred. Bark removal on engulfed trees occurred up to 40cm above the upper surface of the deposit (Fig. 14c). Pumice-and-ash flows that impacted houses caused little structural damage. They flowed in through open doors and windows, without causing any structural damage to walls. In some cases, wire mesh fencing contained or deflected the pumice-and-ash flows within the last few hundred metres before the distal terminations.
Morphology of pumice-and-ash flow deposits Pumice-and-ash flow deposits were sinuous and lobate both within valleys and on relatively flat unconfined plains. Individual lobes were up to 300m long and up to 50m wide (Figs 7 and 14a.d). The thicknesses of individual flow-deposit lobes were typically 20% fine ash. The lower layers were coarser (Mdo = —0.2 to 2.7) than the upper layers (Mdo = 1.8 to 3.5). Sorting coefficients ( o) of both the upper and lower layers have similar ranges of between 1 and 2.3 (Fig. 19). Weight percentage histograms demonstrate that the lower layers, poor in fine ash, generally contain coarser fragments than the overlying deposits (Fig. 18b).
DEPOSITS FROM PYROCLASTIC FLOWS
251
Fig. 16. (continued)
Towards the northern limits of the pyroclastic surge deposit, away from the centre of the valley, the lower layers thinned, became intermittent and richer in fine ash (from 2 to 18wt%), whereas the upper layers showed no systematic grain-size variation (Fig. 19c). The temperature of these pyroclastic surge deposits measured by thermocouple 15 days after emplacement was 326°C.
Deposits 1.7km NNE of the dome (adjacent to the Tar River Estate House) varied between 15 and 30cm in total thickness. The lowermost deposit was a lens, up to 6cm thick, poor in fine ash. Three separate cross-stratified layers above this varied from 4 to 20cm thick and each contained climbing dune structures with wavelengths of up to 30cm. Mantle-bedded fine ash layers up to 4cm thick containing accretionary lapilli were interbedded between the cross-stratified layers (Figs 17d and 18e).
Pyroclastic surge deposits of 3 July 1998 Pyroclastic surges formed during the dome collapse on 3 July 1998 (four months after the dome had stopped growing) impacted the whole of the Tar River valley. Seismic signals related to this dome collapse were sustained as it waxed and waned over at least 2.5 hours. Pyroclastic flows travelled initially in a southeasterly direction, impacting Roche's Mountain, and were then deflected obliquely across the Tar River valley to the NE, spilling out of the valley on its northern margin. The village of Long Ground was impacted by pyroclastic surges for the first time during the eruption (Fig. 1). Pumiceous ballistic blocks were observed on Roche's Mountain and sandblasting and erosion of trees was common 1.7 km east of the dome.
Synopsis Pyroclastic surges of variable extent were associated with many dome-collapse pyroclastic flows. They represent the part of the ash cloud overlying the parent block-and-ash flow that is denser than air. They appear to originate by expansion of gases and entrained air from the block-and-ash flows and were thus analogous to 'ash-cloud surges' described from other volcanoes (Boudon et al. 1993; Abdurachman et al 2000; Kelfoun et aL 2000). Tree felling and building damage associated with flows from the largest dome collapses were extensive in the first 2km of pyroclastic
252
P. D. COLE ET AL.
Fig. 17. (a) View of Tuitt's village following impact by the pyroclastic surges of the 21 September 1997 flows. Note the absence of structural damage to walls of houses and abundance of upstanding trees. Several houses retain partial roofs. (b) House in Streatham village impacted by pyroclastic surge on 25 June 1997. showing the accumulation of boulders 1 mm in the 12 May 1996 pyroclastic surge deposits and nearly 100wt% of clasts > l m m within lower layers of the 25 June 1997 pyroclastic surge deposits. In other cases, material eroded from prehistoric deposits or those formed earlier during the eruption was petrologically indistinguishable from the juvenile dome rock. The 21 September 1997 dome-collapse pyroclastic flow overran and entrained pumice from pumice-and-ash flow deposits formed during the August 1997 Vulcanian explosions. Accidental pumice typically occurred only in the margins of deposits, within 20m of the edge of the flow lobes. Coarse clasts (>3cm) within the central part of flow lobes were composed predominantly of dense dome rock (Fig. 6 inset). We observed that there was a systematic relationship between the volume of pyroclastic flows and the maximum slope angle on which the related deposits were emplaced. Rockfalls and small pyroclastic flows formed a 33 talus slope around the dome. Generally, as the flows increased in size and runout, the slope at which transition occurred between erosion and deposition decreased. Flows reaching between 1 and 2km were erosive on slopes > 10-15 and flows reaching the sea were erosive on slopes >6-82. In the Tar River valley two prominent benches showed no deposition on their steep, seaward-facing scarps (see Fig. 2). In the Tar River valley the largest flows often eroded substantial amounts of earlier deposits and thus the valley contains only a limited amount of preserved deposit. Development of coastal fans Two main coastal fans of pyroclastic flow deposits were formed during the 1995-1999 phase of the eruption. A fan was built first at the mouth of the Hot River (within the Tar River valley) on the east side of the volcano (Figs 1. 2 and 21a.b). An extensive coastal
Fig. 18. Selected measured sections through pyroclastic surge deposits for a number of different dome collapses between 1996 and 1999. See Figure 2 for location of samples from 12 May 1996 and 17 September 1996. TRV = Tar River valley. TREH = Tar River Estate House.
256 P- D. COLE ET AL.
^SB&E^^^^' S£SSES^SS MS3SH??wS for d js^sif zr^r^t r ™ """fr * depoS,lira,s,opesd^/;™S(«- *«ing ' ""»»« °f
October 1997 and H *!
?
™
Vulca
™an explosions in
valleys by acoustic sounding in July 1998. In the Tar River case e Pyr ClaStIC fan WaS traCed ab Ut vef9S900m OOaP water «m ofoffshore °,depth, with an estimated° volume t ooove, at least M J x I O m- confirming observations that substantial parts of pyroclastic flows entered the sea. The submarine deposits were up to 50m thick and deposition took place on slopes 9.1 10.7-12.2 9.1-10.7 10.7-12.2 ~7.6 12.2
Pyroclastic flows
Date
Ta W Tu M Ty G Wr
X
X
X
O
X
X
0
X
X
X
0
X
X
0
X
X
X
X
X
X
X
X
X
X
X
X
X
X
0
X
X
X
X
X
X
X
0
X
X
X
X
X
X
X
X
X
X
X
X
0
0
0
X
X
0
X
X
0
X
X
0
X
X
0
X
X
0
X
X
0
X
X
O
X
0
12.2 12.2 7.6 12.2§ 12.2 12.2 12.2-13.7 9.1-13.7 2 Hz) component is due to fountain collapse, ballistic impact, and pyroclastic flow. (b) Enlargement of the first 1.4 minutes of the same signal. Phase 1 is interpreted as the time for the ballistic blocks and collapsing fountain to first hit the ground. Pulsing of the 0.5-1 Hz component is attributed to resonance of seismic waves in the conduit. (c) Frequency spectra for the different phases of the explosion signal, as well as for the background seismicity. Two spectra are shown for phase 2: one prior to the intensity drop (arrow in (a)) and one after it. The lowfrequency component of the explosion signal is present throughout the explosion, whereas the high-frequency component occurs only in phase 2.
299
EPISODES OF CYCLIC EXPLOSIVE ACTIVITY Table 3. Characteristics of the explosion seismic signals Date
Time (local)
4 Aug. 4 Aug. 5 Aug. 5 Aug. 6 Aug. 6 Aug. 7 Aug. 7 Aug. 7 Aug. 8 Aug. 8 Aug. 1 1 Aug. 12 Aug.
06:30 16:43 04:45 16:57 04:02 14:36 00:34 12:05 21:55 10:32 20:51 11:38 10:12
22 Sep. 22 Sep. 22 Sep. 23 Sep. 24 Sep. 24 Sep. 24 Sep. 25 Sep. 25 Sep. 25 Sep. 26 Sep. 26 Sep. 27 Sep. 27 Sep. 27 Sep. 28 Sep. 28 Sep. 28 Sep. 29 Sep. 29 Sep. 29 Sep. 29 Sep. 30 Sep. 30 Sep. 1 Oct. 1 Oct. 1 Oct. 2 Oct. 2 Oct. 2 Oct. 4 Oct.
00:57 10:45 20:42 07:23 00:34 10:54 17:16 03:54 11:09 20:05 04:25 14:56 00:01 09:46 17:15 04:28 10:34 23:03 06:26 11:23 16:48 21:57 04:43 17:44 05:00 11:34 17:40 01:05 12:53 22:50 08:33
Duration phase 1 (s)
Max. vertical velocity explosion (nms - 1 )*
Duration phase 2 (s)
10
43960
360
18 19 17 19 18
101270 65560 89160 50410 84950 34500
300
Max vertical velocity py flows (nms-')t
330 270 300 240 300
12 10.6 10 8 10 9.6 6 14.5 8.5 14 14 17 17 17 20 22.2 10.4 17.5 12.9 23.5 18 7 34 16 4 14 18 7 17 15
31950 56125 24113 16932 25965 20763 8925 53240 17256 27772 13522 62372 38362 24208 14339 139868 13179 133280 10142 40622 29554 21468 20005 189880 1950 5442 41329 47944 29 551 30904
480 360 270 300
330 240 300 300 300 240 300 270 360 270 240 300 270 240 270 270 270 210 240 300 240 450
123235 155352 107300 40525 42514 64639 19651 103013 42964 65515 35585 50483 57430 45939 32940 78465 17927 115986 30784 29695 82990 54007 21610 75990 35951 32284 38498 47944 12119 161781
Date
Time (local)
Duration phase 1 (s)
Max. vertical velocity explosion (nms- 1 )*
Duration phase 2 (s)
4 Oct. 5 Oct. 5 Oct. 5 Oct. 6 Oct. 6 Oct. 6 Oct. 7 Oct. 7 Oct. 8 Oct. 8 Oct. 9 Oct. 9 Oct. 10 Oct. 10 Oct. 1 1 Oct. 12 Oct. 12 Oct. 13 Oct. 13 Oct. 14 Oct. 14 Oct. 14 Oct. 15 Oct. 15 Oct. 15 Oct. 15 Oct. 16 Oct. 16 Oct. 16 Oct. 16 Oct. 16 Oct. 17 Oct. 17 Oct. 17 Oct. 17 Oct. 18 Oct. 18 Oct. 19 Oct. 19 Oct. 20 Oct. 20 Oct. 21 Oct. 21 Oct.
18:27 02:53 10:41 18:41 02:44 10:42 17:50 04:06 16:02 03:47 15:10 03:03 12:32 04:13 18:40 17:57 07:55 22:24 09:32 15:24 01:36 13:48 23:16 05:47 08:33 14:50 22:20 02:51 06:35 09:44 14:20 18:48 04:01 12:35 16:05 23:18 06:48 15:17 05:13 21:27 05:04 15:13 11:39 19:02
19 13 4 14 21 14 5
5250 42141 5644 31789 41720 28892 9210
270 240 210 300 240 270 300
24 22 15 13 20 16 17 15 14
96213 17342 31663 34058 21303 42172 23 915 39798 15306
300 300 300 300
49257 36828 54754 74980 38356 81507 41035 74243 25726
39 18 13
28399 20828 14414
270 210 240 300 240
15813 30044 48731 25354
10 34 31 19
240
0
44 12 6 26
16538 48463 3252 18674 7061 32847 39357 22517 35000
25185 34041 45914 15536 0 11275 15117 39790 52444
15 26 8 10 22 55 17 18 21
33870 27381 14923 7655 46554 30973 15588 66336 20120
300 240 360 450 480 270 270 480 240 480
52215 32620 45512 42377 1 1 1 028 44074 23411 78478 33765 22438
300 300
Max. vertical velocity py flows (nms- 1 )t 28641 40914 19594 56612 39381 37664 55690
* Maximum vertical component of the velocity spectrum for the explosion component of the seismogram. Values for August were measured from the seismometer at Galway's Estate. Those for September and October were measured on the Windy Hill seismometer. The two data sets are therefore not directly comparable. t Maximum vertical component of the velocity spectrum for the pyroclastic flow component of the seismogram, measured at Windy Hill.
of DRE volumes from 0.01 to 6.6 x 10 5 m 3 , with an average of 1.4 x 105 m3, which is more consistent with the field estimate. The total volume of pumice-and-ash pyroclastic flow deposits generated during the two episodes of Vulcanian explosions was 32 x 10 6 m 3 , equivalent to an average of 1.9 x 105m3 DRE per explosion. This was calculated from volume surveys of the main ghauts carried out throughout the Soufriere Hills eruption. The method involved surveying of the ground surface by helicopter using rangefinder binoculars and GPS (Sparks et al. 1998). Only for one explosion (15:17 on 18 October) was a detailed survey carried out of the pyroclastic flow deposits from a single event in enough detail to calculate a volume. The resulting map is given in Cole et al. (2002). This explosion was average in magnitude (NOAAestimated plume height 9.1 km), and the calculated pyroclastic flow
volume (2.3 x 105 m3 DRE) agrees broadly with the overall average given above. The variation of pyroclastic flow volume with explosion magnitude cannot be determined. There are indications from the data in Table 1 that flows from larger explosions (as indicated by higher plumes) entered more ghauts around the dome, and thus were perhaps more voluminous, although this is not possible to quantify owing to the incompleteness of the observations. On the other hand, there is no obvious correlation between pyroclastic flow runout and plume height. For example, flows from the relatively small explosions at 11:34 on 1 October (plume height 4.6 km) and at 17:57 on 11 October (plume height 6.9 km) had runouts down Tuitt's Ghaut of 4.5 and 5.5 km respectively, which is comparable to, or greater than, those from explosions with higher plumes. Neither is there any
300
T. H. DRUITT ET AL.
correlation between plume height and the amplitude or duration of the high-frequency (pyroclastic flow) component of the explosion seismic signals that might suggest systematic variation of pyroclastic flow volume with explosion magnitude. We conclude that the average Vulcanian explosion during August, September and October 1997 discharged a total of about 3.0 x 10 5 m 3 DRE of magma - 1.1 x 10 5 m 3 as fallout and 1.9 x 10 5 m 3 as pyroclastic flows - although there was a large variation from the smallest to the largest explosions. Since the volume of ash in the co-pyroclastic-flow plumes was c. 10% of the total fallout (Bonaonna et al. 2002b), the partitioning of magma during an average explosion is estimated to have been as follows: central plume, umbrella cloud and waning plume 1.0 x 10 6 m 3 , pyroclastic flows 1.9 x 10 6 m 3 , and co-pyroclastic-flow ash plumes 0.1 x 10 6 m 3 . About two-thirds of the material ejected during an average explosion underwent fountain collapse to form pyroclastic flows. Exit velocities during the explosions Exit velocities during the explosions (the velocity at which the material left the dome summit crater) were estimated from analysis of video footage, distributions of ballistic blocks and, more crudely, from the observed fountain-collapse heights and durations of phase 1 of the explosion seismic signals.
Observed fountain-collapse height and duration of seismic phase 1 A crude estimate of exit velocity (u) at the start of each explosion can be made from the estimated fountain collapse heights (h = 300-650 m above the summit crater rim) during the initial 10-20s of the explosions. Numerical modelling has shown that fountain-collapse height can be approximated by h u2/2g (Dobran et al. 1993; Clarke et al. 2002), which implies a vertical exit velocity at crater-rim level of 80-115m s - 1 . For such velocities, the transit time (t) of the collapsing fountain from initiation to ground impact would be approximately t 2u/g, or 16-23 s, which is consistent with the observed durations of phase 1 of the explosion seismic signals (10-20 s).
Analysis of video footage More accurate estimates of vertical ascent velocities were made for the three main plumes of the explosion at 12:05 on 7 August, using the northern crater rim (940m) as reference level. Video footage from two sites (MVO South and Fleming) was analysed in order to construct height-time curves and thus to estimate ascent velocities. The height of each of the three plumes was measured as a function of time and corrected for perspective effects using the equations of Sparks & Wilson (1982). Distortions were small because the plumes were filmed from distances of several kilometres and scaling corrections were approximately linear. The vertical and horizontal fields of view for the two cameras were determined by filming a known landscape using the same settings as during the explosion. Tracings of the three plumes are shown in Figure 15, and the height-time data are shown in Figure 16a. Plume ascent velocities were calculated using best-fit polynomials to smooth the heighttime curves, then differentiating the polynomials (Fig. 16b). For all three plumes the velocity first decreased, reached a minimum in the range 15 to 3 0 m s - 1 , then increased again. The velocity minimum is interpreted as the transition from momentum-driven to buoyancydriven behaviour and occurred at heights of 450-650 m (plumes 1 and 2) and 1200m (plume 3) above the crater rim. Exit velocities for the three plumes were estimated by backextrapolation of the velocity curves to crater-rim level. Plume 2 left the crater 7 s into the explosion with a vertical velocity of 80 1 0 m s - 1 ; but, since it was initially inclined at about 60° to the horizontal, the absolute exit velocity was about 95 1 0 m s - 1 .
Fig. 15. Tracings of plumes 1 to 3 of the explosion of 12:05 on 7 August (Fig. 4). The data were measured from video footage taken from Fleming and corrected for perspective effects using the equations of Sparks & Wilson (1982). Numbers are the time in seconds after the onset of the explosion and refer to the overlying plume front. The tracing intervals are either 2 or 3s. The star marks the vent location.
The data for plume 1 do not allow accurate extrapolation, but the velocity and emergence angle were about the same as for plume 2. Plume 3 left the vent vertically 17 s into the explosion at a velocity of at least 1 3 0 m s - 1 . Clarke et al. (2002) have analysed the same video footage of the 12:05 explosion on 7 August using a similar method, but without distinguishing the three individual plumes recognized here (i.e. by tracing the progress of the front of the entire plume). Their data are -1 consisted with an initial exit velocity of and a velocity-minimum height of about 450m above the crater rim. They also analysed footage of the 14:36 explosion on 6 August, for which the exit velocity and velocity-minimum height were estimated to be -1 and 650m respectively. A more detailed analysis of video footage of three explosions in October 1997 (17:50 on 6 October, 16:02 on 7 October and 12:32 on 9 October) has been carried out by Formenti & Druitt (in prep.). The analysis involved plotting height-time curves for individual finger jets and extrapolating velocities back to crater-rim level using a fluid dynamic model for a jet. The calculated exit velocities range from 40 to 140m s-1 and in all three explosions increased with time as fragmentation progressed to deeper levels in the conduit. Thus the slowest jets emerged at the start of the explosion and the fastest ones about 10s later. Higher exit velocities may have occurred subsequently, but any such jets were hidden by the billowing clouds of ash.
Analysis of ballistic trajectories The locations and sizes of ballistic blocks also served to constrain exit velocities. Ballistic crater fields were mapped by helicopter following the August explosions using onboard GPS, and block diameters were estimated to within 0 cm (Fig. 17). No accurate measurements were made in September and October due to safety considerations, but observations indicate that they were similar. The August explosions threw blocks out to 1.7 km from the crater centre. The distribution was approximately symmetrical around the dome. Blocks with the largest ranges had diameters of 0.7m (to the north)
EPISODES OF CYCLIC EXPLOSIVE ACTIVITY
301
Fig. 16. (a) Heights and (b) velocities as a function of time for the three plumes of the 12:05 explosion on 7 August (Fig. 15). The lines simply connect the data points.
and 1.2m (to the south). Blocks smaller than 0.4m are not shown in Figure 17, but became increasingly abundant nearer the dome, reflecting the stronger air drag experienced by small blocks (Bower & Woods 1996). Two models (Self et al 1980; Waitt et al 1995) have been used to estimate launch velocities for the ballistic blocks. In the Self et al. (1980) model the maximum range for a block several decimetres in size is achieved for an optimum launch angle of about 35°. Launch angles as low as 30° are observed on video footage, so the optimum angle was assumed in the calculations. Selected launch velocities and travel times calculated using the model are shown on Figure 17. These take into account the elevation difference between the crater rim and landing site. Blocks 0.7m large on Farrell's Plain require launch speeds of about 160ms - 1 to reach their range of 1.6km,
540m below the crater rim. The 1.2m block launched over Galway's Wall requires a velocity of 135ms - 1 . The highest velocities (250 m s - 1 ) are required by 0.4m blocks that landed 1.7km from the vent. However, it was unclear in the field whether these were fragments of larger blocks that had broken on impact, so the result is ambiguous. Launch velocities up to 160m s-1 are therefore required by the Self et al. (1980) model to explain the ballistic data, which is higher than plume exit velocities measured from video footage (up to 140ms - 1 ). Calculated flight times for the ballistics, which range from about 17 to 27 s (Fig. 17), are, however, broadly consistent with observations of the 12:05 explosion of 7 August, in which the first ballistic impacts on Farrell's Plain were observed at 21.6 s (Table 2). The same ballistic data have been modelled by Clarke et al. (2002) using the numerical scheme of Waitt et al. (1995). This
302
T. H. DRUITT ET AL.
Fig. 17. Sizes and ranges of ballistic blocks from the 13 explosions in August 1997. The numbers are the launch velocities and flight times calculated using the model of Self el al. (1980), taking into account air drag and the elevation differences between the impact sites and the crater rim. The calculations assume an optimum eruption angle of 352 to the horizontal. Locations shown by squares are south of the dome, dots between Tar River valley and White's Ghaut, upwardpointing triangles between White's Ghaut and Tuitt's Ghaut, and downward-pointing triangles between Tuitt's Ghaut and Mosquito Ghaut.
Fig. 18. Summary of a single explosive cycle in 1997: (a) immediately prior to explosion onset; (b) during the explosion; (c) during the interval between explosions. See the text for discussion.
requires lower initial velocities for the same 35° launch angle as used above (< 1 2 5 m s - 1 as opposed to < 160ms - 1 ), since the drag coefficients assumed in this model result in lower form drag than in the model of Self et al. (1980). The result is in better agreement with the exit velocities observed. Discussion Explosion cyclicity Activity of Soufriere Hills Volcano in 1997 involved a regime of cyclic explosive behaviour, with an average interval of about 10 hours. Repetitive Vulcanian explosions have been reported from other volcanoes. For example, six explosions occurred regularly at Mount Ngauruhoe, New Zealand, on 19 February 1975 at intervals of between 0.5 and 1 hour (Nairn & Self 1978). Over the period 12 to 14
June 1991, a sequence of four explosions occurred at Mount Pinatubo, Philippines, at intervals ranging from 10 to 28 hours, culminating in the climatic eruption (Hoblitt et al 1996). At Volcan Galeras, Colombia, six explosions took place at intervals of between 8 and 181 days in 1992-1993 (Stix et al. 1997). Twenty-three explosions occurred at Tokachi-dake, Hokkaido, between 16 December 1988 and 5 March 1989, an average of three to four days apart (Katsui et al. 1990). Repetitive explosive behaviour at Montserrat is attributed to the cyclic build-up and release of magmatic pressure beneath a rheologically stiffened plug of degassed magma at shallow levels in the conduit below the dome (Voight et al. 1998, 1999). A type of stickslip effect has been invoked to explain cyclic conduit pressurization at Montserrat, resulting in cyclic deformation of the dome and surrounding terrain (Voight et al. 1999; Denlinger & Hoblitt 1999; Wylie et al. 1999). Hybrid earthquakes are attributed to hydrofracturing and associated gas flow in rock or crystal-rich magma at
EPISODES OF CYCLIC EXPLOSIVE ACTIVITY
303
Table 4. Estimates of physical parameters for the 1997 explosions Parameter
Value*
Plume height (km) Ascent duration of plume (s) Total magma volume discharged (m3 DRE) Duration of peak discharge (s) Exit velocity (ms - 1 ) Fountain-collapse height (m) Mass fraction entering collapse fountain Initial velocity of pyroclastic surges ( m s - 1 ) Runout of pyroclastic flows (km) Runout duration of pyroclastic flows (s) Typical velocity of pyroclastic flows ( m s - 1 ) Fragmentation pressure (MPa) Velocity of fragmentation wave ( m s - 1 ) Conduit withdrawal depth (km) Duration of waning phase of explosion (min) Explosion interval (min) Magma ascent velocity between explosions ( m s - 1 )
c. 10(3 to 15) c.500 3x 10 5 45-70 40-140 300-650 c.2/3 30-60 3-6 c.500 10 5-15 10-50 0.5 to >2 c. 60 (20-190) c. 600 (100-3 700) >0.02
* Average values, with ranges in brackets.
the peak of each pressurization cycle (Neuberg et al. 1998; Voight et al. 1999). Synchronized tilt cycles and hybrid swarms during the August explosive episode provided accurate indicators of the pressurization state of the system, enabling MVO volcanologists to anticipate many of the explosions successfully and to reduce the threat to the population. They also facilitated study of the explosions and their products. Subsequently, in September and October, when there was no tiltmeter and hybrid swarms were weak or absent, the strong periodicity of the explosions themselves played this role.
Explosion mechanisms Figure 18 summarizes schematically the events throughout one cycle of 1997 explosive activity at Soufriere Hills Volcano. Our best estimates of the physical parameters are given in Table 4. Explosive eruption commenced when the conduit overpressure exceeded the strength of the cap of degassed crystal-rich magma. During the initial few seconds, crater rubble and fragments of disrupted, degassed plug were thrown out, forming ballistic showers. A fragmentation wave then descended the conduit into the region of pressurized magma, resulting in a rapid escalation of exit velocities from about 40 to 140ms - 1 . Eruption was highly unsteady, peak discharge lasting just a few tens of seconds with the highest intensity over the first 10-20 s. Each explosion discharged on average about 3 x 105 m3 DRE of magma. Since the conduit diameter during the 1995-1999 phase of the eruption is estimated at 25 to 30m from spine dimensions and the widths of early vents (Watts et al. 2002), and is not believed to have varied greatly with time (Voight et al. 1999), the conduit drawdown during an average explosion was about 500 m below the crater floor (which itself was about 800m above sea level). This is a DRE drawdown; the actual average drawdown of vesicular magma would have been greater, but not more than 1 km. The largest explosions must have emptied the conduit to depths of 2km or more. Since peak discharge lasted a few tens of seconds, the velocity of the fragmentation wave down the conduit is constrained to have been of the order of l0-50ms -1 , although the initial value could have been greater. High-intensity eruption probably ceased once the wave reached a level in which the magma pressure was not sufficiently large to drive fragmentation. Ash then continued to be discharged for 1-3 hours, but at a greatly reduced rate. Numerical models of the plume dynamics (Clarke et al. 2002) and conduit flow (Melnik & Sparks 2002b) reproduce several key features of the explosions, including their highly transient nature, peak exit velocities in the range 80-140 m s - 1 , and discharge durations and conduit drawdowns comparable to those observed. The models are based on the rapid decompression and expansion of
Fig. 19. Explosion magnitude (as measured by plume height) as a function of the time interval between explosions: (a) interval preceding the explosion, and (b) interval following the explosion. See the text for discussion.
gas-rich, pressurized magma beneath a degassed plug, thus supporting this interpretation of the eruption dynamics. The eruption columns were partially unstable in all but one explosion. On average about two-thirds of the erupted material collapsed back to form pyroclastic surges and flows, while the other third, including probably a large proportion of smaller particles, was carried up into the plume. However, these proportions may have varied considerably between individual explosions. Fountain collapse occurred in the first 10-20s of each explosion from a few hundred metres above the crater rim. Vertical velocity profiles in the plumes reveal velocity minima corresponding to the transition from momentum-driven to buoyancy-driven behaviour. This is analogous to the superbuoyant regime of sustained eruption columns, which is intermediate between fully stable (convective) and fully unstable (collapsed) regimes (Bursik & Woods 1991) and is seen in the explosion simulation of Clarke et al. (2002). Once each explosion was over, magma rose in the conduit by viscous flow at a couple of centimetres per second or more (1-2 km in 10 hours). This exceeds the critical ascent speed of about l ^ c m s - 1 for amphibole breakdown and explains the presence of hornblende phenocrysts lacking breakdown rims in the explosion pumices (Devine et al. 1998b). In at least some cases the conduit was totally refilled prior to the next explosion and a small dome appeared in the crater. Repressurization of the conduit then occurred until conditions were right for the next explosion, although the exact mechanism is not well understood. Figure 19 shows that a weak correlation exists between plume height and the intervals between explosions for both the August and September-October episodes. Positive correlations exist between (1) plume height and the interval prior to a given explosion (Fig. 19a) and (2) plume height and the interval following a given explosion (Fig. 19b). Correlation 1 would suggest that large explosions result from long preceding intervals, perhaps allowing the build-up of larger magma pressures in the conduit. Correlation 2 is consistent with a scenario in which large explosions drain the conduit to deeper levels, so that longer
304
T. H. DRUITT ET AL.
intervals are then required to refill the conduit prior to the next explosion. The data do not distinguish between these two mechanisms, although correlation 2 appears visually to be slightly better than correlation 1.
pascals) relative to the overlying plug and surrounding conduit walls. The presence of pressurized, gas-charged magma at very high levels in the conduit immediately prior to each explosion is consistent with the observation that exit velocities in excess of 100 m s-1 or more were achieved only a few seconds after each explosion began (Clarke et al 2002; Melnik & Sparks 2002b).
Conduit pressurization and explosive fragmentation Pressurization of the magmatic conduit at Montserrat is attributed to non-linear vertical pressure gradients caused by large viscosity variations that accompany exsolution of water from magma (Sparks 1997; Massol & Jaupart 1999). Magma viscosity is a strong function of water content, particularly at low pressures (Hess & Dingwell 1996). The estimated viscosity of non-degassed magma at Montserrat is about 106 Pas and that of completely degassed magma about 10 14 Pas (Voight et al. 1999). This is believed to have generated large magma overpressures (magma pressure minus lithostatic pressure) at shallow levels in the conduit. Another effect is the development of high gas pore pressures in the ascending magma due to (1) viscous resistance to vesicle expansion, which increases as the liquid exsolves gas (Massol & Jaupart 1999), and (2) growth of microlites in the degassed, undercooled liquid, which forces further gas into vesicles (Stix et al 1997; Sparks 1997). Tilt amplitudes and far-field deformation measurements at Montserrat are consistent with maximum magma overpressures of about ten to a few tens of megapascals a few hundred metres below the base of the dome (Shepherd et al 1998; Voight et al. 1999). The conduit flow modelling of Melnik & Sparks (2002a) predicts steep pressure gradients and overpressures up to l0MPa in the upper conduit. The angular, platy shapes of many of the 1997 fallout pumices with 55-75 vol% vesicles are consistent with brittle fragmentation of a pressurized magmatic foam present in the upper conduit prior to each explosion. Brittle fragmentation of magma requires steep pressure gradients and fast decompression rates in order to drive the magma through the glass transition limit (Dingwell 1996). This has been observed experimentally by Alidibirov & Dingwell (1996), who showed that pressure differentials across the fragmentation interface of a few megapascals can be sufficient to drive brittle failure, generating platy fragments with shapes very much like those at Montserrat. Recent experimental work has shown that the tensile strength of crystal-rich magma like that at Montserrat may be of the order of 20MPa or more (Martel et al 2001). As the fragmentation wave descended the conduit during each explosion, the pressurized foam broke up into tabular fragments that were then accelerated to the surface. Magma fragments erupting from the vent apparently had sufficiently high viscosities due to gas exsolution to suppress post-fragmentation expansion, enabling pumices to retain vesicularities and angular shapes close to those acquired at fragmentation (the viscosity quench effect; Thomas et al 1994). Pumice incorporated into pyroclastic flows were subsequently rounded by abrasion during transport. The presence of magmatic foam with at least 55% bubbles in the upper conduit can be used to provide an independent estimate of magma pressure prior to each explosion. The bulk water content of the magma prior to ascent was about , based on the water content of glass inclusions ; Barclay et al 1998; Devine et al 1998a) and the estimated crystal content in the magma reservoir at 5-6km depth (60-65vol%; Murphy et al 2000). In the Appendix we show that the total confining pressures required for magmatic foam with 1.6 3 wt% water to have vesicularities of 55-75% are 5-15MPa. One key feature of the explosions is that the products are dominantly pumiceous, with dense clasts making up no more than 5% of those erupted (Clarke et al 2002). Given that an average explosion emptied the conduit to about 500 m (DRE) depth, the plug of degassed magma present in the conduit prior to each explosion can have been no more than about 25m thick, corresponding to an overburden of less than 1 MPa. Given that pressures of 5-15 MPa are required in the magmatic foam just below this cap, this suggests that the foam must have been very significantly overpressured (by at least a few mega-
Initiation of episodes of explosive activity on Montserrat Each of the two episodes of explosive activity in 1997 was triggered by a major dome collapse (3 August and 21 September), as was the explosive eruption on 17 September 1996 (Robertson et al 1998). In each case, sudden removal of part of the dome led to the conditions for explosive fragmentation. This was not immediate, the delays being 2.5 hours (17 September 1996), 10 hours (3 August 1997) and 20 hours (22 September 1997), showing perhaps that time was necessary for the build-up of sufficient conduit pressure for this to occur. Many dome collapses occurred during the 1995-1999 period, but only three are known to have triggered major vertical explosions. We exclude here the relatively weak explosions of late 1998 and 1999, which may have been triggered by slow pressure build-up in the slowly crystallizing lava dome and conduit during the period of virtually no magma extrusion (Norton et al 2002). One important factor was probably that the 17 September 1996 and 21 September 1997 collapses resulted in two of the largest height reductions of the active dome-growth area during the 1995-1999 period (130m and 230m respectively), causing large decompressions of the conduit (at least 3.5 and 6 MPa). The height reduction from the 3 August 1997 collapse was not observed clearly, but it is inferred to have been at least 110m (3 MPa) from the form of the crater observed four days later. Sudden decompression of at least 3 MPa therefore appears necessary to trigger explosive fragmentation at Montserrat. Conduit flow beneath lava domes involves complex feedback effects and sudden decompressions can force systems from effusive to explosive behaviour (Jaupart & Allegre 1991; Woods & Koyaguchi 1994). An additional effect in 1997 may have been the high magma discharge rate. The time-averaged magma discharge rate increased throughout the 1995-1999 period, and by August 1997 had reached 7-8 m3 s-1 (Sparks et al 1998; Sparks & Young 2002). High discharge rates favour explosive fragmentation by limiting the time available for magma degassing during ascent (Jaupart & Allegre 1991; Melnik & Sparks 2002a). High magma flux during August, September and October of 1997 may have helped to prime the conduit for explosive activity once a suitably large dome collapse occurred. Strangely, the largest dome collapse of the 1995-1999 period (26 December 1997; Sparks et al 2002) decompressed the conduit by at least 8 MPa. giving rise to a violent lateral blast and pyroclastic density current, but triggered no vertical explosion from the conduit and produced little pumice. This highlights the complexity of the system and the existence of important effects not considered here. Conclusions Two episodes of cyclic explosive activity occurred at Soufriere Hills Volcano in 1997. Thirteen explosions took place in August and another 75 in September and October. The activity had a major impact on southern Montserrat and triggered northward enlargement of the evacuation zone in mid-August. Like the explosive eruption of 17 September 1996. both episodes in 1997 were preceded by major dome collapses that decompressed the conduit by 3 MPa or more. Delays of 3 to 20 hours then followed before explosive activity commenced. Large gravitational collapses are a prerequisite for vertical explosive eruption at Montserrat. The explosions were highly unsteady, with the most intense phase lasting only a few tens of seconds. Peak discharge was accompanied by ballistic showers, exit velocities up to 140 m s - l , and (in all but one event) fountain collapse from a few hundred metres above the crater rim over the first 10-20s of each explosion. Pyroclastic flows travelled up to 6km down all major drainages around the
EPISODES OF CYCLIC EXPLOSIVE ACTIVITY dome and entered the sea on the south and east coasts. Buoyant eruption plumes with large, bulbous heads rose to 3-15 km in the atmosphere, then spread out as umbrella clouds. After 10 minutes or so, each explosion settled into a waning phase that typically lasted an hour and generated a low, bent-over ash plume. Fallout and pyroclastic flows/surges from the explosions accounted on average for one-third and two-thirds of the magma discharged, respectively. The explosions emptied the conduit to a depth of 0.5-2 km, perhaps more in some cases. Filtering of explosion seismic signals permitted distinction of a low-frequency (c. 1 Hz) component due to the explosion itself and a high-frequency (>2 Hz) component due to ballistic impact, fountain collapse and pyroclastic flow. Relative timing of the onsets of the two components provided information on the flight durations of ballistic blocks and on the transit time for fountain collapse, from inception to first ground impact. Explosions in August were accompanied by cyclic patterns of seismicity and edifice deformation. Repeated slow inflation, followed by rapid deflation, of the volcano recorded cycles of build-up, then release, of pressure beneath the dome. The explosions were driven by rapid decompression and brittle fragmentation of overpressured magmatic foam in the upper conduit and occurred at intervals of 2.5 to 63 hours, with a mean of 10 hours. Synchronized tilt cycles and hybrid earthquake swarms during the August explosions provided accurate indicators of the pressurization state of the system, enabling volcanologists to anticipate many of the explosions and reduce the threat to the population. In September and October, when there was no tiltmeter and hybrid swarms were weak or absent, the strong periodicity of the explosions themselves played this role. We thank the staff of the M VO for their very important contributions in the study of the 1997 explosions. D. Lea, M. Sagot and D. Williams kindly provided us with video footage of the explosions and allowed us to study it. D. Williams helped us in the analysis of video footage. B. Poyer kindly provided the photographs in Figure 7. Careful reviews by T. Koyaguchi, L. Wilson and P. Kokelaar are gratefully acknowledged.
Appendix
Estimation of fragmentation pressures during the explosions We estimate the pressure necessary for the magma with 60-65 vol% crystals (Murphy et al. 2000) and a bulk water content of % to have 55-75 vol% vesicularity. Consider a unit volume of crystal-bearing magmatic foam immediately prior to fragmentation. The volume fraction of bubbles is X and the volume fraction of crystals in the liquid phase is F. The masses of gas, liquid, and crystals are given by:
Mg = pgX
M 1 = A(1-A-)(1-F) Mc = Pc(l - X)F
(Al)
where M is mass, p is density and the subscripts g, 1 and c stand for gas, liquid and crystals respectively. Given the solubility law for water in magmatic liquid, n = P 1 / 2 , where n is a mass fraction and a is approximately 4.1 x 10-6 Pa 1/2 for rhyolite (the composition of interstitial glass in the pumices), we can write the mass balance equation for water in the foam Mg + M 1 aP 1/2 = N(Mg + M1 + Mc)
(A2)
where N is the bulk mass fraction of water in the magma. The density of the gas is given by:
(A3) where T is temperature (about 860°C or 1133 K; Barclay et al. 1998) and r is the gas constant (462 J k g - 1 K-1 for water). Given a bulk water content N, we can use Equation A2 to estimate the vesicularity X of the foam as a function of pressure P. For a bulk water content of , bubble contents of 55-75 vol% require pressures in the range 5-15MPa.
305
References ALIDIBIROV, M. A. & DINGWELL, D. B. 1996. Magma fragmentation by rapid decompression. Nature, 380, 146-148. ASPINALL, W. P., MILLER, A. D., LYNCH, L. L., LATCHMAN, J. L., STEWART, R. C., WHITE, R. A. & POWER, J. A. 1998. Soufriere Hills eruption, Montserrat 1995-1997: volcanic earthquake locations and fault plane solutions. Geophysical Research Letters, 25, 3397-3400. BARCLAY, J., RUTHERFORD, M. J., CARROLL, M. R., MURPHY, M. D., DEVINE, J. D., GARDNER, J. & SPARKS, R. S. J. 1998. Experimental phase equilibria constraints on pre-emptive storage conditions of the Soufriere Hills magma. Geophysical Research Letters, 25, 2437-3440. BONADONNA, C., MACEDONIO, G. & SPARKS, R. S. J. 2002a. Numerical modelling of tephra fallout associated with dome collapses and Vulcanian explosions: application to hazard assessment on Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 517-537. BONADONNA, C., MAYBERRY, G. C., CALDER, E. S., SPARKS, R. S. J., CHOUX, C., JACKSON, P., LEJEUNE, A. M., LOUGHLIN, S. C., NORTON, G. E., ROSE, W. L, RYAN, G. & YOUNG, S. R. 2002b. Tephra fallout in the eruption of Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 483-516. BOWER, S. M. & WOODS, A. W. 1996. On the dispersal of clasts from volcanic craters during small explosive eruptions. Journal of Volcanology and Geothermal Research, 73 19-32. BURSIK, M. I. & WOODS, A. W. 1991. Buoyant, superbuoyant and collapsing eruption columns. Journal of Volcanology and Geothermal Research, 45, 347-350. CLARKE, A. B., NERI, A., VOIGHT, B., MACEDONIO, G. & DRUITT, T. H. 2002. Computational modelling of the transient dynamics of the August 1997 Vulcanian explosions at Soufriere Hills Volcano, Montserrat: influence of initial conduit conditions on near-vent pyroclastic dispersal. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 319-348. COLE, P. D., CALDER, E. S., SPARKS, R. S. J. ET AL. 2002. Deposits from dome-collapse and fountain-collapse pyroclastic flows at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoir. DENLINGER, R. P. & HOBLITT, R. P. 1999. Cyclic behavior of silicic volcanoes. Geology, 27, 459-462. DEVINE, J. D., MURPHY, M. D., RUTHERFORD, M. J. ET AL. 1998a. Petrologic evidence for pre-eruptive pressure-temperature conditions, and recent heating, of andesitic magma erupting at the Soufriere Hills Volcano, MONTSERRAT, W. I. Geophysical Research Letters, 25, 3669-3672. DEVINE, J. D., RUTHERFORD, M. J. & GARDNER, J. C. 1998b. Petrologic determination of ascent rates for the 1995-1997 Soufriere Hills Volcano andesite magma. Geophysical Research Letters, 25, 3673-3676. DINGWELL, D. B. 1996. Volcanic dilemma: Flow or blow? Science, 273, 1054-1055. DOBRAN, F., NERI, A. & MACEDONIO, G. 1993. Numerical simulations of collapsing eruption columns. Journal of Geophysical Research, 98, 4231-4259. DRUITT, T. H. 1992. Emplacement of the 18 May 1980 lateral blast deposit ENE of Mount St. Helens, Washington. Bulletin of Volcanology, 54, 554-572. FAGENTS, S. A. & WILSON, L. 1993. Explosive volcanic eruptions - VII. The ranges of pyroclasts ejected in transient volcanic explosions. Geophysical Journal International, 113, 359-370. HESS, K. U. & DINGWELL, D. B. 1996. Viscosities of hydrous leucogranite melts: a non-arrhenian model. American Mineralogist, 81, 1297-1300. HOBLITT, R. P., WOLFE, E. W., SCOTT, W. E., COUCHMAN, M. R., PALLISTER, J. S. & JAVIER, D. 1996. The pre-climactic eruptions of Mount Pinatubo, June 1991. In: NEWHALL, C. G. & PUNONGBAYAN, R. S. (eds) Fire and Mud: Eruptions and Lahars of Mount Pinatubo, Philippines. University of Washington Press, 457-511. JAUPART, C. & ALLEGRE, C. J. 1991. Gas content, eruption rate and instabilities of eruption regime in silicic volcanoes. Earth and Planetary Science Letters, 102, 413-429.
306
T. H. DRUITT ET AL.
KATSUI, Y., KAWACHI, S.. KONDO, Y. ET AL. 1990. The 1988-89 explosive eruption of Tokachi-dake, Central Hokkaido, its sequence and mode. Bulletin of the Volcanological Society of Japan, Series 2, 35, 111-129. KIENLE, J. & SHAW, G. E. 1979. Plume dynamics, thermal energy and long-distance transport of vulcanian eruption clouds from Augustine volcano, Alaska. Journal of Volcanology and Geothermal Research. 6. 139-164. KOKELAAR, B. P. 2002. Setting, chronology and consequences of the eruption of Soufriere Hills Volcano, Montserrat (1995-1999). In: DRUITT. T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society. London. Memoirs, 21, 1-43. LOUGHLIN, S. C, CALDER, E. S., CLARKE, A. B. ET AL. 2002. Pyroclastic flows and surges generated by the 25 June 1997 dome collapse, Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs. 21, 191-209. MARTEL, C., DINGWELL, D. B., SPIELER, O., PICHEVANT, M. & WILKE. M. 2001. Experimental fragmentation of crystal- and vesicle-bearing silicic melts. Bulletin of Volcanology, 63, 398-405. MARTIN, D. P. & ROSE, W. I., JR. 1981. Behavioural patterns of Fuego volcano, Guatemala. Journal of Volcanology and Geothermal Research, 10,67-81. MASSOL, H. & JAUPART, C. 1999. The generation of gas overpressure in volcanic eruptions. Earth and Planetary Science Letters, 166. 57-70. MATTHEWS, S. J., GARDEWEG, M. C. & SPARKS, R. S. J. 1997. The 19841996 cyclic activity of Lascar Volcano, Northern Chile: cycles of dome growth, dome subsidence, degassing and explosive eruptions. Bulletin of Volcanology, 59, 72-82. MELNIK, O. & SPARKS, R. S. J. 2002a. Dynamics of magma ascent and lava extrusion at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 153-171. MELNIK, O. & SPARKS, R. S. J. 2002b. Modelling of conduit flow dynamics during explosive activity at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society. London. Memoirs, 21, 173-190. MELSON, W. & SAENZ, R. 1973. Volume, energy and cyclicity of eruptions at Arenal volcano, Costa Rica. Bulletin of Volcanology, 37, 416-437. MILLER, A. D., STEWART, R. C., WHITE, R. A. ET AL. 1998. Seismicity associated with dome growth and collapse at the Soufriere Hills Volcano, Montserrat. Geophysical Research Letters, 25, 3401-3404. MORRISSEY, M. M. & MASTIN, L. G. 2000. Vulcanian eruptions. In: SIGURDSSON, H. (ed.) Encyclopedia of Volcanoes. Academic Press. San Diego, 463-476. MORTON, B., TAYLOR, G. I. & TURNER, J. S. 1956. Turbulent gravitational convection from maintained and instantaneous sources. Proceedings of the Royal Society, A234, 1-23. MURPHY, M. D., SPARKS, R. S. J., BARCLAY, J., CARROLL, M. R. & BREWER, T. S. 2000. Remobilization of andesite magma by intrusion of mafic magma at the Soufriere Hills Volcano, Montserrat, West Indies. Journal of Petrology, 41, 21-42. NAIRN, I. & SELF, S. 1978. Explosive avalanches and pyroclastic flows from Ngauruhoe 1975. Journal of Volcanology and Geothermal Research, 3, 39-60. NEUBERG, J. & O'GORMAN, C. 2002. A model of the seismic wavefield in gas-charged magma: application to Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 603-609. NEUBERG, J., BAPTIE, B., LUCKETT, R. & STEWART, R. 1998. Results from the broadband seismic network on Montserrat. Geophysical Research Letters, 25, 3661-3664. NORTON, G. E., WATTS, R. B., VOIGHT, B. ET AL. 2002. Pyroclastic flow and explosive activity at Soufriere Hills Volcano, Montserrat during a period of virtually no magma extrusion (March 1998 to November 1999). In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society. London, Memoirs, 21, 467-481. ROBERTSON, R., COLE, P., SPARKS, R. S. J. ET AL. 1998. The explosive eruption of Soufriere Hills Volcano, Montserrat, West Indies, 17 September, 1996. Geophysical Research Letters, 25, 3429-3432.
SELF. S.. WILSON. L. & N A I R N . I. A. 1979. Vulcanian eruption mechanisms. Nature. 277. 440-443. SELF, S., KEINLE. J. & HUOT. J. P. 1980. Ukinrek Maars. Alaska: II. Deposits and formation of the 1977 craters. Journal of Volcanology and Geothermal Research. 7. 39-65. SHEPHERD. J. B.. HERD. R. A.. JACKSON. P. & WATTS. R. 1998. Ground deformation measurements at the Soufriere Hills Volcano. Montserrat: II: Rapid static GPS measurements June 1996-June 1997. Geophysical Research Letters. 25. 3413-3416. SPARKS. R. S. J. 1997. Causes and consequences of pressurization in lava dome eruptions. Earth and Planetary Science Letters, 150. 177-189. SPARKS. R. S. J. & WILSON. L. 1982. Explosive volcanic eruptions - V. Observations of plume dynamics during the 1979 Soufriere eruption. St. Vincent. Geophysical Journal of the Royal Astronomical Society. 69. 551-570. SPARKS. R. S. J. & YOUNG. S. R. 2002. The eruption of Soufriere Hills Volcano. Montserrat (1995-1999): overview of scientific results. In: DRUITT. T. H. & KOKELAAR. B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society. London. Memoirs. 21, 45-69. SPARKS, R. S. J., YOUNG. S. R.. BARCLAY. J. ET AL. 1998. Magma production and growth of the lava dome of the Soufriere Hills Volcano. Montserrat, West Indies: November 1995 to December 1997. Geophysical Research Letters. 25. 3421-3424. SPARKS. R. S. J.. BARCLAY. J.. CALDER. E. S. ET AL. 2002. Generation of a debris avalanche and violent pyroclastic density current on 26 December (Boxing Day) 1997 at Soufriere Hills Volcano. Montserrat. In: DRUITT, T. H. & KOKELAAR. B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society. London. Memoirs. 21.409 434. STIX. J.. TORRES. R. C.. NARVAEZ. M. L.. CORTES. G. P.. RAIGOSA. J. A.. GOMEZ. D. M. & CASTONGUAY. R. 1997. A model of Vulcanian eruptions at Galeras Volcano. Columbia. Journal of Volcanology and Geothermal Research. 77. 285-304. THOMAS. N.. JAUPART. C. & VERGNIOLLE. S. 1994. On the vesicularity of pumice. Journal of Geophysical Research. 99. 15633-15644. U H I R A , K., YAMASATO. H & TAKEO. M. 1994. Source mechanism of seismic waves excited by pyroclastic flows observed at Unzen volcano. Japan. Journal of Geophysical Research. 99. 17757-17773. VOIGHT. B.. HOBLITT. R. P.. CLARKE. A. B.. LOCKHART. A. B.. MILLER. A. D., LYNCH. L. & MCMAHON. J. 1998. Remarkable cyclic ground deformation monitored in real-time on Montserrat. and its use in eruption forecasting. Geophysical Research Letters. 25. 3405-3408. VOIGHT. B.. SPARKS. R. S. J.. MILLER. R. C. ET AL. 1999. Magma flow instability and cyclic activity at Soufriere Hills Volcano. Montserrat. British West Indies. Science. 283. 1138-1142. WAITT, R. B.. MASTIN. L. G. & MILLER. T. P. 1995. Ballistic showers during Crater Peak eruptions of Mount Spurr Volcano, summer 1992. The 1992 Eruptions of Crater Peak Vent, Mount Spurr Volcano, Alaska. US Geological Survey. Bulletin 2139. WATTS. R. B.. HERD. R. A.. SPARKS. R. S. J. & YOUNG. S. R. 2002. Growth patterns and emplacement of the andesitic lava dome at Soufriere Hills Volcano. Montserrat. In: DRUITT. T. H. & KOKELAAR. B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society. London. Memoirs. 21. 115-152. WOODS. A. W. 1995. A model of vulcanian explosions, Nuclear Engineering and Design. 155. 345-357. WOODS. A. W. & BURSIK. M. I. 1991. Particle fallout, thermal equilibrium and volcanic plumes. Bulletin of Volcanology. 53. 559-570. WOODS, A. W. & KIENLE. J. 1994. The dynamics and thermodynamics of volcanic clouds: theory and observations from the April 15 and April 21 1990 eruptions of Redoubt Volcano. Alaska. Journal of Volcanology and Geothermal Research. 62. 273-299. WOODS, A. W. & KOYAGUCHI. T. 1994. Transitions between explosive and effusive eruption of silicic magmas. Nature. 370, 641-645. WYLIE, J. J.. VOIGHT. B. & WHITEHEAD. J. A. 1999. Instability of magma flow from volatile-dependent viscosity. Science. 285. 1883-1885. YOUNG. S.. SPARKS. R. S. J.. ROBERTSON. R.. LYNCH. L. & ASPINALL. W. P. 1997. Eruption of Soufriere Hills Volcano in Montserrat continues. EOS, Transactions, American Geophysical Union. 78. 404. 408-409. YOUNG. S. R., SPARKS. R. S. J.. ASPINALL. W. P.. LYNCH. L. L.. MILLER. A. D.. ROBERTSON. R. E. A. & SHEPHERD. J. B. 1998. Overview of the eruption of Soufriere Hills volcano. Montserrat. 18 July 1995 to December 1997. Geophysical Research Letters. 25. 3389-3392.
Modelling of conduit flow dynamics during explosive activity at Soufriere Hills Volcano, Montserrat O. MELNIK 1,2 & R. S. J. SPARKS2 1 Department of Earth Sciences, University of Bristol, Bristol BS8 1RJ, UK (e-mail:
[email protected]) 2 Institute of Mechanics, Moscow State University, 1 Michurinskii prosp., Moscow 117192, Russia
Abstract: Magmatic explosive activity at Soufriere Hills Volcano involved a sub-Plinian eruption, on 17 September 1996 and two series of repetitive short-lived (c. 1 min) Vulcanian explosions in 1997. Explosive activity followed major collapses of the dome. We have modelled unsteady conduit flow in explosive eruptions after unloading. Two cases are investigated: (i) equilibrium between gas dissolved in the melt and bubbles for sustained sub-Plinian eruption; and (ii) no mass transfer between pre-existing gas bubbles and melt for Vulcanian explosions. The models for Vulcanian explosions agree with observations of erupted volumes, eruption durations (tens of seconds), typical drawdown depths (a few hundred metres to c. 2 km), exit velocities and discharge rates. Explosive mixtures are predicted to have high densities consistent with the occurrence of fountain collapse. The models for sub-Plinian eruption show good agreement with observed erupted volumes and drawdown depths (c. 4km). Three fragmentation criteria were studied, namely fragmentation at fixed porosity, at a critical gas overpressure, and at a critical elongation strain rate. Results are similar for the three cases, but the critical overpressure and critical strain-rate criteria both predict strong pulsations, whereas the fixed-porosity criterion predicts continuous fragmentation. Pulsations are caused by feedback, with the threshold conditions for magma fragmentation being repeatedly crossed. Pulsations are indicated from seismic and video observations.
Explosive volcanism has been a prominent feature of the eruption of Soufriere Hills Volcano, Montserrat. Such activity has included early phreatic explosions (Young et al. 1998; Bonadonna et al. 2002), sub-Plinian magmatic explosive activity on 17 September 1996 (Robertson et al. 1998), two series of repetitive short-lived Vulcanian explosions from 3 to 12 August 1997 and from 22 September to 21 October 1997 (Druitt et al. 2002) and small sporadic Vulcanian explosions in the period following dome growth after March 1998 (Norton et al. 2002). This paper is concerned with magmatic explosive eruptions, all of which were preceded by major collapses of the dome, unloading pressurized magma in the conduit. This explosive activity has been documented elsewhere (Robertson et al. 1998; Druitt et al. 2002). Clarke et al. (2002) develop a numerical model for decompression and discharge of the explosive mixture into the atmosphere, using a simplified conduit model as an initial condition. This paper focuses on the unsteady magma flows in the conduit during these explosive eruptions and on the influence of magma fragmentation processes on dynamics. Explosive activity, caused either by a sudden decompression or when the gas overpressure in rising magma reaches a threshold value, are modelled as unsteady flows, taking account of vertical viscosity variations and fragmentation conditions. In the case of Soufriere Hills Volcano, all the episodes of explosive activity have occurred shortly after a major dome collapse in which the conduit has been rapidly decompressed. Repetitive individual Vulcanian explosions could be triggered when a critical gas overpressure in growing bubbles is exceeded in rapidly rising magma between the explosions. The Vulcanian explosions of 1997 were short-lived (typically a few tens of seconds). In contrast, the sub-Plinian explosive activity of 17 September 1996 reached a peak after 10 minutes, declining thereafter over a period of 30 minutes. Episodes of explosive activity did not develop into sustained steady eruptions and did not tap directly into the chamber. The models here therefore focus on transient flows, in contrast to published models which have been largely concerned with sustained explosive eruptions where changes with time are sufficiently slow that steady conditions can be assumed (Wilson et al. 1980; Dobran 1992; Barmin & Melnik 1993; Woods & Koyaguchi 1994). We compare different assumptions on fragmentation conditions to establish whether the results are sensitive to the exact mechanism of fragmentation. Three assumptions are compared: fragmentation at a fixed volume fraction of bubbles (VF) (Sparks 1978; Wilson et al. 1980), fragmentation at a fixed overpressure in the growing bubble (OP) which exceeds the tensile strength of the magma (Barmin & Melnik 1993; Melnik 2000), and fragmentation at a threshold where the elongation strain rate (SR) exceeds the magma strength (Papale 1999).
Magmatic explosive eruptions at Soufriere Hills Volcano This section describes the magmatic explosive eruptions at Soufriere Hills Volcano, largely based on the studies of Robertson etal. (1998) and Druitt et al. (2002). The sub-Plinian activity of 17 September 1996 followed a 9-hour (11:30 to 20:30 local time (LT)) period of continuous dome collapse. The course of the activity is depicted in the seismic record (Fig. 1). Explosive activity initiated at 23:42 LT after a relatively quiet period from 20:30 LT. The seismic energy reached a peak after 10 minutes and declined with prominent fluctuations over a 30-minute period thereafter. Various observations (Robertson et al. 1998) indicate that the eruption column reached about 14-15 km high, with a peak discharge rate of 3000m3 s-1 (9 x 106 kgs - 1 ). Ballistic clasts up to 1.5m diameter reached 2.1km from the dome. Robertson et al. (1998) estimated a launch velocity of 180m s-1 and explosion pressure of at least 24 MPa to account for the ballistic ejection. Various observations and inferences from ejecta volumes, conduit dimensions, post-eruption seismicity and re-establishment of dome growth two weeks later are consistent with draining of the conduit down to a depth of about 4 km. The ejecta include several components with contrasted densities and textures. The main pumice is moderately well vesiculated (7001200 kg m - 3 ) and is inferred to derive from the conduit by continuous discharge after the explosive disruption of more consolidated degassed rocks in the uppermost parts of the conduit. The ballistic blocks include dense, glassy non-vesicular blocks, a suite of poorly vesicular dense pumice (1200-2000 kg m - 3 ), and various lithified volcanic breccias. These rocks are interpreted as being derived form the uppermost few hundred metres of the conduit and represent vesiculated magma in the core of the upper conduit, chilled and fully degassed magma at the conduit walls (the glassy rocks) and conduit wallrock breccias. These blocks are inferred to have been discharged ballistically in the first several minutes of eruption when explosive intensity was at a maximum. It is estimated that about 11 x 106 m3 of dome rock collapsed, leaving a scar 200m deep over the conduit and decompressing the magma conduit by at least 4 MPa (Robertson et al. 1998; Calder et al. 2002). The two series of repetitive Vulcanian explosions followed major dome collapses on 3 August 1997 and 21 September 1997. They differed from the 17 September 1996 sub-Plinian explosive activity in a number of ways. They took place as a series of explosions with 13 explosions between 3 and 12 August and 75 explosions between 22 September and 21 October. Each series involved quasi-periodic behaviour, with rise of magma occurring between each explosion. The average interval between explosions
DRUITT, T. H. & KOKELAAR, B. P. (eds) 2002. The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 307-317. 0435-4052/02/$15 © The Geological Society of London 2002.
307
308
O. M E L N I K & R. S. J. SPARKS
Fig. 1. One-minute realtime seismic amplitude measurement (RSAM) seismic record over 8 hours during sub-Plinian explosive activity of 17 September 1996. Three peaks of seismic activity are marked (1-3).
from 22 September to 21 October was 9.5 hours, with a range of intervals from 4 to 33 hours. Repeated Vulcanian explosions were short-lived, with the most intense activity occurring over a few tens of seconds (Druitt et al. 2002), followed by periods of strong ashventing lasting typically 30 minutes to an hour. The explosions involved fountain collapse and generation of pumice-and-ash flows. Ballistic clasts (up to 1.2m diameter) were ejected distances up to 1.6km with observed velocities of 110-140m s - 1 . Eruption column heights ranged from 3km to a maximum of 15km, with inferred peak discharge rates of hundreds to a few thousand of cubic metres per second. Druitt et al. (2002) estimated a typical volume of ejecta as 3 x 10 5 m 3 (dense rock equivalent), although volumes of the largest individual explosions may have exceeded 10 6 m 3 . If the conduit has a diameter of 30m (see Melnik & Sparks 2002) then typical drawdown depths are 200-1000m. Pumice clasts in the Vulcanian explosions have angular and platy character (Druitt et al. 2002). Alidibirov & Dingwell (1996) have reproduced similar angular, platy shapes in laboratory experiments in which natural dome samples and pumice were decompressed from fluid pressures of a few megapascals. The observations on Vulcanian ejecta from Soufriere Hills Volcano support brittle fragmentation of vesicular magma under conditions where the overpressure exceeded the tensile strength of the magma. We infer that a fragmentation wave penetrated into already vesiculated magma.
Description of the physical model for explosive volcanic eruptions
Fig. 2. Schematic representation of flow regimes during explosive eruptions.
the conduit. At the outlet of the conduit the exit pressure is higher than atmospheric pressure due to the weight and viscous resistance of the overlying dome. At the moment t = 0 at the top of the conduit we assume a sudden pressure drop down to the atmospheric value caused by dome collapse, and we study the dynamics of the flow in the conduit caused by this pressure drop. We will examine the influence of fragmentation criteria, pre-eruption parameters and assumptions on the mechanism of mass transfer between magma and growing bubbles on the eruption behaviour. The unsteady model is based on the model developed by Melnik (2000) and is used to describe the motion of the multiphase mixture in the conduit, with some modifications to make the numerical code easier and faster, and to account for the presence of crystals in the ascending magma.
System of equations We consider the following mechanism of explosive eruption. As the overlying dome collapses, the pressure at the top of the conduit decreases rapidly. Conditions soon thereafter exceed a critical condition and explosive activity begins. A rarefaction wave propagates into vesiculated magma. In the rarefaction wave the mixture accelerates and fragmentation occurs when some critical condition, which will be discussed below, is reached. Fragmented magma forms a gasparticle dispersion, which propagates to the exit of the conduit to form a volcanic column in the atmosphere. Figure 2 shows schematically the processes which occur in the conduit. The problem of unsteady magma flow dynamics in a volcanic conduit is considered. Steady-state solutions for the flow of magma during lava-dome extrusion (Melnik & Sparks 1999) are used to define the initial porosity, velocity and pressure distribution along
The mechanical description of the conduit flow during explosive eruptions has been discussed in Barmin & Melnik (1993) and Melnik (2000). For unfragmented magma the main assumptions are: (i) the ascent velocity of bubbles and the gas velocity in permeable media (Melnik & Sparks 2002) are negligible in comparison with the velocity that the mixture attains between the rarefaction wave and fragmentation level (Fig. 2); (ii) temperature variations are small due to the high thermal capacity of magma; and (iii) no additional bubble nucleation is assumed as the rarefaction wave propagates into initially vesiculated magma. We also neglect changes in crystal content due to microlite crystallization (Melnik & Sparks 2002), as the time scale of this process is much longer than the duration of explosive eruptions.
MODELLING OF EXPLOSIVE ERUPTIONS
We now estimate the degree of non-equilibrium of mass transfer between the melt phase in the magma and the bubbles. In an explosive eruption two end-member situations can be considered. First, magma in the conduit has some distribution of vesicularity and gas pressure at the onset of explosive flow, and the propagation of the fragmentation front is too fast for any further diffusive mass transfer from the melt into vesicles. This end-member case is completely non-equilibrium with no mass transfer. Second, propagation of the fragmentation wave is sufficiently slow that diffusion has time to maintain equilibrium between the melt phase and the free-gas phase. This end-member is an equilibrium case. In principle the system can move from one end-member to the other during an explosive eruption, as discussed further below. At the initial stage of eruption, when the rarefaction wave is strong, fragmentation occurs just after the decompression and therefore mass transfer can be neglected. At later stages the fragmentation wave stops, or decreases its velocity significantly, and mass transfer can become important, with equilibrium attained if the flow conditions become sufficiently slow. A dimensionless criterion can be defined to evaluate the degree of equilibrium of mass transfer. Alidibirov & Dingwell (1996) estimated the velocity of a fragmentation wave, VF, to be in the range of tens to over one hundred metres per second. The speed of sound for the mixture, C = (dp/dp)1/2, is in the range of a hundred to several hundred metres per second, depending on the volume concentration of bubbles. Therefore, the length of the region between the front of the rarefaction wave and fragmentation wave (see Fig. 2), LF, is equal to (C - VF)t. The characteristic time of fragmentation is then tF = L F / (V F + V), where V is an average velocity of unfragmented magma. This time can be compared with the characteristic time of diffusion of dissolved gas tD = h s 2 /D (here hs is the thickness of the liquid shell surrounding the bubble and D is a diffusion coefficient) to establish whether mass transfer is significant. The parameter PeD = h s 2 (V F + V)/DLF determines the degree of nonequilibrium, with PeD > 1 being the case where diffusion can be neglected and PeD V and PeD = h s 2 (D(C/V F - 1)t) - 1 . The value of N is in the range of 1010 -1014 m-3 (Navon & Lyakhovski 1998) and the volume fraction of bubbles before the fragmentation is estimated to be between 0.3 and 0.6 (Melnik 2000). Even for an extreme set of parameters (N = 10 1 4 m - 3 ,D=10 - 1 1 m2s-1, = 0.6 and C/VF = 1.5) the condition of equilibrium mass transfer (PeD < 0.1) requires t > 40 s. During this time a fragmentation wave will travel several kilometres. Thus during magma fragmentation, mass transfer between melt and bubbles is negligible. In the second situation, after the fragmentation wave stops (VF = 0), we can rewrite PeD = hs2V/(DCt). Average velocity can be estimated by means of the Pousieulle law as V = pd 2 /32uL F where A/? is a pressure drop in the rarefaction wave, d is the conduit diameter and u is magma viscosity. Thus the final form of the criterion is: PeD = (3/4 N) 2/3 (l - 1/3)2 pd 2 /32uD(Ct) 2 . As magma degasses, the product uD is expected to remain approximately constant since melt viscosity increases and water diffusivity decreases as gas is lost. A typical value of uD is 10 - 4 N for u = 107Pa s and D= 1 0 - 1 1 m 2 s - 1 for Montserrat andesite at 850°C with 5wt% dissolved water in the rhyolitic melt (Barclay et al 1998). The upper estimate of PeD for N= 10 10 m- 3 and a = 0.3 gives PeD = 11.7/t 2 . Thus mass transfer reaches an equilibrium state after approximately 10s. The lower estimate for PeD (N = 10 1 4 m - 3 and a = 0.6) indicates equilibrium mass transfer when t > 0.23 s. Therefore the system can rapidly transform from negligible mass transfer to the equilibrium case when the fragmentation wave stops. There are several processes which have not yet been incorporated into models of conduit flow during explosive eruptions. These include continuous nucleation generating a size distribution of bubbles, interaction of bubbles at high concentrations, and a full
309
analysis of diffusion of gas in the melt shells surrounding growing bubbles. Incorporation of the diffusive bubble growth in particular requires elaborate numerical models (Proussevitch et al. 1993; Navon & Lyakhovski 1998). Here we consider only the two endmember cases: equilibrium mass transfer and total absence of mass transfer between melt and bubbles to bound the range of behaviours. For the gas-particle dispersion we assume that the particle temperature is constant and equal to the temperature of the gas phase. This is justified by a high heat capacity, large mass concentration of particles and small size of particles as in previous models (e.g. Wilson et al. 1980). We also neglect the variation of the gas temperature since there is rapid heat exchange between the gas and small particles. At the outlet of the volcanic conduit, where the particle concentration is low and the expansion rate of the gas is substantial, the gas temperature will decrease by no more than 3-5% (Barmin & Melnik 1993). We also consider that relative velocity between particles and the gas is small in comparison with the mixture velocity. With these assumptions the system of flow equations for both unfragmented magma and gas-particle dispersion (with u = 0) can be written in the following form:
(la) (1b) (1c) (1d) (1e) (1f)
Here the following notations are used: q are densities (subscript: g, gas; c, condensed phase, crystals plus melt; no subscript, mixture), p0g is a density of pure gas, V is the mixture velocity, pg and pc are pressures, n is number density of bubbles, a is the bubble radius, dis the conduit diameter, u is the mixture viscosity, T is magma temperature, R is gas constant, t is time and x is a vertical coordinate. The system consists of the continuity equations for gas and condensed components (Equations la and 1b) and number density of bubbles (Equation 1c), the momentum equation for the mixture as a whole (Equation 1d), as well as the Rayleigh-Lamb equation (see Scriven 1959) for bubble growth (Equation le). Equations 1f and lg are the perfect gas law and the volume fraction of bubbles. Gravity forces and the conduit resistance (in Pousieulle form) are taken into account in the momentum Equation 1d. We can introduce densities in a way to avoid mass transfer terms in the mass conservation Equations la and 1b. For the case of equilibrium mass transfer the densities of the gas and the condensed phase can be written as follows:
(2a) (2b) (2c) Here a is volume concentration of bubbles, (3 is volume concentration of crystals in condensed phase, pm and px are the densities of melt and crystals, c is mass fraction of dissolved gas, and Cf is a solubility coefficient. In Equations 2 the densities are defined in such a manner that Equation 2b incorporates both bubbles and dissolved gas, and Equation 2a incorporates only melt and crystals.
310
O. M E L N I K & R. S. J. SPARKS
In the case of no mass transfer, pg is the density of the gas in the bubble phase and pc includes the remaining dissolved gas in the melt phase:
(3a) (3b) We constrained magma rheological properties for the case of the Soufriere Hills eruption, as discussed by Melnik & Sparks (2002). Magma viscosity is a strong function of the volume concentration of dissolved gas and crystal content. In the case of equilibrium mass transfer, viscosity is calculated for equilibrium concentration of dissolved gas, and in the case of no mass transfer the conservation of viscosity in the melt phase is incorporated into Equations 1: (4)
The system of Equations 1 and 2 can be simplified due to the assumption of equal component velocities and equilibrium of mass transfer. From Equations la, 1b and 1c we can obtain the following integral relationship between the densities of components: (5)
which allows density and volume fraction of bubbles to be calculated as functions of pressure. The Rayleigh-Lamb Equation le can be developed in the form: (6)
In the case of no mass transfer = 0. Taking into account Equations 5 and 6, Equations 1 can be reduced to the Euler equations of gas dynamics with a complicated equation of state and an additional term due to pressure non-equilibrium:
Initial and boundary conditions We use the steady-state solution of the lava dome extrusion model (Melnik & Sparks 2002) to determine the initial distribution of parameters along the conduit. We assume that there is no further crystallization during the explosive eruption and that the crystal content (3) is constant. For the case of equilibrium mass transfer, volume concentration of bubbles is a function of pressure only, therefore gas loss due to the development of magma permeability is assumed to be negligible. This limitation is not assumed in the case of no mass transfer, where the volume fraction of bubbles is treated independently of pressure. We assume that the top 200m of the conduit is occupied by the gas-particle dispersion with pressure equal to atmospheric pressure: this avoids difficulties with boundary conditions at the initial stages of eruption before the flow structure is developed. Particular choices of the length of this zone make no difference to the outlet flow parameters a few seconds after the beginning of eruption. Values of parameters after Melnik & Sparks (2002) used for the calculations are listed in Table 1. Exit pressure at the top of the conduit is determined by the hydrostatic weight of the dome and its viscous resistance, and therefore depends on discharge rate and initial dome height before the collapse. We use a value of discharge rate (about 3 m3 s-1) that gives an exit pressure of 12MPa and a volume fraction of bubbles equal to 0.6 to compare the three fragmentation criteria. Figure 3 shows the pre-eruptive distributions of volume fraction of gas and pressure along the conduit for the two end-member cases. Only one boundary condition at the bottom of the conduit is needed in the case of equilibrium mass transfer and three boundary conditions in the case of no mass transfer. As the volume erupted in an individual explosion is much smaller than the chamber volume, it is reasonable to assume fixed pressure in the chamber throughout the eruption. The vesicularity and viscosity of the magma feeding into the base of the conduit are also assumed to be constant. If outflow from the conduit is supersonic or sonic, no boundary condition is needed because disturbances related to the atmospheric part of the flow cannot propagate into the conduit. In contrast, subsonic exit conditions require the conduit flow to be coupled with the flow in the atmosphere by assumption of continuity of all parameters at the top of the conduit. In this case, an artificial boundary condition is developed. We assume that, when the exit pressure is higher than atmospheric pressure and velocity is subsonic velocity remains sonic and
(7)
Table 1. Parameters for the Soufriere Hills andesite eruption used in modelling
Here A is a coefficient that indicates the flow regime: = 1 for bubbly liquid and A = 0 for gas-particle dispersion. In the case of no mass transfer, the volume fraction of bubbles remains an independent variable and pressure is a function of both density and a (Equation la). Because the conduit resistance differs strongly between the regimes of unfragmented magma and gas-particle dispersion (Fig. 2), magma discharge rate depends on the velocity of the fragmentation wave. Different criteria can be developed to characterize fragmentation. The earliest suggestion was that fragmentation occurs if the volume fraction of bubbles exceeds some critical value (Sparks 1978; Wilson et al. 1980). However, recent studies (Barmin & Melnik 1993; Melnik 2000; Alidibirov & Dingwell 1996; Papale 1999) suggest that fragmentation occurs when a critical overpressure (OP, defined as the difference between the pressure in a growing bubble and in the melt) or a critical elongation strain rate (SR) of magma is reached. In spite of different criteria for fragmentation, these latter two models suggest that the product of viscosity and elongation strain rate will reach a threshold value at fragmentation. We will study numerically the influence of different fragmentation conditions on the discharge rate.
Parameter
Symbol Value range
Magma chamber depth
L
5km
PC
l0MPa
Melt water content
C0
5%
Barclay et al. (1998)
Magma temperature
T
850 C
Barclay et al. (1998): Murphy et al. (2000)
Magma crystal content
3
0.6
Murphy et al. (2000)
Conduit diameter
D
30m
Dimensions of spines and early crater: hornblende reaction rims (Devine et al. 1998: Watts et al. 2002)
Density of melt
Pm
2300 kg m-3
Density of crystals
Px
2700 kg m-3
Solubility coefficient
cf
Magma chamber overpressure
4.1 x 10 - 6 Pa - 1
Information sources
Earthquakes (Aspinall el ill. 1998) and phase equilibria (Barclay el al. 1998)
2
Stolper (1982)
MODELLING OF EXPLOSIVE ERUPTIONS
311
Table 2. Calculation sets Run number
Fragmentation criteria
Mass transfer
1 2 3 4 5 6 7 8 9
VF (0.60) OP(l0MPa) SR(250MPa) VF (0.60) VF (0.70) VF (0.80) OP(l0MPa) OP(50MPa) SR (250 MPa)
Equilibrium Equilibrium Equilibrium No No No No No No
VF, fixed volume fraction of bubbles; OP, critical overpressure in growing bubble; SR, critical elongation strain rate. Critical values are given in parentheses.
Figure 4 shows profiles of (a) discharge rate, (b) volume fraction of bubbles and (c) pressure for run number 1 which adopts the volume fraction (VF) fragmentation criterion and equilibrium mass transfer, with initial conditions shown by curve 1 on Figure 3. A rarefaction wave propagates into the conduit, accelerating the mixture and decompressing it. The fragmentation wave, which follows the rarefaction wave, splits the flow domain into two different regions: bubbly magma and gas-particle dispersion. As the magma viscosity and volume fraction of bubbles are functions of pressure only, all the properties are fixed at the fragmentation level. High magma viscosity before fragmentation prevents rapid acceleration of the mixture in the rarefaction wave; therefore disturbances cannot propagate far into unfragmented magma. There is a zone of
Fig. 3. Distributions of (a) volume fraction of bubbles and (b) pressure in the conduit used as initial conditions in the calculations of unsteady explosive eruption. Curve 1 corresponds to conditions of impermeable magma; curve 2 takes into account gas loss due to the development of permeability in the ascending magma. Pressure at the top of the conduit is 12MPa.
pressure decreases in a way such that discharge rate at the outlet is equal to the discharge rate in the nearest mesh point in the conduit. This condition dampens oscillation of parameters during the transition from supersonic to subsonic regime and gives a monotonic decrease of exit pressure to the atmospheric value. This boundary condition can be overcome in a coupled model taking into account both conduit flow and atmospheric plume dynamics. This model is under development, but is not considered here. A detailed technical description of the numerical model is given in the Appendix.
Results of numerical modelling We present the results of simulations of the explosive eruptions generated by a sudden pressure drop at the top of the conduit. We focus on the role of fragmentation criteria and mass transfer between growing bubbles and the surrounding melt in the eruption dynamics. Parameters in the models are listed in Table 1. Table 2 summarizes the set of calculations for reference.
Fig. 4. Typical profiles of (a) discharge rate, (b) volume fraction of gas, and (c) pressure in the conduit during an explosive eruption. Fragmentation wave propagates down the conduit. Corresponding times (in seconds) are marked on the curves.
312
O. MELNIK & R. S. J. SPARKS
sharp gradients near the fragmentation level. As the velocity of the fragmentation wave decreases, gradients become smoother and the rarefaction wave separates from the fragmentation wave. Similar parameter distributions were observed for all other runs. Figure 5 shows (a) magma discharge rate at the outlet of the conduit and (b) position of fragmentation level versus time for runs 1 to 3 which investigate different assumptions for fragmentation with the same initial conditions and physical properties of magma as described in Figure 3. In case of the VF criterion, there is always a level in the conduit where the porosity reaches the fragmentation threshold, so fragmentation occurs continuously throughout the eruption. With time the system comes to a steady state, but this evolution is very slow, as the relaxation time in highly viscous magma is long. In the case of the OP and the SR criteria, fragmentation occurs as a series of discrete events. As the fragmentation wave propagates downwards, the pressure in the gas-particle dispersion after the fragmentation increases because its inertia prevents rapid evacuation of fragmented material. This leads to decrease in magma viscosity before fragmentation; consequently the critical fragmentation conditions cannot be reached and fragmentation stops. Between fragmentation events pressure above the fragmentation level decreases as material is evacuated and viscosity
Fig. 5. Variations of (a) discharge rate and (b) fragmentation depth with time for the case of equilibrium mass transfer between melt and bubble phase. Fragmentation criteria: VF, fixed volume fraction of bubbles; OP, critical overpressure in growing bubbles; SR, critical elongation strain rate.
increases again. Each fragmentation event leads to rapid inflow of newly fragmented material into the gas-particle dispersion zone, which causes sharp increases in flow rate. This increase propagates up the conduit and causes large changes in discharge rate at the surface. Thus the models incorporating more realistic fragmentation criteria produce strong fluctuations in discharge rate, although the broad evolution of the eruption is similar to all three cases. For run number 1 (VF fragmentation criterion. Fig. 5), the velocity at the top of the conduit is initially supersonic and reaches a maximum value of 2 5 0 m s - 1 after about 10s with a maximum Mach number value of 2.5. The maximum pressure at the top of the conduit is 2.7MPa. To estimate the characteristic timescale of discharge rate decrease we approximated the calculated values (after the maximum in discharge rate) by means of exponential functions. This approximation gives the characteristic timescale (when the current value is l/e times smaller then the initial value) of 140 s over the first 200 s of eruption, 1140 s from 200 s until the end of the explosion, and an average of 890 s for the whole discharge interval. The initial fragmentation wave velocity is 9 2 m s - 1 . but it decreases to 2 0 m s - 1 after 23s. After 10 minutes of eruption the velocity of the fragmentation wave is only 1 . 2 m s - 1 . The whole duration of eruption is about an hour (if explosive discharge is assumed to stop as exit velocity becomes less then 5 m s - 1 ) ; therefore only previously fragmented magma evacuates from the conduit during the last 50 min. Total erupted mass is 4.5 x 109 kg, which gives an erupted volume of approximately 1.7 x 10 6 m 3 dense rock equivalent of lava. The fragmentation penetrates to a depth of about 4km in run 1. In the case of OP and SR fragmentation criteria, the velocity of the fragmentation wave is much higher, reaching 260 and 2 0 0 m s - 1 respectively. As the influx of fragmented material is higher, the mixture accelerates up to velocities of 370 and 2 8 0 m s - 1 respectively (corresponding Mach numbers are 3.8 and 3.1). Fragmentation stops abruptly as critical values of OP or SR become unreachable. The duration of the first fragmentation event is 9s in the case of OP and 4.5 s in the case of SR. Then the boundary between the bubbly unfragmented magma and gas-particle dispersion starts ascending with a velocity of about O . l m s - 1 . As the pressure at the fragmentation surface is not fixed, the bubble fraction decreases to values of 0.25-0.3 as the fragmentation wave propagates downwards. After a series of several fragmentation events, the fragmentation level reaches a depth of 3500m for SR and arrives at the magma chamber for the OP criteria. Corresponding total erupted masses are 4.9 and 5.2 x 109 kg. In the initial stage of eruption the fragmentation wave almost follows the rarefaction wave and the assumption of equilibrium mass transfer is not applicable close to the fragmentation level. Thus, as discussed further below, the models are not likely to be realistic in the initial stages. We therefore examine the other endmember case of no mass transfer and the explosive eruption of a vesiculated magma column with a distribution of porosity and pressure given by the dome extrusion dynamics (Fig. 5; Melnik & Sparks 2002). Figure 6 shows the variation in (a) discharge rate and (b) the position of the fragmentation level with time for the case of no mass transfer and the VF fragmentation criterion for initial distribution of parameters represented by curve 2 on Figure 3. Runs 4-6 differ in the critical bubble fraction for the VF fragmentation criterion. For comparison, the results from run 1 are also plotted (dashed line). The velocity of the fragmentation wave is much smaller then for the case of run 1: 49, 27 and 1 4 m s - 1 for a = 0.6. 0.7 and 0.8 respectively. Corresponding maximum velocities at the top of the conduit are 130, 118 and 9 4 m s - 1 , with Mach numbers of just over unity. As we assume no mass transfer, the fragmented mixture is denser because it contains less gas (minimum gas fractions at the vent are 0.85, 0.89 and 0.92, respectively). Therefore the discharge rate (density x velocity x cross-sectional area of conduit) is similar to the case of equilibrium mass transfer (22.2 x l 0 6 k g s - 1 for run 1 and 19.2 x 1 0 6 k g s - 1 for run 4). Discharge rate decreases more rapidly than in the case of equilibrium mass transfer, as the rate of magma fragmentation is smaller and decreases faster as
313
MODELLING OF EXPLOSIVE ERUPTIONS
1 Fig. 6. Variations of (a) discharge rate and (b) fragmentation depth with time for the case of no mass transfer between melt and bubbles. VF fragmentation criterion: critical value of VF is labelled on curves. Dashed curve represents solution of equilibrium (e) mass transfer equations for comparison, with VF = 0.6. the fragmentation wave comes into regions of low porosity. Maximum depth of fragmentation is significantly smaller (1700m for run 4 instead of 4100m for run 1). Increase in the critical volume fraction value decreases the maximum value of the discharge rate by about two, but the discharge rate decreases asymptotically with the same speed, as it is controlled by the inertia of the fragmented gas-particle dispersion. Fragmentation depth also decreases as the threshold porosity for fragmentation increases. Figure 7 shows variations of (a) discharge rate and (b) depth of fragmentation level with time for the case of OP (10 and 50MPa) and SR (250 MPa) fragmentation criteria in the case of no mass transfer (runs 7, 8 and 9, Table 2). The use of a higher fragmentation threshold (50 MPa) in run 8 is justified by high volume content of crystals (about 60%) in Montserrat magma, which should increase the overall strength of magma. The curve for run 4 using the VF fragmentation criterion is shown for comparison (dashed lines). The initial velocity of the fragmentation wave is higher than in the VF case criteria (200 and 150m s-1 in the case of OP for runs 7 and 8, 1 1 0 m s - 1 in the case of SR). As the viscosity of magma in the conduit depends only on its preemptive distribution for the case of no mass transfer, its value decreases rapidly with depth and fragmentation stops quickly. As all
10 100 time (s)
1000
Fig. 7. Variations of (a) discharge rate and (b) fragmentation depth with time for the case of no mass transfer between the melt and growing bubbles. dissolved gas in this case remains in the melt, viscosity does not increase after the fragmentation stops. Maximum values of bubble overpressure or elongation strain rate decrease monotonically with time after the end of the initial fragmentation phase, as the intensity of the rarefaction wave decreases. Fragmentation processes cannot therefore start again. The assumption of no mass transfer must break down at later stages of the eruption and the model does not predict the asymptotic behaviour of the eruption correctly. There are two possible developments. First, diffusive mass transfer becomes sufficiently fast that gas-melt equilibrium is attained, so that much greater depths of the conduit are tapped, as in runs 1 to 3. Second, mass transfer may be too slow for explosive conditions so lava-dome extrusion resumes. In the case of the OP criterion, fragmentation stops after 6 and 4.5s and the depth of fragmentation is 870 and 580m for runs 7 and 9, respectively; in the case of SR criterion these values are 3 s and 250m respectively. When the fragmentation stops there is an outflow of previously fragmented material from the conduit. Discharge durations are 300, 130 and 80s, respectively. Figure 8 shows calculated (a) gas fraction, (b) exit velocity, and (c) bulk density of the eruption mixture for runs presented in Figure 7. Although the models differ in detail according to the fragmentation criteria, they all show common features. The maximum in velocity occurs in the first few seconds and declines thereafter. The maximum in density of the discharging mixture occurs several
314
O. M E L N I K & R. S. J. SPARKS density at later times will therefore tend to change conditions towards convective uprise. Summarizing the results of the calculations, assumptions on the intensity of mass transfer between melt and bubbles, and fragmentation criteria make significant differences to the eruption behaviour. In the case of no mass transfer, the velocity of the fragmentation wave and the duration of eruption both decrease compared to the case with mass transfer. Application of the VF criterion of fragmentation generates more moderate and longer-lived eruptions. In the case of equilibrium mass transfer, the velocity of the fragmentation wave reaches a few hundred metres per second, which is substantially higher than velocities observed in experiments (Alidibirov & Dingwell 1996). As discussed earlier, these high calculated velocities are unrealistic, because the timescales are too short for mass transfer to occur. Because of rapid influx of fragmented material into the gas-particle dispersion, the discharge rate at the top of the conduit is also overestimated. A totally non-equilibrium model produces fragmentation rates comparable with experimental data. A prominent feature of the equilibrium model is that fragmentation occurs in a series of separate pulses separated from each other by periods of magma ascent without fragmentation. Each fragmentation episode generates a sharp increase in eruption intensity at the outlet of the conduit.
Comparison of the results with field observation data
Fig. 8. (a) Volume fraction of bubbles, (b) exit velocity and (c) bulk density at the top of the conduit as a function of time for the case of no mass transfer. Abbreviations as in Figure 5. seconds later. The bulk densities and volume fraction of particles in discharging gas-particle dispersions are very high (100-250 kg m-3 and 0.1-0.3 respectively) and so the conditions for fountain collapse are strongly favoured (Clarke et al. 2002). Declining mixture
We have developed models for unsteady explosive eruptions using different fragmentation criteria, assumptions of the initial distribution of porosity, and assumptions of the importance of mass transfer between melt and gas phases. As summarized in Table 1, the models are illustrated with parameters thought to be relevant to conditions during the explosive activity of Soufriere Hills Volcano. We now compare features of the models with observations during this eruption. We suggest that the short-lived, repetitive Vulcanian explosions are examples close to the end-member case of no mass transfer. The timescales of the Vulcanian explosions were a few tens of seconds and are comparable to those predicted by the model. In both models and observed Vulcanian explosions, peak intensity occurs very early (order of 10 s). Peak discharges of several thousand cubic metres per second are predicted and are comparable to those estimated from column heights (Druitt et al. 2002) for some of the larger Vulcanian explosions. The natural explosions merged into a longer waning stage of vigorous degassing, which may represent the discharge of fragmented material and escape of gas from the top of the rising, but non-fragmenting magma below. The drawdown depths of Vulcanian explosions are estimated at a few hundred metres to c.2 km (Druitt et al. 2002) and are consistent with those estimated in the models. The models predict peak velocities at the early stage followed by a maximum in erupted mixture density after the order of 10s. Conditions for fountain collapse are strongly favoured in this first stage, and conditions favourable for convective column formation develop later as mixture density decreases, and also as mass transfer between melt and gas phases becomes more important. The broad sequence of fountain collapse, then vertical convective column development, is in accord with observations (Druitt et al. 2002). Figure 9 summarizes our interpretation of the two series of repetitive Vulcanian explosions in 1997. Magma rising in the conduit increases its internal pressure, resulting in inflation as recorded by tiltmeters and hybrid seismicity (Voight et al. 1999). The system reaches a condition where the internal pressure equals the tensile strength of the magma and an explosion is triggered. Melnik & Sparks (2002) have shown that steady solutions for lava extrusion show overpressures comparable to. or greater than, the tensile strength of crystal-rich magma (a fewmegapascals) at a fewhundred metres depth in the conduit. Thus the system will attempt to reach steady-state flow conditions, but will be interrupted when the overpressure exceeds the tensile strength. The system fails
MODELLING OF EXPLOSIVE ERUPTIONS
Fig. 9. Schematic view of the sequential (1-3) processes in the conduit during one cycle in the series of repetitive Vulcanian explosions. See text for explanation.
repeatedly in explosions. The field evidence of angular platy pumice supports brittle fragmentation and development of a fragmentation wave, as envisaged by Alidibirov & Dingwell (1996). These observations also indicate that the OP fragmentation criterion is to be preferred to the SR criterion, since there is no textural evidence for strain elongation in Montserrat pumice. An explosion ensues, building up quickly in peak intensity and then waning after about a minute. Most of the Vulcanian explosions resulted in fountain-collapse conditions. This is consistent with the no-masstransfer end-member, which releases much less gas, so that the erupted mixture densities are high. Fragmentation stops abruptly and unfragmented magma starts to rise. Although mass transfer must begin again, it is evidently too sluggish to re-establish or maintain conditions for explosive eruption. The downturn of the cycle involves deflation almost sufficient to recover all the strain stored during inflation (Voight et al. 1999). During this period there is vigorous ash-venting. We interpret this stage as due to gas being lost from the top surface of the rising magma column. Eventually the mass transfer and gas exsolution become sufficiently important that pressure once again starts to rise in the magma, with the system moving towards an overpressured state and towards the critical condition for another explosion several hours later. The sub-Plinian activity of 17 September 1996 differs from the Vulcanian explosions in several ways that suggest it is better described by the equilibrium case. The eruption was longer-lasting and more energetic, as shown by greater ranges of ballistics and by the total mass of ejecta being much greater than that of the largest Vulcanian explosion (Robertson et al. 1998). The eruption is estimated to have tapped down to 4km below the top of the conduit, and the subsequent rise of magma took two weeks with no development of repetitive explosive activity. There was also only a vertical convective column with no fountain collapse, suggesting low densities of the discharging mixture and substantial gas exsolution. The equilibrium mass transfer models show comparable features with larger exit velocities, drawdown to near the base of the conduit and comparable peak discharge rates of a few thousand cubic metres per second. A prominent feature of the models using the OP and SR fragmentation criteria is marked pulsations. We interpret the strong pulsations in the RSAM record of the seismic energy (Fig. 1) as recording the predicted pulsations. The equilibrium model is almost certainly unrealistic early in this eruption and so the prediction of the maximum intensity peak at about 10s is not thought to be realistic. Rather, we anticipate that the first few tens of seconds involved clearing of a rheologically
315
stiffened magma due to degassing and crystallization in the upper part of the conduit, before a discharge of deeper vesiculated magma could develop. This early stage is better modelled by the no-masstransfer model. Subsequently, each later fragmentation event would be better described by the no-mass-transfer model, as the timescale of these events is much shorter than that of diffusive bubble growth. The notion of an initial vent-clearing phase is consistent with strong ballistic ejection of denser lithologies from the uppermost conduit (Robertson et al 1998), followed by more vesiculated pumice. We interpret this eruption as evolving from the no-mass-transfer case to the equilibrium case, so a more comprehensive model which incorporates mass transfer will be required for complete description. An interesting feature of the 17 September 1996 explosive activity is the 2.5 hour gap between the end of dome collapse and the onset of the explosion (Robertson et al. 1998). This suggests that the unloading was not quite sufficient for the fragmentation conditions to be reached. We suggest that the unloading triggered gas exsolution in the conduit and this led to a build-up of critical conditions of overpressure in the uppermost parts of the conduit. In this sense the triggering mechanism is similar to that of the two series of Vulcanian explosions and is not explicitly unloading. Exsolution of gas in stiffened magma resulted in attainment of a critical overpressure threshold. Thus major dome collapses favour development of conditions for explosive eruption, but are not the specific trigger of the explosions themselves. Conclusions We have developed models of unsteady explosive eruptions following decompression of a magma column. The full problem would require incorporation of mass transfer as gas exsolves from the melt to the bubbles and would involve many complications and uncertainties. However, progress can be made by exploring two end-member cases. In the equilibrium case the distribution of gas between melt phase and bubbles is everywhere in equilibrium and so porosity is only a function of pressure. This case attains where the timescale for diffusion is small compared to the timescale of magma decompression and ascent. In the no-mass-transfer case only gas exsolved prior to onset of explosive discharge can expand and there is no further mass transfer. This case is attained when the timescale for diffusion is large compared to the timescale of explosive eruption. Our exploration of typical values of parameters suggests that both end-members can occur in nature. In particular we have interpreted the explosive activity at Soufriere Hills Volcano in terms of the two end-members. (1)
The Vulcanian explosions are envisaged as close to the nomass-transfer case where slowly rising magma develops a porosity and overpressure structure in the conduit that evolves to critical conditions for explosive fragmentation. The short durations, energetics and tendency for fountain collapse are consistent with the model, which predicts parameters broadly comparable to those observed. The associated patterns of repetitive explosions are also consistent with development of overpressures to trigger onset of explosive activity (Voight et al. 1999), followed by an abrupt halt of fragmentation and slower degassing. (2) The sub-Plinian activity of 17 September 1996 is interpreted as close to the equilibrium case, although the earliest stages and each discrete fragmentation event are likely to be closer to the no-mass-transfer case. Again the model predicts conditions consistent with observations such as discharge rates, eruption duration, volumes and inferred deep drawdown depth. (3) We have also studied the influence of fragmentation criteria on eruption dynamics, comparing the cases of constant-porosity criterion (VF), a critical overpressure criterion (OP) and a critical strain rate criterion (SR). Overall the differences in terms of averaged properties of mixture at the outlet of the conduit are small, as they are strongly controlled by the inertia of the gas-particle dispersion. The OP and SR criteria predict
316
O. MELNIK & R. S. J. SPARKS
strong pulsations for the equilibrium mass transfer conditions. Feedback processes cause fragmentation to pause and then resume as the erupted mixture is discharged. There are only subtle differences between the OP and SR results, although the former is favoured for Soufriere Hills Volcano on the basis of observations of shapes and textures of pumice clasts. The introduction of strong pulsations is in good agreement with seismic energy fluctuations (Fig. 1) and also observed pulsations in Vulcanian explosions. The authors acknowledge a grant from the Royal Society for the visit of O. Melnik to Bristol University, NERC Research Grants GR3/11683, GR3/11020 and GR3/10679. R.S.J.S. acknowledges support from the Leverhulme Trust (F/182/AL) and NERC through a Research Professorship. O.M. acknowledges support from the grant of Russian Foundation for Basic Research (99-01-01042). Comments by A. Barmin and A. Woods and reviews by A. Neri and T. Druitt were much appreciated.
Appendix
else end if
else where
k
— max(| -(u)|)
end if
Marquina's flux formula F M (u l . u r ) is then:
where r p (u l ), r p (u r ) are the right (normalized) eigenvectors of the Jacobian matrices A(u l ), A(u r ). The first-order scheme based on Marquina's flux formula is thus:
Numerical method Due to the presence of a moving boundary separating unfragmented magma and gas-particle dispersion with very different physical properties, computation of the unsteady problem for Equations 1 has some difficulties. For example, in the bubbly flow zone, the pressure gradient is governed by mixture weight and conduit resistance with the inertial term being small. In the gas-particle dispersion, conduit resistance is negligible and inertia is dominant. Near the fragmentation surface all terms in Equations 1 are of the same order. The fragmentation wave is a zone of very steep gradients of the main variables, and has a thickness much smaller than the conduit length (Barmin & Melnik 1993; Melnik 2000). Another difficulty arises from the differential form of fragmentation criteria (Equation 6). All complications listed above produce strong requirements on the numerical method in terms of accuracy and stability. We have used numerical methods of'large particles' (Davidov & Belotzerkovskii 1980) and the Lax-Wendroff method with additional smoothness of solution (Ramos 1995). Results were satisfactory only for weak explosive eruptions and the condition of fixed bubble fraction as a fragmentation criterion. Finally, we used the flux-splitting algorithm developed by Donat & Marquina (1996). The idea of this method can be explained if we rewrite the governing equations in vector form, keeping only the convective terms:
The values of unknowns u are calculated in the middle of the mesh cells; u l and ur represent the values on the left and right boundaries of the cell respectively. These values will be used for the calculation of fluxes (mass and momentum) between cells respectively. The algorithmic description of Marquina's flux formula is as follows. Given the left and right states, we compute the 'sided' local characteristic variables and fluxes:
for p= 1 , 2 , . . . , m . Here 1 p (ul), l p (u r ), are the (normalized) left eigenvectors of the Jacobian matrices A(u l ), A(u r ) (A(u) = f/ u). Let l (u l ),... , m (u/) and / ( u r ) , . . . , m (u r ) be their corresponding eigenvalues. We proceed as follows: For k = 1 , . . . , m, If (u) does not change sign in [ul ur], then If (u l ) > 0 then
As the magma viscosity is very high in the bubbly flow regime, viscous terms [4/3>( / x)] f( )( V x) and (32 d2) require high accuracy. They are treated implicitly. Therefore the momentum Equation 1d leads to a three-point implicit scheme which is solved by means of the Thomas method (Tannehill et al. 1997). Convective terms in this equation are taken in explicit form as given by Marquina's flux. The usual Courant-Friedriehs-Levy (CFL) condition and CFL number of 0.5 are used in all calculations to define the time step. The code was tested on the analytical solution of a shock tube problem for a perfect gas and gives an accuracy about 1 % for 500 nodes. As the fragmentation wave is an area of steep gradients we used finer grid (2500 points or 2m in space) to solve the flow accurately. A 500-point grid gives an error of 5% in the calculated discharge rate at the top of the conduit and 15% in evaluation of the position of the fragmentation level. The difference in solution using 2500 and 5000 grid points is negligibly small.
References ALIDIBIROV. M. A. & DINGWELL, D. B. 1996. Magma fragmentation by rapid decompression. Nature, 380, 146-148. ASPINALL, W. P.. MILLER, A. D., LYNCH, L. L.. LATCHMAN. J. L., STEWART, R. C. WHITE, R. A. & POWER. J. A. 1998. Soufriere Hills eruption. Montserrat: 1995-1997: volcanic earthquake locations and fault plane solutions. Geophysical Research Letters, 25, 3397-3400. BARCLAY. J., CARROLL, M. R., RUTHERFORD. M. J., MURPHY, M. D., DEVINE. J. D., GARDNER, J. C. & SPARKS. R. S. J. 1998. Experimental phase equilibria: constraints on pre-eruptive storage conditions of the Soufriere Hills magma. Geophysical Research Letters, 25. 3437-3440. BARMIN. A. A. & MELNIK. O. E. 1993. Features of eruption dynamics of high viscosity gas-saturated magmas. Fluid Dynamics, 28. 195-202. BONADONNA, C., MAYBERRY. G. C.. CALDER, E. S. ET AL. 2002. Tephra fallout in the eruption of Soufriere Hills Volcano, Montserrat. In. DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society. London, Memoirs, 21,483-516. CALDER, E. S., LUCKETT, R., SPARKS. R. S. J.. & VOIGHT, B. 2002. Mechanisms of lava dome instability and generation of rockfalls and pyroclastic flows at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London. Memoirs. 21, 173-190. CLARKE, A. B., NERI, A., VOIGHT. B.. MACEDONIO, G. & DRUITT, T. H. 2002. Computational modelling of the transient dynamics of August 1997 Vulcanian explosions at Soufriere Hills Volcano. Montserrat: influence of initial conduit conditions on near-vent pyroclastic dispersal. In: DRUITT, T. H. & KOKELAAR. B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs. 21. 319-348.
MODELLING OF EXPLOSIVE ERUPTIONS DAVIDOV, Y. M. & BELOTZERKOVSKII, O. M. 1980. Method of Large Particles in Gas Dynamics. Mir, Moscow. DEVINE, J. D., RUTHERFORD, M. J. & GARDNER, J. C. 1998. Petrologic determination of ascent rates for the 1995-1997 Soufriere Hills volcano andesite magma. Geophysical Research Letters, 25, 3673-3676. DOBRAN, F. 1992. Non-equilibrium flow in volcanic conduits and application to the eruption of Mt. St. Helens on May 18 1980 and Vesuvius in AD79. Journal of Volccanology and Geothermal Research, 49, 285-311. DONAT, R. & MARQUINA, A. 1996. Capturing shock reflections: an improved flux formula. Journal of Computational Physics, 125, 42-58. DRUITT, T. H., YOUNG, S. R., BAPTIE, B. ET AL. 2002. Episodes of cyclic Vulcanian explosive activity with fountain collapse at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 281-306. MELNIK, O. E. 2000. Dynamics of two-phase conduit flow of high-viscosity gas-saturated magma: Large variations of Sustained Explosive eruption intensity. Bulletin of Volcanology, 62, 153-170. MELNIK, O. E. & SPARKS, R. S. J. 1999. Non-linear dynamics of lava dome extrusion. Nature, 402, 37-41. MELNIK, O. E. & SPARKS, R. S. J. 2002. Dynamics of magma ascent and lava extrusion at the Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 153-171. MURPHY, M. D., SPARKS, R. S. J., BARCLAY, J., CARROLL, M. R. & BREWER, T. S. 2000. Remobilization of andesite magma by intrusion of mafic magma at the Soufriere Hills volcano, Montserrat, West Indies Journal of Petrology, 41, 21-42. NAVON, O. & LYAKHOVSKI, V. 1998. Vesiculation processes in silicic magmas. In: GILBERT, J. S. & SPARKS, R. S. J. (eds) Physics of Explosive Eruptions. The Geological Society, London, Special Publications, 145, 27-50. NORTON, G. E., WATTS, R. B., VOIGHT, B. ET AL. 2002. Pyroclastic flow and explosive activity at Soufriere Hills Volcano, Montserrat, during a period of virtually no magma extrusion (March 1998 to November 1999). In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 467-481.
317
PAPALE, P. 1999. Strain-induced magma fragmentation in explosive eruptions. Nature, 397, 425-428. PROUSSEVITCH, A. A., SAHAGIAN, D. L. & ANDERSON, A. T. 1993. Dynamics of diffusive bubble growth in magmas: isothermal case. Journal of Geophysical Research, 98, 22283-22308. RAMOS, I. J. 1995. One-dimensional, time-dependent, homogeneous, 2-phase flow in volcanic conduits. International Journal for Numerical Methods in Fluids, 21, 253-278. ROBERTSON, R. E. A., COLE, P., SPARKS, R. S. J. ET AL. 1998. The explosive eruption of Soufriere Hills Volcano, Montserrat, West Indies, September 17, 996. Geophysical Research Letters, 25, 3429-3432. SCRIVEN, L. E. 1959. On the dynamics of phase growth. Chemical Engineering Science, 10, 1-13. SPARKS, R. S. J. 1978. The dynamics of bubble formation and growth in magmas - a review and analysis. Journal of Volcanology and Geothermal Research, 3, 1-37. STOLPER, E. 1982. Water in silicate glasses: an infrared spectroscopic study. Contributions to Mineralogy and Petrology, 81, 1-17. TANNEHILL, J. C., ANDERSON, D. A. & PLETCHER, R. H. 1997. Computational Fluid Mechanics and Heat Transfer. Taylor & Francis, London. VOIGHT, B., SPARKS, R. S. J., MILLER, A. D. ET AL. 1999. Magma flow instability and cyclic activity at Soufriere Hills Volcano, Montserrat. B.W.I. Science, 283, 1138-1142. WATTS, R. B., HERD, R. A., SPARKS, R. S. J. & YOUNG, S. R. 2002. Growth patterns and emplacement of the andesitic lava dome at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 115-152. WILSON, L., SPARKS, R. S. J. & WALKER, G. P. L. 1980. Explosive volcanic eruptions - IV. The control of magma properties and conduit geometry on eruption column behaviour. Geophysical Journal of the Royal Astronomy Society, 63, 117-148. WOODS, A. W. & KOYAGUCHI, T. 1994. Transitions between explosive and effusive eruption of silicic magmas. Nature, 370, 641-645. YOUNG, S. R., SPARKS, R. S. J., ROBERTSON, R., LYNCH, L., MILLER, A. D., SHEPHERD, J. & ASPINALL, W. A. 1998. Overview of the Soufriere Hills Volcano and the eruption. Geophysical Research Letters, 25, 3389-3392.
Computational modelling of the transient dynamics of the August 1997 Vulcanian explosions at Soufriere Hills Volcano, Montserrat: influence of initial conduit conditions on near-vent pyroclastic dispersal A. B. CLARKE 1 , A. NERI 2 , B. VOIGHT1, G. MACEDONIO3 & T. H. DRUITT4 1 Department of Geosciences, Penn State University, University Park, PA 16802, USA (e-mail:
[email protected]) 2 CNR-CSGSDA, Department of Earth Sciences, Pisa, Italy 3 Osservatorio Vesuviano, Napoli, Italy 4 Laboratoire Magmas et Volcans, Universite Blaise Pascal et CNRS, Clermont-Ferrand 63038, France
Abstract: This paper presents numerical models of the Vulcanian explosions that occurred in 1997 at Soufriere Hills Volcano. Plume evolution and velocities were calculated for the well-documented and typical explosions of 6 and 7 August 1997, and these data and other observations were compared to transient, axisymmetric, multiphase flow simulations of coupled conduit evacuation and pyroclastic dispersal. Pre-explosion conduit conditions were estimated from Montserrat data, using a simple gas solubility law and assuming that conduit magma flow had stagnated with a constant overpressure prior to the explosions. Reference simulation input parameters include conduit diameter of 30m, crater diameter of 300m, meltwater content of , grain sizes of 30, 2000 and 5000um, and conduit overpressure of l0MPa. The numerical simulations of the explosions resolved highly unsteady vent exit conditions such as velocity, pressure and mass flux, and the spatial and temporal dispersal of pyroclasts during the initial few minutes was investigated using one gas phase and two or three solid phases representing pyroclasts of different size. Our simulations produced transitional eruptive regime behaviour, dividing the erupted mass into a portion that generated a radial pyroclastic current fed by a collapsing column, and a convective portion that generated a buoyant plume. This behaviour generally mimicked the observed explosions. The movement of different particle sizes was tracked, with fine particles dominantly influencing the convective behaviour of the central plume and ash plume thermals generated above the pyroclastic currents. Simulated initial vent velocities ranged from 85 to 120 m s - 1 , collapse heights ranged from 450 to 1370m above the vent, initial pyroclastic current velocities ranged from 40 to 6 0 m s - 1 with surge runouts to 1.8km, drawdown depths in the conduit were a few hundred metres, and simulated pyroclastic current deposit temperatures ranged between 135 and 430°C. Subsets of these results are in reasonable agreement with observed and measured parameters of the 1997 explosions. The best match was intermediate between our reference simulation, which assumed no loss of volatiles from the conduit during rise from the magma reservoir and which appeared too energetic, and another simulation in which much volatile leakage was assumed. The results suggest that volatile depletion in the conduit was an important factor in influencing the dynamic behaviour of the Vulcanian explosions on Montserrat.
A total of 88 short-duration Vulcanian explosions, nearly all accompanied by radial fountain collapse, occurred at Soufriere Hills Volcano, Montserrat in 1997 (see Fig. 1). Thirteen occurred in August and 75 more occurred during September and October. These explosions provided an unprecedented opportunity for repeated observation and monitoring (Fig. 2; Druitt et al 2002), as only a few events of this type have been observed closely (Nairn & Self 1978; Sparks & Wilson 1982; Hoblitt 1986). Because of the intensive real-time monitoring of precursory seismicity and tilt, it was possible over the short term to forecast (on the timescale of several hours with an uncertainty of tens of minutes) the onset of many of these explosions (Voight et al 1998, 1999; Druitt et al 2002), thus facilitating scientific preparations for impending explosions. Therefore, many explosions were documented in unusual detail by video and repetitive still photography, by theodolite surveying of eruption plumes, by tephra sampling, and by monitoring of broadband seismicity and deformation. Numerical models of short-duration Vulcanian explosions with time-varying vent flux have not received much attention, and no eruption model to date has tried to combine highly unsteady vent dynamics with explosive dispersal of pyroclastics. Recent improvements in numerical code development provided new opportunities for our study. Therefore, we have used an axisymmetric, multipleparticle-size, numerical code (Neri 1998; Neri et al. 2001b) to calculate the unsteady and transient vent flux following vent cap failure and resulting pyroclast dispersal (explosion) for a number of assumed pre-explosion conduit profiles. The observational data fall into two categories: (1) those which constrained initial conduit conditions and input parameters for our numerical models of explosive pyroclast dispersal; and (2) those which describe the explosions and provide a basis for comparison with the results of the numerical models.
In this paper, we first present a general description of the 1997 Montserrat explosions, and follow this by a brief review of eruption modelling, focusing on what is known about fountain-collapse pyroclastic current generation and Vulcanian explosions. Next, we summarize the pyroclast dispersal model we used (Neri 1998; Neri et al. 2001b) and our methods for determining the pre-explosion conduit conditions. The solutions focus on the near-vent region (within a few kilometres) over the first 90 to 150 s of the explosions. Then we use several numerical simulations to investigate the influence of key input parameters on results, namely conduit overpressure, mass fraction of water, and fines fraction. Finally, we compare our results against observations and inferred parameters of the 1997 explosions at Soufriere Hills Volcano in order to test the pyroclast dispersal (explosion) model and its assumptions. Characteristics of the cyclic Vulcanian explosions An overview of the 1997 Vulcanian explosions and their physical parameters is given in Druitt et al. (2002). The main features relevant to our numerical simulations are summarized here. The Vulcanian explosions of 1997 occurred in two episodes, the first between 4 and 12 August, and the second between 22 September and 21 October. The explosions began with an initial phase of ballistic throwout followed by a phase of powerful fountaining of a pyroclastic mixture, which lasted 10-20s. Condensation of atmospheric moisture caused by shock waves was noted on video footage of some explosions. The condensation patterns propagated radially away from the vent ahead of the expanding fountain of gas and pyroclasts during the first few seconds of the explosions. The vertical jets collapsed to generate fountains and high-velocity radial pyroclastic surges that moved to distances of about 2 km (Fig. 3),
DRUITT, T. H. & KOKELAAR, B. P. (eds) 2002. The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 319-348. 0435-4052/02/$15 © The Geological Society of London 2002.
319
320
A. B. C L A R K E ET AL.
Fig. 1. Map of southern Montserrat and Soufriere Hills Volcano, showing the active lava dome (vent) location inside English's Crater, the principal drainages (ghauts) about the dome, and the area affected by pyroclastic surges and flows during 1995-1999 (after Druitt et al. 2002). Photograph and video observation points include Montserrat Volcano Observatory (MVO South) and Fleming.
and also high-concentration pumice-and-ash flows that were generally confined to channels and ran out to distances of 3 to 6 km. Subsequently, buoyant-convecting plumes developed from the fountains and rose to heights of 3 to 15 km, where they generally spread out as umbrella clouds. The explosions were dominated by the buoyant plumes for roughly 500 s. Buoyant ash plume thermals also formed above the surges and channelled flows and tended to rise and move inward to join the main plume. The explosions ended with approximately an hour of waning exhalations characterized by a bent-over plume. Collapse of the fountain generally occurred 10 to 20s after explosion initiation, from heights of 300-650 m above the crater rim (400-750 m above the vent). The individual explosions expelled on average 3.0 x 10 5 m 3 of magma 1.1 x 10 5 m 3 as fallout and 1.9 x 105 m3 as surges and pumice-and-ash flows, evacuating the conduit to depths of 500 to 2000 m. Vent exit velocities ranged from 40 to 140ms - 1 . The pyroclastic surges developed slope-parallel velocities of 30-60 m s - 1 , whereas the channelled pumice-and-ash flows typically had velocities of 10 m s-1. The emplacement temperatures of pumice-and-ash flows ranged from 180 to 2200C (Druitt et al. 2002; Cole et al. 2002), with air entrainment cooling the mixture with respect to its eruption temperature of about 8600C (Barclay et al. 1998). The explosions occurred from a conduit 30m ( 5m) in diameter (Voight et al. 1999) into a flared crater, approxi-
mately 300m 0 m) in diameter. Our paper focuses mainly on two well-documented events at 14:35 local time (LT) on 6 August and at 12:05 LT on 7 August 1997 (all times given are local times), with simulations attempting to reproduce the near-vent behaviour over the first 90 to 150s. Background on explosion modelling and overview of our model Fountain collapse occurs when an ejected pyroclast-gas jet does not have enough momentum to continue rising, and, having failed to become positively buoyant, falls back toward and across the ground surface as a dispersion of hot particles (fragmented magma and lithic clasts) and gas (Sparks & Wilson 1976; Sparks et al. 1978; Neri & Macedonio 1996a). Early understanding of plume behaviour was derived from numerical work on turbulent gravitational convection (Morton et al. 1956). Experimental and steady-state, pseudogas models Sparks et al. (1978) described the physics of column collapse and pyroclastic flows by applying steady-state conservation of mass
MODELLING DYNAMICS OF VULCANIAN EXPLOSIONS
321
Fig. 2. Photographs of the explosion at 14:35 LT on 6 August 1997. Times after the low-frequency onset of the explosion signal at 14:35:12 LT: (a) 22s (b) 37s (c) 55 s (d) 91 s (e) 96 s. In (a) the column was about 800m high and collapse had begun, but pyroclastic currents had not clearly penetrated the veil of fallout. Gages Mountain (Fig. 1) is in foreground below plume. In (b), pyroclastic currents were descending Mosquito Ghaut (left) and Gages valley (right), and other major drainages around the volcano. In (c), the pyroclastic currents had thickened by development of dilute ash plumes, and the bulbous central plume, depleted in coarse particles, was rising buoyantly. By (d), active tan-coloured ash-rich thermals that developed over the pyroclastic current, and rose upward and inward, were being incorporated into the stalk of the buoyant central plume. A full view of the central plume is shown in (e), which rose ultimately to about 12km above sea level, where it spread to form an umbrella cap that was subsequently sheared from the main column and transported NE by high-altitude winds. Photographs taken from 7 km NW at MVO South by B. Voight.
and momentum to a gas-particle mixture in which the particles and gas are in thermal and kinetic equilibrium. Bursik & Woods (1996) added the conservation of energy expression to the description of the ash flows but did not develop the fountain model further. In all these studies the gas-particle mixture, or pseudogas, is treated as a single-phase fluid with bulk properties determined by the volumetric proportion, size and temperature of the particles. The main parameters controlling fountain formation were found to be the vent radius, gas content and initial vent velocity (or a combination of exsolved gas content and vent velocity, i.e. vent mass flux), whereas flow runout is also controlled by air entrainment and particle sedimentation (Bursik & Woods 1996). Reasonable agreement has been claimed between theory and observations (Wilson et al 1978; Turner 1979; Sparks & Wilson 1982; Sparks 1986); however, as noted by the original authors, these relationships are valid only within the context of one-dimensional steady-state pseudogas theory, where pyroclasts are very fine-
grained, and they were generally developed for steady Plinian-type eruptions. They have been used to estimate first-order time-averaged column behaviour, but the required simplifications impede detailed comparison with the transient, unsteady, multidimensional, and multiphase nature of real eruptions. Shock-tube and gas-particle experiments also suggest that the pseudogas, steady-state, onedimensional approximation may not be a good assumption for highvelocity two-phase volcanic flows (Anilkumar et al. 1993; Sparks et al. 1997; Neri & Gidaspow 2000).
Steady discharge, multiphase models Computer codes that were originally designed for atomic explosion simulation or fluidization studies have been adapted and further developed for the study of volcanic eruptions (Harlow & Amsden
322
A. B. C L A R K E ET AL.
Fig. 2. (continued)
1971, 1975; Amsden & Harlow 1974; Valentine & Wohletz 1989; Valentine et al. 1992; Dobran et al 1993; Gidaspow 1994). These codes enabled solutions to time-dependent, two-phase (one solid, one gas) compressible Navier-Stokes equations. They addressed thermal and kinetic disequilibrium between the solid and gas phases and time-dependent behaviour of the plume and the flows, while holding vent conditions steady. In addition to the previous findings of Sparks et al (1978) and Woods (1988) these results showed that plume behaviour is sensitive to the ratio of conduit pressure to atmospheric pressure at the vent (Valentine & Wohletz 1989). The codes also led to recognition of important relationships between plume behaviour and simulation particle size. Under otherwise identical vent conditions, simulations with larger particle sizes (larger Rouse numbers) are more likely than those with smaller particle sizes to fall back and form pyroclastic currents (Valentine & Wohletz 1989). This is caused by two phenomena, namely slower heating of the surrounding gas by the larger particles and their higher settling velocity.
Neri & Macedonio (1996b) made a first attempt to account for the grain-size distribution of the eruptive mixture by adding a second particle phase to the numerical solution of Dobran et al. (1993). Different drag terms between gas and particles were included for different particle sizes, and a drag term was added to account for collisions between particles of different sizes. Their model results indicate that the particles of different sizes have considerably different dynamics and affect one another's behaviour, thus changing the resulting plume behaviour and runout dynamics of the pyroclastic currents. Although the aforementioned numerical models provide insight into many of the factors that control eruption column behaviour, like the pseudogas models described above, they specifically address Plinian-style eruptions where the vent conditions are quasi-steady. None directly treats the sudden decompression of a pressurized conduit or chamber, or resulting short-pulse explosions, and therefore they do not directly apply to the Vulcanian explosions witnessed at Soufriere Hills Volcano in 1997.
MODELLING DYNAMICS OF VULCANIAN EXPLOSIONS
Fig. 2. (continued)3
323
324
A. B. CLARKE ET AL.
Fig. 3. Schematic representation of selected features of Vulcanian explosions.
Vulcanian models Several studies have explored the physics of short-duration Vulcanian explosions, with important contributions by Self et al. (1979), Kieffer (1981), Turcotte et al. (1990), Fagents & Wilson (1993) and Woods (1995). Self et al. (1979) treated the mixture as a pseudogas, and compared results to Vulcanian explosions observed at Ngauruhoe, New Zealand. Kieffer (1981) developed equations to describe the sudden decompression of a high-pressure magma chamber due to the failure of the north flank of Mount St Helens on 18 May 1980. Fagents & Wilson (1993) examined ranges of pyroclasts ejected. Turcotte et al. (1990) developed a set of equations based on the one-dimensional shock-tube problem, using unsteady conservation laws for mass and momentum and treating the mixture as a pseudogas. Woods (1995) extended the Turcotte et al. model by allowing the mixture to cool adiabatically, and by including a factor to account for particles not in thermal equilibrium with the gas. In addition, Wohletz et al. (1984) used the KACHINA numerical code (Amsden & Harlow 1974) to simulate a caldera-forming eruption through a sudden decompression of a non-homogeneous magma chamber. The code treats one solid phase and one gas phase, and focuses on the initial unsteady blast phase, and on shock wave phenomena.
Our model The previous Vulcanian models addressed transient vent conditions, and related the initial conduit pressure to exit conditions. In this paper, we advance the work of our predecessors by (1) applying a two-dimensional model with multiple particle sizes, which solves for unsteady conduit flow and resulting pyroclast dispersal, (2) constraining pre-eruptive conduit conditions with analyses of observational data, (3) comparing explosion model simulations with well-documented explosions, and (4) varying some key conduit parameters in order to better understand their effects on explosion simulation results.
The pyroclast dispersal model, named PDAC2D (Neri 1998; Neri et al. 200la,b), is an extension of the three-phase flow code used by Neri & Macedonio (I996b) and solves a set of equations expressing the conservation of mass, momentum and energy for one gas phase, and a number of solid phases representative of particles of different sizes. The gas phase consists of a mixture of water vapour and atmospheric air. The fundamental transport equations are solved on an axisymmetric computational grid (5m to 40m grid spacing), as discussed in detail in the Appendix.
Input conditions for explosion model For our simulations of short-duration Vulcanian explosions, several input parameters were required. The most important of these were the conduit, crater and plugging cap geometry, the twodimensional (axisymmetric) topography of the region surrounding the vent, the initial conduit gas pressure, gas mass fraction and gas volume fraction as functions of depth in the conduit, and the sizes and densities of solid particles. Information acquired on Montserrat was used to constrain these input parameters, as described in the following sections.
Conduit and crater diameters, cap thickness and ground topography The conduit diameter of approximately 30m ) was constrained by spine dimensions, magma ascent rates and volume extrusion rates (Voight et al. 1999). The crater diameter of roughly 300m 0 m) was measured using dual-position theodolite measurements approximately one half hour prior to the 12:05 event on 7 August. The crater depth was approximately 100 m. The cap thickness was assumed to be 20m, which is a size roughly consistent with the estimated total volume of dense clasts per explosion (roughly 1-5% by volume of pumice-and-ash flow deposits).
Fig. 4. Topography used in the simulations. The grid is defined for an axisymmetric solution, thus the crater and conduit dimensions are radii.
MODELLING DYNAMICS OF VULCANIAN EXPLOSIONS
We used the ground topography representative of the north side of the volcano (Fig. 4). The north sector is characterized by a relatively smooth, low-slope fan with average slope approximately 22°, bordered by channels which, by mid-summer 1997, were nearly filled by pyroclastic debris. The model sloping-fan topography thus was more or less realistic for a broad sector of the near-vent region. The model assumption of axisymmetric flow is a reasonable approximation to reality, because the Vulcanian events in August of 1997 are described as axisymmetric (Druitt et al 1998, 2002), with fountain collapses and radial pyroclastic surges occurring simultaneously in all sectors and with pyroclastic current runouts similar in all directions.
Pressure and gas fraction profiles along the pre-explosion conduit For this study, the pre-explosion conduit was modelled quite simply, mostly on the basis of Sparks (1997). We assumed that magma flow in the conduit immediately prior to the explosions had stagnated due to viscous plugging at the vent. Specifically we assumed constant overpressure with depth, since for magma systems with large vertical viscosity gradients, almost all excess pressure drop from the chamber to the surface is concentrated near the top of the conduit (due to the higher viscosity of the magma near the surface), with excess pressure being roughly constant below the upper reaches of the conduit. There is some basis for assuming stagnated flow, because the Vulcanian explosions occurred in association with oscillatory flow (Voight et al. 1999), and modelled conduit flow rates were relatively small just prior to explosions (Wylie et al. 1999; Denlinger & Hoblitt 1999). The actual flow conditions and overpressure distributions were undoubtedly more complicated (Melnik & Sparks 2002a), but our assumptions were practical for a first approximation. We estimated gas mass and gas volume fractions as functions of depth in the conduit using the following solubility and hydrostatic laws. Symbols are summarized in Table 1. The mass fraction, ne, of exsolved water vapour in the bulk magma (melt + crystals + vesicles) at a given conduit depth was taken according to Henry's law as: (1)
where n0 is the mass fraction of water in the melt, z is depth in the conduit, Pg(z) is the total gas pressure in the conduit at depth z, s and (3 are experimentally determined constants for the appropriate melt composition, and 0 is the total volume fraction of crystals in the upper conduit magma (excluding vesicles). The water dissolved in the melt in the chamber, n0 was taken as 4.3 0.5 wt% (Devine et al. 1998; Barclay et al. 1998) and the total crystal volume fraction, was taken to a constant 0.65 (Murphy et al. 2000). The chemistry of the melt phase is rhyolitic (Murphy et al. 1998), for which s = 4.1 x l0 -6 N1/2m -1 and 0 = 0.5 (Wilson et al. 1980). The pressure distribution, Pg(z), is given by:
Jo
(2)
where AP is the overpressure at a given depth, which we assume to be constant with depth for each of the simulations presented in this paper, p is the bulk density of the overlying gas-magma mixture, and g is acceleration due to gravity. The water vapour is treated as an ideal gas with density, pg(z), as follows:
325
Table 1. Symbols and variables Symbol Description Ah aatm aj aw CD..k Cpgik Cs dk,j Dgk Dkj e Ev g G(Eg) hc hg Kk M Ms ne n0 Nuk Patm Pg Pg(z) P2 Pr Qk R R Rek s t Tg,k VDREA VDREE Vmax Vj vg vkj z a (3 P £g
Cross-sectional area of ballistic block Acoustic velocity in atmospheric air Acceleration of ballistic block in j direction Acoustic velocity in water vapour Drag coefficient for kth solid particle size Specific heats of gas and kth solid particle size Smagorinsky's constant (assumed to be 0.1) Particle diameter of kth and jth solid particle sizes Gas-solid drag coefficient Drag coefficient between kth andy'th solid particle sizes Restitution coefficient for a particle-particle collision Specific expansion energy of conduit water vapour Acceleration due to gravity Solid elastic modulus Fountain collapse height (above the vent) Gas enthalpy Thermal conductivity of kth solid particle size Mass of ballistic block Mach number of shock wave Mass fraction of exsolved water vapour in the bulk magma Mass fraction of water in the melt Nusselt number for kth particle size Atmospheric pressure Total gas pressure Total gas pressure in the conduit (at depth z) Pressure on high-pressure side of shock wave (see Fig. 5) Prandtl number Heat transfer coefficient between the gas and the kth particle size Ideal gas constant for water vapour Radius of conduit Reynolds number of kth particle size Solubility constant for Henry's law Time from onset of explosion Viscous stress tensor for the gas and kth solid particle size DRE volume available in the conduit DRE volume ejected from the conduit Maximum vent velocity for a given simulation Velocity of ballistic in j direction Velocity of water vapour Velocity of kth and j3th solid particle sizes Depth in conduit Restitution coefficient for non-head-on collisions Exponent for Henry's law Gas overpressure Gas volume fraction E Volume fraction of kth solid particle size k Ekj Maximum solids volume fraction for particles of size j and k ES Solid volume fraction in conduit Specific heat ratio for atmospheric air atm m Specific heat ratio for water vapour-magma mixture w Specific heat ratio for water vapour Gas molecular viscosity g Effective gas viscosity ge Effective gas turbulent viscosity gt k Viscosity of kth solid particle size Particle volume fraction of kth or jth size at maximum packing k.j p Bulk density of gas-magma mixture pa Atmospheric air density Pg Gas density Pkj Density of kth and jth solid particle sizes ps Density of melt-crystal mixture (solid) in the pre-explosion conduit Gas deformation tensor g Coulombic repulsive component among solid particles Solid stress tensor for kth particle size
(3)
where R = 4621kg -1 K - 1 i s the universal gas constant divided by the molecular weight of water, and T is the temperature of the water vapour, which we assume to be isothermal with the meltcrystal mixture (from here on referred to as 'solid') at 1133K
(860°C). If the density of this solid phase, ps, is constant with depth, then the bulk density of the gas-solid mixture is: (4)
326
The solid volume fraction
A. B. CLARKE ET AL. s
at a given depth is: (5)
And the gas volume fraction is simply: (6)
At the depth where the solid fraction s — 0.70 (30% vesicularity), which represents a bulk density slightly below 2000 kg m - 3 , a solid boundary is assumed in our pyroclastic dispersal model. Although this assumption is arbitrary, it is necessary to specify a solid boundary at the base of the conduit in order to effectively simulate the dispersal of the particles. This does not appear to be an unreasonable assumption because our pumice density measurements show that only 13% of randomly sampled clasts have vesicularities less than 30%. Because magma viscosity was very high, bubble expansion after fragmentation was probably minimal (Thomas et al. 1994; Druitt et al, 2002). Therefore, we assume that the vesicularity of pumice represents essentially the pre-explosion state of magma in the conduit, allowing us to claim that roughly 87% of material ejected had pre-explosion solid volume fraction, s < 0.70 (>30% vesicularity). The depth at which s = 0.70 (30% vesicularity) is referred to as effective conduit depth, DE. The dense rock equivalent (DRE) volume of solid material above this boundary is called DRE volume available, VDREA. The DRE volume of solid material ejected permanently from the conduit, that is, that which does not fall back into the conduit during the simulation, is termed DRE volume ejected, VDREE .
Pre-explosion specific expansion energy It should be noted that changing a single conduit parameter (e.g. overpressure or bulk H2O mass fraction), changes several characteristics of the modelled pre-explosion conduit, including VDREA and the effective depth of the simulation conduit. Because of this, it is difficult to attribute differences in simulated explosion behaviour to a single changed conventional conduit parameter. Therefore, to assist interpretation, we combined total pressure, exsolved gas mass fraction, DRE volume available, and conduit depth into a single parameter. This parameter represents the expansion energy due to gas overpressure per unit mass of solid material and is called specific expansion energy, Ev, which provides a convenient way to compare simulation results. Due to the very fast decompression of the system, we assumed adiabatic conditions (Woods 1995) and used the following equation to calculate this energy (Self et al. 1979; Neri et al. 1998):
(7)
where is the specific heat ratio for water vapour (the specific heat at constant pressure divided by the specific heat at constant volume: . = 1.25), Patm is atmospheric pressure, Pg(z) is the total gas pressure, DE is the effective depth of the conduit, r is the radius of the conduit, and is the mass of the material for a segment of the conduit of depth dz.
Ballistic analysis Numerous ballistics were observed during the initial seconds of the Vulcanian explosions of early August 1997. These ballistics were not affected by the motion of the plume because, in general, they were very large and dense, with fall velocities close to the initial explosion gas velocity, making them Type I ballistics according to Self et al. (1980). Documentation of the resulting crater size and position, and clast diameters, estimated from a helicopter, resulted
in a set of 24 ballistic range and size data (Druitt et al. 1998, 2002). We have used these data, along with basic dynamics equations and drag relationships, to estimate the initial velocities of the ballistics (Wilson 1972; Fagents & Wilson 1993; Waitt et al. 1995). The basic equation adopted for the ballistic analysis is dv j /dt = aj where aj is the acceleration (or deceleration) of the ballistic due to components of gravity and drag in the j direction. This acceleration is approximated by the methods of Wilson (1972) and Waitt et al. (1995) as follows: (8) and (9)
where vx and vv are the ballistic velocities in the horizontal and vertical directions respectively, pa is the density of the surrounding atmosphere, v is the total velocity of the ballistic, Ab is the ballistic cross-sectional area, CD is the drag coefficient for the ballistic, M is the ballistic mass, and g is acceleration due to gravity. Calculations for M and Ab assume a spherical clast and dome rock density of 2250 kg m-3 (Sparks et al. 1998). The coefficient of drag, CD, for smooth spheres was assumed. CD varies from 0.07 to 0.20 and is a function of Reynolds number, Re, such that CD = 0.11 log(Re) - 0.55 over a range 4 x 105 < Re < 6 x 106, estimated from Achenbach (1972) and Waitt et al. (1995). The drag coefficient is poorly constrained because the shapes of the ballistics vary greatly. Others, such as Wilson (1972), Fagents & Wilson (1993), Fudali & Melson (1972) and Self et al. (1979), have assumed that irregularly shaped ballistics have CD ranging from 0.7 to >1.0. However, in a more recent analysis, Waitt et al. (1995) suggest that these high values of drag coefficient result in unrealistically high launch velocities, particularly for smaller ballistics. As theoretical justification, they offer the evidence that a dimpled golf ball achieves a greater range than a smooth one, illustrating that the rough surface of ballistics can contribute to reduced drag. In general, although roughness increases skin-friction drag, it delays flow separation and reduces form drag, which is the largest component of total drag. Equations 8 and 9 were solved for initial velocities (vx.i and vy.i) by a fourth-order Runge-Kutta numerical scheme as done by Wilson (1972) and Waitt et al. (1995). We used an average ground slope of 22°. Drag was neglected in the first 1 0 - 2 s because the initial velocity and therefore the initial drag were not known. There was no significant difference in calculated initial velocity when drag was neglected during the first 1 0 - 2 s compared to when drag was neglected during the first 10 - 4 s, indicating that neglecting drag for the first 1 0 - 2 s did not significantly affect results. Total initial ballistic velocity is referred to as Vi. Launch angle is a significant factor. The angles for any specific ballistic clasts and associated craters are unknown. Therefore we executed the method above using angles at 5C intervals, from 30° to 70°. Values of initial velocity ranged from 93 to 1 1 6 m s - 1 for a 30C launch angle and from 131 to 1 6 9 m s - 1 for a 70 0 launch angle. The optimum launch angle was between 30 and 350.Druitt et al. (2002) carried out similar calculations using the ballistic model of Self et al. (1980), which adopts higher drag coefficients than those used here. For an optimum launch angle of about 350, they estimated launch velocities of up to 1 6 0 m s - 1 . Analysis of video footage gives exit velocities of between 40 and 1 4 0 m s - 1 (our work, and Druitt et al. 2002), suggesting that the Self et al. (1980) model overestimates exit velocity and that the values calculated in this paper are probably more realistic.
Estimation of overpressure Estimation of the gas overpressure beneath the conduit cap has been approached in several ways. Robertson et al. (1998) suggested explosion pressures of 10-27MPa for the sustained explosion of
MODELLING DYNAMICS OF VULCANIAN EXPLOSIONS
327
17 September 1996. Voight et al (1999) suggest that tilt data could be used to calculate pre-explosion overpressures, but simple halfspace elastic models resulted in only upper-bound estimates of several tens of megapascals. In the same paper, an extrusion model led to an estimate of overpressure approximately 11 to 25 MPa. The conduit flow models of Melnik & Sparks (2002a) predict overpressures up to 10 MPa. A lower-bound estimate of overpressure is the rock tensile strength, which is on the order of 4 MPa (Voight et al 1999). Druitt et al. (2002) use pumice vesicularities (55-75%) to place total confining pressures between 5 and 15 MPa. The highest of these vesicularities probably represent the upper parts of the conduit immediately beneath the cap. Our assumed 20m thick cap represents less than 0.5 MPa of overburden; therefore the calculated overpressure falls between 4.5 and 14.5 MPa, which is in reasonable agreement with the aforementioned values determined by independent methods. We have therefore chosen the middle of the range of estimates, roughly 10 MPa, as a reasonable estimate of overpressure for our reference simulation. In one of our variations, we used an overpressure of 7 MPa to illustrate the effect of reduced overpressure on results.
Fragmentation description Models of magma flow unsteadiness developed by Melnik & Sparks (2002b) test the significance of different fragmentation assumptions, such as fragmentation at a fixed volume concentration of bubbles (Sparks 1978; Wilson et al 1980), at a bubble overpressure threshold in excess of the magma tensile strength (Melnik 1999), or at a critical elongation strain rate (Papale 1999). The field evidence at Montserrat includes angular platy pumice that lacks bubble elongation texture (Druitt et al 2002). These observations suggest brittle fragmentation produced by a fragmentation wave (Alidibirov & Dingwell 1996) and support the overpressure threshold fragmentation criterion of Melnik (1999). At Soufriere Hills Volcano the sudden decompression of the conduit, due to fracture and disruption of the conduit cap, occurred when the strength of the cap rock was exceeded by the overpressure in the conduit. This disruption of the cap greatly increased the pressure difference between the vesicles near the top of the conduit and the surrounding environment, which was suddenly reduced to atmospheric pressure (Fig. 5). This new pressure state exceeded the fragmentation threshold, thus initiating the fragmentation wave. The following paragraphs describe how our model represents this condition and enumerate the associated assumptions. In our numerical model, the pre-explosion conduit at time t=0 was represented by a two-phase mixture whose properties (pressure, gas mass fraction and gas volume fraction) were described as a function of depth by the equations presented above. The 30m diameter conduit was discretized in a number of cells 10 m deep and 7.5m wide and no radial variation in the initial conditions was considered. At t = t1 > 0, the pressure disequilibrium between the overpressured conduit and the confining pressure produced a decompression wave which travelled down the conduit. As the decompression wave reached a particular depth, the mixture of gas and particles at that depth began to flow out of the conduit, whereas the undisturbed portion of the mixture below the decompression wave remained in magmastatic equilibrium. Such a representation of the conduit embodies three key assumptions. We first assumed that the melt phase at any depth quenched and fragmented nearly instantaneously when the decompression wave reached it, allowing the mixture of gas, crystals and melt to be represented as a gas-particle flow. This is supported by the results of Melnik & Sparks (2002b) which indicate that fragmentation occurs immediately after the decompression wave has reached a given depth. The second assumption is that the fragmentation process involved no energy loss, which in any case was probably relatively small (1-5% according to Alidibirov 1994). Third, we assumed that no further water exsolution occurred once
Fig. 5. Schematic representation of the numerical conduit, which defines the initial conditions for explosion simulations. The conduit is subdivided into grid cells, with input parameters that include solid volume fraction, S(Z), water vapour volume fraction, g(z), diameters of particles, and total pressure of water vapour, Pg(z}. Input values are discussed in the text and summarized in Table 2. The figure on the left also represents a shock tube at time zero, representing the pre-explosion, capped conduit, and on the right at time t1, representing the condition after disruption of the caprock.
disruption of the cap and decompression began. The propagation of a fragmentation front, and acceleration of the fragmented gasparticle mixture to the surface, is too rapid for significant diffusive mass transfer to occur during Vulcanian explosions (Gardner et al. 1999; Melnik & Sparks 2002b), and thus only gas that had exsolved before the cap was disrupted participated in the explosion.
Particle sizes The sizes of solid particles chosen for our simulations (up to three sizes) have been constrained by our grain-size analyses of the pumice-and-ash flow deposits, which considered the full range of particle sizes observed in the outcrops. Our estimates of the clasts ranged from a maximum size of 50cm to a minimum size of 0.9 m. The mean grain size of the pumice-and-ash flow deposits, accounting for a 30 wt% loss of fines (5m 3 s - 1 ), although it slowed during the later part of December (Sparks et al. 1998). The dome reached its maximum elevation (c. 1030m above sea level) for any time during this first phase of the eruption (before November 1999), and loading of the dome and talus in the southern sector led to a condition that was increasingly unstable. Seismic activity was generally low, although large hybrid earthquakes occurred in late October and at the beginning of November. A gradual build-up in hybrid earthquake energy began on 22 December, and developed into a hybrid swarm on 24 December. The swarm included periods when individual events merged into continuous tremor. With increasing edifice loading and increasing seismicity, the potential for sector collapse was enhanced to levels similar to those late in 1996, although the geometry of the dome and edifice complex was now substantially different. Sector collapse took place at 07:01 GMT (03:01 local time) on 26 December (Sparks et al. 2002). A volcanic debris avalanche was generated by a landslip that involved the edifice between Galway's Soufriere and parts of Galway's Wall, and included substantial amounts of dome talus (Voight et al. 2002). The sector collapse facilitated rapid depressurization of the lava dome and explosive initiation of a high-energy pyroclastic density current (Woods et al. 2002; Ritchie et al. 2002). The entire southwestern sector of the volcano was devastated, involving an area of 10 km2 (Sparks et al. 2002).
Hazard management and anticipation of critical events Detailed hazard maps and hazard management on Montserrat concentrated until September 1996 on pyroclastic flow and ashfall hazards. The explosive activity of 17 September 1996 led to a reappraisal of volcanic hazards to include vertical explosive eruption phenomena. However, it was not until the onset of deformation of the southern edifice wall that assessment of sector collapse and associated lateral explosion hazards was made. The likelihood of sector collapse was informally assessed by MVO staff in late November 1996 as representing approximately a 30% probability for collapse within a three-month period. A revised risk map was
357
issued to the public on Montserrat at that time (Kokelaar 2002), with the zone of highest risk being around the southwestern flanks of the volcano and in the White River valley. A small but resolute population had remained in the village of St Patrick's (Fig. 1) until late November, but, with the signs of instability visible, they were persuaded to evacuate. A formal assessment of the probability of sector collapse using techniques described by Aspinall & Cooke (1998) was made at MVO on 21 December 1996. The likelihood of a sector collapse and subsequent lateral blast in the Galway's sector was then assessed at 15% in three months, suggesting a reduced risk in comparison to the previous assessment. An identical result was found in a subsequent formal elicitation on 7 March 1997. Identification of the potential hazard prompted efforts to upgrade the monitoring systems specifically to provide warning of an imminent sector collapse. Five real-time indicators were watched closely for signs of rapid increase: RSAM, hybrid earthquake magnitude, frequency of occurrence of Galway's Wall rock avalanches, crack extension and inflationary ground tilt. Earthquake types and locations were also closely monitored in real time. The upgraded systems proved very useful, but were recognized as not necessarily definitive. The problems of instrument site accessibility, the risk to staff involved in monitoring, and the limitations of available instrumentation, were severe and provided practical limits to what could be accomplished. A major managerial problem remained in that confident short-term (hours to days) warning for a sector collapse could not be guaranteed. Limit-equilibrium models were conducted in December 1996 for the purpose of investigating the stability of the Galway's Wall and Chances Peak area, and understanding the nature and significance of the rock avalanches that were occurring at that time (B. Voight, unpublished data). The analyses suggested that the shallow-slab failures of the wall that generated rock avalanches were primarily due to the dynamic loading imposed by the shallow strong hybrid earthquakes, rather than due to any fundamental instability of the wall itself. Loading from the static weight of dome rock that then existed behind the wall was viewed as subcritical, although a possible future influence of more complicated pressure patterns due to endogenous behaviour or changes in edifice geometry could not be precluded. The results were communicated to MVO by telefax on 14 December. Finite-element modelling of the Galway's Wall area was undertaken in late 1996 at the request of MVO (Wadge et al. 1998) in order to obtain some insight into possible failure mechanisms for that sector. Several models were tested, including a pressurized conduit model and one involving direct loading by the lava dome. All models suggested that the lower part of the wall and its base around Galway's Soufriere were the most likely failure points, should failure occur as a result of conduit pressure or dome loading. The models were necessarily idealized and homogeneous materials were assumed. The initial delay in identifying the hazard, the many months required to construct and run the models, and the difficulties in communication of the findings, all conspired to decrease the value of the modelling to real-time hazard assessment. Secondary hazards were also considered. A preliminary assessment of the tsunami threat from a sector collapse was made at MVO, using approximate methods developed in the literature. An improved assessment was later made with the assistance of a team of French scientists. The French island of Guadeloupe was most at threat from any tsunami generation, and ties between scientific staff on Montserrat and Guadeloupe were strengthened at this time. Numerical modelling of potential tsunami generation was rapidly undertaken using estimates of possible debris avalanche volumes, runout distances and velocities (Heinrich et al. 1998). These models inferred potentially hazardous wave heights along parts of the coastline of Montserrat, should a collapse occur, and also suggested only very small wave heights in Guadeloupe. Other Caribbean islands were not expected to be affected, although they were alerted to the potential at an early stage through the regional disaster agency (CDERA). The models proved useful and small (1.5 to 2 m) waves did come ashore along the central west coast of Montserrat following the 26 December sector collapse (Calder et al.
358
S. R. YOUNG ET AL.
1998; Sparks el al. 2002). It is interesting to note, however, that the waves were not generated by the debris avalanche, as had been modelled, but by the main pyroclastic density current. Monitoring and visual observations during December 1996 had not revealed signs of rapid deterioration of the situation that might conceivably have led to a short-term forecast of sector collapse. Nevertheless, a partly explosive dome collapse did occur on 19 December after cessation of the main phase of deformation associated with the Galway's Wall and Chances Peak area. Established procedures developed by MVO, based on conventional eruption indicators, led to a red-alert advisory from MVO during this dome collapse (Aspinall et al. 2002). It is impossible to know how close Galway's Wall was to failure during this period. With the burial of Galway's Wall after March 1997 and the burial and devastation caused by the explosions of August to October 1997, the opportunities for effective monitoring were much diminished. The visual indications of deformation were effectively lost, along with the loss of proximal monitoring equipment and telemetry; sites available for effective new monitoring stations did not exist. Consequently, no specific deformation precursors to the sector collapse of 26 December 1997 could be identified. The only relevant precursory information came from the seismic network, with the build-up in hybrid earthquake activity over the 36 hours prior to the collapse event. This build-up was noted by MVO, although the nature of the culmination of the seismicity could not be known. Two other factors influenced MVO thinking at the time. The first was an awareness of long-term cyclicity (six to seven weeks) for
periods of heightened activity, evident since May-June 1997 (Voight et al. 1999). Indeed, the occurrence of a series of explosive events after mid-September 1997 had been anticipated a month earlier. The last 'peak' of intensive activity had been in early November, so that there was some expectation by MVO of heightened activity during the last week of the year. The second was the relationship between hybrid earthquake swarms and dome collapses (Miller et al. 1998; Voight et al. 1998). Often a period of increasingly intense hybrid swarm activity preceded a dome collapse and, at the time, no major event had occurred during the first such swarm. From these factors, MVO considered that another dome collapse and runout of pyroclastic flows down the southwestern flanks of the volcano were the most likely culmination of these precursors, probably occurring after a few days of distinctive hybrid earthquake swarms of growing intensity. In order to analyse retrospectively the precursor seismicity, an analysis of the RSAM record from the single-component Windy Hill seismic station (MWHZ. Fig. 1) was undertaken using the materials failure forecast method (FFM; Voight & Cornelius 1991). We used the graphical method of Cornelius & Voight (1995) at three different time points during the generally accelerating build-up of hybrid earthquake activity prior to 07:01 GMT (03:01 local time) on 26 December. The 10-minute RSAM values from the Windy Hill station were averaged to hourly data and the inverse then plotted against time (Fig. 8). We have chosen three points at which the RSAM record shows a distinctive trend that might have been interpreted as build-up towards a climactic event.
to
Fig. 8. Application of the materials failure forecast method for the period prior to the sector collapse of 26 December 1997. Each plot shows the inverse RSAM record for MWHZ. All times are GMT. (a) Record as at 19:00 on 24 December 97 with graphical solution of FFM from the previous 8 hours of data, (b) Record as at 17:00 on 25 December 97 with graphical solution of FFM from the previous 12 hours of data, (c) Record as at 00:00 on 26 December 97 with graphical solution of FFM from the previous 5 hours of data, (d) Record as at 00:00 on 26 December 97 with graphical solution of FFM from the previous 19 hours of data. The collapse of Galway's Wall actually occurred at 07:01 GMT (03:01 LT) on 26 December.
SECTOR COLLAPSE HAZARD
The first analysis is at 19:00 (all times GMT) on 24 December (Fig. 8a), when RSAM had been increasing steadily for 8 hours. Taking the least-squares fit of the previous nine hourly RSAM figures, a conservative failure criterion (infinite RSAM) is considered to be met at 07:30 on 25 December, about 23.5 hours prior to actual sector collapse. The second analysis point is at 17:00 on 25 December (Fig. 8b), when RSAM had been increasing sporadically for the previous 12 hours. Extrapolation of the least-squares fit of the previous 13 points predicts infinite RSAM at 12:30 on 26 December, about 5.5 hours after actual failure. The final analysis point is at 00:00 on 26 December, when RSAM had been generally increasing for 19 hours and rapidly increasing for the previous 5 hours. The least-squares fit for the rapid build-up (Fig. 8c) predicts failure at 00:30, whereas the same fit for the slower, 19 hour build-up (Fig. 8d) gives a prediction of failure within half an hour of the actual failure at 07:01 GMT. The FFM method as applied to volcanic edifice collapse (Voight 1988) is probably more straightforward when deformation data such as displacements (from global positioning system or EDM), crack monitoring, or tilt are available. These kinds of data were only available at Soufriere Hills for the late 1996 period, when no accelerating build-up towards a potential event occurred. The actual collapse occurred when only seismic data were available, and the analysis above demonstrates the difficulty in using FFM for predictions, based on a secondary, noisy monitoring parameter (in this case RSAM). However, judicious use of the methodology, with suitable appreciation of its frailties and limitations, can potentially assist in real-time, systematic assessment of complex data in certain situations.
Discussion Small-scale sector collapse with associated magma depressurization and lateral blast generation constitutes a significant potential hazard at most dome-forming volcanoes. Where the volcanic edifice is inherently weak, loading of crater walls or flank areas by dome growth can cause destabilization and the potential for a sector collapse that includes both cold and hot material. In the case of Soufriere Hills Volcano, stressing of a steep flank by a growing dome and loading of a hydrothermally altered area at the base of the flank led to a sector collapse. This collapse then prompted further catastrophic disintegration of much of the remaining dome. The hazard implications for sector collapses are severe. Emplacement of a devastating debris avalanche in proximal areas with probable distal lahars or tsunami waves can be followed immediately by generation of violent pyroclastic density currents. All these events can occur in sequence, with little or no direct warning. Detailed monitoring of the lava dome and edifice, emphasizing parameters bearing on stability, are required to provide any chance of qualitative or quantitative forecasting for such events. However, the identification of the hazard and the confirmation of symptoms of potential collapse should prompt evacuation of any population at risk as soon as possible.
Influences on hazard assessment Due to the paucity of historical information on sector collapse and lateral blast eruptions, appraisals of sector-collapse hazard for Montserrat were necessarily based primarily on the improved understanding gained from geological investigations during the eruption. Nevertheless, experience elsewhere was also considered. The 3100 year BP eruption of la Soufriere on Guadeloupe (Boudon et al. 1984, 1987) is a well documented example at a similar volcano, which was considered as a basis for hazard assessment. Likewise, the published (and personal) experience from Mount St Helens and Bezymianny provided analogues useful for perspective. The sector collapse and lateral blast of Mount St Helens was also
359
used to provide an example in public lectures on Montserrat, to inform the public on the concern for Galway's Wall in late 1996, and to generate public support for evacuation zoning. Appraisal of edifice stability by engineering criteria (Voight 1996; Wadge et al. 1996, 1998; B. Voight unpublished data) suggested that the southern sector was the most likely to fail, and it explained the causes of the rock avalanches of 1996. All of the theoretical analyses were limited by the modelling simplifications required, and by the uncertainties involving distribution of materials and their properties. Although the models were useful in providing perspectives on various mechanisms, they were not as reliable as the information obtained from systematic observations and monitoring. The hydrothermal alteration in the area of Galway's Soufriere was an important factor in hazard assessment and is likely to have been important to the final collapse geometry, and indeed to its occurrence. The role played by increased pore fluid pressure within the hydrothermal system (Lopez & Williams 1993; Day 1996) remains uncertain. However, the process of progressive sealing of the hydrothermal system (Boudon et al. 1998) is favourable for raising fluid pressure and thus weakening the area. Qualitative monitoring of pore fluids seeping from Galway's Wall was undertaken during late 1996, which assisted with hazard assessment at that time, but no quantitative methods were attempted. Nevertheless, the existence of saturation within parts of the edifice is considered to be significant as regards both mechanical stability (Wadge et al. 1998), and mobility of the debris avalanche following collapse (Voight et al 2002). Despite the small number of well documented small-scale sector collapses worldwide, estimates of the various collapse and hazard parameters made during hazard assessments on Montserrat in late 1996 proved reasonable in retrospect. Both the collapse volume and the extent of the devastated area on 26 December 1997 were similar to those forecast on hazard maps in December 1996. Tsunami generation by small-scale edifice collapse has been documented for only a few similar volcanoes, e.g. Mount St Augustine in Alaska (Beget & Kienle 1992; Waythomas & Waitt 1998) and Unzen-dake (Latter 1981; Siebert 1984), but modelling undertaken for Soufriere Hills demonstrated the potential threat. Again, the actual phenomenon occurred on a scale quite comparable to that forecast in hazard assessments.
Volcano monitoring applied to sector collapse Beginning in early 1996, the Galway's Wall area was recognized as susceptible to sector collapse should substantial dome growth continue. Preliminary stability assessments were made at this time, materials were sampled from the base of the wall (by S. R. Young and B. Voight), and strength tests were conducted (Voight 1996). By late 1996, the stability of the southern sector was clearly an issue, and we attempted a number of methods of monitoring deformation in this sector. The success of this venture cannot easily be assessed due to the significant changes in overall geometry and circumstances that occurred between this period and the time of eventual sector collapse at the end of 1997. We did, however, collect useful data by various methods. Ground deformation monitoring by tiltmeter and EDM, and crack measurements by both manual methods and extensometer, proved to be the most effective methods in providing quantitative data that could be indicative of an imminent failure. We also found that visual observations combined with innovative use of routine seismic data assisted in tracking landslide frequency from the unstable sector. An acoustic emission approach (e.g. Stateham & Merrill 1979) was not attempted, although its implementation in such a potentially hazardous situation would have been problematic. We believed that the operating seismic arrays provided adequate information of this general type. The eventual sector collapse was monitored most effectively by the seismic network, although the exact nature of the deformations and culminating volcanic activity could not be interpreted from the
360
S. R. YOUNG ET AL.
collected seismic data. Because the collapse and eruption occurred at night, no visual or photographic documentation was available (as at Mount St Helens) to link the seismic record to observed events. Finally, we found field assessment of the nature of the volcanic edifice and potential debris-avalanche path most useful in parameterizing the quantitative models of avalanche runout and tsunami formation. Detailed digital terrain models of base topography and changes in dome morphology assisted both in the estimation of potential collapse volume and, in the case of dome bulging, in the assessment of hazard potential.
Conclusions Sector collapse may represent the most severe volcanic hazard during many andesitic dome-forming eruptions. Debris avalanche, tsunami and either lateral or vertical volcanic explosions (or both) are likely, with potentially devastating consequences. The geological record of Caribbean volcanoes and the recent history of the Soufriere Hills eruption on Montserrat suggest that sector collapse is a common feature of dome-forming eruptions in the region and probably worldwide, not the rare event it was once considered. Identification of potentially weak sectors and investigation of their character should be a part of any hazard assessment, preferably prior to eruption onset. Monitoring of seismic and deformation signals related to sector instability was possible on Montserrat in late 1996, although definitive criteria for imminent sector collapse could not be ascertained. By the time that the sector collapse occurred a year later, changes in volcano morphology and danger levels for monitoring meant that only seismic signals were recorded. Education of the public and crisis management officials in the dangers of sector collapse proved particularly challenging. The lack of visible signs of a worsening situation gave no indication of the potential catastrophe to the casual onlooker. The inability of observatory staff to guarantee any warning added to the difficulties of management for sector collapse hazards. We are grateful to the many members of the MVO staff who assisted in collection and interpretation of data used in this work; we thank especially M. Davies for crack measurements, R. Luckett and L. Pollard for seismological analysis, D. Williams and L. Lynch for technical support, and B. Darroux, L. Luke, A. Grouchy and J. McMahon for logistical support. We also thank USGS-CVO for logistical, instrumental and cerebral support, especially R. Hoblitt and A. Lockhart. The French modelling group provided rapid assistance in hazard appraisal, and we thank J.-L. Cheminee (IPGP) and Y. Caristan (CEA) for support. Comments from our reviewers and editors greatly improved the manuscript. Successful hazard management was achieved with the support of crisis managers on Montserrat and thanks to the spirit of the Montserratian population. Funding for MVO was mainly provided by DFID. Important support for B.V. was provided by the US National Science Foundation. For relevant authors, published by permission of Director, BGS (NERC).
References ASPINALL, W. P. & COOKE, R. M. 1998. Expert judgement and the Montserrat Volcano eruption. In: MOSLEM, A. & BARI, R. A. (eds) Proceedings of the 4th International Conference on Probabilistic Safety Assessment and Management PSAM4, 13-18 September 1998, New York City, USA, 3, 2113-2118. ASPINALL, W. P., MILLER, A. D., LYNCH, L. L., LATCHMAN, J. L., STEWART, R. C., WHITE, R. A. & POWER, J. A. 1998. Soufriere Hills eruption, Montserrat 1995-1997: Volcanic earthquake locations and fault plane solutions. Geophysical Research Letters, 25, 3397-3400. ASPINALL, W. P., LOUGHLIN, S. C., MICHAEL, F. V. ET AL. 2002. The Montserrat Volcano Observatory: its evolution, organization, role and activities. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 71-91.
BEGET, J. E. & KIENLE, J. 1992. Cyclic formation of debris avalanches at Mount St Augustine volcano. Nature. 356. 701-704. BOUDON. G., SEMET, M. P. & VINCENT, P. M. 1984. Flank failure and directed blast at Soufriere. Guadeloupe. French West Indies. Geology, 12, 350-353. BOUDON, G.. SEMET, M. P. & VINCENT. P. M. 1987. Magma and hydrothermally driven sector collapses: the 3100 and 11 500 Y. B. P. eruptions of La Grande Decouverte (La Soufriere) Volcano. Guadeloupe. French West Indies. Journal of Volcanology and Geothermal Research, 33, 317-323. BOUDON, G., VILLEMANT, B., KOMOROWSKI. J.-C. ILDEFONSE. P. & SEMET, M. P. 1998. The hydrothermal system at Soufriere Hills Volcano, Montserrat (West Indies): Characterization and role in the on-going eruption. Geophysical Research Letters. 25. 3693-3696. BOUDON, G., KOMOROWSKI. J.-C.. SEMET, M.. LE FRIANT. A. & DEPLUS, C. 1999. Frequent volcanic flank-collapses in the Lesser Antilles Arc: Origin and hazards. EOS, Transactions of the American Geophysical Union, 80, F1142. CALDER, E. S., YOUNG. S. R.. SPARKS, R. S. J. ET AL. 1998. The Boxing Day Collapse, 26 December 1997. MVO Special Report. 6. CHOUET, B. A. 1996. Long-period volcano seismicity: its source and use in eruption forecasting. Nature, 380, 309-316. CORNELIUS, R. R. & VOIGHT, B. 1995. Graphical and PC-software analysis of volcano eruption precursors according to the Material Failure Forecast Method (FFM). Journal of Volcanology and Geothermal Research, 64, 295-320. DAY, S. J. 1996. Hydrothermal pore fluid pressure and the stability of porous, permeable volcanoes. In: MCGUiRE. W. J.. JONES, A. P. & NEUBERG. J. (eds) Volcano Instability on the Earth and Other Planets. Geological Society. London, Special Publications. 110, 77-93. DENLINGER, R. P. & OKUBO. P. 1995. Structure of the mobile south flank of Kilauea volcano, Hawaii. Journal of Geophysical Research, 100. 24499-24507. DEPLUS, C.. FRIANT, A. L., BOUDON, G. ET AL. 1999. Large-scale debris avalanche deposits on the western flank of the Lesser Antilles arc revealed by the Aguadomar Cruise. EOS, Transactions of the American Geophysical Union, 80, F1152. DRUITT, T. H.. YOUNG, S. R.. BAPTIE. B. ET AL. 2002. Episodes of cyclic Vulcanian explosive activity with fountain collapse at Soufriere Hills Volcano. Montserrat. In: DRUITT, T. H. & KOKELAAR. B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London. Memoirs. 21, 281-306. ENDO, E. T. & MURRAY, T. 1991. Real time seismic amplitude measurement (RSAM): a volcano monitoring and prediction tool. Bulletin of Volcanology, 53, 533-545. HARFORD, C. L. 2000. The Volcanic Evolution of Montserrat. PhD thesis. University of Bristol. HARFORD. C. L., PRINGLE, M. S., SPARKS. R. S. J. & YOUNG. S. R. 2002. The volcanic evolution of Montserrat using 40 Ar 39 Ar geochronology. In: DRUITT, T. H. & KOKELAAR. B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society. London, Memoirs. 21. 93-113. HEINRICH, P., MANGENEY, A., GUIBOURG, S., ROCHE, R., BOUDON, G. & CHEMINEE, J.-L. 1998. Simulation of water waves generated by a potential debris avalanche in Montserrat. Lesser Antilles. Geophysical Research Letters, 25, 3697-3700. KOKELAAR, B. P. 2002. Setting, chronology and consequences of the eruption of Soufriere Hills Volcano. Montserrat (1995-1999). In: DRUITT. T. H. & KOKELAAR. B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society. London. Memoirs. 21, 1-43. LATTER, J. H. 1981. Tsunamis of Volcanic Origin: Summary of Causes, with Particular Reference to Krakatoa. 1883. Bulletin of Volcanology. 44, 467-490. LIPMAN, P. W. & MULLINEAUX. D. R. 1981. The 1980 Eruptions of Mount St. Helens, Washington. United States Geological Survey. Professional Papers 1250. LOPEZ, D. L. & WILLIAMS, S. N. 1993. Catastrophic volcanic collapse: relation to hydrothermal processes. Science. 260, 1794-1796. McGuiRE, W. J. 1996. Volcano instability: a review of contemporary themes. In: MCGUIRE, W. J.. JONES, A. P. & NEUBERG, J. (eds) Volcano Instability on the Earth and Other Planets. Geological Society, London, Special Publications, 110. 1-23.
SECTOR COLLAPSE HAZARD MATTIOLI, G. S., JANSMA, P. E., JARAMILLO, L. & SMITH, A. L. 1995. Sector collapse in island arc volcanoes: A digital topographic and bathymetric investigation of the Qualibou depression, St Lucia, Lesser Antilles. Caribbean Journal of Science, 31, 163-173. MILLER, A. D., STEWART, R. C, WHITE, R. A. ET AL. 1998. Seismicity associated with dome growth and collapse at the Soufriere Hills volcano, Montserrat. Geophysical Research Letters, 25, 3401-3404. NORTON, G. E., WATTS, R. B., VOIGHT, B. ET AL. 2002. Pyroclastic flow and explosive activity at Soufriere Hills Volcano, Montserrat, during a period of virtually no magma extrusion (March 1998 to November 1999). In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 467-481. RITCHIE, L. J., COLE, P. D. & SPARKS, R. S. J. 2002. Sedimentology of deposits from the pyroclastic density current of 26 December 1997 at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 435-456. ROBERTSON, R., COLE, P., SPARKS, R. S. J. ET AL. 1998. The explosive eruption of Soufriere Hills Volcano, Montserrat, West Indies, September 17 1996. Geophysical Research Letters, 25, 3429-3432. ROOBOL, M. J. & SMITH, A. L. 1998. Pyroclastic stratigraphy of the Soufriere Hills volcano, Montserrat: Implications for the present eruption. Geophysical Research Letters, 25, 3393-3396, ROOBOL, M. J., WRIGHT, J. V. & SMITH, A. L. 1983. Calderas or gravityslide structures in the Lesser Antilles island arc. Journal of Volcanology and Geothermal Research, 19, 121-134. SIEBERT, L. 1984. Large volcanic debris avalanches: characteristics of source areas, deposits, and associated eruptions. Journal of Volcanology and Geothermal Research, 22, 163-197. SIEBERT, L. 1992. Threats from debris avalanches. Nature, 356, 658-659. SIEBERT, L. 1996. Hazards of large volcanic debris avalanches and associated eruptive phenomena. In: SCARPA, R. & TILLING, R. (eds) Monitoring and Mitigation of Volcanic Hazards. Springer, Berlin, 541-572. SIEBERT, L., GLICKEN, H. & UI, T. 1987. Volcanic hazards from Bezymianny- and Bandai-type eruptions. Bulletin of Volcanology, 49, 435-459. SPARKS, R. S. J. 1997. Causes and consequences of pressurisation in lava dome eruptions. Earth and Planetary Science Letters, 150, 177-189. SPARKS, R. S. J., YOUNG, S. R., BARCLAY, J. ET AL. 1998. Magma production and growth of the lava dome of the Soufriere Hills Volcano, Montserrat, West Indies: November 1995 to December 1997. Geophysical Research Letters, 25, 3421-3424. SPARKS, R. S. J., BARCLAY, J., CALDER, E. S. ET AL. 2002. Generation of a debris avalanche and violent pyroclastic density current on 26 December (Boxing Day) 1997 at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 409-434. STATEHAM, R. M. & MERRILL, R. H. 1979. Acoustic emission techniques applied to slope stability problems. In: VOIGHT, B. (ed.) Rockslides and Avalanches, 2: Engineering Sites. Elsevier Scientific, Amsterdam, 577-593. STOOPES, G. R. & SHERIDAN, M. F. 1992. Giant debris avalanches from the Colima Volcanic Complex, Mexico: implications for long runout landslides (>1OO km) and hazard assessment. Geology, 20, 299-302. VINCENT, P. M. & BOURDIER, J. L. 1989. The primitive volcano of Mt Pelee: its construction and partial destruction by flank collapse. Journal of Volcanology and Geothermal Research, 38, 1-16. VOIGHT, B. 1988. A method for prediction of volcanic eruptions. Nature, 332, 125-130.
361
VOIGHT, B. 1996. Strength of wall rock of English's Crater, Soufriere Hills Volcano, Montserrat. In: WADGE, G. (ed.) The Soufriere Hills Eruption, Montserrat (Abstracts). Geological Society of London, 27 Nov 1996,29. VOIGHT, B. & CORNELIUS, R. R. 1991. Prospects for eruption prediction in near real-time. Nature, 350, 695-698. VOIGHT, B. & ELSWORTH, D. 1997. Failure of volcano slopes. Geotechnique, 47, 1-31. VOIGHT, B., HOBLITT, R. P., CLARKE, A. B., LOCKHART, A. B., MILLER, A. D., LYNCH, L. L. & McMAHON, J. 1998. Remarkable cyclic ground deformation monitored in real time on Montserrat and its use in eruption forecasting. Geophysical Research Letters, 25, 3405-3408. VOIGHT, B., SPARKS, R. S. J., MILLER, A. D. ET AL. 1999. Magma flow instability and cyclic activity at Soufriere Hills volcano, Montserrat, British West Indies. Science, 283, 1138-1142. VOIGHT, B., KOMOROWSKI, J.-C., NORTON, G. E. ETAL. 2002. The 26 December (Boxing Day) 1997 sector collapse and debris avalanche at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 363-407. WADGE, G. & ISAACS, M. C. 1988. Mapping the volcanic hazards from Soufriere Hills Volcano, Montserrat, West Indies using an image processor. Journal of the Geological Society, London, 145, 541-551. WADGE, G., FRANCIS, P. W. & RAMIREZ, C. F. 1995. The Socompa collapse and avalanche event. Journal of Volcanology and Geothermal Research, 66, 309-336. WADGE, G., WILLIAMS, C. & VOIGHT, B. 1996. Stability of the Gages Wall MVO Open-file Report 96/11. WADGE, G., WOODS, A. W., JACKSON, P., BOWER, S., WILLIAMS, C. A. & HULSEMANN, F. 1998. A hazard evaluation system for Montserrat. In: IDNDR, U. N. C. C. (ed.) Forecasts and Warnings, UK National Coordination Committee IDNDR, Project 3. Thomas Telford, London, 32. WATTS, R. B., HERD, R. A., SPARKS, R. S. J. & YOUNG, S. R. 2002. Growth patterns and emplacement of the andesitic lava dome at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 115-152. WAYTHOMAS, C. F. & WAITT, R. B. 1998. Preliminary Volcano-Hazard Assessment for Augustine Volcano, Alaska. United States Geological Survey, Open File Report 98/106. WHITE, R. A., MILLER, A. D., LYNCH, L. & POWER, J. 1998. Observations of hybrid seismic events at Soufriere Hills Volcano, Montserrat: July 1995 to September 1996. Geophysical Research Letters, 25, 3657-3660. WOODS, A. W., SPARKS, R. S. J., RITCHIE, L. J., BATEY, J., GLADSTONE, C., & BURSIK, M. 2002. The explosive decompression of a pressurized volcanic dome: the 26 December 1997 collapse and explosion of Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 457-466. YOUNG, S. R., HOBLITT, R. P., SMITH, A. L., DEVINE, J. D., WADGE, G. & SHEPHERD, J, B. 1996. Dating of explosive volcanic eruptions associated with dome growth at the Soufriere Hills volcano, Montserrat, West Indies. In: AHMAD, R. (ed.) Proceedings of the Second Caribbean Conference on Natural Hazards and Hazard Management, October 1996, Kingston, Jamaica. MVO Open-file Report 96/22. YOUNG, S. R., BARCLAY, J., MILLER, A. D., SPARKS, R. S. J., STEWART, R. C., DAVIES, M. A. & MVO STAFF. 1997. Deformation of the Galway's Wall and Related Volcanic Activity, November 1996 to March 1997. MVO Special Report 2. YOUNG, S. R., SPARKS, R. S. J., ASPINALL, W. P., LYNCH, L. L., MILLER, A. D., ROBERTSON, R. E. A. & SHEPHERD, J. B. 1998. Overview of the eruption of the Soufriere Hills volcano, Montserrat, July 18 1995 to December 1997. Geophysical Research Letters, 25, 3389-3392
The 26 December (Boxing Day) 1997 sector collapse and debris avalanche at Soufriere Hills Volcano, Montserrat B. VOIGHT1, J-C. KOMOROWSKI2, G. E. NORTON3, A. B. BELOUSOV4, M. BELOUSOVA4, G. BOUDON5, P. W. FRANCIS6, W. FRANZ7, P. HEINRICH 8 , R. S. J. SPARKS9 & S. R. YOUNG10 1 Geosciences, Penn State University, University Park, PA 16802, USA (e-mail:
[email protected]) 2 Observatoire Volcanologique de la Soufriere (IPGP), Le Houelmont, Gourbeyre 97113, Guadeloupe 3 British Geological Survey, Key worth, Nottingham, NG12 5GG, UK 4Institute of Volcanic Geology and Geochemistry, Petropavlovsk-Kamchatsky, 683006, Russia 5 Institut de Physique du Globe de Paris (IPGP), 4 Place Jussieu, B 89, 75252 Cedex 05 Paris, France 6 Department of Earth Sciences, Open University, Milton Keynes MK7 6AA, UK (deceased) 1 Gannett-Fleming Engineers, Harrisburg, PA 17110, USA ! Laboratoire de Detection et de Geophysique, Commisariat a I'Energie Atomique,Bp 12, 91680 Bruyeres-le-Chatel, France 9 Department of Earth Sciences, Bristol University, Bristol, BS8 1RJ, UK 10 Montserrat Volcano Observatory, Montserrat, West Indies
Abstract: The southern sector of Soufriere Hills Volcano failed on 26 December 1997 (Boxing Day), after a year of disturbance culminating in a devastating eruptive episode. Sector collapse produced a c. 50 x 106 m3 volcanic debris avalanche, and depressurized the interior of the lava dome, which exploded to generate a violent pyroclastic density current. The south-directed growth of a lava lobe and build-up of lava-block talus, since early November 1997, brought the hydrothermally weakened sector to a condition of marginal stability. Limit-equilibrium stability analyses and finite-difference stress-deformation analyses, constrained by geomechanical testing of edifice and debris avalanche materials, suggest that the sector collapse was triggered by a pulse of co-seismic exogenous lava shear-lobe emplacement. Slip-surface localization was influenced by strain-weakening. The source region fragmented into avalanche megablocks, and further disruption generated a chaotic avalanche mixture that included variably indurated and coloured hydrothermally altered material, and much talus. The avalanche consisted of several flow pulses that reflected complexities of source disruption and channel topography. In the proximal zone, within 1.5km from source, many megablocks preserve pre-collapse stratigraphy. At major bends the avalanche separated into channelled and overspill flows. In the distal region, >2.5km from source, stacked sets of the main lithologies occur with a hummocky surface and abrupt flowage snouts, beyond which sparse hummocks occur in a thinly spread deposit. Textures suggest emplacement by laminar mass transport of partly saturated debris riding on a frictionally sheared base. Three-dimensional numerical simulations of emplacement governed by a Coulomb-type (Pouliquen) basal friction law imply low values of friction (25 % (Medley 1997), although the reported values may be reasonable estimates of the local strength at avalanche or megablock boundaries. The values may also provide lower-bound strength estimates for full-scale slip-surface failure of Fumarolic Unit fragmental deposits during the Boxing Day sector collapse.
Edifice materials Block samples of relatively weak intact tuff had been collected by B.V. and S.R.Y. at the base of Galway's Wall in March 1996, in anticipation of future instability (Voight 1996). Towards the end of 1996, the sampling site (near Fig. 5b) was buried by rockfall debris and dome talus (Fig. 6). Additional samples of weathered tuff were taken in March 1996 from the Gages Wall area of English's Crater (see Fig. 2), and samples of fresh (1995 lava) and altered (prehistoric lava) dome material were also acquired. In order to provide constraints for stability assessments, the block samples were subjected to laboratory triaxial rock-mechanics testing to provide the shear strength of intact material as a function of confining pressure, using procedures discussed in the Appendix. Additional testing was carried out to obtain intact tensile strengths and specimen dry bulk density. Note that strengths reported below are for intact materials, and values appropriate for the in situ jointed rock-mass would generally be less (Voight 2000). In general, the experimental Mohr-circle data indicated curved strength envelopes as a function of confining pressure. For Galway's Wall tuff, with dry bulk density of 1980 kg m-3 and tensile strength of 1.21 MPa, results are shown in Figure 33. The data gave good agreement with the Hoek & Brown (1980) failure criterion that predicts a parabolic envelope for brittle rocks subjected to compressive stress conditions, and approximates classic Griffiths theory in the region of tensile effective normal stresses.
Fig. 31. Shear stress versus displacement for reversal direct-shear tests on Soufriere facies debris avalanche samples. Note strength loss with displacement. (a) Remoulded sample 15-2A, clayey sand with gravel, normal stress 0.86 MPa. (b) Block sample 15-2C, highly plastic inorganic clay, normal stress 1.72 MPa.
Considering the full Galway's Wall dataset, the Coulomb parameters yielded a linear fit with cohesion = 2.19 MPa and friction angle = 33.0°, although this set of parameters underestimated strength at intermediate confining pressures, and overestimated strength in the tensile region (see Voight 2000). A piece-wise linear fit can provide reasonable Coulomb parameter approximations for different ranges of normal stress. Thus, for confining pressures >2.07MPa, the parameters cohesion = 3.65 MPa and friction angle = 31° provide an adequate fit. The weathered Gages Wall tuff was less dense and weaker, with dry bulk density of 1840 kg m-3 and tensile strength of 0.41 MPa. Coulomb parameters for moderate compressive confining pressures were, approximately, cohesion = 1.7 MPa, friction angle = 32°, although, as noted, the envelope was non-linear (Fig. 33). Corresponding data for fresh dome lava collected in the southwest moat inside English's Crater, and mildly altered lava from prehistoric dome-derived breccia exposed in Galway's Wall, were obtained. Approximate Coulomb friction angles are 55° and 44°, respectively. In addition, disaggregated materials were collected from weathered tuffs presumed to be representative of the outer-slope base of Galway's Wall, and also from weathered portions of Gages Wall (Fig. 1). Both materials were classified as grey-brown silty sand with gravel, USCS group SM (Fig. 30). Laboratory shear strength tests were conducted, using methods summarized in the Appendix. At all normal stresses, higher resistance was obtained on second-cycle shearing, probably reflecting enhanced interlocking and dilatation between sliding or rotating angular grains, whereas further displacements caused some reduction in strength (Fig. 34). Linear Coulomb
THE BOXING DAY SECTOR COLLAPSE
395
Fig. 32. Shear strength versus effective normal stress for Soufriere facies debris avalanche materials subjected to direct-shear tests, (a) Remoulded sample 15-2A, clayey sand with gravel. (b) Block sample 15-2C, highly plastic inorganic clay. Open symbols show peak shear strength; filled symbols show residual shear strength.
Fig. 33. Shear stress versus effective normal stress for intact samples of edifice materials subjected to triaxial and indirect-tension tests. Data plotted as Mohr circles, with curved envelope according to Hoek-Brown criterion, (a) Gages Wall tuff, altered. (b) Galway's Wall tuff.
effective-stress strength relations applied approximately over the normal-stress range examined (Fig. 35). For both sites, a single nominal set of parameters seemed appropriate, namely zero cohesion, peak friction angle = 35°, residual friction angle = 31°. These parameters may be appropriate for disaggregated material, or blocks bounded by uncemented joints or joint gouge within the tested range of normal stresses.
pated dome growth. A trial slip surface was postulated. and the shearing resistance required to equilibrate the mass of material on this surface (and other applied forces) was calculated by statics. The calculated resistance was compared with available shear strength to yield a factor of safety F, defined as the quotient of shear strength of the material and the shear stress required for static equilibrium. The value F= 1 represents a condition of incipient failure, and F > 1 represents stability. The procedure was repeated for other postulated slip surfaces, and the lowest F was found by iteration (Duncan 1996; Voight 2000). Laboratory tests on block samples of indurated tuffs and dome rock, and samples of disaggregated materials, as described above, were used along with literature data to constrain the stability analyses (Voight 1996): scaled reductions were used for rock-mass properties (Hoek 1983; Voight 2000). Seismic loading was simulated simply by a coefficient that represented the potentially destabilizing earthquake force due to horizontal acceleration of the material (Kramer 1996). The results of these stability analyses suggested that the outer part of Galway's Wall was marginally stable with respect to static loading for the dome geometry existing at that time. Stability decreased when pseudo-static earthquake loading was added, and for horizontal accelerations on the order of O.lg. shear failure on the upper south face of Galway's Wall was predicted by the analyses. The failures predicted were shallow, not deep-seated. Successive slope failures triggered by successive earthquake shocks were predicted to reduce wall thickness by exterior slabbing, rather than by massive deep-seated failure. The results did not preclude the latter possibility, given the uncertainties involving the forces caused by a shallow intrusion (Young et al. 2002). potential
Stability assessments and failure mechanisms
Studies made before the sector collapse Growth of the andesitic lava dome at Soufriere Hills caused decreasing structural stability of the southern sector of the volcano. The possibility of a future instability problem in this sector was recognized in early 1996, and clear warning signs of growing instability were recognized in November-December 1996 (Young el al 2002). Catastrophic sector collapse occurred a year later on 26 December 1997, during a period of enhanced seismicity and after two months of southward expansion of the lava dome and talus apron over the hydrothermally weakened area. Preliminary quantitative stability assessments, using two-dimensional limiting-equilibrium analyses, were made during hazards evaluations at Soufriere Hills in March 1996 and December 1996 (B. Voight, unpublished data). In such analyses, a representative topographic profile and cross-section was drawn, to represent existing conditions and also some future conditions related to antici-
396
B. VOIGHT ET AL.
Fig. 34. Shear stress versus displacement for reversal direct-shear tests on disaggregated edifice tuffs. Note small strength loss with displacement, (a) Sample 1: Gages Wall tuff, altered, (b) Sample 2: Galway's Wall tuff.
Fig. 35. Shear strength versus effective normal stress for edifice materials subjected to direct-shear tests, (a) Sample 1: Gages Wall tuff, altered, (b) Sample 2: Galway's Wall tuff. Open symbols show peak shear strength; filled symbols show residual shear strength.
through-going structural discontinuities within the wall, and assumed material properties. Indeed some radial cracks were observed and monitored (Young et al. 2002). However, it was recognized that deterioration of the wall by slabbing was most likely, and this could cause reduction of the wall height and consequently lead to overtopping by the growing dome, with rapid erosion of the friable, jointed wall rock. The deterioration of the wall in general followed the lines suggested by the analyses. Wall failures involved release of shallow slabs of jointed rock, triggered by the larger felt shocks in repetitive seismic swarms. Then, after several shifts in the locus of dome growth had reduced pressure against the south sector, lava overtopped the low point on Galway's Wall in March-April 1997, block-and-ash flows severely eroded parts of Galway's Wall, and Galway's Soufriere was partially buried by talus. In addition to these analyses, axisymmetric finite element modelling was conducted to explore the stress changes associated with dome growth and conduit pressurization, for the geometry existing in late 1996 (Wadge et al. 1998). The analysis assumed homogeneous elastic media, and was somewhat limited inasmuch as pore pressures, seismic loading, material zonation, and plastic deformation were not treated. Relative stability was interpreted from the ratios of elastic-media shear stresses to an assumed Coulomb relation.
outlined above. There are various ways in which these procedures can be manipulated, and the analyst needs to consider which methods are most accurate, and which of the accurate methods can be applied most easily (Duncan 1992, 1996; Fredlund 1984). A good comparative summary of the various methods is given by Bromhead (1986). The modified Bishop method was used in this work; the method satisfies moment and vertical force equilibria, and gives values of factor of safety F that fall within the range of equally correct solutions as determined by so-called exact methods (Duncan 1996; Bromhead 1986; Lambe & Whitman 1969). The method assumes a slip surface shaped as a circular arc, with material above this surface subdivided in a series of vertical slices for computation of body forces. As indicated previously, results for a specified slip
Limiting-equilibrium analyses made in 1998 Limiting-equilibrium stability analyses were carried out in January 1998, soon after the sector collapse, using the general procedures
Table 1. Assumed material properties for limiting equilibrium analyses Unit weight (kNm- 3 )
Cohesion (kPa)
Friction angle (deg)
Model A 1 2 3 4
20 20 20 20
1 10 1 5
39 42 39 45
Model B 1 2 3 4
20 20 20 20
1 10 1 5
37 42 38 45
Material
THE BOXING DAY SECTOR COLLAPSE
Fig. 36. North-south cross-section for Boxing Day collapse, showing materials and slip plane boundaries assumed in limit-equilibrium stability analyses. Material properties are given in Table 1. The water table, indicated by the dotted line at 'w', joins the ground surface at the toe of the slope.
surface are summarized as a factor of safety, with the value F= 1 implying incipient failure. In general, the F for three-dimensional analysis is slightly greater than the F for two-dimensional analysis (Duncan 1992). Table 1 lists material properties assumed, specific unit weights, and strength properties given by cohesions and effective-stress friction angles, varied selectively according to material type. The distributions of materials, and piezometric surface assumed, are shown in Fig. 36. A centre of rotation and radius was specified to represent the actual Boxing Day failure surface (assuming it to be circular), as inferred from construction of cross-sections (Fig. 12). The slip surface was assumed to be constrained by the south edge of the partly eroded Galway's Wall (Figs 36 and 9b). Results of analyses for two similar sets of model properties (Table 1) are discussed here. The parameters represented bulk, aver-
Fig. 37. Seismic data for 24-26 December 1997, for MBWH (Windy Hill, vertical component) seismic station (Fig. 2). (a) Number of earthquakes per hour, (b) Amplitudes of hybrid earthquakes as a measure of relative energy in units of m s - 1 . (c) Real-time seismic amplitude measurements (RSAM), in arbitrary units.
397
age properties assumed for these zones. However, we recognized that each zone was in fact heterogeneous and that local properties within a zone might differ substantially from the averages assumed. For model A the calculated safety factor F for the Boxing Day surface was 1.22 under static loading, and 1.01 for applied seismic loading with pseudo-static seismic acceleration coefficient of 0.1. Similarly, slightly weaker model B suggested F= 1.15 under static loading, and F=0.96 for a seismic coefficient of 0.1. These results could be interpreted to imply marginal stability for the static conditions considered. Small perturbations of applied loads, whatever the cause, could then bring the slope to a condition of failure. Such perturbations may be considered as the trigger mechanism, but the overall causes of failure are more complex and indeed include all the various factors that brought the slope to its condition of marginal static stability (Voight & Elsworth 1997, table 1). Prominent among these are the major topographic changes reflecting dome growth, wall erosion and talus deposition that occurred since 1996. as well as the history of hydrothermal alteration and possible episodes of prior slip. Potential triggers included the direct effects of seismic shaking, pore-fluid pressure changes induced by seismic shaking or other mechanisms, strain-weakening by deep-seated creep, and loading by lava extrusion. A combination of these cannot be excluded. For example, increase of pore-fluid pressure in Galway's Soufriere materials (Fumarolic Unit source materials) to an artesian condition could have resulted in instability. This magnitude of pore pressure would require an average piezometric surface above the original ground, but such a circumstance cannot be discounted given the local presence in or under the soufriere of weak lowpermeability clay layers, pre-existing shallow slip surfaces associated with landslides active in 1996 and earlier times, and the rapid loading of this area by dome-block talus in the weeks preceding the collapse. A low F (near unity) could also have promoted localized creep of weak materials, and the deformation tests previously described indicated that some Fumarolic Unit materials were indeed characterized by relatively low frictional strength and profound strain-weakening. Thus, the static value of F may have gradually decreased in the weeks preceding the failure. However, in addition, the coincidence of the Boxing Day sector collapse with a brief period of enhanced seismicity, from 14:30 on
398
B. VOIGHT ET AL.
Fig. 38. Numerical model of north-south cross-section for Boxing Day collapse, seismic loading, (a) Material zones: 1, dome talus and rockfall debris (purple); 2, fresh lava (yellow); 3, altered soufriere materials (brown); 4, old edifice materials (tan). White arrow marks the water table. Coulomb parameters assumed in zone 1: c' = lOkPa, = 40°; zone 3: c' = 50kPa, = 40°; other zones strong. Seismic loading coefficient, 0.15. Movement indicated by velocity vectors, arrows (max 0.037 m s . ) . (b) As (a), with deformation accentuated from strain weakening. Zone 3: c' = 20kPa, - 20°.
24 December onwards, was probably not accidental (Fig. 37). To judge from previous periods of enhanced hybrid swarm activity, the seismicity was probably accompanied by an enhanced pulse of effusion of gas-charged lava (Voight et al 1999). The hybrid swarm began with events every 20 minutes or so, and increased in intensity until the late evening on 25 December, after which the signal was effectively a tremor (Fig. 37; Sparks et al 2002). The individual hybrid events increased in amplitude as the swarm progressed, and, although even the largest of these events was smaller in magnitude than similar events recorded in November 1997, the sector geometry and its static loading was now at a more critical level. The amplitude of tremor peaked around 23:00 on 25 December, declined briefly, and then continued with ascending amplitude (roughly doubling the previous maxima) to the onset of the sector collapse at 03:01 (Fig. 37). The near-coincidence of distinct hybrid seismicity with the sector collapse was suggestive of a possible causative relation, and
supported the inclusion in the modelling of transitory earthquake loading by equivalent static forces. Strong-motion data on volcanoes are rare, although Voight et al (1983) reported information from Mount St Helens, and such data were acquired at Montserrat in May-June 1997, when ground accelerations measured on Chances Peak from shallow hybrid earthquakes were as much as O . l g (B. Voight, unpublished data). The accelerations associated with earthquakes in December 1997 may not have been this severe, but we nevertheless have explored the use of pseudo-static seismic coefficients of about 0.1 for analyses of the 26 December 1997 sector collapse.
Numerical deformation models The modelling work of 1996-1998, considered above, was of limited applicability in that stress-strain behaviour, development
THE BOXING DAY SECTOR COLLAPSE
399
Fig. 39. Numerical model of lava shearlobe emplacement at Soufriere Hills. (a) Lava in the upper part of the slope (yellow) is pushed southwards (to left) at a constant rate, and creates a velocity field (vector arrows, max 0 . 6 5 m s - 1 ) . (b) The velocity field creates a shear zone that extends to the toe of the slope and includes altered soufriere material. The deformed grid illustrates the displacements: the shear strain-rate field is shaded. Grid distortion scale is 190x. Zone at toe of slope is soufriere material that is susceptible to strain-weakening, with effects shown in Figure 40.
of plastic zones and strain-weakening, and dynamic loading, among other aspects, could not be adequately represented by the methods employed. It is preferable that the mechanism of failure should be elucidated by modelled strain localizations, rather than pre-specified as in limiting-equilibrium analyses, or vaguely interpreted from elastic models. The previous results did not preclude seismic shaking as a trigger mechanism, but the methods did not facilitate consideration of some other potential causes, such as enhanced conduit pressure, strain-weakening, and the invigorated emplacement of a fresh shear-lobe of lava. In order to illuminate these issues, we evaluated a series of two-dimensional explicit plain-strain finite difference models of the Soufriere Hills slope, using the Fast Lagrangian Analysis of Continua procedure (Cundall & Board 1988; Coetzee et al. 1998; Detournay & Hart 1999). The slope was subdivided into four zones, representing (1) fresh dome talus, (2) fresh lava, (3) altered soufriere material, (4) old
edifice, as shown in Figure 38, with the actual material parameter values different in various models. In general the bulk modulus was taken as 20GPa and shear modulus was l0GPa. The models could undergo plastic deformation when Coulomb yield limits were reached, which are different in the various zones. In some cases pseudo-static loads were used to simulate earthquake loading. Figure 38a shows assumed material zonation and displacement velocity vectors for the slope subjected to seismic loading, with Figure 38b showing enhanced deformation after strain-weakening. As the velocity and strain fields appeared to be more or less consistent with the deduced position of the actual failure surface at Montserrat (Fig. 12), the results did not preclude the hypothesis of seismic shaking alone as a trigger of the Boxing Day sector collapse. The lava shear-lobe emplacement hypothesis was treated next, whereby lava in the upper part of the slope was displaced southwards at a steady rate, creating a shear zone that extended to the toe
400
B. VOIGHT ET AL.
Fig. 40. (a) As Figure 39, with material in outer slope strain-weakened to: c' = 20 kPa, = 30° in zone 1 (talus and rockfall debris); and c' = 20 kPa, — 20° in zone 3 (soufriere material). Lava-lobe emplacement still continues as in Figure 39 (note grid distortion at upper right), but gravity-driven movements on outer slope dominate the grid distortion, shear strain-rate and velocity vector field. Grid distortion scale 24x, max. vector 5.0ms - 1 . (b) Gravity collapse of the outer part of the slope results in an oversteepened and decompressed face of fresh lava, which begins to collapse (grid distortion scale l . l x ) . Generation of the decompression-induced explosions is beyond the scope of the model.
of the slope and included altered soufriere material (see velocity field, Fig. 39a, and shear zone, Fig. 39b). In this model, shear-zone materials in zones 1 and 3 were assumed to strain-weaken, so that spontaneous, localized gravitational collapse might then occur in the outer slope (Fig. 40a). The deformation rates due to gravitational loading exceeded those associated with the southerly displacement due to continuing lava shear-lobe emplacement (Fig. 39b), thus promoting the gravity collapse. This, in turn, resulted in an oversteepened and decompressed face of fresh volatile-rich dome lava, which began to collapse, in part explosively, and to generate a violent pyroclastic density current (Fig. 40b). The details of the explosive initiation are beyond the scope of this type of model, but are explored by Fink & Kieffer (1993), Alidibirov & Dingwell (1996), Voight & Elsworth (2000) and Woods et al (2002). In general, these models suggest that increased loading and strain-weakening
deformation induced by lava lobe emplacement are viable trigger mechanisms for the Boxing Day sector collapse; some influence of superposed seismic loading is not excluded. Edifice stresses resulting from conduit pressure could also be modelled, although these stresses dissipate radially and a twodimensional model only provides an upper bound to the stresses. There were no field observations that suggested that propagation of brittle shear fractures occurred in a fashion that would connect an outer-flank collapse with stressed rock surrounding the pressurized conduit. Overall the analyses seemed to favour (1) collapse triggered by a pulse of rapid lava shear-lobe emplacement, with subsequent gravity-driven slip-surface localization influenced by enhanced loading, creep and strain-weakening, and/or (2) seismic shaking. We favoured a combination of the two mechanisms, inasmuch as lobe emplacement was accompanied by seismicity which may have
THE BOXING DAY SECTOR COLLAPSE
changed the distribution of stresses, and the seismicity itself was not particularly strong in comparison with previous seismic episodes and thus may not have been the exclusive trigger. Numerical modelling of dynamics of debris avalanche emplacement Assumptions
where 1 , 2 and D are characteristics of the material that in principle could be measured from deposits. Extrapolating Pouliquen's results to real avalanches, very approximately D = 7d. where d is the 'effective mean diameter' of flowing material. The equation provides a friction angle ranging between two values 8\ and 82 (61 < 62), depending on the instantaneous values of the velocity and thickness of the flow. The higher the velocity, the higher is the friction. Numerical emplacement models
The dynamics of the debris avalanche motion have been evaluated by numerical modelling (Heinrich et al. 2001). In these studies: (1) the avalanche was idealized as the flow of a homogeneous incompressible continuum due to its observed, macroscopic fluidlike behaviour; (2) bed erosion was neglected; (3) mass and momentum conservation equations were depth-averaged over the flow thickness, since the characteristic length of the avalanche was much larger than the thickness; (4) energy dissipation within the flow was neglected and the slope-parallel velocity was assumed approximately constant over the thickness (Savage & Hutter 1989); and (5) longitudinal gradients of the deviatoric stress were neglected throughout the flow. The governing equations used a slope-parallel and slope-normal co-ordinate system (Heinrich et al. 2001). Basal friction was modelled by Coulomb-type friction laws, where 6 is the friction angle between the rough bed and the mass. The friction angle could be assumed to be constant, independent of the shear rate (Savage & Hutter 1989; Naaim et al. 1997) or defined as a function of both the velocity (u) and the height of the flow (/?) (Pouliquen 1999). In the case of a constant friction angle, a constant ratio of the shear stress to the normal stress at the base was assumed, similar to a friction law for a rigid block on an inclined plane. Pouliquen (1999) argued that the constant friction assumption failed for granular flows over rough bedrock, and proposed an empirical friction coefficient = tan as a function of the Froude number (u/-\/(gh)) and the thickness h of the granular layer: (u,h ) = tan
401
+ (tan
- tan ) exp
The numerical model was based on a shock-capturing method, similar to those used to simulate shock waves in compressible flows, which appeared to be stable and accurate in other applications (Mangeney et al. 2000; Heinrich et al. 2001; Assier et al. 2000). The method is a one-dimensional Lagrangian approach, based on a highorder Godunov-type scheme as discussed by Heinrich et al. (in press). The landslide is initialized in the simulation by a parabolic shaped solid with volume of 50 x 10 6 m 3 , released from rest. A first series of numerical simulations was performed using the simple Coulomb friction law. for different values of the basal friction angle. The best agreement was obtained for 13 < < 15 (Heinrich et al. 2001). However, with this model a steep flow snout was absent, contrary to the field observations, and the model overspill areas located west of the White River valley did not correspond well to the field observations. In a second series of simulations, using Pouliquen's friction law. the observed phenomena were approximately reproduced for around 10 , 62 around 20 and assumed mean diameters of the order of 1 m (Fig. 41). By trial and error, the friction angles 8\ and 62 were chosen respectively at 11 and 25 . with d, the theoretical mean particle diameter, ranging from 1.35m to 2.6m. Model velocities were comparable with the field estimates as discussed above. For effective mean diameters d < 2.2m the avalanche overtopped the west bank of the White River valley and flowed for an excessive distance (Fig. 41c). For d > 2.2 m the deposit area was somewhat smaller than observed. The calculated runout distance was partly dependent on the mean diameter, with a difference of 300m obtained between solutions for d= 1.35m and 2.6m.
Fig. 41. (a, b) Flow evolution for Pouliquen's friction law with = 11 and 62 = 25 , and a mean particle diameter d=2.2m. (c) Flow position at / = 200 s for the same angles and and d = 1.35m. (d) Flow position at t = 200 s for the same rheological parameters (a) and (b) but using a new topography with a partial filling of the White River valley.
402
B. VOIGHT ET AL.
Fig. 42. Deposit thicknesses (m) calculated by Pouliquen's friction law with = 11° and 62 = 25° and a particle diameter d— 2.2m.
A good general agreement was obtained with deposit thicknesses (Fig. 42). West of the Fergus Mountain bend, the maximum model deposit thickness was about 10m, fairly close to that observed. In the valley, deposit thicknesses ranged from 7 to 60m for distances of 300 to 800 m from the sea, with a more pronounced frontal snout (c. 20m) than for models calculated by simple Coulomb friction. The simulated avalanche stopped 200m from the shoreline. A final simulation was carried out to estimate the influence of topographic changes in the valley. Starting with pre-eruption topography, the valley was then arbitrarily filled to 0.75 depth before releasing the avalanche volume. Numerical results are presented in Figure 41d for Pouliquen's law with friction angles between 11° and 25° and D = 2.2m. The results at t = 200s showed that overspill flow at the Fergus Mountain bend was accentuated, and resulting overspill deposit areas were very large. The volume passing the bend in the valley was then smaller, and so smaller thicknesses were obtained in the lower part of the White River valley. As a result of thickness reduction, the angle 82 was activated, which finally led to shorter runout. Thus, in this case, partly filling up the valley and smoothing topography did not lead to greater downstream velocities and runout, and smaller deposit thicknesses were obtained at the lower end of the valley. However, this counterintuitive result cannot be generalized, as less extreme filling might have reduced localized roughness of the channel floor, and yet not have facilitated an excessive overspill; in this case, intuition would have correctly suggested an increased runout.
Discussion of model results The best agreement was obtained for a model using Pouliquen's friction law, with an assumed effective mean diameter of 2.2m and
friction angles varying from 11° to 25°. This best-fit model-based particle diameter was greater than the median size of small-volume samples, and was perhaps also larger than field values even if much larger sample sizes were considered, although we must also recognize that fragmentation occurred during the flowage process (cf. Ui et al. 1986). In part this result may reflect the very small range of sizes considered in the original Pouliquen analysis, and also the need to extrapolate the size calibration procedure to very poorly sorted materials. Comparisons between flows calculated by Coulomb and Pouliquen's friction laws have shown the importance of the dependence of the friction angle on the Froude number and the flow thickness. The Pouliquen model gave results in better agreement with thicknesses observed in the field than those calculated by a simple Coulomb law. This result suggests a shear-rate dependence in the mechanical behaviour of debris avalanches, or at least the action of a second parameter. Irrespective of the friction law used, the empirical value of the apparent friction angle required to reproduce the significant mobility of the Boxing Day avalanche in Montserrat was low ( < 15°), in general agreement with the widely observed but poorly understood apparent excess mobility of large debris avalanches (Voight et al. 1985). At Montserrat, as at Mount St Helens (Voight et al. 1983), pore-water pressure in disaggregated, fumarolically altered debris probably partly accounted for enhanced mobility. For water-saturated materials, the apparent friction coefficient is a function of the actual friction coefficient and the pore-fluid pressure (Voight 1978, pp. 154-155; Sassa 1988; Voight & Elsworth 1997, pp. 11-14). Geotechnical testing of avalanche debris at Montserrat suggested effective-stress friction angles for sandy textured debris of 25-35° (peak) and 17-25° (residual), with minor cohesive strength. The apparent friction values for these materials in a flowing avalanche associated with high pore-water pressures would be much reduced, and more consistent with modelled values. In addition, local seams of clay-rich materials have measured low frictional strength (3-14°). Such materials are only found locally and probably do not account for bulk avalanche mobility, although it may be noted that failures in nature preferentially select the weakest materials. The ratio of fall height to runout distance (H/L) for the avalanche was about 0.22 (Fig. 25). This ratio can be crudely interpreted in terms of the apparent friction coefficient of the avalanche (Pariseau & Voight 1979), and thus the apparent friction angle is around 12-13°. This result suggests that much of the initial potential energy was consumed by basal friction (Hutter 1996), and that local topography variations did not too strongly influence runout. These results contrast with those of Dade & Huppert (1998), who suggest from scaling relationships that a constant-stress resistance law applies to long-runout mass movements. Our experimental and modelling results suggest that this resistance is not constant, but instead is related by friction coefficient to the varying thickness of the avalanche during emplacement. The Pouliquen relation suggests further that resistance may also be related to velocity. The Dade & Huppert result possibly can be rationalized by the view that an average resisting stress can be defined, and is approximately proportional to the average thickness of an avalanche. However, in detail, this thickness will vary considerably from point to point and from time to time during the emplacement process, and the resisting stresses will also vary in a corresponding way. Comparisons and contrasts Comparisons of the Boxing Day sector collapse and dome explosion with the well documented May 1980 edifice collapse and cryptodome explosion of Mount St Helens, USA, and the March 1956 collapse and explosion of Bezymianny in Russian Kamchatka, seem particularly relevant. All three complex events, edifice failures and explosions, are of Bezymianny-type (Gorshkov 1962; Siebert et al 1987; Voight & Elsworth 1997), meaning that an injection of magma had preceded and probably had prompted the sector collapse. A similar event at c. 3000 BP has been postulated for Soufriere Volcano in neighbouring Guadaloupe (Boudon et al. 1984).
THE BOXING DAY SECTOR COLLAPSE
In all these events, a sector collapse preceded and triggered the generation of energetic pyroclastic currents. At Mount St Helens, dacite magma was intruded asymmetrically early in the eruption, as a cryptodome into the interior of the stratovolcano (Lipman & Mullineaux 1981). The high-viscosity intrusion caused marked deformation of the north slope at rates of c. 2m day - 1 (Lipman et al 1981), which caused much concern for the USGS observatory team. In an interpretation one month in advance of the collapse, prompted by qualitative observations of the deformation, the possibility of a massive sector collapse and an accompanying pyroclastic explosion was recognized (Voight 1980, 2000; Decker 1981). On 18 May 1980, after several months of significant deformation and seismicity - but without recognized short-term precursors such as accelerating deformation or seismicity - sector collapse occurred retrogressively to generate a complex debris avalanche. This in turn facilitated a devastating laterally directed explosion that evolved contemporaneously with the later movement stages of the avalanche (Voight et al. 1981, 1983; Glicken 1986, 1998; Fisher et al. 1987; Sousa & Voight 1995; Hoblitt 2000). A moderately large (M = 5) earthquake occurred at the time of slope collapse and was generally interpreted to have triggered it. In the case of Bezymianny in 1956, andesitic magma was intruded into the edifice as much as six months before slope collapse, and frequent minor magmatic explosions had occurred. Massive deformation of the volcano slope by intruding magma preceded sector collapse of the edifice (Gorshkov 1959). Failure there too was associated with an earthquake, and collapse probably occurred retrogressively (Voight & Elsworth 1997) to generate a massive volcanic debris avalanche, and a violent laterally directed explosion (Belousov & Bogoyavlenskaya 1988; Belousov 1996). No detailed monitoring had been attempted, but large earthquakes were detected by the regional seismic network. In contrast, at Soufriere Hills Volcano a dome of pressurized andesitic lava built up over the south flank, and this, with talus shed by the dome, loaded a weakened (and eroded) area of altered rock that included active hydrothermal springs. Seismic precursors preceded the failure in the short term. The sector collapse undermined the pressurized dome, and thus the debris avalanche was succeeded by a destructive, laterally directed pyroclastic current. The collapse occurred at a mature stage in the eruption, which started about 2.5 years earlier, but signs of potential structural instability had developed over a year before and recognition of these signs had led to effective risk mitigation. The common factors in these three events are a relatively weak, partly altered edifice further destabilized by emplacement of silicic magma containing pressurized volatiles, and subjected to dynamic seismic loading. However, the differences are also substantial. It appears that the Soufriere Hills example falls into a separate class involving exogenous shear loading by extruding lava above a hydrothermally weakened flank, in contrast to endogenous stressing of the edifice by an intruding cryptodome (Donnadieu & Merle 1998; Voight 2000). Further, although large earthquake shocks occurred at Mount St Helens and Bezymianny, no such large individual triggering shocks were detected at Soufriere Hills; instead the sector collapse was preceded by a seismic swarm and tremor of many hours' duration. The Boxing Day events illustrate the potential for highly dangerous activity in all dome eruptions where hydrothermal alteration related to earlier episodes of volcanic activity has created a weakened edifice, and where high gas pressures in the interior of a dome lava and/or conduit can contribute to destabilization (Voight & Elsworth 2000), and can also generate violent pyroclastic currents if the edifice fails. The Socompa debris avalanche, northern Chile, as studied by our deceased co-author and friend Peter Francis and his associates (Francis et al. 1985; Francis & Self 1987; van Wyk de Vries & Francis 1996), offers other significant comparisons. The Socompa sector collapse (c. 7000 BP) reflected loading of weakened base material by the growing volcano (van Wyk de Vries et al. 1999), which caused outward spreading in the substrate prior to collapse. A somewhat similar process may have occurred at the base of the
403
rapidly growing talus cone at Montserrat. It is conceivable that loading under undrained conditions there (in saturated strata of low permeability) promoted creep movement and strain-weakening that influenced the flank failure. Some volcanoes, such as Mount St Augustine (Alaska), Shiveluch (Kamchatka, Russia) and Egmont (New Zealand) have collapsed repeatedly (Beget & Kienle 1992; Belousov et al. 1999; Ui et al. 1986; Palmer et al. 1991). Work by Boudon et al. (1999) and Deplus et al. (1999) has shown that volcanoes in the Carribbean arc also have been prone to sector collapse in the past. At least eight collapses have been recognized at Soufriere Volcano, Guadeloupe, where continuous vigorous hydrothermal activity, over thousands of years, has considerably modified the structure and mechanical properties of an edifice periodically affected by phreatic eruptions and less frequent magmatic dome eruptions. Indeed, the Boxing Day sector collapse at Soufriere Hills Volcano repeats, in general terms, a phenomenon that occurred there several times before, with English's Crater representing the scar of a major prehistoric debris avalanche, and with extensive hummocky avalanche deposits offshore (Boudon 2001). Discussion: evolution of the sector collapse Monitoring and mitigation The 26 December sector collapse at Soufriere Hills Volcano illustrated that a major edifice failure and directed explosion could occur, despite relatively intensive monitoring, without precursors sufficiently diagnostic to enable reliable short-term forecasting. Nevertheless, the potential for a collapse and explosion of this general kind at Montserrat was recognized over a year before its occurrence, which led to a precautionary evacuation. Consequently there were no casualties, despite the unmistakable violence of the events (Sparks et al. 2002). Further, the recognition of a pattern of enhanced volcanic activity every six to seven weeks from May 1997 onwards (Voight et al. 1998, 1999) enabled anticipation by the Montserrat Volcano Observatory team that a significant event or series of events might take place towards the end of December 1997. Further discussion of the hazards assessments and their implications are reported by Young et al. (2002). Causes of the collapse The 26 December sector collapse was caused by some combination of long-term alteration-weakening by hydrothermal activity, subsequent dome lava and talus loading of the (partly) water-saturated materials, and shaking associated with volcanic seismicity. Static conduit-magma pressurization seems unlikely as a major direct influence on failure, but might have been indirectly related, for instance as a factor in the production of hybrid seismicity (Voight et al. 1999). The rapid south-directed exogeneous growth of a shear-lobe of dome lava, and shedding of lava-block talus over the Galway's Wall and Galway's Soufriere area since early November 1997 provided quasi-static loading and perhaps also pore-fluid pressure enhancement that strongly contributed to failure. These loads may have been applied in undrained fashion (in the geotechnical sense; see Lambe & Whitman 1969) to pockets of watersaturated, weak, low-permeability, fines-rich alteration materials. Seismic loading may have augmented the destabilization forces in the 24 hours (especially the last few hours) prior to the sector collapse, in concert with the load changes caused by rapid effusion and endogenous dome lava emplacement. Thus, although the sequence of night-time failure events was not observed and is imprecisely known, an episode of rapid effusion accompanied by the shallow seismicity of 24-26 December probably provided the trigger for localized base shear-failure of the hydrothermally weakened hill of tuff between Galway's Soufriere and the deteriorated Galway's Wall. Possibly, this hill may have comprised a toreva block from a prehistoric sector collapse in the
404
B. VOIGHT ET AL.
White River valley, such that a partial slip surface already existed and merely required reactivation. Failure was probably progressive. With strain-weakening causing the shear resistance in the altered material to approach residual strengths, the slip surface propagated northwards through talus to the top of the lava dome, and broke out southwards through a toe at Galway's Soufriere while becoming detached on lateral margins. Strain-weakening behaviour is indicated by the laboratory tests, with the residual friction angle results varying widely, from 3° in clay to 31° in tuff. A typical friction angle for the avalanche debris may be between these extremes (e.g. sample 2A), but it should be recognized also that failure processes are selective, and thus the weakest units could strongly influence the failure, even if they did not represent a large volumetric proportion of the materials that failed.
is a function of both the actual friction coefficient and the pore-fluid pressure (Voight 1978; Sassa 1988; Voight & Elsworth 1997). Further, as discussed above, avalanche transport dynamics are strongly influenced by weak lithologies and weak zones. The best-fit dynamic model suggested an avalanche emplacement time of less than three minutes and typical maximum velocity of about 40 m s - 1 . These data are consistent with estimates based on field constraints, which suggest a probable emplacement time of 138s or more, dependent on location, and an average speed in excess of 19-35 m s - 1 . The numerical results illustrate the potential of such models to evaluate the capability of debris avalanches to surmount relief and to affect areas beyond the obvious drainage channels. Such modelling could be useful in anticipating potential effects of avalanches in future volcanic crises.
Flow behaviour
Conclusions
Most of the detached mass accelerated under the impetus of gravity, and moved rapidly as a disintegrating, shearing avalanche down the White River valley. In these movements the source region tended to fragment along weak layers and boundaries, forming avalanche megablocks, which subdivided further at weak boundaries during transport. The strength of the avalanche was effectively governed by the weakest lithologies and weak zones. Much of the megablock-facies material was more or less cohesive, with the overlying talus non-cohesive. This difference probably resulted in non-uniform avalanche flow behaviour, although the possibility is enigmatic to assess because of difficulty in identifying some talus materials from blocky pyroclastic flow deposits. The debris avalanche flow was unsteady, with several flow pulses implied by channel overtopping relations and by lobate forms on the deposit surface. Movement of following material was arrested as, at several locations, debris slowed in front, and transverse steps were formed similar to those described for volcanoes Socompa (Wadge et al 1995), Avachinsky (Castellana et al. 1995) and Shiveluch (Belousov et al. 1999). In turn, the sector collapse resulted in an oversteepened front of the fresh lava dome, which caused it to disintegrate, partly by gravity but also by release of external loads on the gas-pressurized lava, and to form a pyroclastic current that ravaged the south flank. The shear tended to be focused towards the base of the avalanche, accounting for the basal shear fabric, and for the transport of little-deformed megablock facies. Nevertheless, both compressional and dilational strains affected the avalanche debris at different places and times, and with increasing distance there was enhanced clast shattering, clast-size reduction, and mingling of coloured units. The textures mainly suggest emplacement by laminar flow, with interaction between megablock domains dominated by near-neighbour compression and shearing.
(1)
Emplacement dynamics Insights into the dynamics of avalanche emplacement were provided by three-dimensional numerical simulations, with the avalanche idealized as a flow of a homogeneous continuum governed by a basal friction law. The dominance of basal friction is fully justified by the fabric and structural observations discussed above. Numerical results showed that the observed distribution of debris, including overspill deposition, is well reproduced for an assumed Coulombtype (Pouliquen) friction law with a friction coefficient dependent upon the thickness and the velocity of the flowing mass. The low values of friction implied by numerical modelling are consistent with the fact that large avalanches travel farther than expected from overly simple models of slope failure. The mobility of this avalanche may have been enhanced by the presence of water-saturated, highly altered material from the Galway's fumarolic system, which constituted a large portion of the avalanche. For such water-saturated materials, the apparent friction coefficient
(2)
(3) (4)
(5)
(6)
Failure of the southern sector of Soufriere Hills Volcano on 26 December 1997 culminated in a devastating eruptive episode. Sector collapse produced a volcanic debris avalanche, and exposure of the depressurized face of the lava dome then resulted in generation of a powerful pyroclastic density current. A precautionary evacuation resulted in avoidance of casualties. The south-directed growth of a lava lobe and shedding of lavablock talus over the Galway's Wall and Galway's Soufriere area, particularly since early November 1997, provided static loading on hydrothermally weakened materials that brought the flank to a condition of marginal stability. Collapse was triggered by a pulse of co-seismic, exogenous, lava shear-lobe emplacement, with subsequent gravity-driven slip-surface localization influenced by strain-weakening. The source region tended to fragment along weak layers and boundaries to form avalanche megablocks, which were then transported and further disrupted by shear along the megablock boundaries and weak internal zones. Thus the strength of the avalanche was effectively governed by the locally weakest lithologies and weak zones. Shear was focused towards the base of the avalanche, accounting for the basal shear fabric, and for the transport of less-deformed megablock facies within the body of the flow. The resulting debris avalanche was sustained and consisted of several flow pulses that reflected complexities of the source disruption, varying flow properties of the older rocks and dome talus, and channel topography. At major bends, the avalanche separated into channelled and overspill flows, and in the distal region, stacked sets of the main lithologies occur with a hummocky surface and with abrupt flow-unit snouts. Numerical simulations support field observations that the avalanche dynamics were governed by basal friction, and suggest that this friction was dependent upon the materials, and also the thickness and the velocity of the flowing mass. The low values of friction implied by the simulations are consistent with geotechnical test data and the inferred localized presence of pore-water pressures.
We are primarily grateful to our colleagues of Montserrat Volcano Observatory (MVO) for considerable, continued intellectual and fieldwork support. We thank the UK Department for International Development for generous financial support of the volcano monitoring on Montserrat. Likewise the assistance of British Geological Survey (BGS) is acknowledged and appreciated. B.V., R.S.J.S. and S.R.Y. acknowledge support from BGS as Senior Scientists affiliated with MVO, and B.V. notes in addition, important assistance from several grants from the US National Science Foundation. Work by A.B.B. and M.B. was partly supported by grant RG1-172 of the US Civilian Research and Development Foundation awarded to A.B.B. and B.V., and also by the Alexander von Humboldt Foundation. Fieldwork by J.-C.K. was undertaken during time spent as a Senior Scientist with MVO. Supporting funds for J.-C.K. and G.B. were also obtained from the French PNRN (INSU-CNRS) research programmes, and the Observatoires Volcanologiques of the Institut de Physique du Globe de Paris (IPGP). Ph.H. was supported by the Commissariat a PEnergie Atomique, France.
THE BOXING DAY SECTOR COLLAPSE We recall with sadness and appreciation our colleague Peter Francis, a charter member with H. Glicken of Friends of Volcanic Debris Avalanches, with both being contributors of seminal works. We thank the helicopter pilots, particularly J. McMahon, A. Grouchy, and pilots from Bajan Helicopters (Montserrat Air Support Unit), for their skill and help, often beyond the call of duty. K. West and M. Feuillard offered photographs for our use, and many other photographs used were provided by staff of MVO. B.V. acknowledges helpful interchanges with colleagues D. Elsworth, H. R. Hardy, E. Kimball. H. W. Shen, and E. Oh. D. Hidayat, R. Herd, F. Donnadieu, and M. Volero helped with drafting. Critical reviews by R. E. A. Robertson and S. Self, and the exceptional editorship of P. Kokelaar, led to significant improvement of the paper. The assistance of J. K. McClintock in many matters is much appreciated. Published by permission of the Director, BGS (NERC).
Appendix Geotechnical tests For most direct shear tests on Sample 15-2A, well-mixed moist samples were pushed through a 2 mm sieve and moulded into the shear frame, which had a diameter of 6cm. A normal stress of 0.43 MPa was applied, the specimens were consolidated in a watersaturated state, and slow shearing accomplished at rates of, typically, 1.3x 10 - 4 mms - 1 . The specimens were then reverse-sheared through a number of cycles to, usually, cumulative displacements of 90-40 mm or more, to provide a measure of displacementweakening. These procedures were repeated for a normal pressure range of 0.86-1.72 MPa. Similar tests were carried out on clayrich block sample 15-2C, but in this case the block was cored with a thin-walled tube, and core specimens were extruded, trimmed, and inserted into the direct shear frame. Similar tests were also conducted on disaggregated, weathered tuffs from the Galway's Wall and Gages Wall areas of English's Crater (Fig. 2). For the block samples of tuff collected at Galway's Wall and Gages Wall, the blocks were cored (2.5 cm diameter) and duplicate right-cylinder specimens were subjected to laboratory triaxial testing to determine the shear strength of intact material as a function of confining pressure. Tests were conducted dry at room temperature, using loading rates of 138kPas - 1 , with confining pressures varied from 0.62-9.93 MPa. Additional testing was carried out to obtain intact tensile strengths by the method of cylinder edge-loading.
References ALIDIBIROV, M. & DINGWELL, D. B. 1996. Magma fragmentation by rapid decompression. Nature, 380, 146-148. ASSIER, S., HEINRICH, P.. SABATIER, P. C, SAVOYE, B. & BOURILLET, J. F. 2000. Numerical modeling of a landslide-generated tsunami: the 1979 Nice event. Pure and Applied Geophysics, 157. 1707-1727. BEGET, J. & KIENLE, J. 1992. Cyclic formation of debris avalanches at Mount St. Augustine volcano. Nature. 356, 701-704. BELL, F. G. 1993. Engineering Geology. Blackwell, London. BELOUSOV, A. B. 1996. Pyroclastic deposits of March 30 1956 directed blast at Bezymianny volcano. Bulletin of Volcanology, 57. 649-662. BELOUSOV, A. B. & BOGOYAVLENSKAYA, G. 1988. Debris avalanche of the 1956 Bezymianny eruption. Proceedings Kagoshima International Conference on Volcanoes, 460-462. BELOUSOV, A. B.. BELOUSOVA. M. & VOIGHT, B. 1999. Multiple edifice failures, debris avalanches, and associated eruptions in the Holocene history of Shiveluch volcano, Kamchatka, Russia. Bulletin of Volcanology, 61, 324-342 BOUDON, G. 2001. Instabilites gravitationnelles de l'are volcanique des petites Antilles. Atelier sur les Aleas Volcaniques les volcans antillais - des processus aux signaux. Paris, 18-19 Janvier. Institut National des Sciences de l'Univers, 4-5. BOUDON, G.. SEMET. M. P. & VINCENT. P. M. 1984. Flank failure-directed blast eruption at la Soufriere de Guadaloupe, French West Indies: a 3.000 year old Mount St Helens? Geology. 12. 350-353. BOUDON, G., VILLEMANT. B., KOMOROWSKI. J-C, ILDEFONSE. PH., HAMMOUYA. G. & SEMET. M. P. 1996. The hydrothermal system of Soufriere
405
Hills volcano. Montserrat (West Indies): Chemical, mineralogical. and microtextural signatures in fluids, altered rocks, and 1995-96 tephra. ///: WADGE. G. (ed.) The Soufriere Hills Eruption, Montserrat. Discussion Meeting of the Volcanic Studies Group of the Geological Society, 27 November 1996. Montserrat Volcano Observatory and Geological Society of London. 31-34. BOUDON. G., VILLEMANT. B.. KOMOROWSKI. J.-C.. ILDEEONSE. P. & SEMET. M. P. 1998. The hydrothermal system at Soufriere Hills volcano. Montserrat (West Indies): Characterization and role in the ongoing eruption. Geophysical Research Letters, 25. 3693-3696. BOUDON. G., KOMOROWSKI. J-C.. SEMET. M.. LE FRIANT. A. & DEPLUS. C. 1999. Frequent volcanic flank-collapses in the Lesser Antilles arc: origin and hazards. Eos, Transactions, American Geophysical Union. 80(46). F1142. BROMHEAD. E. N. 1986. The Stability of Slopes. Surrey University. London. CALDER. E. S.. YOUNG. S. R.. SPARKS. R. S. J. ET AL. 1998. The Boxing Day collapse, 26 December 1997. Montserrat Volcano Observatory Special Report 6. CASTELLANA. B.. DAVIDSON. J. P.. BELOUSOV. A. B. & BELOUSOVA. M. 1995. Milestones of geology of the Avachinskiy volcano. Kamchatka. Russia. Eos, Transactions, American Geophysical Union. 76. 537. CHIODJNI, G.. CIONI. R.. FRULLANI. A.. GUIDI. M.. MARINI. L.. PRATI. F. & RACO. B. 1996. Fluid geochemistry of Montserrat Island. West Indies. Bulletin of Volcanology. 58. 380-392. COETZEE. M. J.. HART. R. D . VARONA. P. D. & C U N D A L L . P. A. 1998. FLAC
Basics. Itasca. CRANDELL. D. R.. MILLER. C. D.. GLICKEN. H. X.. CHRISTIANSEN. R. L. & NEWHALL. C. G. 1984. Catastrophic debris avalanche from ancestral Mount Shasta volcano. California. Geology. 12. 143-146. CUNDALL. P. & BOARD. M. 1988. A microcomputer program for modeling large-strain plasticity problems. In: Numerical Methods in Geomechanics (Proceedings 6th International Conference. Innsbruck. Austria. April 1988). Balkema. Rotterdam. 2101-2108. DADE, W. B. & HUPPERT. H. E. 1998. Long-runout rockfalls. Geology. 26. 803-806. DECKER. R. W. 1981. The 1980 activity: a case study in volcanic eruption forecasting. ///: LIPMAN. P. W. & M U L L I N E A U X . D. R. (eds) The 1980 Eruptions of Mount St Helens. Washington. US Geological Survey. Professional Paper. 1250. 815-820. DEPLUS. C., BOUDON. G.. CHEMINEE. J-L. ET AL. 1999. Large-scale debris avalanche deposits on the western flank of the Lesser Antilles arc revealed by the Aguadomar Cruise. Eos. Transactions, American Geophysical Union. 80(46). 1152. DETOURNAY. C. & HART. R. (eds) 1999. FLAC and Numerical Modeling in Geomechanics. Balkema. Rotterdam. DONNADIEU. F. & MERLE. O. 1998. Experiments on the indentation process during cryptodome intrusions: new insights into Mount St Helens deformation. Geology. 26. 79-82. DRUITT. T. H. 1998. Pyroclastic density currents. ///: GILBERT. J. S. & SPARKS. R. S. J. (eds) The Physics of Explosive Volcanic Eruptions. Geological Society. London. Special Publications. 145. 145-182. DRUITT. T. H.. CALDER. E. S.. COLE. P. D. ET AL. 2002. Small-volume, highly mobile pyroclastic flows formed by rapid sedimentation from pyroclastic surges at Soufriere Hills Volcano. Montserrat: an important volcanic hazard. In: DRUITT. T. H. & KOKELAAR. B. P. (eds) The Eruption of Soufriere Hills Volcano. Montserrat, from 1995 to 1999. Geological Society. London. Memoirs. 21. 263-279. DUNCAN, J. M. 1992. State-of-the art: static stability and deformation analysis. American Society of Civil Engineers. Geotechnical Engineering Division Special Publication. 31. 222-265. DUNCAN. J. M. 1996. State-of-the-art: limit equilibrium and finite-element analysis of slopes. American Society of Civil Engineers. Journal of Geotechnical Engineering. 122. 577-596. FINK. J. H. & KIEFFER. S. W. 1993. Estimates of pyroclastic flow velocities resulting from explosive disintegration of lava domes,Nature. 363. 612-614. FISHER. R. V., GLICKEN. H. X & HOBLITT. R. P. 1987. May 18 1980. Mount St Helens deposits in South Coldwater Creek. Washington. Journal of Geophysical Research. 92(B10). 10267-10283. FRANCIS, P. W. & SELF. S. 1987. Collapsing volcanoes. Scientific American. 287. 90-99. FRANCIS, P. W. & WELLS. G. L. 1987. Landsat thematic mapper observations of debris avalanche deposits in the Central Andes. Bulletin of Volcanology, 50. 258-278.
406
B. VOIGHT ET AL.
FRANCIS, P. W., GARDEWEG, M., RAMIREZ, C. F. & ROTHERY, D. A. 1985. Catastrophic debris avalanche deposit of Socompa volcano, northern Chile. Geology, 13, 600-603. FREDLUND, D. G. 1984. Analytical methods for slope stability analysis. Proceedings 4th International Symposium on Landslides, Toronto, Vol. 1, 229-250. GIGGENBACH, W. F. 1992. Magma degassing and mineral deposition in hydrothermal systems along convergent plate boundaries. Economic Geology, 87,1927-1944. GLICKEN, H. 1986. Rockslide-debris avalanche of May 18 1980, Mount St Helens Volcano, Washington. PhD dissertation, University of California at Santa Barbara. GLICKEN, H. 1991. Sedimentary architecture of large-volume debris avalanches. In: FISHER, R.V. & SMITH, G. A. (eds) Sedimentation in Volcanic Settings. Society of Sedimentary Geologists (SEPM), Special Publications, 45, 99-106. GLICKEN, H. 1998. Rockslide-debris avalanche of May 18 1980, Mount St. Helens volcano, Washington. Bulletin Geological Society of Japan, 49 (2/3), 55-106. GORSHKOV, G. S. 1959. Gigantic eruption of the volcano Bezymianny. Bulletin of Volcanology, 21, 77-109. GORSHKOV, G. S. 1962. On the classification and terminology of Pelee and Katmai type eruptions. Bulletin of Volcanology, 24, 155-165. HEINRICH, P., MANGENEY, A., GUIBOURG, S., ROCHE, R., BOUDON, G. & CHEMINEE, J-L. 1998. Simulation of water waves generated by a potential debris avalanche in Montserrat, Lesser Antilles. Geophysical Research Letters, 25(19), 3697-3700. HEINRICH, P., PIATANESI, A., OKAL, E. & HEBERT, H. 2000. Near-field modelling of the July 17 1998 tsunami in Papua New Guinea. Geophysical Research Letters, 27(19), 3037-3040. HEINRICH, P., PIATANESI, A. & HEBERT, H. 2001. Numerical modelling of tsunami generation and propagation from submarine slumps: the 1998 Papua New Guinea event. Geophysics Journal International, 145(1), 97-111. HEINRICH, P., BOUDON, G., KOMOROWSKI, J-C., SPARKS, R. S. J., HERD, R. & VOIGHT, B. 2001. Numerical simulation of the December 1997 debris avalanche in Montserrat, Lesser Antilles. Geophysical Research Letters, 28, 2529-2532. HOBLITT, R. P. 2000. Was the 18 May 1980 lateral blast at Mt St Helens the product of two explosions? Philosophical Transactions of the Royal Society, London, A358, 1639-1661. HOEK, E. 1983. Strength of jointed rock masses. Geotechnique, 33, 187-223. HOEK, E. & BROWN, E. T. 1980. Empirical strength criterion for rock masses. American Society of Civil Engineers, Journal of Geotechnical Engineering, 106, 1013-1035. HOEK, E., KAISER, P. K. & BAWDEN, W. F. 1994. Support of Underground Excavations in Hard Rock. Balkema, Rotterdam. HUTTER, K. 1996. Avalanche dynamics, a review. In: SINGH, V. P. (ed.) Hydrology of Disasters. Kluwer, Dordrecht, 317-394. INMAN, D. L. 1952. Measures for describing the size distribution of sediments. Journal of Sedimentary Petrology, 22,125-145. KOMOROWSKI, J-C., GLICKEN, H. X & SHERIDAN, M. 1991. Secondary electron imagery of microcracks and hackly fracture surfaces in sandsize clasts from the 1980 Mount St Helens debris-avalanche deposit: implications for particle-particle interactions. Geology, 19, 261-264. KOMOROWSKI, J-C., NORTON, G., BOUDON, G. ET AL. 1999. The 1997 flankcollapse of Soufriere Hills volcano, Montserrat: Constraints on emplacement mechanisms from debris avalanche morphological and textural data. Eos, Transactions, American Geophysical Union, 80(46), 1152. KRAMER, S. L. 1996. Geotechnical Earthquake Engineering. Prentice Hall, Upper Saddle River, New Jersey. LACROIX, A. 1904. La Montagne Pelee et ses eruption. Masson et Cie, Paris. LAMBE, T. W. & WHITMAN, R. V. 1969. Soil Mechanics. John Wiley, New York. LIPMAN,P.W.&MuLLiNEAUX,D. R.(eds) 1981. The 1980 Eruptions of Mount St Helens, Washington. US Geological Survey, Professional Paper, 1250. LIPMAN, P. W., MOORE, J. G. & SWANSON, D. A. 1981. Bulging of the north flank before the May 18 eruption: geodetic data. In: LIPMAN, P. W. & MULLINEAUX, D. R. (eds) The 1980 Eruptions of Mount St Helens, Washington, US Geological Survey, Professional Papers, 1250, 143-156. MACGREGOR, A. G. 1952. Eruptive mechanisms: Mt. Pelee, the Soufriere, St. Vincent, and the Valley of Ten Thousand Smokes. Bulletin Volcanologique, 12, 49-74.
MANGANEY, A., HEINRICH, PH. & ROCHE, R. 2000. Analytical and numerical solution of dam-break problem for application to water floods, debris and snow avalanches. Pure and Applied Geophysics, 157(6-8), 1081-1096. MAYBERRY, G. C., ROSE, W. I. & BLUTH, G. 1998. Analysis of the ash cloud from the December 26 1997 (Boxing Day) dome collapse of Soufriere Hills Volcano, Montserrat using GOES-8 Imagery. Eos, Transactions, American Geophysical Union, 79, 1014. MEDLEY, E. 1997. Geological engineering of bimrocks (abstract). Association of Engineering Geologists, Annual Meeting. MILLER, A. D., STEWART, R. C., WHITE, R. A. ET AL. 1998. Seismicity associated with dome growth and collapse at the Soufriere Hills Volcano, Montserrat. Geophysical Research Letters, 25, 3401-3404. NAAIM, M., VIAL, S. & COUTURE, R. 1997. Saint-Venant approach for rock avalanche modeling. Saint-Venant Symposium, August 28-29, Paris. PALMER, B. A., ALLOWAY, B. V. & NEALL, V. E. 1991. Volcanic-debrisavalanche deposits in New Zealand - lithofacies organization in unconfined, wet-avalanche flows. In: FISHER, R. V. & SMITH, G. A. (eds) Sedimentation in Volcanic Settings. Society of Sedimentary Geologists (SEPM), Special Publications, 45, 89-98. PARISEAU, W. G. & VOIGHT, B. 1979. Rockslides and avalanches: basic principles and perspectives in the realm of civil and mining operations. In: VOIGHT, B. (ed.) Rockslides and Avalanches, 2: Engineering Sites. Elsevier, Amsterdam, 1-92. POULIQUEN, O. 1999. Scaling laws in granular flows down rough inclined planes. Physics of Fluids, 11(3), 542-548. REA, W. J. 1974. The volcanic geology and petrology of Montserrat, West Indies. Journal of the Geological Society, London, 130, 341-366. REICHE, P. 1937. The toreva-block - a distinctive landslide type. Journal of Geology, 45, 538-548. RITCHIE, L. J., COLE, P. D. & SPARKS, R. S. J. 2002. Sedimentology of deposits from the pyroclastic density current of 26 December 1997 at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 435-456. ROBERTSON, R. E. A., COLE, P., SPARKS, R. S. J. ET AL. 1998. The explosive eruption of Soufriere Hills Volcano, Montserrat, West Indies, 17 September, 1996. Geophysical Research Letters, 25, 3429-3432. ROOBOL, M. J. & SMITH, A. L. 1998. Pyroclastic stratigraphy of the Soufriere Hills volcano, Montserrat - Implications for the present eruption. Geophysical Research Letters, 25, 3393-3396. SASSA, K. 1988. Geotechnical model for the motion of landslides. Proceedings 5th International Symposium on Landslides, Lausanne, Vol. 1, 37-55. SAVAGE, S. B. & HUTTER, K. 1989. The motion of a finite mass of granular material down a rough incline. Journal of Fluid Mechanics, 199, 177-215. SIEBERT, L. 1984. Large volcanic debris avalanches: characteristics of source areas, deposits, and associated eruptions. Journal of Volcanology and Geothermal Research, 22, 163-197. SIEBERT, L. 1996. Hazards of large volcanic debris avalanches and associated eruptive phenomena. In: SCARPA, R. & TILLING, R. I. (eds) Monitoring and mitigation of volcano hazards. Springer-Verlag, Berlin, 541-572. SIEBERT, L., GLICKEN, H. & UI, T. 1987. Volcanic hazards from Bezymiannyand Bandai-type eruptions. Bulletin of Volcanology, 49, 435-459. SIEBERT, L., BEGET, J. E. & GLICKEN, H. 1995. The 1883 and late-prehistoric eruptions of Augustine volcano, Alaska. Journal of Volcanology and Geothermal Research, 66, 367-395. SLINGERLAND, R. L. & VOIGHT, B. 1979. Occurrences, properties, and predictive models of landslide-generated water waves. In: VOIGHT, B. (ed.) Rockslides and Avalanches, 2: Engineering Sites. Elsevier, Amsterdam, 317-397. SOUSA, J. & VOIGHT, B. 1995. Multiple-pulsed debris avalanche emplacement at Mount St Helens in 1980: evidence from numerical continuum flow simulation. Journal of Volcanology and Geothermal Research, 66, 227-250. SPARKS, R. S. J., BARCLAY, J., CALDER, E. S. ET AL. 2002. Generation of a debris avalanche and violent pyroclastic density current on 26 December (Boxing Day) 1997 at Soufriere Hills Volcano, Montserrat. In: DRUITT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 409-434. UI, T. 1983. Volcanic dry avalanche deposits identification and comparison with non-volcanic debris stream deposits. Journal of Volcanology and Geothermal Research, 18, 135-150.
THE BOXING DAY SECTOR COLLAPSE UI, T. & GLICKEN, H. 1986. Internal structural variations in a debrisavalanche deposit from ancestral Mount Shasta, California, USA. Bulletin of Volcanology, 48, 189-194. UI, T., KAWACHI, S. & NEALL, V. E. 1986. Fragmentation of debris avalanches during flowage. Evidence from the Pungarehu Formation, Mount Egmont, New Zealand. Journal of Volcanology and Geothermal Research, 27, 255-264. UI, T., TAKARADA, S. & YOSHIMOTO, M. 2000. Debris avalanches. In: SIGURDSSON, H. (ed.) Enclyclopedia of Volcanoes. Academic Press, San Diego, 617-626. VAN WYK DE VRIES, B. & FRANCIS, P. W. 1996. Catastrophic collapse at stratovolcanoes induced by gradual volcano spreading. Nature, 387, 387-390. VAN WYK DE VRIES, B., SELF, S., FRANCIS. P. & KESZTHELYI, L. 1999. Development of the Socompa debris avalanche (N Chile) from a spreading volcanic edifice. Eos, Transactions, American Geophysical Union, 80(46), 1142. VOIGHT, B. 1978. Lower Gros Ventre slide, Wyoming, USA. In: VOIGHT, B. (ed.) Rockslides and Avalanches, 1: Natural Phenomena. Elsevier, Amsterdam, 113-166. VOIGHT, B. 1980. Slope stability hazards, Mount St Helens volcano, Washington. File report, US Geological Survey Cascades Volcano Observatory, Washington (reproduced as appendix in: VOIGHT, B. 2000. Stability of andesite volcanoes and lava domes. Philosophical Transactions of the Royal Society, London, A 358, 1694-1703). VOIGHT, B. 1996. Strength of wall rock of Englishes Crater, Soufriere Hills volcano, Montserrat. In: WADGE, G. (ed.) The Soufriere Hills Eruption, Montserrat. Discussion Meeting of the Volcanic Studies Group of the Geological Society, 27 November 1996. Montserrat Volcano Observatory and Geological Society, London, 29-30. VOIGHT, B. 2000. Stability of andesite volcanoes and lava domes. Philosophical Transactions of the Royal Society, London, A 358, 1663-1703. VOIGHT, B. & ELSWORTH. D. 1997. Failure of volcano slopes. Geotechnique, 47, 1-31. VOIGHT, B. & ELSWORTH, D. 2000. Stability and collapse of hazardous gaspressurized lava domes. Geophysical Research Letters, 48, 1-4. VOIGHT, B., GLICKEN, H., JANDA, R. J. & DOUGLASS, P. M. 1981. Catastrophic rockslide avalanche of May 18. In: LIPMAN, P. W. & MULLINEAUX, D. R. (eds) The 1980 Eruptions of Mount St Helens, Washington, US Geological Survey, Professional Papers. 1250, 347-378.
407
VOIGHT. B., JANDA, R. J.. GLICKEN. H. & DOUGLASS. P. M. 1983. Nature and mechanics of the Mount St Helens rockslide-avalanche of 18 May 1980. Geotechnique, 33. 243-273. VOIGHT. B., JANDA, R. J., GLICKEN. H. & DOUGLASS. P. M. 1985. Nature and mechanics of the Mount St Helens rockslide-avalanche of 18 May 1980: Reply to discussion. Geotechnique, 35, 357-368. VOIGHT, B., HOBLITT. R. P., CLARKE, A. B.. LOCKHART, A. B.. MILLER, A. D.. LYNCH, L. & McMAHON, J. 1998. Remarkable cyclic ground deformation monitored in real time on Montserrat and its use in eruption forecasting. Geophysical Research Letters, 25. 3405- 3408. VOIGHT, B., SPARKS, R. S. J., MILLER, A. D. ET AL. 1999. Magma flow instability and cyclic activity at Soufriere Hills Volcano. Montserrat, B.W.I. Science, 283, 1138-1142. WADGE, G. & ISAACS, M. C. 1988. Mapping the volcanic hazards from Soufriere Hills volcano, Montserrat. West Indies using an image processor. Journal of the Geological Society, London. 145, 541-551. WADGE, G., FRANCIS. P. W. & RAMIREZ. C. F. 1995. The Socompa collapse and avalanche event. Journal of Volcanology and Geothermal Research. 66,309-336. WADGE, G., WOODS, A., JACKSON, P., BOWER. S.. WILLIAMS, C. & HULSEMANN, F. 1998. A hazard evaluation system for Montserrat. In: Forecasts and Warnings. UK Coordination Committee for IDNDR. Thomas Telford, 3.1-3.32. WALKER, G. P. L. 1971. Grain-size characteristics of pyroclastic deposits. Journal of Geology, 79, 696-714. WOODS. A. W., SPARKS, R. S. J., RITCHIE, L. J., BATEY, J., GLADSTONE, C. & BURSIK. M. I. 2002. The explosive decompression of a pressurized volcanic dome: the 26 December 1997 collapse and explosion of Soufriere Hills Volcano, Montserrat. In: DRUITT. T. H. & KOKELAAR. B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society. London. Memoirs. 21. 457-466. YOUNG, S. R., SPARKS, R. S. J.. ASPINALL, W. P.. LYNCH. L. L.. MILLER. A. D., ROBERTSON, R. E. A. & SHEPHERD, J. B. 1998. Overview of the eruption of Soufriere Hills Volcano, Montserrat, July 18 1995. to December 1997. Geophysical Research Letters, 25. 3389-3393. YOUNG, S. R., VOIGHT. B., BARCLAY, J. ET AL. 2002. Hazard implications of small-scale edifice instability and sector collapse: a case history from Soufriere Hills Volcano, Montserrat. In: DRUITT. T. H. & KOKELAAR. B. P. (eds) The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London. Memoirs. 21. 349-361.
Generation of a debris avalanche and violent pyroclastic density current on 26 December (Boxing Day) 1997 at Soufriere Hills Volcano, Montserrat R. S. J. SPARKS1, J. BARCLAY 2 , E. S. CALDER 1 , R. A. HERD 3 , J.-C. KOMOROWSKI 4 . R. LUCKETT 5 , G. E. NORTON 3 , L. J. RITCHIE 6 , B. VOIGHT 7 & A. W. WOODS8 1 Department of Earth Sciences, University of Bristol, Bristol, BS8 1RJ, UK 2 Department of Environmental Sciences, University of East Anglia, Norwich, NR4 7JT, UK 3 British Geological Survey, Keyworth, Nottingham, NG12 5GG, UK 4 Observatoire de Guadeloupe, Institut de Physique du Globe, Guadeloupe, French Antilles 5 British Geological Survey, Murchison House, West Mains Road, Edinburgh, EH9 3LA, UK 6 Centre for Volcanic Studies, University of Luton, Park Square, Luton, LU1 3JU, UK 7 Department of Geosciences, Penn State University, University Park, PA 16802, USA 8 BP Institute, Madingley Rise, Cambridge University, Cambridge CB3 OEZ, UK
Abstract: Growth of an andesitic lava dome at Soufriere Hills Volcano, Montserrat, beginning in November 1995. caused instability of a hydrothermally altered flank of the volcano. Catastrophic failure occurred on 26 December 1997. 14 months after the instability was first recognized. Two months before failure a dome lobe had extruded over the unstable area and by 25 December 1997 this had a volume of 113 x 10 6 m 3 . At 03:01 (local time) the flank rocks and some dome talus failed and generated a debris avalanche (volume 46 x 106 m 3 ). Between 35 and 45 x 106 m3 of the dome then collapsed, generating a violent pyroclastic density current that devastated 10km 2 of southern Montserrat. The failure of the flank and dome formed two adjacent bowl-shaped collapse depressions. The most intense activity lasted about 11.6 minutes. The hummocky debris avalanche deposit is composed of a mixture of domains of heterolithic breccia. The pyroclastic density current had an estimated peak velocity of 80-90 m s - 1 , and minimum flux of 10 8 k g s - 1 . The current was largely erosional on land with most deposition out at sea. Destructive effects included removal of houses, trees and large vehicles, and formation of a scoured surface blackened by a thin (3-4 mm) layer of tar. Two discrete depositional units formed from the pyroclastic density current, each with a lower coarse-grained layer and an upper fine-grained stratified layer. These deposits are overlain by an ashfall layer related to buoyant lofting of the current. Flank failure is attributed to loading of hydrothermally weakened rocks by the dome. The generation of the pyroclastic density current is attributed to failure and explosive disintegration of the dome, involving release and violent expansion of gases initially at high pore pressures.
In October 1996 the upper southern flank of Soufriere Hills Volcano, Montserrat, started to show signs of instability, by development of open fractures and rock avalanches associated with intense earthquake swarms. This instability raised concerns about the possibility of a sector collapse and an associated violent lateral blast. The anticipated phenomena took place 14 months later at 03:01 local time (LT) on 26 December 1997. This date is a holiday in the UK known as Boxing Day. Failure of the southern flanks of the volcano was followed by collapse of about 50% of the andesitic lava dome. As a consequence a debris avalanche formed and 10 km2 of southern Montserrat was devastated by a highly energetic pyroclastic density current (PDC). This paper documents these events, their effects and their products. There was no loss of life, as the area had been evacuated when the instability was first recognized in October 1996. The flank failure of 26 December 1997 contrasts with the larger flank failure of Mount St Helens in 1980, where sliding of the northern flanks of the volcano unroofed a pressurized cryptodome that immediately exploded and generated a laterally directed high-energy PDC or volcanic blast (Christiansen & Peterson 1981). In the case of Soufriere Hills Volcano, a large pressurized andesitic lava dome had built above a hydrothermally altered and unstable flank of the volcano. Failure of the flank rocks undermined the dome, which partially collapsed, disintegrated and generated a high-energy PDC. This paper complements other contributions describing the instability of the southern walls of the volcano and the debris avalanche (Voight et al. 2002), the sedimentology of the PDC deposits (Ritchie et al. 2002), modelling of the blast dynamics (Woods et al. 2002), remote sensing data (Mayberry et al. 2002) and hazards aspects (Young et al. 2002). Precursory activity Long-term evolution The events on 26 December 1996 were the culmination of long-term destabilization of the upper flank of the volcano and dome growth,
which can be traced back to October 1996. This section places the eventual failure of the southern flank in the context of the overall evolution of the eruption. The eruption of Soufriere Hills Volcano involved the growth of an andesitic lava dome within English's Crater (Fig. 1: Robertson et al. 2000). English's Crater was partly infilled by a young lava dome known as Castle Peak, which was extruded about 350 years BP (Young et al. 1998). A semi-circular depression or moat existed between the Castle Peak dome and the walls of English's Crater (Fig. la). Lava extrusion began in mid-November 1995 and throughout 1996 lava dome growth was confined to within English's Crater, with pyroclastic flows generated by dome collapse being discharged to the east down the Tar River valley. The area of eventual failure occurred between Galway's Wall and Galway's Soufriere. Galway's Wall constitutes the southern margin of English's Crater and extends about 600m from Chances Peak to Galway's Mountain (Fig. 1b). The outer, southward-facing side of Galway's Wall was precipitous and composed of pyroclastic breccias and tuffs related to the formation of the Chances Peak and Galway's domes (Voight et al. 2002). Galway's Soufriere was an area of active fumaroles beneath Galway's Wall, at an altitude of 400-500m above the White River valley (Fig. 1b). Faulting and fracturing of the Galway's Soufriere area had occurred prior to and in the early stages of the eruption. The exact date of the onset of this deformation is not known, but substantial changes in the area were first noted in October 1995. One prominent fault, trending NE. cut and down-faulted the road to Galway's Soufriere by about a metre to the SE. The inner side of Galway's Wall was not affected by the dome until June 1996 when active dome growth was focused in the SW side of English's Crater. Talus rapidly infilled the moat between Castle Peak and Galway's Wall. Activity from July to mid-September 1996 focused in the northern and eastern areas of English's Crater. A new period of dome growth, which began on 1 October 1996 following the explosive activity of 17-18 September 1996. had almost refilled the crater by the end of October. In late October 1996, small
DRUITT, T. H. & KOKELAAR, B. P. (eds) 2002. The Eruption of Soufriere Hills Volcano, Montserrat, from 1995 to 1999. Geological Society, London, Memoirs, 21, 409-434. 0435-4052/02/S15 r The Geological Society of London 2002.
409
410
R. S. J. SPARKS ET AL.
Fig. 1. (a) Topographic map of southern Montserrat showing main locations and features described in the text. The location of the seismic stations at Windy Hill (MBWH) and St Patrick's (MSPT) are also shown. Contours in hundreds of metres. (b) Detailed topography of English's Crater and upper flanks of Soufriere Hills Volcano with names of geographic features mentioned in the text.
avalanches on Galway's Wall first indicated instability. Fractures opened on Galway's Wall in late November. Dome growth slowed markedly in November 1996 (Sparks et al. 1998) to 0.5-1 m3 s - 1 . A remarkable pattern of alternating hybrid earthquake swarms and sluggish aseismic dome growth characterized activity, culminating in an eight-day intense hybrid swarm at the beginning of December 1996. During this period, there was pronounced instability of Galway's Wall. Fractures opened on the Chances Peak end of the wall and in the south-facing outer wall. Periods of avalanching occurred frequently across the entire wall, mostly concurrent with swarms of hybrid earthquakes. Direct observations showed that avalanches occurred simultaneously in several places during an intense hybrid earthquake. Avalanches from Galway's Wall were uncommon during aseismic periods, but increased rockfalls from the dome indicated that extrusion had accelerated then. Surveys in early December 1996 showed that part of the dome erupted in June 1996 had risen by at least 30m along an east-west zone adjacent to Galway's Wall. These observations suggested either a period of endogenous dome growth or shallow intrusion adjacent to Galway's Wall. These periods of deformation of Galway's Wall alternated with aseismic periods and dome
extrusion (Voight et al 1999). There was considerable concern about the possibility of wall failure, sector collapse and a lateral volcanic blast (Young et al. 2002). During the period December 1996 to late March 1997 the area of active dome growth moved to the north and east. Avalanche activity at Galway's Wall simultaneously diminished. Fractures developed on the Galway's Mountain end of Galway's Wall. In late March 1997, active growth shifted back to the south and a new lobe of lava started to extrude towards Galway's Wall. At this stage, the lava overlapped the lowest point on Galway's Wall and domecollapse pyroclastic flows spilled over the wall into the White River valley. Numerous pyroclastic flows eroded a gully a few tens of metres deep into Galway's Wall (Fig. 2a) and began burying Galway's Soufriere. From May 1997 to the end of October 1997 dome activity shifted to the north and east, away from Galway's Wall. Two series of repetitive Vulcanian explosions occurred in the August to October period (Druitt et al. 2002b). There were minor changes in the Galway's area, with occasional landslips. At the beginning of November 1997 dome growth shifted to the south. Two major dome collapses occurred on 4 and 6 November 1997. The associated pyroclastic flows
GENERATION OF DEBRIS AVALANCHE AND PYROCLASTIC DENSITY C U R R E N T
further infilled the White River valley, and formed a delta extending 400 m from the former coastline. The collapses in early November 1997 together produced about 8 x 10 6 m 3 of pyroclastic deposits.
411
This minor change cannot be linked to any inflation or deflation. A vigorous gas plume was observed above the dome on clear days in the weeks preceding the eruption, but no correlation spectroscopy measurements were being made in this period.
Short-term precursors Dome growth during November and December 1997 was rapid, at 7-8 m3 s-1 (Sparks et al. 1998), and was focused in the southern area adjacent to Galway's Wall. A lobe extruded laterally over Galway's Wall (Fig. 2b). There were numerous rockfalls, which formed a large apron of dome talus between the base of the lobe and Galway's Soufriere area. By 21 December 1997 (Fig. 2c) the dome summit had reached 1030m altitude and Galway's Wall and Galway's Soufriere were completely buried. The massive and blocky upper parts of the dome extended over and slightly beyond the position of the crest of Galway's Wall (Fig. 2c). Surveys (Fig. 3) indicated that 32 x 106 m3 of dome material extended beyond Galway's Wall and that the dome had reached a volume of 110 x 10 6 m 3 by 21 December 1997. The dome volume is estimated to have been 113 x 10 6 m 3 by 25 December. Volumes in this paper are given as bulk volumes uncorrected to dense rock values. During late November and December 1997 seismicity was low: rockfalls averaged about 70 per day, a few hybrid earthquakes occurred per day, and there were occasional long-period earthquakes (Fig. 4). Long-period earthquakes were typically short bursts of harmonic tremor. These lasted up to a few tens of seconds and were nearly monochromatic at approximately 2 Hz. Cyclicity in the seismicity, which had been pronounced earlier in the eruption (Voight et al. 1999), was barely discernible. There was, however, an expectation at the Montserrat Volcano Observatory (MVO) that activity would become more vigorous towards the end of December. A pattern of enhanced activity every six to seven weeks had been recognized since May 1997, from abrupt changes in tilt patterns, clustering of swarms of hybrid earthquakes, and occurrence of major collapses (Voight et al. 1998, 1999). The last such episode of enhanced activity had occurred on 4-6 November, so a major collapse towards the end of December was anticipated. The build-up to the flank failure and dome collapse of 26 December 1997 was rapid. The real-time seismic amplitude measurement (RSAM) chart and earthquake count data (Fig. 5) show seismic activity only marginally above background with a slight increase that can be traced back to 22 December. At 14:30 LT on 24 December a hybrid swarm began with events approximately every 20 minutes. The frequency of events slowly increased until approximately 20:00 LT on 25 December. The individual events also generally increased in amplitude as the swarm progressed (Fig. 5), but even the largest earthquakes were relatively small in amplitude, an order of magnitude smaller than those recorded in early November 1997. At 20:00 LT hybrid earthquakes were occurring too frequently to trigger the network and from this time onwards the signal was effectively continuous tremor. The amplitude of this tremor peaked at 23:00 LT on 25 December and declined until 00:00 LT when individual events could again be discerned. These individual events merged back into tremor at 01:30 LT on 26 December, which then built in amplitude up to the onset of the signal corresponding to the main event at 03:01 LT. Other monitoring data, gathered in the weeks between the end of explosive activity on 21 October and 26 December, do not reveal any marked changes. Deformation detected by electronic distance measurement and global positioning system did not show any deviation from established trends in this period, although the sampling frequency was insufficient to pick up short-term trends (days). About 40 mm of northward movement of a station on the western flanks of the volcano was recorded between mid-November and mid-January, but it is not known how this movement was partitioned between the periods before and after the 26 December activity. An electronic tiltmeter at the village of Long Ground showed irregular fluctuations during the precursor hybrid swarm on 25 December. There was a permanent positive offset on the x-axis of c. 3 rad, but no offset on the y-axis after 26 December.
Chronology Seismicity Since the flank failure of 26 December 1997 occurred during darkness, seismicity provides the main constraint on the timing of the activity. Seismic data reported here are taken mainly from the Windy Hill (MBWH) 1 Hz single-component station, which is part of the digitally telemetered "broadband' network (Fig. la). The seismic signal from the failure and ensuing events is shown in Fig. 6. Additional information is taken from the 'short-period', analogue
Fig. 2. (a) The lava dome in April 1997 viewed from the SW towards Galway's Wall. The new lobe can be seen in the centre of the photograph (L) and there is a deep erosional gully (G) incised into Galway's Wall by pyroclastic flows related to dome collapse in early April, (b) The 4 November lobe of the dome actively growing over Galway's Wall on 8 November 1997. with the shoulder of Galway's Mountain to the far right. The new spine-like lobe is growing in the collapse scar excavated in the dome during the major collapses of 4 and 6 November 1997. (c) The dome on 21 December 1997 viewed from the SW. The summit of the dome is at an altitude of about 1030m with the top of Chances Peak at 914m on the left. partly obscured by condensed steam drifting from the dome. Galway's Wall has been covered by the new dome and its talus (compare with (a)), and Galway's Soufriere area, in the foreground, has been buried by dome talus.
412
R. S. J. SPARKS ET AL.
Fig. 2. (continued).
network. The St Patrick's (MSPT) station (Fig. la) on this network stopped transmitting at 03:03.3 LT when destroyed by the pyroclastic density current. The main seismic signal was emergent and pulsatory and can be divided into six pulses (Fig. 6 and Table 1). Onset of the signal was gradual so that a starting time cannot be defined precisely. We have taken the start time of the initial collapse as 03:01.0 LT and the finish time as 03:16.2 LT. This main period was followed by less intense high-amplitude signals until the seismicity dropped to background levels at 03:32.1 LT. During this latter period, intervals of monochromatic tremor were recorded at 1.9 Hz, including three higher amplitude pulses. This gives a total duration of the main activity of 15.2 minutes, although the period of strongly pulsed high amplitude seismicity is only 11.6 minutes. The whole seismic anomaly lasted 31.1 minutes. The frequency of the main seismic signal was similar to that of the hybrids and tremor prior to the collapse, with dominant frequency below 2.8 Hz. The dominance of lower frequency energy was associated with the final two pulses of the main activity prior to a period of near-monochromatic tremor at 1.9 Hz. Monochromatic
tremor, best developed between 03:21 LT and 03:25 LT, was punctuated by three high amplitude signals of the same dominant frequency (Fig. 6). Each successive signal, occurring at 03:18.3 LT, 03:23.8 LT and 03:26.7 LT, was smaller than the previous one. The last high amplitude signal was followed by a rockfall signal. The tremor signal and the high amplitude peaks were similar to the signals generated by ash-venting and Vulcanian explosions respectively in the August to October 1997 period. They are therefore interpreted as post-collapse ash-venting and Vulcanian explosions.
Witness information People in several places on Montserrat heard roaring. The timing of these noises is consistent with the roaring being associated with ashventing, as interpreted from the seismic signals. Police at the village of Salem reported hearing two 'explosions' and then seeing flashes of light between 03:15 LT and 03:25 LT. The exact timing of these observations is unknown, and there were also separate reports of thunder and lightning. Darkness and low cloud cover limited direct
GENERATION OF DEBRIS AVALANCHE AND PYROCLASTIC DENSITY CURRENT
413
Fig. 3. Contour map of the dome (shaded area) on 17 December 1997 from a dome survey. The Montserrat grid is shown for reference and contours are in metres. The thick dashed line between Chances Peak and Galway's Mountain marks the position of Galway's Wall. observations. Buildings in the Trials area were burning at about 03:45 LT. At this time, an ash plume was seen from Garibaldi Hill, while a low-level ash plume drifted SW. A small amount of ash fell at Garibaldi Hill but none further north.
Satellite data The eruption plume was estimated to rise to 14.9km altitude, from GOES-8 satellite observations (Mayberry et al. 1998, 2002). The height estimates are constrained by comparing wind directions from Radiosonde data above Puerto Rico and Guadeloupe with
the plume dispersal direction, and by estimates of cloud-top temperature. Piarco FIC (Trinidad) reported ash at 12.2km at 03:55 LT and a British West Indies Airline pilot reported seeing ash at 11 km at 06:05 LT. The ash plume was rapidly dispersed to the SSE and SE at 20-28 km hr-1 (Mayberry et al. 1998, 2002), and passed over the central part of the eastern Caribbean and then out into the Atlantic. A lower level ( -5.0). moderately sorted ( =1.9) layer 1 breccia. It is up to 2.7m thick with a drape, up to 30cm thick, of layer 2, which thins to 10cm at the southern end of the structure. Clasts within layer 1 are up to 1 m in diameter, but are typically 60cm or less. They are of fresh, grey juvenile dome rock, together with abundant (>50% volume) hydrothermally altered rock. Imbrication of elongate large clasts and wood fragments is well developed within the mound, and long axes of imbricated clasts are oriented approximately 200 as arrow on Fig. 3). In Region 3, west of the debris avalanche overspill lobe at Morris' (locality 48, Fig. 3), is an isolated elongate mound 40m long, 7m wide and 3m high, with its long axis oriented at 222 (Fig. 15a). It thins at both ends (NE and SW). with the downcurrent end thinning as a tapered wedge. The sides slope at 30 to the SE
Fig. 13. (a) Deposit accumulated downcurrent of a boulder in Region 3 locality 46 (Fig. 3). Unit I, layer 1 rests upon pre-eruption ashfall deposits associated with previous block-and-ash flows that travelled down the White River valley in 1997. Note the scattering of Vulcanian fallout clasts on the surface of the deposit, (b) Accumulation of Unit I. layer 1 upcurrent of a boulder in Region 3 near locality 48 (Fig. 3). (c) Accumulation of Unit II. layer 1 up- and downcurrent of a boulder in Region 3 near locality 48 (Fig. 3). (d) Accumulation of Unit I, layer 2 in the lee of a demolished building, locality 47. Morris' (Fig. 3) (person for scale).
SEDIMENTOLOGY OF DEPOSITS FROM THE PYROCLASTIC DENSITY CURRENT
Fig. 13. (continued)
447
448
L. J. R I T C H I E ET AL.
Fig. 14. Relationships of Unit I, layers 1 and 2 and Unit III in the lee of boulders. Layer 1 is preserved directly behind the boulder or within a metre downcurrent. Sections perpendicular to the long axes of the features show that erosion may have occurred obliquely to the long axis of the structure.
and 18 to the NW. It is composed of coarse layer 1 breccia with subangular to subrounded clasts of grey, juvenile andesite and hydrothermally altered dome rock up to 60cm, but predominantly 10-30 cm. The whole mound is draped by up to 25cm of layer 2 deposit. The surrounding area is devoid of the coarse layer 1 material, and layer 2 occurs on the pre-eruption surface. Unit I, layer 1 breccia accumulations are typical in the lee of obstacles, such as severed tree trunks and building debris (Fig. 13d). The dimensions of the accumulation vary according to the size of the obstacle, but are generally 2-3 m long, 1-1.7m wide and 5070 cm high (Fig. 13). Layer 1 is typically 0-25 cm thick, composed of poorly sorted ( = 2.5), coarse ash (Md = 2.0). Layer 2 is 0-4 cm thick, well sorted ( = 1.7), fine ash (Mdo = 3.0). Two different occurrences in the stratigraphy of the deposits downcurrent of the boulder are identified. First, Unit I, layer 1 occurs downcurrent of
the boulder with Unit I. layer 2 directly above (Fig. 14a). Second. Unit I. layer 2 occurs directly downcurrent of the boulder and rests upon the pre-eruption substrate. An erosional contact with Unit I. layer 1 occurs about 30cm downcurrent of the boulder (Fig. 14b). The edge of the feature that is facing the volcano is typically truncated, exposing both layers 1 and 2. and Unit III is absent from this side of the mound. Clasts up to 11 cm in size protrude from the surface of the truncated edge.
Granulometry Grain-size distributions were determined for 270 samples collected from 76 localities (Fig. 3). Samples were dry sieved at 1 intervals
Fig. 15. (a) Symmetrically streamlined breccia of Unit I. layer 1 in Region 3. Morris' area, locality 48 (Fig. 3). Flow direction is from top right to bottom left. Note the absence of coarse blocks in the vicinity of the feature. Foreground consists of layer 2 deposits resting directly on eroded surface (person for scale), (b) Northern end of the large streamlined breccia mound in Region 5 on the south side of the White River valley, locality 69 (Fig. 3) (helicopter on far right for scale).
SEDIMENTOLOGY OF DEPOSITS FROM THE PYROCLASTIC DENSITY CURRENT
449
Fig. 15. (continued)
to 63 m (4 ). Those samples with a fine ash (15% were further analysed with a laser particle sizer to 12 . The grain sizes of coarse breccias (coarser than —5 ) were determined from photographs. Grain-size statistical parameters of Inman (1952) are used. Median diameters (Md ) of deposits range from -5 to 5 . Md versus sorting coefficients ( ) show good distinction between layers 1 and 2 of Unit I (Fig. 17). At any one locality layer 1 is always coarser than layer 2, but samples of layer 2 in axial areas can be coarser than layer 1 samples from peripheral areas, resulting in the overlap in Figure 17a and b. Unit III is finer grained than the other layers. Grain-size data from the Mount St Helens blast deposit of 18 May 1980 were chosen for comparison with 26 December 1997 samples. They are plotted with Walker's (1983) fields for pyroclastic flows and surges in Figure 18. The 26 December 1997 samples are similar to the Mount St Helens Md and data, plotting within Walker's pyroclastic surge and flow fields.
Fig. 16. Strongly reverse-graded fine to coarse breccia deposited in Region 6 on the coastal fan edge.
Mean values of the grain-size parameters Md and were calculated for layers 1 and 2 for each region (Fig. 19a). Layers 1 and 2 become finer grained, and the contrast in Md between layers 1 and 2 decreases, from axial to peripheral regions. Tie-lines connecting layers 1 and 2 for each region (Fig. 19b) show that the difference in grain size (Md and ) between layers 1 and 2 decreases between Regions 4 (axial) and 1 (peripheral). A transect 5km from the dome, traversing (from axial to peripheral) across the White River valley to Aymer's Ghaut in a NW direction, shows the variation in Md of Unit I layers 1 and 2 (Fig. 20a). This graph illustrates that within each region Md of layer 1 remains similar, and that the most pronounced change occurs across Gingoes Ghaut. Generally the grain size of layer 1 increases markedly towards the White River valley, while layer 2 shows little systematic change (Fig. 20a). Transects from vent to coast, parallel to Aymer's and Gingoes Ghauts for Regions 2 and 3 respectively, show little change in median diameter (except Region 1, which becomes
450
L. J. RITCHIE ET AL.
Fig. 17. (a) Median diameter (Md ) versus sorting coefficients ( ) for all samples with fields for Unit I, layers 1 and 2 and Unit III. (b) Median diameter (Md ) versus wt% of fine ash ( 106Pa
Fig. 4. Pressure as a function of radius in a permeable dome of radius 10 < r < 300 m. Curves are given for dome permeabilities of value K= 10 -12 ,10 -13 and 10 -14 m2.
A. W. WOODS ET AL.
460
throughout the dome, except in a narrow boundary layer, of width 5-10 m, near to the surface, where it decreases towards atmospheric pressure. As a result of this pressure distribution, if there is a collapse or landslide, regions of very high pressure just below the outer part of the dome may be exposed to much lower pressures. This may trigger a fragmentation wave as these large local pressure gradients can disrupt and fragment the dome rock. We will explore this process in more detail later in the paper. Note also that the pressure near the centre of the dome increases rapidly as the radius decreases; the precise value depends on the radius at the top of the conduit, but typically has a value 5-10 MPa. We now examine the gas mass and associated energy stored within the model dome in order to develop some insight into the scale of the phenomenon. Gas and energy available for dome fragmentation We now use the above model of the gas pressurization in the dome to determine its energy and gas content, and hence determine the potential for explosive fragmentation and disintegration following a trigger such as a landslide or collapse. The gas mass in the dome, G is given by the integral: G=
pr2
dr
(5)
Here, we take the porosity of the dome to be = 0.1, based on measurements presented in Robertson et al. (1998) and Melnik & Sparks (2002). The density p may be expressed in terms of the pressure, using the radial profiles calculated above (Fig. 4). For the three examples shown in Figure 4, the total gas content of the dome is estimated to be 1.7 x 107, 5.2 x 107 and 1.6 x 10 8 kg respectively, for the three different cases in which the permeability has values 10 - 1 2 , 10 - 1 3 and 10 - 1 4 m 2 . In each of these cases, the mass of solid magma in the dome is 1.3 x 10 11 kg, and so the net mass fraction of gas in the dome ranges from about 1.3 x 10 - 4 to 1.2 x 10 -3 . This is a key result in terms of the subsequent dynamics of the fragmented mixture, as we shall establish later in the paper. Such pressurized domes have one or two orders of magnitude less exsolved volatile per unit mass of magma than is typically present in a steady Plinian explosive eruption, and this can lead to a substantially greater the density of the pyroclastic gravity currents produced. The energy E that is available to fragment the dome materials and subsequently to disperse the rock fragments following disintegration of the dome, may be expressed in terms of the thermal energy of the gas together with that fraction of the solid dome material which remains in good thermal contact with the gas. For the fragmentation wave, the mass fraction of solid remaining in good thermal contact with the gas is small owing to the short time available for solid-gas heat transfer during the passage of the fragmentation wave. However, the flow that subsequently develops following fragmentation evolves over times of order tens of seconds. A much more substantial mass of fine-particulate material, corresponding to particles smaller than about l00 m, can remain in thermal equilibrium with the gas during the flow (Woods & Bursik 1991). Since this quantity is poorly constrained by field observations, we take it to be a parameter in the model. The internal energy of the gas stored in the dome may be expressed as
E=
pc Tr2 dr
(6)
where pc T is the energy per unit volume. In this calculation the mass of solid material that remains in thermal equilibrium with the gas during the explosion is included through the definition of the bulk specific heat (Wilson et al. 1978; Woods 1988): c =fcp + cvw
(7)
where cp denotes specific heat of the particulate, cvw is the specific heat at constant volume of the water vapour, and is the mass of
solid per unit mass of gas. Equation 6 may be rewritten in the simpler form: E=
dr
(8)
J
where 7, the ratio of specific heats at constant volume and constant pressure, depends on the mass fraction of the solid, , which remains in thermal equilibrium with unit mass of the gas: 7 = (fcp + cpw)(fcp + cvw)
(9)
Here cpw is the specific heat at constant pressure of the water vapour. For simplicity, in these calculations, we assume that 7 is constant, consistent with our approximation that the water vapour behaves as a perfect gas (Equation 2): this simplification is justified because any differences in the relevant properties of water vapour within the dome have only a small effect compared to the uncertainty in the value of / (Equation 9). For the evolving decompressed flow, after passage of the fragmentation wave, values of may lie in the range 104-102, depending on the number of fine fragments in the flow. However, later in the paper, we show that during the passage of the fragmentation wave, a much smaller mass of solid is in contact with the decompressing gas. We therefore examine values of in the range 1 3 3 >3 3 3 >3 3 >3
n/a 3 9 2.5 6 n/a 3 6
20 Jan. 97
2.3
n/a
TR
>3
9
0.8 2.3 0.4 4.9 7.0 11.0 4.6 35-45 19.2 2.3 3.8
0.02 n/a 0.02 n/a n/a n/a n/a 4.5 n/a n/a n/a
WR WR TR MG FG TG, WG WR WR TR TR TR
3.6 4.1 3 6.7 >5.6 >6.0 >5 >5 >3 >5 >3
31 Mar. 971 1 1 Apr. 97 27 May 97 25 Jun. 97 3 Aug. 97 21 Sep. 97 06 Nov. 97 26 Dec. 97 03 Jul. 98 12 Nov. 98 20 Jul. 99
sw
Low SW High SE Low SW High NE NW n/a n/a W WNW NW na SW ENE NW Low W High NW
4 n/a n/a 11.8 7 n/a 10.8 15 14 8 10
* Pyroclastic-flow deposit volumes include surge deposits and were surveyed throughout the eruption by MVO staff. They were estimated using a combination of GPS surveys, maps (calculated areas and thickness estimates), visual observations and simple geometry of the collapse scar (Calder et al. 2002; see text for details on DRE-volume calculations). Tephra-fallout deposit volumes were calculated using the method of Pyle (1989) apart from that of 26 December 1997, which was obtained from analytical calculations (Sparks et al. 2002; see text for details on DRE-volume calculations). \ Collapse valleys are shown on Figure 1. § Values of pyroclastic-flow runout preceded by '>' indicate that the flow entered the sea. The runout for 31 March 1997 refers to flows shed down the White River valley in the morning of that day. || Plume maximum heights were estimated from direct observations, with the exception of those of 25 June, 6 November and 26 December 1997. which were calculated from satellite data. Dome collapses with DRE volume