Montserrat Volcano Observatory

Assessment of the Status of the
Soufriere Hills Volcano, Montserrat and its Hazards

18 December 1997

Executive Summary

  1. The eruption of the Soufriere Hills Volcano, Montserrat has continued for over two and a half years. A large volume of lava has been extruded, forming a sequence of summit domes. Activity has slowly increased in vigour and has been at its highest levels over the last six months. The scale of the eruption combined with the small size of the island and the desire to maintain a population presents unprecedented challenges for volcanological monitoring and risk assessment. Compared with other volcanic crises, there is reduced scope for including broad safety margins in delineating areas of different degrees of hazard.

  2. The principal hazards are pyroclastic flows, and fragments of volcanic rock ejected by explosions. Pyroclastic flows are avalanches of hot lava fragments and volcanic ash which can move at speeds of over 100 kph and are highly destructive and invariably fatal to people in their path. Explosions cause rock fragments to be dispersed around the volcano and can cause damage to property and injury to people.

  3. Fine respirable volcanic ash poses a potential long-term health hazard wherever large amounts of volcanic ash accumulates. The ash contains a component of toxic silica known as cristobalite. The geographical distribution of cristobalite-rich ash is not uniform because of prevailing wind directions and volcanic factors. The north of the island has very much lower exposure to fine ash than communities closer to the volcano.

  4. The most likely outcome of the present crisis is that the eruption will progress at levels of activity comparable to those observed so far. This scenario allows for some further modest increases in activity which, if not exceeded, would pose no additional significant safety problems for those living outside the current official exclusion zone north of the Nantes River. This assessment is based on improved understanding of the dynamics of the volcano through the monitoring programme; current understanding of the geological history of this volcano; and comparison with other volcanoes, particularly in the Caribbean. The eruption has not yet stepped beyond the boundaries indicated by the recent geological record of past eruptions of the Soufriere Hills volcano.

  5. The outlook is that the eruption will continue for at least several more months and more likely a few more years. Continuation for more than 15-20 years is considered unlikely. This suggests that scientific input to planning for the island might recognise two stages: one relating to the period while the volcano is active and the second relating to the period when activity is either clearly in decline or stops. It is not feasible to place specific time constraints on the active period and it could take up to a year after the cessation of manifest activity for scientists to feel assured that the eruption had finished. The volcano will then need to be monitored closely for the foreseeable future in order to detect signs of reawakening and to identify hazards that can occur after the eruption.

  6. Larger magnitude and more hazardous eruptive events are possible that would widen the areas affected. The area north of Nantes River through to Woodlands could be affected by such more powerful eruptions, although their probability is believed to be low.

  7. An important issue is the occurrence of eruptions that are sufficiently large that they could affect the north of Montserrat. While this possibility exists, the likelihood is very low and should be viewed in the context of other natural hazards to which the island is continuously exposed. For example, in terms of annual probabilities of occurrence, the chance of such a large magnitude eruptive event during the next six months is probably less than that of Montserrat being struck by a devastating major regional earthquake in the same period.


  1. The scientists at the Montserrat Volcano Observatory (MVO) have been directed by HMG to carry out an updated assessment of the Soufriere Hills volcano to provide the best possible information for the Governments of Montserrat (GoM) and the United Kingdom for decision-making and planning. Thus the assessment is required to include evaluations of the current scientific understanding of the volcano as it relates to the potential hazards and the threat that these hazards pose to the people of the island and its infrastructure. The assessment includes the outlook over the short-term (up to 6 months) and long-term (up to 15 years). This scientific information provides a major input into a risk analysis of the volcano. A companion report will deal with the risk analysis and includes consideration of monitoring, warning and mitigation measures.

  2. To meet these requirements a scientific meeting was held in Antigua from 2 to 5 December 1997. The meeting consisted of the Chief Scientists of the MVO and a number of other experienced scientists, most of whom have worked extensively on many other volcanoes and on Montserrat as senior scientific advisers. A list of attendees and brief description of their expertise are provided in Appendix A.

  3. The Terms of Reference (ToR) for the meeting are provided in Appendix B. The main body of the report is written to inform decision-making and aims to keep scientific technical terms to a minimum. A glossary is provided to help the non-technical reader in Appendix C. The methods used at the meeting are the traditional approach of scientific debate and on-the-spot peer review together with the application of a formal expert elicitation method that is commonly used in some industrial contexts related to hazards and risks. The elicitation method is described in Appendix D. Details of the scientific resources that underpin the report and technical descriptions of the science critical to hazards assessment are provided in Appendix E.

  4. The report is structured as follows. The eruption is briefly outlined and the main hazardous phenomena are described and explained. The current scientific assessment is then summarised in terms of the information that has been gathered for volcanic hazard assessment. Forecasts of future activity are provided, which include various scenarios, and these are followed by a hazards assessment. To facilitate discussion of the hazards, Figure 1 shows a map with the main place names, geographical features and population areas.

    Figure 1: Key features of Montserrat and population zones (900x635 GIF, 6.5K -- 1225x865 GIF, 10K

  5. An erupting volcano is a complex natural phenomenon. The scientists urge that politicians and officials who will use this document to inform their decisions should work as closely as possible with the scientists to avoid misunderstandings and oversimplifications. The experience of the MVO during this crisis is that decisions are always likely to be better when there is a close working relationship between scientists and civil authorities. In view of the sensitivity of the matters in this document it is important to be aware of the problems from casual or intentional selective quotation, particularly by the media.

  6. There have been several previous assessments of the status of the volcano and hazards by the MVO during the crisis. The last major assessment was provided for HMG and GoM on 14 August 1997 with a supplementary document on the safety of the north of the island on 3 September 1997. This document supersedes all previous assessments.

The Eruption So Far

  1. The eruption began on 18 July 1995 within English's Crater, which is a structure about 1 km in diameter with walls 100 to 150 m high, open to the east. The first four months of the eruption involved intense earthquake swarms and vigorous steam explosions, caused by rapid heating of the groundwater by rising magma. The magma reached the surface by mid-November 1995 and a new lava dome began to form. The lava is typical of many Caribbean volcanoes and is known as andesite. Such lava is so viscous that it piles up around the vent to form a dome; a steep-sided rubbly mound hundreds of metres high. The Soufriere Hills dome has been growing ever since.

  2. As a lava dome grows it becomes unstable and parts of the dome can suddenly avalanche away and simultaneously disintegrate to form a flow of fragments and volcanic ash known as a pyroclastic flow. The flows vary from small avalanches down the sides of the dome to major failures of the dome in which millions of tonnes of fragmented lava move at devastating speeds of over 100 kph and temperatures up to 800°C.

  3. In April 1996 the first major pyroclastic flows moved down the Tar River valley to the east of the volcano. By May 1996 pyroclastic flows entered the sea on the east coast and there were further large flows in July, August and early September. A major shift in the volcano's behaviour occurred around 20 July 1996 which heralded an escalation of activity in the following months. The first explosive eruption of the volcano occurred on 17 September 1996, generating an eruption column about 14 km high and ejecting 1 metre diameter rocks to about 2 km from the volcano. The escalation of activity and new explosive behaviour indicated more rapid flow of gas-rich magma to the surface. The explosive eruption was triggered by about 30% of the dome avalanching away in the previous 12 hours which decompressed gas-rich magma deeper in the volcano. This explosion is used as a reference event later in this assessment.

  4. Dome growth recommenced two weeks after the explosive eruption on 17 September 1996. Both the rate of growth and size of the dome increased over the next several months, interrupted repeatedly by many episodes of pyroclastic flow generation. Eventually the dome became so large that it filled up English's Crater. The walls of the crater had protected the southwestern, western and northern flanks of the volcano from pyroclastic flows, but by March 1997 the southwestern wall was overwhelmed and from June 1997 onwards the northern wall was overtopped. The major pyroclastic flow eruption of 25 June 1997 killed at least 19 people and nearly reached the airport 5.5 km northeast of the volcano. About 8 million cubic metres of the dome avalanched in less than 20 minutes. In late July 1997 large pyroclastic flows went down valleys on the west, resulting in the partial destruction of Plymouth.

  5. Following major dome collapses in early August 1997 explosive eruptions occurred at fairly regular (12 hourly) intervals over an 8 day period. These eruptions introduced an additional kind of hazard: pyroclastic flows formed by explosions rather than by avalanching of the unstable dome. Although these pyroclastic flows are similar in behaviour and consequent hazard to the dome collapse type, they are less constrained by topography, as the explosion can eject the materials in all directions around the volcano. By this time the scientists at MVO had recognised regular patterns of pressure build-up which allowed quite accurate prediction of when explosive events or pyroclastic flows would occur. This enabled emergency operations to take place in potentially dangerous areas in the periods when the internal pressure of the volcano was estimated to be low.

  6. The largest pyroclastic flow so far occurred on 21 September 1997 and destroyed the airport terminal. A prolonged period of quite regularly spaced explosive eruptions followed. Between 22 September and 21 October 1997 there were 75 explosions spaced at a mean interval of 9.5 hours. The explosions produced eruption columns of 5 to 12 km height and the largest events were only slightly less energetic than the 17 September 1996 explosion. Since then dome growth has continued and further dome collapses have generated more pyroclastic flows.

  7. The general trend of the eruption so far has been slow escalation (Figure 2). The mean flux of magma in the first 6 months was less than 1 m3/s; it rose to 2.3 m3/s in 1996 and 5 to 8 m3/s in the past 6 months. Superimposed on this trend are many pulsations. The dome growth rate and activity can be well below average for days to many weeks and then increase quite rapidly to well above average. The activity is punctuated by episodes of major dome collapse and pyroclastic flow generation. Each of the three major periods of explosive activity have occurred after one of the major dome collapses. Measurements of the flux of sulphur dioxide gases from the volcano also show a slow baseline increase with time.

    Figure 2: Volume of rock erupted from the Soufriere Hills volcano. (670x1000 GIF, 23K -- 980x1460 GIF, 38K

  8. Measurements of the deformation of the ground provide an important monitoring method. The movements around the volcano have mirrored the general changes in activity. The changes over the last 6 months have been the largest since the dome first appeared, but recent measurements have been hampered by destruction of some equipment and the increasing difficulties in scientists getting close to the volcano safely.

Volcanic Hazards on Montserrat

  1. Pyroclastic flows are the most serious hazardous phenomena so far encountered. Pyroclastic flows have been formed both by avalanches from the dome and explosions. The flows consist of three parts. First the main avalanche of hot boulders and ash is confined to valleys and totally destroys most structures and kills people. Second the flows are forced to mix with air as they move and thus form an expanded turbulent hot cloud of ash and air above the main avalanche. These clouds are usually lethal to most people caught in them (typical mortality rates of 90%). They can move at hurricane speeds (about 100 kph), have measured temperatures of 150 to 250°C and badly damage most buildings. People that survive suffer severe burns. The clouds are up to 200 m high and so can spill out of valleys and more energetic clouds can move uphill. Third a pyroclastic flow forms buoyant plumes of very fine ash and hot air that rise above the flows to heights typically between 1 and 6 km. The ash plumes drift in the wind to deposit fine layers of ash.

  2. Tephra fall results when explosions and pyroclastic flows form high buoyant plumes of fragmented material in the atmosphere and these materials fall onto the surrounding countryside. The fragments are known by volcanologists as tephra and they range in size from large blocks of rock to the finest volcanic ash. When a big explosive eruption occurs the eruption column can rise high into the atmosphere. The 17 September 1996 eruption, for example, reached about 14 km. Fragments include rocks up to 1 m diameter, lumps of pumice and volcanic ash. Substantial accumulations of tephra can cause roofs to collapse and impact by large pumice fragments of about 10 cm in diameter or more can cause injuries and even fatalities. The ash high in the atmosphere poses a hazard to regional aviation and a passenger airliner reportedly had one engine cut out when it entered the ash plume from the 17 September 1996 eruption. Explosive eruptions are measured by both their intensity (the mass flux of material and a measure of the power) and their magnitude (the total mass of material erupted). These two parameters, together with the wind conditions, determine where and how much tephra falls.

  3. Fine volcanic ash poses a potential long-term threat to health. The pyroclastic flows generate unusually large amounts of fine respirable ash (10 to 15% of the ash). The respirable ash contains 10 to 25% cristobalite, a form of silica which is thought to be about twice as toxic as the more common form of silica (quartz). The levels of respirable ash in the atmosphere have sometimes exceeded the maximum safe levels of particulates in UK industrial settings in communities close to the volcano where there are large amounts of ash. Air quality tends to be poorer where human activity disturbs the ash (e.g. vehicles and children playing). In contrast to the ash from pyroclastic flows, the ash from the explosions contains lower levels of respirable ash (2 to 5%) with lower cristobalite content (3 to 6%). The cristobalite forms from the hot gases in the dome and is selectively concentrated by fragmentation in the pyroclastic flows.

  4. Pyroclastic surges are a flowing mixture of turbulent hot gas with suspended volcanic ash. They form by particularly intense volcanic explosions that cause the hot mixture to move at speeds of up to 400 kph. Large pyroclastic surges are relatively rare and their causes are unfortunately rather poorly understood. They are exceptionally destructive and far less constrained by topography than normal kinds of pyroclastic flow. Large magnitude surge eruptions that have occurred on other volcanoes (e.g. Mont Pelee on Martinique, 1902 and Mount Lamington in Papua New Guinea, 1951) can cover large areas and the more extreme events of this kind are known to have surmounted topographic barriers as high as the Centre Hills (Figure 1). The deposits from these events are thin and are easily eroded, making it hard to prove that such events have never happened at the Soufriere Hills volcano.

  5. Lahars (or volcanic mudflows) are flows of volcanic debris and water that occur after heavy rain as run-off mingles with and erodes loose material accumulated around a volcano Several lahars have occurred already on Montserrat. They can be very destructive and threaten life in the valleys around a volcano. The main lahar threat at the moment is confined to the south, within the current total exclusion zone. If heavy tephra fall occurs in other parts of the island then lahars could also pose problems elsewhere on the island. When the eruption has finished this hazard may well remain for a considerable time.

  6. Large volcanic landslides occur when a sizeable part of a volcanic edifice becomes unstable and slides catastrophically. Such a landslide is believed to have occurred 4,000 years ago on Montserrat to form English's Crater and has occurred on other Caribbean islands. The landslides can be triggered by major regional earthquakes or by weakening of the volcano by hot fluids. A large volcanic landslide can cause tidal waves if it reaches the sea and may trigger an explosion or renewed volcanic activity. A related problem on Montserrat is that a large volume of pyroclastic flow material (at least 15 million cubic metres) has accumulated on the east and southwest shores of the island. Slumping of this coastal material is another potential cause of tidal waves.

  7. Other hazards include lightning and volcanic gases. Explosive eruptions can generate lightning which has been a significant cause of death on other volcanoes. So far lightning displays have been modest from Soufriere Hills eruptions. Volcanic gases do not pose a serious threat from the Soufriere Hills volcano. The gases contain very little carbon dioxide. Apart from water, the other main gases are sulphur dioxide, hydrochloric acid and minor amounts of hydrofluoric acid. Weekly air monitoring indicates that exposure to these acid gases in the currently populated areas has been negligible and there has been no risk to health. Fluoride adsorbed to the surface of ash particles has been monitored and has not so far posed a risk to the health of grazing animals. Water supplies are from spring sources around the Centre Hills and monitoring has confirmed that they have not been adversely affected.

Gathering of Information for Volcanic Hazard Assessment

  1. There are a number of ways in which information for making volcanic hazard assessments is gathered. In the case of the Soufriere Hills volcano, three main methods have been used. These methods are outlined below, along with a summary of the volcanological data pertinent to the current situation on Montserrat. More detailed technical matters are contained in Appendix E.

History of Soufriere Hills Volcano and Other caribbean Volcanoes

  1. Looking at the history of any particular volcano by inspection of volcanic deposits (which are related to processes at the volcano) and by review of historic accounts provides some qualitative constraints for hazard assessment. No volcanic eruptions have been recorded since the island was first colonised in 1632. However, a dome erupted about 400 years ago in English's Crater and there have been significant earthquake swarms and associated fumarolic activity in 1897-98, 1933-37 and 1966-67. Knowledge of the geological history of the volcano is inevitably incomplete and patchy. The oldest known rocks of the volcano are about 100,000 years. There is some evidence for large magnitude explosive eruptions near the beginning of its history. Five previous major episodes of dome growth and collapse are known in its more recent geological history over about the last 30,000 years. These lava domes and their associated deposits are similar in character to the products of the current eruption. Pumice-rich flow deposits and minor airborne tephra layers close to the volcano also indicate that the dome eruptions have involved explosive activity comparable in size to the explosions observed in this eruption.

  2. Domes and deposits from past eruptions are somewhat more voluminous than the current dome and deposits. Volumes of past dome eruptions are consistent with events lasting several years for current eruption rates at the Soufriere Hills volcano. Since about 30,000 years the geological evidence suggests predominantly dome forming eruptions with no evidence for large magnitude explosive eruptions. Deposits from surge eruptions are not well preserved in the tropical environment and so occurrence of such violent events cannot be precluded. However airborne tephra deposits from eruptions with magnitudes about ten times or more greater than the largest explosive event in the current eruption are well preserved on other Caribbean volcanic islands, such as Mont Pelee, Martinique and La Soufriere, St Vincent. It is inferred that the deposits would likely be preserved had such significantly larger eruptions happened on Montserrat over the last 30,000 years or so. Deep-sea cores from the areas around Montserrat support the on-island evidence that there have been no large magnitude explosive eruptions during the last 30,000 years.


  1. 25 Monitoring undertaken by MVO at the Soufriere Hills volcano and key supporting research combine to provide a good working knowledge of many aspects of the current eruption which can be used to guide the hazard assessment. There have been significant improvements in understanding the eruption and assessing the hazards, particularly over the last few months.

  2. The source of the andesite magma feeding the current eruption is located approximately 5-6 km beneath the crater of the Soufriere Hills volcano, as constrained by the mineralogy of the lava, the depth of some earthquakes and ground deformation. The magma comprises a mixture of 60 to 70% crystals and a silica-rich liquid at a temperature of around 830°C. The eruption is thought to have been prompted by the heating of the rocks in the source at about 5 km depth through addition of much hotter magma from deeper in the earth. Such reheating events appear to have prompted eruptions throughout the 30,000 year history of Soufriere Hills Volcano. It is likely that conditions of the eruption are similar to those during past eruptions.

  3. The magma contains 1.5 to 2.5% dissolved water (the main volcanic gas). Much of this gas is lost as the magma rises from over 5 km depth through the feeder conduit and into the dome. The high viscosity of the magma controls flow through the conduit. Loss of gas from the rising magma in the upper part of the conduit and within the dome gives rise to most of the seismic activity which has been recorded from the volcano over the past 2 years. Evidence from explosive eruptions in August to October 1997 suggests that conduit dimensions have not changed substantially since the explosion on 17 September 1996, and this provides limiting control on the size of likely future eruptions (see Appendix E.2).

  4. Deformation and seismic data collected from the volcano are consistent with a source of pressurisation in the upper conduit. Pressurisation and increase in viscosity of magma in the upper conduit provide conditions for the strongly cyclic behaviour, with typical cycle periods of several hours, witnessed on many occasions at the Soufriere Hills volcano. Formation of a resistant plug in the upper conduit through degassing leads to pressure build up beneath the plug. When the pressure reaches a threshold, the plug fails leading to a period of either rapid dome growth and pyroclastic flow generation or to an explosion.

  5. Although spectacular to witness, none of the explosions so far witnessed are large by global standards. The explosions are short-lived (a few minutes) and originate at very shallow depth, although the 17 September 1996 explosion tapped deeper levels (perhaps as much as 4 km) in the feeder conduit.

  6. The volume of erupted magma has been monitored throughout the eruption (Figure 2). Long term trends show that activity has slowly increased with time. Other well-documented eruptions are most vigorous in the first few days or weeks, declining thereafter. The Soufriere Hills volcano is thus unusual. Pulsations and uncertainties in the measurements mean that several months data are required to establish long-term trends. For example, data since June 1997 show the highest extrusion rates but, considered without reference to earlier data, no major change over the six month period can yet be recognised.

Empirical Data and Theoretical Models

  1. Empirical data from broadly similar volcanic eruptions around the world can be combined with the theory of volcanic eruptions to provide a basis for modelling the behaviour of the volcano. Theories for volcanic activity can be tested by comparison with data from eruptions. Eruptions, such as those at Mont Pelee, Martinique in 1902-1905 and 1929-1932, Mount Lamington, Papua New Guinea in 1951 and Mount Unzen, Japan in 1991-1995, can be used as comparisons for the activity on Montserrat.

  2. During the assessment presented here, theoretical and empirical modelling of individual phenomena has been utilised to assess the degree of hazard posed by the volcano in a number of defined scenarios. Models have differing complexity and degree of evolution, and must be used with care. Information on the scientific resources and technical issues is given in Appendix E. The basic theory is simple. Magma that rises slowly loses its dissolved gas en route to the surface and erupts as lava. Magma that rise sufficiently fast retains dissolved gas and erupts explosively. Behaviour at the Soufriere Hills volcano is in the awkward transitional regime between these two styles of behaviour.

  3. The possible occurrence of much larger magnitude explosive eruptions requires comparison with other volcanoes and models for such eruptions. Volcanoes that do have large explosive eruptions usually show evidence of the magma being more mobile (low proportions of crystals or higher temperatures). Volcanoes like the Soufriere Hills may have magmas too viscous to allow the rapid flow of magma to the earth's surface required to sustain a major explosive eruption (see Appendix E).

  4. While monitoring of the volcano has led to a much richer understanding of the eruption and its possible future course, the dynamical interactions of temperatures, pressures and rock properties means that the system as a whole is complex and at some levels its behaviour is inherently unpredictable with possible rapid switches from one state to another.

Scenarios for Future Activity

  1. For planning and decision-making, authorities need to focus on what are the possible future outcomes of the current eruption. Volcanology is not an exact science and ability to predict has considerable limitations because of our incomplete understanding of the phenomena and inherent dynamic variability. In common with many other natural systems (like the weather), a volcano's behaviour is controlled by complex interactions which mean that its future behaviour cannot be completely known. There are also considerable difficulties in evaluating the probabilities of more extreme events (see Appendix D). Furthermore the monitoring programme is restricted by the hazards themselves, which have destroyed some instrumentation and make measurements more difficult or impossible on safety grounds.

  2. The scientific team has identified different eruptive scenarios to which can be attached some estimate of likelihood. These scenarios form the basis of the hazard assessment. The methods of determining probabilities are described in Appendix D and relevant technical information in Appendix E. To facilitate placing these scenarios in the context of risk, Figure 1 shows Montserrat divided into 6 areas. Areas 6 and 5 are currently evacuated. Area 4 is at the boundary of the current total exclusion zone and includes the village of Salem where a significant number of people still reside against official advice. Area 3 is residential and where many government offices are currently located. Areas 1 and 2 together constitute northern Montserrat and are north of the Lawyer's Mountain ridge (the boundary between areas 2 and 3), which is deemed to provide significant protection from all but the most extreme kinds of volcanic event. Probabilities have been assessed for the next six months only. Longer term prognosis is necessarily less detailed. We recommend that the figures provided for probabilities are not extrapolated beyond 6 months.

  3. The current eruption seems typical of the previous eruptions within the last 30,000 years. Major changes in eruptive behaviour therefore seem less likely. Likely scenarios for this eruption can thus be defined as those which do not take the eruption outside the general pattern of the known recent behaviour of the volcano.

  4. Less likely scenarios are those that are either rare occurrences worldwide or which require the volcano to act outside its known behaviour pattern. Such events, although of low probability, deserve some attention because they include violent phenomena which could threaten much of the island adversely. Given that at least some of these more extreme events cannot be altogether excluded, it becomes a matter of determining what levels of risk are acceptable and investigation of mitigation measures to reduce vulnerability. The themes of vulnerability and risk are the subject of the companion report.

  5. This section addresses the following issues: the duration of the eruption, likely scenarios, bigger lower probability events which would have moderate to high impact and implications for serious consequences, and the situation after the eruption.

Duration of the Eruption

  1. The expert elicitation gives a central value of 5 to 6 years for total eruptive duration. Opinions indicate that the eruption is very unlikely to cease within the next six months or to go on longer than 15 years. These estimates are based on empirical evidence from other dome eruptions around the world and comparison of the size of this eruption with past Soufriere Hills volcano eruptions. Dome eruptions have lasted a few months in some cases to many decades in others. Scientific input into planning could be conveniently divided into the period while the volcano is still erupting vigorously and the period when the eruption has greatly declined or stopped.

  2. The estimates for various outcomes are for the next 6 months. Central values are quoted from the elicitation, but we note that the ranges, reflecting expert uncertainty, are used in the risk analysis. Although the likely scenarios identified below may continue for a few more years, estimation of longer-term probabilities is not simply a matter of adding a pro-rata factor for the duration extension. This would presume steady state dynamics of the volcanic system. While activity continues, there will always be an underlying concern that the eruption may change course to more dangerous levels.

Likely Scenarios

  1. Three main possibilities for activity over the next six months were considered: (i) noticeable decline in activity, (ii) activity approximately at the same level and (iii) a noticeable increase in activity. Criteria for recognising which of these outcomes is happening include growth rate of the dome, ground deformation, overall level and character of the seismicity and sulphur dioxide gas flux. None of these indicators are necessarily sufficient to define the trend of the eruption in isolation and all information would be used to make the assessment. For example a substantial decrease in dome growth rate accompanied by an increase in ground deformation and seismicity might well be interpreted as an increase in activity and signs that the eruption was moving towards something far more serious.

  2. The expert elicitation process revealed the central opinion that the next six months would see approximately the same level of activity as the last six months (60% chance) and that the chances of a significant decline or increase were about equal (20% each). The most likely scenario for future activity over the next six months is for continued dome growth. Intermittent periods of pyroclastic flow generation would lead progressively to greater areas being inundated by flows, as the dome gets bigger and valleys around the volcano fill up with debris and smooth the path for future flows. Explosive activity is also likely to continue intermittently, but major increases in the scale of explosions are much less likely than a continuation of explosions of similar size to those already experienced. Runout from pyroclastic flows generated during explosions is unlikely to be greater than that from the large dome collapse flows. Ash hazards will continue and depend on which valleys pyroclastic flows move down, the occurrence of further explosive eruptions and the direction and speed of the wind.

  3. A specific scenario is a repeat of explosive eruptions comparable in size to that of 17 September 1996. The probability of a repeat of this explosion during the next six months is about 45% (nearly 1 in 2). This event is the largest explosion to have taken place so it has been used as a reference to help estimate the chances of larger explosive eruptions. For comparison of this reference event with larger explosive eruptions, there are two principal measures of the size of an eruption: the intensity (the flux of mass and a measure of its power) and magnitude (the total mass of material erupted). Empirical data from volcanoes worldwide indicate that these two measures of explosion size are related (see Appendix E). Explosions somewhat larger than the largest yet seen during the current eruption can also be envisaged. An explosive eruption 3 times more intense and 7.5 times greater magnitude is estimated to have a probability around 6% (1 in 17) during the next six months. Such an event would be within the bounds of expected behaviour for this volcano.

  4. There is the question of how much more active could the volcano get for its behaviour to be regarded as outside the "expected" pattern. Limited empirical evidence indicates that large magnitude explosive events in similar volcanoes are preceded by dome growth rates of 10 to 20 m3/s or more. This suggests that some significant increases in growth rate might be used as an indicator of moving beyond the bounds of "expected" behaviour. Large ground deformation, deep earthquake swarms and major increases in the production of SO2 are other possible indicators of significantly escalating activity.

  5. In terms of hazards implications the most likely scenarios are that areas 5 and 6 (Figure 1) are likely to remain highly dangerous over the next six months and parts of areas 3 and 4 will remain threatened. In the longer term, even if activity does decline, caution is recommended in considering reoccupation of the present exclusion zone.

Bigger Events

  1. A number of scenarios have been constructed to embrace bigger and in some cases extreme events. These scenarios are regarded as low probability at the Soufriere Hills volcano. Because such events are outside the recent history of the volcano, they are more difficult to assess in terms of the likelihood of their occurrence. Empirical evidence of recurrence intervals from global datasets forms part of the basis for constraining probabilities.

  2. An explosive eruption 10 times greater in intensity and 25 times greater in magnitude is estimated to have a probability of occurrence in the next six months of 0.2% (1 in 500). Such explosive eruptions occur once per year on average in the world, but there are many limiting factors at the Soufriere Hills volcano which gives it a low probability. There have been several such eruptions on Martinique over about 30,000 years, and the consequences of such eruptions can be readily transferred to the case of the Soufriere Hills volcano. An improved understanding of some of these limiting factors (see Appendix E) has led to a significant reduction in the estimated probability of this event since the last elicitation of opinion over 3 months ago.

  3. An explosive eruption 30 times greater in intensity and 75 times greater in magnitude is considered to be the maximum plausible explosive event at the Soufriere Hills volcano. The probability of such an event at the Soufriere Hills volcano has been estimated at 0.05% (1 in 2000) in the next six months.

  4. A more difficult issue is the occurrence of large pyroclastic surges caused by a major explosion. There are less global data because such eruptions are not preserved easily in the geological record and few examples have been well documented. They are not well understood phenomena and have been a major cause of casualties in a number of eruptions, including Mont Pelee on Martinique in 1902. Probabilities for such an event are discussed further below.

  5. Other large magnitude but low probability events include a large volcanic landslide (which may or may not include lateral blast) and a tidal wave generated by a landslide or the collapse of unstable volcanic sediment. Another possibility is the occurrence of a large-magnitude earthquake close to Montserrat during the current eruption. Estimates of the likelihood of such events are further discussed in the following section.

After the Eruption

  1. The eruption will stop, but it will be difficult to define when this happens as volcanoes typically wane slowly. It may be a year or more after surface manifestations have ceased before the volcanologists can conclude that the eruption has stopped. Other volcanoes have apparently almost stopped and then started again. So monitoring will need to be continued after the eruption.

  2. There are potential hazards after the eruption ceases, including lahars, late explosions of the slowly cooling dome, and landslides. Ash will continue to be a threat to health until it is washed away or fixed by re-growth. The dome, the whole volcanic edifice and coastal volcanic sediments will need investigation for their stability.

  3. Vegetation is expected to recover quite quickly in the Caribbean climate as judged from experience of other volcanoes in similar climates. On other volcanoes seed has been sprayed over devastated areas to promote recovery.

  4. There is potential for renewed eruptive activity on Montserrat on a decade scale. A thirty-year recurrence of increased seismic activity was evident prior to the current eruption and this pattern might continue after the current eruption.

Hazard Assessment

  1. This section presents the likely impact areas for a variety of volcanic hazard types within each of the scenarios presented in the previous section. This information, when combined with estimated probabilities of occurrence of each scenario, feeds directly into the risk assessment.

Dome and Edifice Collapse Hazards

  1. For the most likely scenario of continued activity at about the same level, the hazard from dome collapse increases. As the lava dome increases in volume, pyroclastic flows will tend to be generated more often and be larger in volume. The continued filling of valleys around the volcano, as well as increased flow volumes, means that pyroclastic flows are likely to run out further and inundate greater areas even if the rate of eruption does not increase significantly. Continuation will however only take the flows into the sectors that have already been impacted or have been identified as vulnerable (e.g. areas 4 and 5).

  2. For larger dome collapses than have already been encountered there are limiting controls to the areas impacted by gravity-driven pyroclastic flows even with order of magnitude changes in dome collapse volume (Appendix E.5).

  3. Impacts of pyroclastic flows on areas 4 and 5 are entirely conceivable in the next six months even at the same level of activity. Collapses which have already taken place could have reached the lower parts of the Belham valley (boundary of areas 5 and 4) had they travelled in that direction. Increased volumes of collapses provide a higher chance of sufficient material being partitioned into the Belham valley for flows to reach the sea. Impacts in area 4 may be limited to hot drifting ash, although sustained growth on the northern flank of the dome coupled with infilling of the Belham River valley could make area 4 directly vulnerable to pyroclastic flows.

  4. In summary, irrespective of whether dome growth stays the same or increases, the hazardous phenomena associated with dome collapse will tend to increase in their impact.

Tephra Fall Hazards

  1. Controls on falling tephra include the explosion energy, the height of the eruption plume, the total mass erupted, the distance from the volcano and the wind velocity (Appendix E4). The largest fragments of 1 metre diameter have already been ejected to distances of up to 2 km, well within the current exclusion zone. Blocks this size have been thrown up to 5 km on other volcanoes, still within the current total exclusion zone, but area 5 would be reachable. Although larger distances are theoretically possible they have never been observed on any terrestrial volcano. Thus transport of such large blocks into the currently populated areas is considered highly unlikely.

  2. The deposit thickness and fragment sizes at a particular place for an eruption of given intensity and magnitude can be estimated by quantitative methods (see Appendix E.4 for details). The methods are based on a combination of theoretical models and systematic empirical relationships observed on a large number of volcanoes. Knowledge of the local wind speeds and directions and the probability distributions of these parameters are required to make an assessment of whether life threatening accumulations are likely. The models have been reasonably well validated against many examples. Pumice fragments of 10 cm diameter are a convenient size to consider, as fragments of this size or above can cause serious injuries or even death. Also fragments in the 15 to 20 cm range may be able to penetrate the most common types of roof.

  3. With respect to the reference eruption (17 September 1996) an explosive eruption of three times the intensity only poses a potential threat to area 5, which is currently evacuated. If the wind directions are to the north-west at values significantly above typical Caribbean wind speeds, then area 4 could be threatened by a few potentially dangerous fragments (10 cm diameter). Such meteorological conditions are only statistically present on several days per year so that at most times such an eruption would not seriously impact on the populated areas. Few if any roofs would be expected to collapse due to the weight of accumulated tephra.

  4. An eruption of ten times the intensity of the reference explosion would have an impact that depended on the wind direction and strength. The probability of this event affecting the various areas then depends both on the probability of the event and the conditional probability of the meteorological conditions being appropriate. For the mean regional high altitude wind speed this eruption has a probability of 1 in 800 and would impact areas 4 and 5 with thicknesses likely to cause some roof collapse and dangers from large fragments (pumices above 10 cm). Area 3 would be marginally affected but problems are not expected to be serious. A wind blowing to the north-west maximises the impact on the currently populated area but the probability decreases to less than 1 in 8,000 as a consequence of winds in this direction being unusual. All areas would be impacted and estimates of the actual effects and consequent risk are given in the companion risk analysis report.

  5. An eruption thirty times the intensity would have a major impact on all population areas under most meteorological conditions. For the mean wind direction and speed around Montserrat the impact on areas 3, 4 and 5 will be high. Areas 1 and 2 will be more marginally affected.

Fine Ash Hazard

  1. It is expected that further ash falls from pyroclastic flows and explosions will occur. A combination of meteorological conditions and volcanic factors mean that the exposure to this hazard varies greatly over Montserrat.

  2. The fine ash from pyroclastic flows is dispersed in low plumes to the west and occasionally west-north-west. Thus the ash rich in cristobalite and fine respirable dust is largely confined to areas 4, 5 and 6. Area 4 has often had poor air quality over the last several months. Area 3 could be affected by fine ash if there is substantial pyroclastic flow activity on the northern flanks of the volcano. Ash falls are infrequent in areas 1 and 2 (northern Montserrat).

  3. Explosions inject ash to much higher altitudes, and winds to the north in the 6 to 18 km altitude range are more frequent (about 10% of the year). However, this ash does not contain as much respirable dust and cristobalite contents are low. Monitoring of suspended dust concentrations in the north (areas 1 and 2) shows generally good air quality in contrast to area 4 where air quality is commonly poor. These results clearly imply that conditions in northern Montserrat are very much better than in communities closer to the volcano and downwind of ash plumes from the pyroclastic flows.

Pyroclastic Surge Hazards

  1. Pyroclastic surges are less constrained by topography than pyroclastic flows. Larger examples are able to climb over topographic obstacles hundreds of meters high. This combination of high temperature, high speed and high mobility makes this an exceedingly hazardous phenomenon, and survivability of life or property in an area strongly impacted by surges is negligible.

  2. Volcanoes can produce surges over a wide size range. We consider here the areas likely to be affected by surges produced in eruptions 3, 10, and 30 times the power of the 17 September reference event. These estimates are based on a combination of empirical observations of surges that have occurred at other volcanoes and mathematical models (see Appendix E.6).

  3. A surge generated by an explosion three times more powerful than the reference event is estimated to move out radially from the volcano to about 4 km distance. The likelihood of an explosion of this power producing a surge of this magnitude is estimated at 1 in 35. The event could affect the eastern part of area 5, along the upper Belham valley. and this consequence has an estimated probability of about 1 in 60. Given the uncertainty in the calculation the affected zone might stretch across area 4 to the margins of area 3, but at a decreased probability of 1 in 3,000. If the surge-producing explosion was directed, it might move more than 3 to 4 km.

  4. A radial surge produced by an explosion ten times more powerful than the reference event has an estimated probability of about 1 in 800. The surge is estimated to reach 3.5 to 4.5 km toward the Centre Hills, but would not surmount this obstacle. The valleys would probably channel the surge to both the east and west to greater distance. This eruption is expected to extend into areas 5 and 4. Given the uncertainties the event could impact area 3 and the margins of area 2, but the probability of this is regarded as very low (1 in 10,000).

  5. A surge produced by an explosion thirty times more powerful than the reference event would probably send parts of the surge over the Centre Hills, well into population area 3, and perhaps marginally into area 2. Areas 4 and 5 would be completely inundated. An event of such severity has been estimated at 1 in 3,000.

Lahar Hazards

  1. The lower part of the Belham valley from the Belham Bridge (boundary of areas 5 and 4) to the coast is currently at risk from lahars during heavy rain storms, due to the high volume of loose debris in the upper parts of the catchment area. Low-lying ground in area 5 is particularly vulnerable.

Volcanic Landslide Hazards

  1. A large volcanic landslide on Montserrat, although regarded as an unlikely event (1 in 500), is not beyond the known behaviour of the volcano. A large landslide could occur on any flank of the volcano, but is less likely to occur in sectors buttressed by massive walls. A volcanic landslide would involve a very variable amount of material and could prompt a laterally directed pyroclastic surge. This might, through secondary effects, generate tidal waves.

  2. In terms of impact, only a sector collapse on the northern flank followed by a lateral blast would impact significantly on the central and northern parts of the island and, in such a scenario, all areas would be devastated. The overall probability of this event is around 1 in 2,000. Sector collapses on flanks leading to the coast (east, southwest, west) could generate tidal waves on entering the ocean, especially if they dislodged or overloaded the relatively unstable submarine and subaerial fans of pyroclastic flow material at the end of White River and Tar River valleys. Tidal wave hazards to populated coastal areas of Montserrat is minimal.

Very Large Magnitude Explosive Events

  1. Explosive eruptions larger than those considered so far have a very small likelihood of occurrence, but cannot be ruled out. However, there was agreement that the event has a less than 1 in 100,000 chance of occurring and should thus be discounted on the grounds that equally devastating phenomena (total island destruction) have a far higher likelihood than this through different processes such as large magnitude earthquakes and one in a 1,000 year hurricanes.

Earthquake-Generated Volcanic Phenomena

  1. Montserrat has a history of influence by magnitude 6 plus earthquakes within 50 km or so of the island and there is empirical data to suggest that earthquakes can influence erupting or meta-stable volcanoes. The most likely result of a large (say magnitude 8) earthquake close (< 50 km) to Montserrat would be to prompt sector collapse or large-scale dome collapse. It may also prompt onset of explosive activity either independently or not. The recurrence interval for magnitude 8+ earthquakes close to Montserrat is not well known but such earthquakes occurred in 1692 and 1843, suggesting a recurrence interval of such earthquakes perhaps every few hundred years. This equates to a 1 in several hundred chance of this occurring in the next six months. Smaller earthquakes have the potential to induce similar effects: their higher occurrence rate is balanced by the lesser chance of influence, so that the probability given above may not vary much.

Concluding Remarks

  1. The main points that are identified as critical for decision-making from this scientific report are provided in the executive summary. Here some concluding remarks are made to place the situation on Montserrat in a broader context of hazard and risk. The Soufriere Hills volcano is a relatively small volcano set in a global context. It has had a small number of broadly similar eruptions since the volcano started about 30,000 years ago. There is some evidence for larger explosive eruptions that are significantly older, however. The signs are that a new eruption is now happening that will prove to be similar to those of the recent geological past. The eruption has slowly increased in vigour, but has not moved outside the bounds of behaviour anticipated from its recent geological record. In this case the eruption is unlikely to affect areas in the north of the island. A great deal of progress has been made in understanding the volcano and it is now becoming one of the best documented examples of an andesite dome eruption. However the volcano is on an island with an area of only 100km2 and the eruption has already impacted over half the island. There will always be a background threat of the eruption escalating into more severe eruptions while it remains active.

  2. Volcanology is not such an advanced science that large magnitude events can be excluded and the first large magnitude eruption of any volcano by definition represents an unprecedented event. The Caribbean region is subject to other natural hazards and so an acceptance of hazard is an implicit aspect of communities in the region. Volcanic events that could affect northern Montserrat are at a level that is comparable to or less than these other kinds of known hazards. However these other hazards are intrinsically different to volcanic hazards. Earthquakes cannot yet be predicted while behaviour of hurricanes can be anticipated and mitigated against. Even though the estimated probabilities for these hazards are similar to the more extreme volcanic events, there must naturally be particular concern while the eruption continues.

Appendix A: List of Attendees, Their Affiliations and Expertise

Dr.W.P. Aspinall (Aspinall and Associates, UK) is a seismologist with extensive experience in the Caribbean on earthquakes and volcanoes, having worked for the Seismic Research Unit, Trinidad for over 15 years. Since 1985 he has been an independent consultant on hazard and risk assessment and the use of expert judgement, advising UK nuclear, civil engineering and aviation industries. He has been one of the senior scientific advisers to HMG throughout the crisis on Montserrat and one of the MVO Chief Scientists*.

Dr. P.J. Baxter (Dept. of Occupational Medicine, Cambridge University, UK) is consultant physician at Addenbrooke's hospital Cambridge and specialises in occupational and environmental medicine. He has worked at the UK Health and Safety Executive and the Centers for Disease Control, Atlanta, USA, where he developed a special interest in technological and natural disasters, including the medical aspects of volcanic eruptions. He has been providing advice on the medical dimensions throughout the crisis.

Dr. J.D. Bennett (British Geological Survey, Nottingham, UK) represented the British Geological Survey who manage the formal parts of the contract for the monitoring work on behalf of DFID. He undertook the arrangements for the meeting and assisted in its implementation.

Dr. P.W. Francis (Dept. of Earth Sciences, Open University, UK) is an expert in volcanic gases, remote sensing as applied to volcanoes and large volcanic landslides. He has co-ordinated the monitoring of sulphur dioxide and chlorine gases during the eruption. He also has extensive experience of many other volcanoes around the world.

Dr. R.P. Hoblitt (Cascades Volcano Observatory, US Geological Survey) is a senior geologist from the Volcano Hazards Program of the US Geological Survey and part of the USGS VDAP (Volcano Disaster Assistance Programme). He has experienced many volcanic disasters, including Mount St Helens and Mount Pinatubo and has made frequent visits to Montserrat to assist MVO since the volcano reawakened in 1995.

Mr. L. Lynch (Seismic Research Unit, University of the West Indies, Trinidad) is a seismic technician. He is an expert in electronics and the installation of monitoring systems. He has over 20 years of experience working on Caribbean volcanoes and earthquakes. He is one of the Chief Scientists* at the MVO.

Dr. G. Mattioli (Dept. of Earth Sciences, University of Puerto Rico) is a geophysicist studying Caribbean plate boundary processes using GPS and other techniques. His primary research focus over the last several years has been in Montserrat, so that his data have been a very valuable complement to the MVO ground deformation monitoring.

Mr. R.A.E. Robertson (Seismic Research Unit, University of the West Indies, Trinidad) is a geologist who has been one of the Chief Scientists* at the MVO. He has carried out hazards-related studies on several Caribbean islands and has a special interest in ground deformation and physical volcanology.

Dr. K. Rowley (Independent scientist and LANDATA Associates, Trinidad) is a geologist and former employee of the Seismic Research Unit, Trinidad with experience of the St Vincent and Guadeloupe volcanic crises. He was head of the SRU from 1989 to 1991. He then entered politics and became the Minister of Agriculture, Land and Marine Resources in the last Trinidad Government and is now an opposition member of the Trinidad parliament. He is contracted by the BGS to act as one of the Chief Scientists* at MVO and continues geological consulting work.

Dr. J.B. Shepherd (Institute of Environmental Sciences, Lancaster University, UK) is a Reader in Geophysics at Lancaster University. Formerly he worked at the Seismic Research Unit, Trinidad from 1964 to 1989 and he was head of the Unit from 1980 to 1989. His main areas of expertise are in the seismicity of the Caribbean, earthquake hazards and their mitigation and in ground deformation. He has been the principal scientist involved in the design and implementation of the ground deformation monitoring on Montserrat.

Professor R.S.J. Sparks FRS (Dept. of Geology, Bristol University, UK) is a volcanologist who has extensive experience of active volcanoes around the world and has specialised in investigations of the dynamics of volcanic processes. He has been one of the senior scientists involved in the monitoring of and research on the Soufriere Hills volcano and in advising HMG. He has also been one of the Chief Scientists*.

Professor B. Voight (Dept. of Geosciences, Pennsylvania State University, USA) is an engineering geologist who has spent much of his career applying his expertise to volcanic phenomena. He is an acknowledged authority on large volcanic landslides and associated phenomena such as tidal waves and pyroclastic surges triggered by volcanic landslides. He holds an adjunct appointment with the U.S. Geological Survey. He was awarded the George Stephenson Medal of the Institution of Civil Engineers (London). He has worked at MVO as a senior scientist.

Professor G.P.L. Walker FRS (Dept. of Geology and Geophysics, University of Hawaii, USA: retired) was formerly at Imperial College and James Cook Professor at University of Auckland. From 1980 to 1996 he held the Gordon MacDonald Chair in Volcanology at the University of Hawaii from which he has just retired. He is one of the most eminent volcanologists in the world. He has had no previous connection with Montserrat and was involved to gain benefit from his worldwide knowledge of volcanoes.

Dr. G. Woo (independent consultant, London, UK) is a Cambridge-trained mathematician who completed his PhD on theoretical physics as a Kennedy scholar at MIT, after which he became a member of the Society of Fellows of Harvard University. Since 1980 he has been a risk consultant on Natural Hazards, and has undertaken numerous risk assessments involving the elicitation of expert judgement. Currently he is the principal risk analyst for the world's largest catastrophe reinsurance brokers and has worked extensively with Dr. Aspinall.

Dr. S.R. Young (British Geological Survey, Edinburgh, UK) was trained in volcanology through a PhD at Lancaster University from where he joined the British Geological Survey. He has experience of volcanic hazard assessment work in Central and South America. He has spent more time on Montserrat than any other UK scientist and has played the leading role in the BGS contribution to the monitoring effort. He is one of the Chief Scientists* at MVO.

*Note: The Chief Scientists post is rotated between UK scientists and the Seismic Research Unit scientists. Each Chief Scientist typically spends between 4 and 6 weeks on duty.

Appendix B: Terms of Reference

The terms of reference here are a summary of various correspondence between the British Geological Survey, DFID and the Foreign Office. The tasks are divided into two stages: a scientific assessment (stage 1) and a risk analysis (stage 2).

    Stage 1

  1. To review the current status of the Soufriere Hills volcano, the hazards that the volcano poses and the scientific monitoring programme.

  2. To consider the prognosis for behaviour of the volcano over the short-term (next 6 months) and long-term (up to 15 years).

  3. To assess the probabilities of future outcomes and hazardous phenomena.

  4. To provide the scientific information necessary for an analysis of the risks posed by the volcano.

    Stage 2

  5. To complete a risk analysis of the volcano based on the best available geoscientific and health information.

  6. To report the findings in timely fashion to HMG and the Government of Montserrat.

Appendix C: Glossary of Terms

Andesite: The name given to the type of magma erupted in Montserrat.

Conduit: In a volcano magma flows to the earth's surface along a pathway known as a conduit. The conduit is usually thought to be a cylindrical tube or a long fracture.

Intensity: The standard measure of the power of an explosive eruption is the intensity and is defined as the mass flux of material coming out of the volcano.

Lahar: A flow of rock debris, ash and mud that occurs on many volcanoes particularly during eruptions. Large lahars can be very destructive and can be a major cause of death.

Lava: Once magma gets to earth's surface and extrudes it can be called lava. Below ground it is always called magma.

Magma: The material that erupts in a volcano is known as magma. It is not simply a liquid, but a mixture of liquid, crystals and volcanic gases. Magma must contain enough liquid to be able to flow.

Magnitude: The magnitude of an explosive eruption is the total mass of material erupted in an event.

Plume: A buoyant mixture of hot air and volcanic particles that rises above the volcano high into the atmosphere.

Pumice: Pumice is bubbly frozen magma. Fragments of pumice are common on Montserrat and are essentially frothy light rock with a density similar to water (about 1 g/cc).

Pyroclastic flow: These are flows of volcanic fragments similar to avalanches of rock in landslides and snow avalanches. They can be formed both by explosions and by parts of an unstable lava dome avalanching.

Pyroclastic surge: These are also flows, but they are dilute clouds rather than dense avalanches. A surge is a rapidly moving mixture of hot particles and hot gas and their behaviour can be compared to a very severe hurricane. Surges can be formed above pyroclastic flows or directly by very violent explosions.

Tephra: A general term for all fragmented volcanic materials, including blocks of rock, pumice and volcanic ash.

Viscosity: This is measure of how easily a fluid flows. The andesite of Montserrat has a viscosity about one million times greater than the fluid lavas at volcanoes like Hawaii.

Volcanic ash: Ash particles are defined as less than 4 millimetres in diameter. Respirable ash consists of particles less than 10 microns (a micron is one thousandth of a millimetre) in diameter.

Appendix D: Procedure for Probability Assignment

At the core of any quantitative risk assessment is the procedure for probability assignment. Montserrat happens to be the first volcanic crisis in which a formalised elicitation methodology for probability assignment has been applied. The reasons for this precedent are several. The methodology itself has developed to its present technical maturity and decision-making applicability over the past two decades, during which only a few volcanic crises have arisen. The prolonged duration of the Soufriere Hills Volcano crisis has allowed a sufficient window of opportunity for the elicitation methodology to be utilised, and the constructive and willing co-operation of Caribbean, British and American volcanologists has permitted its practical implementation. Not all political environs would necessarily be as conducive to such international co-operation.

It has been traditional for volcanological advice to civil authorities to be presented in essentially scientific terms, rather than expressed quantitatively in the language of risk which facilitates political decision-making. Compared with conventional volcanological deliberations, no less discussion, debate and peer-review have taken place in the current context, but a substantial volume of supplementary work has been carried out to translate volcanological judgement into risk terms.

The value of achieving a quantitative expression of risk lies in the possibility of evaluating explicitly the consequences of alternative decisions. Vague expressions such as 'unlikely event' and 'severe casualties' can be replaced by numerical estimates of event frequency and loss probability. In addition, because risk results can be traced back to original premises through an arithmetically defined event tree, the chain of reasoning is much more transparent and explicit than in a purely qualitative argument.

Quantification of risk demands a numerical representation of uncertainty. For a complex natural phenomenon such as a volcano, which is partly subterranean and beyond direct observation, uncertainty takes several forms. There is epistemic uncertainty arising from lack of knowledge of the internal geometry and dynamical characteristics of the volcano. There is aleatory uncertainty associated with inherent random variability in the dynamic evolution of a multi-state stochastic system.

The state-of-the-art in volcanology has not yet reached the stage where detailed dynamical models can be constructed to simulate many eruptive processes reliably. Therefore, the task of assessing the likelihood of alternative future event scenarios has depended on the elicitation of expert judgement of volcanologists experienced not only with the Soufriere Hills volcano, but with worldwide knowledge of volcanic eruptions. The use of expert judgement is standard practice in quantitative risk assessment pertaining to geological issues, ranging from earthquake and landslide exposure to groundwater pollution. Observational results are rarely so abundant that a substantial measure of data inference is avoidable, and the manner in which earth scientists draw these inferences is open to a certain degree of subjectivity.

The elicitation of expert judgement has been conducted with care to maximise the advantages gained from the pooling of data, the sharing of ideas and interactive discussion, whilst respecting the diversity of views and strong convictions held by individuals. Thus a meeting of principal volcanologists was convened in early December so that all volcanological issues could be addressed within the framework of a decision conference, at which an open interchange of knowledge and opinion could be encouraged and facilitated during extended periods of discussion.

But rather than narrow the outcome of any discussion topic to a consensus view, attention was paid to eliciting individual responses. These responses have then been analysed using specialist software developed for eliciting a full spectrum of expert judgements. The algorithms underlying the software are rooted in the formal mathematical theory of scoring rules.

In the formal elicitation of expert opinion, each participant is calibrated by a series of questions to ascertain his level of expertise and confidence with which he makes judgements. The calibration allows an expert's opinion to be weighted against the opinion of other experts, and the mean result of an elicitation is weighted in favour of those who demonstrate the greatest depth of knowledge and those who provide accurate opinions or assessments. As part of the calibration, experts are asked to give a range of answers or predictions to a series of questions to indicate their level of proficiency. Those who display less informativeness by providing a wider range of answers are weighted lower than those who give narrower and more accurate answers. In a full elicitation the scientific factors behind an issue are first debated before answering a question (e.g. how much longer will the Soufriere Hills eruption continue?). This prior debate ensures that all participants are as well-informed about the topic as possible. A suitable piece of information for the example question is that the global mean length of 167 historic dome eruptions like Montserrat is 5 years and that dome eruptions in other Caribbean volcanoes have lasted between 2 and 4 years. Another piece of information is that the Soufriere Hills has had 5 major prehistoric dome eruptions of about 0.5 km3 volume each. At the present rate of eruption it will take another 3 years to erupt comparable volumes.

There can be more than one elicitation cycle so that, if widely divergent views become evident from the answers, then the issue can be further debated and the elicitation revised. The debate for example might reveal that the question was misunderstood by some participants or one line of argument or evidence was not properly comprehended. This further debate might lead to consensus being achieved or to a hardening of opinions, which would be recognised in the probability assignments.

At the meeting, the formal elicitation of opinion proved to be a very useful exercise and has provided an effective mechanism for focussing debate and classifying opinions on key questions relating to the volcano. An important virtue of elicitation exercises is the effect it has of concentrating minds on new aspects of risk. Moreover, the working group of scientists were able to combine the method with some of the more traditional approaches. Having identified key issues, sub-groups were delegated to discuss particular topics and to bring their findings to the main group for discussion.

The tasks were divided on the basis of specialist expertise within the group (for example a seismologist was asked to consider the occurrence of large magnitude earthquakes in the Caribbean and their possible effects on the volcano during the eruption were they to happen). The deliberations of a sub-group were then debated by others (a form of peer-review). Elicitations require for their credibility that questions are very well understood by all respondents; a requirement which is not always so easy to achieve. Iterations of the elicitation procedure were undertaken to resolve difficulties of this kind, where time allowed, or else there was resort to more traditional approaches of open discussion.

The efforts required to ensure comprehension of all questions underscores the need for the careful structuring of an elicitation procedure. As a general principle, in so far as resources permitted, elicited variables were decomposed into primary components which might more readily be estimated. Furthermore, where numerical models were available for gauging the effects of a particular hazard, e.g. tephra fall or pyroclastic flow, they have been used in assessment. Judgement are then restricted to estimating input parameter uncertainty in the models.

The greatest difficulty for the scientific assessment comes with the estimate of probabilities of the more extreme high impact events. There is often very limited empirical data or scientific basis for estimating the probability of an extreme event and this is best illustrated by reference to a particular example: the probability of an eruption 100 times more intense and 250 times the magnitude of the largest explosive eruption so far. An eruption of this scale has not taken place on Montserrat and there have been very few eruptions which have reached or exceeded this size in the entire Caribbean over the last 30,000 years. These eruptions have only occurred on the two largest volcanoes (Dominica and St Lucia) which have extensive records of large explosive eruptions. There are also global statistical figures for eruptions of this scale (1 eruption every 25 years for about 1500 volcanoes). There is also some information on how the probability of an eruption changes with increasing intensity and magnitude. However, many in the scientific group were of the opinion that the Soufriere Hills volcano is a relatively small volcanic system with magma freely flowing to the surface and with petrological features atypical of the magma bodies that produce large plinian eruptions and typical of volcanoes which produce dome eruptions. Thus the elicitation results reflect not only constraints from global or regional data, but the opinion that it is hard to see the volcano may not be capable of having such an enormous eruption.

Respondents' answers stopped short of saying that the eruption was impossible, as the first really big eruption of any volcano will have no previous record of such large eruptions and because all the working group recognise limitations in understanding. Because singular events are a key concern of the risk assessment, the practical limits of standard frequency-based classical statistics are apparent. From an observational stand-point, silence might be maintained on the long-term frequency of events not yet recorded. However, within the current decision-making context, the framework of modern probability provides an appropriate logical basis for representing volcanological expert judgement on the degree of implausibility of extreme events.

Appendix E: Scientific Resources

It is incumbent on the scientific team at MVO to provide some basic documentation of the scientific methods and information used to make the assessment. This is done here by reference to the scientific data available at MVO and by a brief and more technical description of key scientific issues and information used in the hazards assessment.

E.1 Data and analysis at MVO

A large amount of scientific data on the volcano has accumulated through the monitoring programme. The MVO produce a weekly scientific report summarising the main observations. There have been a number of special scientific reports on particular topics such as the 17 September 1996 explosive eruption, the seismicity, the dome volume calculations and the hazards relating to pyroclastic flows. These reports are available to members of the scientific community on request. The large amount of data gathered by a modern observatory means that the detailed analysis and research work lags behind the basic data collection and immediate interpretation for assessing the status of the volcano. Thus there are few publications. Two preliminary articles summarising the scientific work have appeared in Science (Vol. 276, pp. 371-372, 1997) and the Transactions of the American Geophysical Union (Vol. 78, pp. 404, 408-409, 1997). About 26 scientific papers, largely from the work of the MVO, were submitted to a scientific journal (Geophysical Research Letters) in September 1997 for a special issue on the eruption. These papers are currently under peer-review and have been circulating freely within the scientific community.

E.2 Interpretation of the eruption dynamics

The detailed documentation of this eruption resulting from the monitoring programme and allied research work are leading to constraints and new ideas on how andesite lava dome eruptions work. The interpretations may have considerable significance for hazards evaluations, but need to be very carefully applied. Many of the ideas emerging from Montserrat are scientifically exciting, but should be seen as working hypotheses, not well established facts. Many have not yet been through the full rigour of peer review and the scientific team is well aware of the pitfalls of placing too much weight on the latest exciting idea in a hazards assessment on a dynamical system as complex as a volcano. This section draws attention to some of the issues relating to the interpretation of the eruption and draws a distinction between well established information and areas of uncertainty relevant to the scientific assessment.

Many of the most important processes in a volcano occur too deep for direct observation. The source region is known as the magma chamber and the pathway to the earth's surface is known as the conduit. Volcanic theory attributes key controls to the properties of this source region and the conduit. The volume, pressure conditions, magma physical properties and dynamical processes in the magma chamber and the geometry and dimensions of the conduit are major factors in determining eruption processes. Were all these parameters well known, much more confident forecasts and interpretations could be made. Unfortunately many of the key properties are very poorly constrained.

The depth of the source is perhaps the best known parameter. It is estimated at about 5 to 6 km on the basis of phase equilibria and solubility laws for volcanic gases and is also consistent with geophysical data (earthquake depths and ground deformation results). Experimental studies and thermo-dynamical calculations demonstrate that the minerals in the lava can only be formed at pressures equivalent to these depths and the water dissolved in melt trapped within crystals demonstrates that the amount of gas dissolved is at these same pressures. There are few good constraints on the size of the chamber. Ground deformation in principle could be used to constrain magma chamber size, but in practice the deformations are dominated by shallow processes, masking the signature of the deep source.

The chamber must have excess pressures driving the flow to the surface, but there are few constraints. It must be greater than the pressure required to create a dome of height 250 m (about 4 MegaPascals (MPa)) and probably less than the strength of the crustal rocks at depth (about 20 MPa). The changes in driving pressure as the eruption has gone on are not known. There is good evidence that the magma has been reheated recently by hotter magma derived from much deeper in the earth. However we do not know for sure when this happened or indeed if it is still happening. If this hotter material is still invading the chamber to drive the eruption then the implications for the eruption dynamics, the potential for larger eruptions and the eruption duration are quite different than would have been the case had the invasion happened a few decades or centuries ago. We simply do not know. Tied in with these considerations are the rheological properties of the magma in the chamber and the size of the conduit. If these parameters were well-known then it might be possible to state whether a much larger eruption is possible. The viscosity of the magma may be too high and the conduit too narrow to allow a large magnitude explosive eruption. In general volcanic theory indicates that this is a major reason why lava dome eruptions like the Soufriere Hills volcano often do not develop into large magnitude explosive eruptions. However occasionally dome eruptions do evolve into significant explosive eruptions years after dome formation initiates (see below). Experimental work is being carried out to better constrain the viscosity of the Soufriere Hills magma. However viscosity is very sensitive to magma properties such as temperature, gas content and crystal content and these may never be known sufficiently well to provide definitive answers. Conduit dimensions can be constrained from flow models and there are some observational constraints. At present these models imply a conduit cross-sectional area of the order of 1,000 to 1,500m2.

Uncertainties are as much as a factor of two however and it is still not clear whether this conduit is a cylindrical tube or a long narrow crack. The strong dependence of flow rate on the conduit width (to the third or fourth power) makes the uncertainties considerable. Furthermore the viscosity may vary by over nine orders of magnitude as it rises up the conduit, loses gas and solidifies. The flow laws for material with such very large pressure-dependent changes in rheological properties, including transitions from newtonian to highly non-newtonian behaviour, are themselves not well established. In this context for example the changes in flow rate in the eruption (Figure 2) can be explained either by driving pressure or viscosity or very small relative changes in conduit dimensions or more likely some combination of these three factors.

Considerable progress has been made in understanding the high level processes in the conduit and dome. The cyclic patterns of pressurisation that have been identified are believed to be related to the interactions between gas loss and the consequent changes of rheology of the magma. The indications are that many of the main variables being monitored are the consequence of shallow processes (within 2 km of the surface). While this progress is satisfying and exciting from a scientific perspective it has implications for the assessment. It is increasingly apparent that many of the signals and measurements relate to shallow processes and that the deeper processes are being masked or are difficult to detect.

In conclusion while scientific progress on understanding eruption dynamics has been considerable there remain a large number of unanswered questions and uncertainties.

E.3 Serious explosive events in dome-forming eruptions

A critical question on Montserrat is whether there is a possibility that much larger explosive eruptions can take place. The experts are all of the view that such eruptions cannot be excluded. Two kinds of phenomena need to be clearly distinguished: vertical eruptions and pyroclastic surges. From the point of view of hazards these two phenomena are easily distinguished but, in terms of the occurrences and circumstances of them happening, the scientific issues are a good deal more complex. The primary volcanic processes that can produce these two hazardous phenomena are such that they can occur independently (that is one occurs without the other) or they can occur together. In the case of pyroclastic surges in particular, understanding of the causes of the explosions is poor.

Sustained vertical explosive eruptions (known often as plinian eruptions) are relatively well understood. They require quite rapid flow of magma from the chamber into the conduit over the course of a few tens of minutes or hours. Plinian eruptions usually erupt predominantly low density pumice. The pumice on Montserrat is significantly denser (typically by 30 to 60%) than the pumice typically produced by plinian eruptions. World-wide, magmas that erupt in this way usually have low crystal contents, although examples with up to 35% crystals are known, compared to the 60 to 70% crystals in the Soufriere Hills magma. Thus the Soufriere Hills volcano is not typical of volcanoes where such eruptions occur.

As discussed in the main text, there is no evidence that a much larger explosive eruption has happened in the recent geological past (about the last 30,000 years). However, as discussed in E.2, we also cannot be sure that there are not suitable conditions in the volcano for this to happen this time. There are some examples of quite large magnitude eruptions happening several years after a dome has appeared. An example is Lascar volcano in Chile which had a substantial explosive eruption in 1993, 9 years after the andesite dome appeared. This eruption produced pyroclastic flows travelling to 10 km and a 25 km high eruption column, equivalent to an intensity over 10 times the reference event on Montserrat. This is a different volcano which may have different features which it make it possible to happen on Lascar, but not possible on Montserrat. However we cannot be sure. It is much more common for major explosive eruptions to happen in the first few days or weeks of an eruption and such late major explosive eruptions seem to be relatively uncommon. However they do happen.

Pyroclastic flows can occur in a number of circumstances. A minority of sustained explosive eruptions of the plinian type (perhaps 10 to 20%) produce large surges at the same time as the vertical column. Pyroclastic surges can be triggered by landslides as on Mount St Helens in 1980. They can also occur as high level explosions of great intensity lasting for a few minutes as at Mount Lamington in 1951. The surge of 1902 on Mont Pelee, Martinique is another example. In May of this year a large pyroclastic surge moving over 7 km occurred on Bezymianny volcano, Kamchatka Peninsula, 41 years after the andesite dome first started to grow. While these violent events are not common on dome eruptions, they do occur. Some pyroclastic surges are definitely triggered by large failures of a dome or volcanic edifice, but the circumstances that cause events like Mount Lamington are not understood. This lack of knowledge makes it particularly hard to estimate probabilities.

E.4 Modelling of hazards related to vertical explosive eruptions

The last 20 years has seen progress in the quantitative understanding of explosive eruptions and the tephra fall deposits. This work has recently been summarised in a book by Sparks and colleagues (Volcanic Plumes, John Wiley Press, pp. 557, 1997). The dispersal of tephra has been modelled from the fluid dynamics of volcanic plumes and their interactions with the wind field. These models predict distances that the range of rock fragments travel and have been validated by comparison with field observations. The modelling has also established a simple power-law relationship between the height of a volcanic eruption column and the intensity (mass flux) of an explosive eruption which is a major control on fragment range. Comparison of theoretical calculations with an extensive data set from historic eruptions yield similar power law coefficients from theory and from an empirical approach. The distance a fragment goes is also controlled by the wind and there is a reasonably good and validated understanding of wind effects. As a result it is possible to estimate the maximum range of fragments of interest (say 10 cm fragments of pumice) if the wind conditions and column height are known.

Estimates of thickness at a locality have to be based on a more empirical approach. Tephra fall deposits characteristically display exponential thinning away from the volcano and so can be characterised by a distance over which the thickness halves, known as the thickness half distance. There are systematic correlations between the thickness half distance and the height of the eruption column, reflecting the fact that in small eruptions most of the tephra falls close to the volcano and the deposits thin rapidly whereas in large eruptions the tephra falls over a much wider area. These relationships are supported by theory, but the theory is not sufficiently advanced to be comfortable with direct application, so that an empirical approach is preferred until the modelling work advances further. The thickness pattern is distorted by the wind with thinning occurring more rapidly normal to the wind direction and more slowly downwind. Again the degree of distortion can be established empirically and we have used unpublished data from a study of the tephra fall deposits of Mont Pelee, Martinique by G.P.L. Walker to constrain the distortion. These eruptions seem to have occurred under fairly standard Caribbean meteorological conditions. If the mass of erupted material (the eruption magnitude), the eruption intensity and the wind conditions are known then an estimate of thickness and the number of large clasts falling at a particular place can be estimated. These results can be then compared to criteria for roof collapse or injury in a risk analysis. Validation can be carried out by comparison of results with historic eruptions of comparable scale and conditions. There are of course uncertainties in such models but the parameters of interest are very strong functions of the main controlling conditions so that to first order the results stand up to scrutiny.

The hazards assessment uses the 17 September 1996 explosive eruption as a reference. From various observations column height was between 11.5 and 14 km for that eruption. Taking an average of 13 km gives an intensity of 7.5 x 106 kg/s. The duration of the eruption is not well-constrained, but it lasted more than 10 minutes and the intensity decayed over about 40 minutes. A volume of 3 x 106 m3 is estimated as the reference volume.

The meteorological conditions in the Caribbean provide some special conditions as the wind patterns are unusually complex with trade and anti-trade winds moving in opposite directions and seasonal reversals in direction. The high altitude wind speeds are weak compared to other parts of the world. In our models we have used data for Montserrat for the 1979 to 1995 period provided by the National Center of Environmental Prediction in Washington, USA. These data allow probabilities to be assigned to different wind directions and speeds based on the variance so that probabilities can be fed into the risk analysis. Winds are usually east and west and it is quite common for eruption columns to be split into two different directions. These complexities mean that there are some difficulties of comparison with tephra deposits from other parts of the world and that, where possible, other Caribbean examples should be used. Data from the 17 September 1996 eruption of the Soufriere Hills volcano and from Mont Pelee have helped in this respect.

A final problem is that eruptions can vary in both intensity and magnitude and each of these parameters has different influences on the level of hazard. This potentially makes a large number of permutations of conditions. However, empirically, the two parameters are not independent and a large global data set indicates a systematic increase in magnitude with intensity. Regression through this data set indicates that for every increase in intensity by a factor X the magnitude increases by a factor 2.5X.

E.5 Dome collapse hazards

As the lava dome grows, its associated hazards will increase. These hazards include the following: gravitational collapse of large portions of the dome (up to 30%) to produce pyroclastic flows and associated ash clouds; larger collapses, involving major portions of the dome and possibly also parts of the volcano edifice (sector collapse) to produce volcanic debris avalanches; tidal waves generated by avalanches running into the sea or by slumping of the offshore fans comprised of pyroclastic flow deposits; and explosive eruptions triggered by unroofing depressurisation. As the dome has increased in size, and its rate of growth has (on average) accelerated, the frequency and average size of pyroclastic flows has tended to increase. The current dome volume is about 80 x 106 m3. The current average lava production rate of 5 to 8 m3/s produces a net dome increase of about 40 x 106 m3 per 6 month increment, including consideration of attrition by rockfalls. Thus the dome volume after 6 months could be about 120 x 106 m3, but less if there have been significant pyroclastic flows over the period. These volumes place constraints on processes over the period of interest.

It is convenient to consider for comparison the larger dome failures, such as those of 17 September 1996 and 21 September 1997. These involved failure of roughly 10 x 106 m3, and we select this size as a reference event against which possible future collapses may be compared. Thus an event 3 times larger would comprise 25% of the anticipated dome volume after 6 months, an event 10 times larger would comprise virtually the complete dome, and an event about 30 times larger would involve the full lava dome and part of the surrounding edifice as well; the last event would comprise a sector failure.

Pyroclastic flows. All valleys around the volcano are now being affected by pyroclastic flows and accompanying dilute ash currents. The principal issue involves entry of pyroclastic flows generated by dome collapse into potentially populated areas on the southwest, areas 4 and 5, and in the north (area 3 and beyond). The runout may be estimated by volume-runout relations, where the volume refers to the material entering a given channel. MVO has collected data on the relationship between flow volumes and run-out distances. The empirical relationship is in broad agreement with theoretical expectations that the run-out distance increases as a cube of the volume of the avalanche. However local topography can have an important influence on run-out distance.

In the Belham valley, which passes between Cork Hill and Salem, small mobile pyroclastic flows have managed to run 5.6 km down valley. The prospects for larger flows in this drainage are increased by the filling of Mosquito Ghaut with debris. Should north flank collapses of the scale of the reference event occur, the possibility exists for portions of this debris, on the order of 5 x 106 m3, to enter the Belham valley. In that event, runouts of around 7 km are possible, suggesting that the flows could approach the sea. An event 3 or 10 times larger could achieve the same runout more readily, e.g. with a lesser fraction of material partitioned to Belham valley. With larger volumes of material passing through the same channel, the larger are the affected areas adjacent to the channel, and the possibility exists for scorching the region around Old Towne, particularly when winds are strong.

The likelihood of pyroclastic flows generated by gravity collapse entering area 3 by direct run-up over Centre Hills (elevation ~700 m) is practically nil. The mobility is established by empirical data on the relationship between ratio of fall height H to travel distance L, and volume of deposits. For events 10 to 30 times larger than the reference event, comprising complete dome failure or edifice collapse, we use the data for Quaternary volcanic avalanches between 0.1 and 1 cubic kilometre in volume; for these data, the H/L ratios range from 0.09 to 0.18 (average 0.13), and the minimum value was used. These relations apply to dome-related pyroclastic flows. The same conclusions apply to the dilute ash currents (ash-cloud surges) that commonly are associated with pyroclastic flows, and whose travel distance may be safely approximated by a factor (say 1.1 to 1.2) multiplied by the pyroclastic flow runout distance.

Tidal wave hazards. The larger collapses may enter the sea, and the conditional probability for edifice collapse is probably greatest towards the northeast, where the last event of this type is thought to have occurred about 4000 years ago. Such an event would not by itself affect populated areas, but it could produce a tidal wave. Such a tidal wave has been modelled, assuming an avalanche with a front height of 50 m and an impact speed of 50 m/s, using numbers drawn from Mount St. Helens avalanche. Resulting waves are 15 m high within the Tar River area, but less than a few meters high more than 5 km from the source. The harbour area at Little Bay is unaffected.

Eruptions triggered by dome collapse. Dome collapse depressurises magma contained in deeper parts of the dome and the conduit, and this rapid loss of pressure can result in an explosion. The explosions can be of two types, vertical or lateral (with a large horizontal component). The vertical explosions produce ballistic, fallout, or fountain-collapse surge hazards; these are discussed in other sections. The lateral explosion, or lateral blast, is discussed here.

Laterally directed pyroclastic surges are produced when a pressurised magma body is exposed by a slope failure. Only two examples of these have been well-documented in historic times, with blasts having been recorded extending up to about 30 km from the volcano, over an arc of 90 to 180 degrees. The above range applies to Mount St. Helens and Bezymianny volcanoes, where the volume of pressurised magma involved in the eruptions was about 0.1 km3, stored high within the edifice. This size is of the same order as the entire dome of the Soufriere Hills volcano at present. Lateral blasts at low-volume volcanoes are substantially smaller.

The likelihood of such an event is established by the probability for deep slope failure, with conditional probabilities on such a failure triggering a lateral blast, and the blast facing the hazard zones (i.e. facing north or northwest).

E.6 Analysis of surge hazard

There has already been some discussion in E.3 and E.5 about the chances of an event that generates a pyroclastic surge. Here the likely run-out distance is discussed if such an event happens.

Pyroclastic surges are produced by a number of mechanisms. The analysis here first considers surges produced by the collapse of vertical eruption fountains, because this process produces that largest and potentially most extensive surges. Surges produced in this fashion move out radially from the volcano. If the volcano itself is symmetrical and the surrounding terrain is relatively flat, the surge will go about the same distance in all directions. However if the volcano is asymmetrical, or the local terrain is irregular, the surge will go farther in some directions than others. Fountain-collapse surges are infrequent, short lived, and among the most lethal phenomena associated with explosive volcanism. Consequently little raw data are available to help constrain mathematical models. The best known examples are the 1951 eruption of Mount Lamington, Papua New Guinea, the 1982 eruption of El Chichon, Mexico, and the 1991 eruption of Mount Pinatubo, Philippines. All these affected roughly circular areas to distances of 6 to 14 km and overwhelmed topographic barriers several hundred meters high; the Lamington and El Chichon surges resulted in substantial loss of life.

To analyse this potential hazard, we have relied heavily on a recent model by Bursik and Woods (Bursik, M.I., and Woods, A.W., 1996. The dynamics and thermodynamics of large ash flows. Bull. Volcanol., Vol. 58, pp. 175-193). This is a two-dimensional model that assumes a radially-symmetric volcano surrounded by level terrain-assumptions that are unrealistic for Soufriere Hills volcano. At the request of MVO, one of the authors of this paper (A.W. Woods) modified the basic model sufficiently to address a narrow question; would a surge of a given magnitude overtop the Centre Hills? The modified model considers the behaviour as the surge moves outward from the volcano to a distance of 2 km (Farrell's Yard), then moves up an inclined surface (Centre Hills). Model runout estimates have a substantial uncertainty, perhaps ±30%.

Scenario probabilities are calculated as: (Prob. of an eruption of given intensity) x (Prob. of a radial surge) x (Prob. that runout will be mid, high, or low range). The conditional probability of a radial surge is estimated as 0.5 in each case. This rather high value was chosen partly because of compelling petrologic evidence that the magma has a rather low volatile content and high crystal content; both factors will favour the formation of eruption fountains. The high value was also chosen to include a margin of safety.

We used the unmodified Bursik and Woods model to consider an eruption 3 times more intense (or 7.5 times more voluminous) than the largest eruption that has occurred to date during the current eruptive episode. The model predicts that a radial surge produced by "3 times" eruption would move about 3.5 km to the north. With a model uncertainty of ±30%, the runout to the north is likely to be within 2.5 to 4.6 km from the vent. The lower values are more likely than the higher values. We estimate that the conditional probability of the 2.5 km runout (3x - 30%) is 0.9, the conditional probability of 3.5 km runout (3x) is 0.5, and the conditional probability of a 4.6 km runout (3x + 30%) is 0.1. Such an eruption would thus not overwhelm the Centre Hills. The terrain slopes downhill to both the east and west so runout distances in those directions would be somewhat greater. To estimate the runout in these directions, we used a simple flow model called an energy cone (Hsu, K.J., 1975. Catastrophic debris streams (sturzstroms) generated by rockfalls. Bull. Geol. Soc. Am., Vol. 86, pp. 129-140). This model predicts that a flow will affect the area inside a cone whose apex is directly over the volcano and whose sides are projected onto the terrain around the volcano. We constructed cones by placing their apexes just over the top of the dome (at 1,000 m a.s.l.) and chose cone angles that passed through northern runout distance estimates from the Bursik and Woods model. The resulting cones predict that, at the low end of the range (3x - 30%), the surge would not get out of zone 6. A mid-range (3x) surge would affect the southeast side of zone 5. At the high end of the range (3x + 30%), the surge would affect all of zone 5, the inhabited part of population zone 4, and both the eastern and western margins of zone 3. The total estimated probabilities over the next six months are: 1 in 40 (low-range runout), 1 in 70 (mid-range runout) and 1 in 300 (high-range runout).

The same procedure was used to consider an eruption 10 times more intense (or 25 times more voluminous) than the largest eruption that occurred thus far in the current eruptive episode. In this scenario, the Woods model predicts that a radial surge produced by a "10 times" eruption would move about 4.5 km to the north. With ±30% uncertainty, the runout to the north is likely to be within 3.2 to 5.9 km from the vent. Energy cones apexes were placed 1200 m a.s.l. The energy cones predict that, at the low end of the range (10x - 30%), the surge would affect the eastern part of population zone 5. A mid-range (10x) surge would affect all of population zone 5, the southern part of zone 4, and perhaps the extreme eastern part of zone 3. At the high end of the runout range (10x + 30%), the surge would marginally overtop the Centre Hills; it would affect all of zones 5, 4, and 3, and the southeast and southwest margins of zone 2. Note that predictions of areas to be affected are considered to be less trustworthy for scenarios in which the Centre Hills is overtopped. The total estimated probabilities over the next six months are: and 1 in 1,000 (low-range runout), 1 in 2,000 (mid-range runout) and 1 in 10,000 (high-range runout).

Finally, the same procedure was used to consider an eruption 30 times more intense (or 75 times more voluminous) than the largest eruption that occurred thus far in the current eruptive episode. In this scenario, the Woods model predicts that a radial surge produced by a "30 times" eruption would move about 6.8 km to the north. With ±30% uncertainty, the runout to the north is likely to be within 4.7 to 8.8 km from the vent. Energy cone apexes were placed above the vent at 1600 m a.s.l., because the eruption fountains in the Mount Lamington and Pinatubo examples are known to have been at least this high. The energy cones predict that, at the low end of the runout range (30x - 30%), the surge would not overwhelm the Centre Hills; it would affect all of population zone 5, the southeast margin of zone 4, and the extreme eastern margin of zone 3. A surge of mid-range runout (30x) would overwhelm the Centre Hills; it would affect all of zones 5, 4, and 3, as well as the southwest and southeast margins of zone 2. At the high end of the runout range (30x + 30%), the surge would overwhelm the Centre Hills; it would affect all of zones 5, 4, 3, and 2, as well as the southeast margin of zone 1. The total estimated probabilities over the next six months are: 1 in 4,000 (low-range runout), 1 in 8,000 (mid-range runout) and 1 in 40,000 (high-range runout).

For the directed surge the flux is focused on a sector so the effective flux is some factor greater than an equivalent flux for a radially symmetric flux within the same sector. Here we take a factor of 3. Following this reasoning, a flux of 8 x 107 kg/s (the "10 times" event) has a magnified effective flux of 2 x 108 kg/s, resulting in an enhanced runout (from 3.5 km to about 6 km, thus just beyond the wall of Centre Hills). The above is based on the Woods model, using a 10 degree average uphill slope from Farrell's Yard to the Centre Hills. In fact, the terrain is organised in a much more complex way than the model, so that for a wall of about 30 degrees, sedimentation is likely and the flow may become buoyant as it strikes the topographic barrier. For the "100 times" event, modelling suggests a runout of about 6.5 km on a 10 degree uphill slope. In this case the result appears insensitive to flux variation, but there is the suggestion that part of the flow could filter through low places along the Centre Hills ridge and then flow down the various drainages into northern Montserrat (areas 1 and 2). The least favourable direction is toward the northwest, where the flow could move toward Old Towne and then spread northward along the coast. Thus directional pyroclastic surges which can be conceived might impact the northern areas of the island. However, the probabilities associated with these scenarios are very low (estimated at 1 in 4,000 to 1 in 20,000).

Montserrat Volcano Observatory