Soufriere Hills Volcano

Soufriere Hills volcano is a dominantly andesitic structure located on the small Caribbean island of Montserrat [MAP] which is located in the north section of the Lesser Antilles Volcanic Arc. The summit of the volcano had a pre-eruption height of 915m and is composed of the remains of a series of lava domes emplaced by previous eruptions over the last 300,000 years (Zellmer et al., 2003. J. Petrology 44(8), p.1349-1374). The base of the volcanic complex is nearly 1km below sea level and has a diameter of over 25km. No historical activity had been documented, apart from the 17C eruption which emplaced Castle Peak lava dome. Increased volcanoseismic activity damaged buildings in the 1890s, 1930s and 1960s (Kokelaar, 2002. Geol. Soc. London Memoirs 21, p.1-43). Renewed seismic activity in 1992 and then from late 1994 onwards marked the onset of the current unrest. Since 1995, Soufriere Hills volcano has been in a phase of nearly continuous activity characterized by episodes of dome building, rockfalls, pyroclastic flows (PFs) and occasional major dome collapses and explosive events. Activity has forced the evacuation of the south part of Montserrat including the capital Plymouth, much of which is today destroyed and buried in lahar deposits. The sequence of eruptive events is summarized below. After this, a detailed scientific analysis of the eruption as a whole and of specific notable events follows.

Soufriere Hills Volcano in March 2006. The active lava dome is growing in the crater left by collapse of the previous dome.	West flank of Soufriere Hills volcano with extensive lahar flow field. The remains of Plymouth are just off the right hand side of the picture.

Close-up of Soufriere Hills lava dome in March 2006.	Nighttime view revealing incandescence of same dome as picture on left.

On July 18 1995, ash and steam venting was first observed. Activity was largely confined to the NW flank of Castle Peak lava dome which was formed in the 17th century in "English crater" (a 1km wide crater left by a massive collapse event during an eruption about 4000 years ago (Roobol and Smith, 1998. Geophys. Res. Lett. 25, p.3393-3396). Largely phreatic activity continued, reaching a first climax in a large phreatic explosion on 21. August which threw ash over much of Plymouth and initiated a first evacuation. Signs of dome growth were observed in the NW sector of the dome towards the end of September 1995.

Pyroclastic flows were first observed on March 29, 1996, in Tar River Valley (TRV). On the 12th of May, PFs reached the sea and this happened repeatedly in the following months. By July, extrusion rates reached 10 cubic meters / sec and repeated small dome collapses accompanied by large pyroclastic flows were observed. On 17 September 1996, a series of dome collapses was followed by a massive magmatic explosive eruption with a 13km high ash column. Over 500000 Tonnes of ash fell on Montserrat. Dome growth restarted in December.

On March 30, 1997, PFs flowed southwards into White River valley, destroying the touristic sites of Galways Soufriere and Great Alps waterfall. Increasing PF activity to the N and NE culminated in PFs reaching to within 50m of Blackburne Airport during a partial dome collapse on the 25 June 1997. These flows also destroyed the settlements of Streathhams, Riley, Harris, Bramble, Bethel, Spanish Point, Trants and Farms. Although these had been officially evacuated, 19 fatalities were recorded. On 1 July 1997, a PF reached the Catholic Church in central Plymouth. Further flows destroyed parts of Plymouth on the 3rd and 4th. A major PF flowed through the center of Plymouth to the port on the 3rd of August. From the 4th to 12th August, numerous vulcanian explosive events resulted in ash clouds up to 13km high. Column collapses of the explosion clouds were observed with associated pumiceous PFs. On 21 September, a major dome collapse resulted in PFs which reached the sea at several points to the E and NE of the volcano and set fire to the terminal building at Blackburne airport. These led to the complete destruction of Tuitts, Bethel and Spanish Point. Further about 10-hourly vulcanian explosions occurred between 22 September and 21 October with column heights of over 10km. On 4 November 1997, a southerly oriented dome collapse sent PFs down White River valley to the sea. Smaller PFs had reached the sea at this point in the previous month. Destruction of many settlements at the SW side of the volcano occurred on 26 December 1997, when a highly energetic PF surge probably associated with a lateral blast blew many houses off their foundations in the St Patricks and Morris settlements near the White River valley and set fire to many houses nearby.

Pyroclastic Flow emerging from cloud cover and travelling down Tar River Valley. Viewed from old airport control tower in January 2006.	Small pyroclastic flow travelling down flank of Soufriere Hills lava dome, March 2006

1998 witnessed a period of dome degradation with only sporadic minor activity. Occasional collapses of the dome continued to generate PFs. Passage of Hurricane George in September caused massive lahars which swept huge amounts of loose volcanic debris down the flanks of the volcano, burying parts of Plymouth and other locations in up to several meter thick deposits. Further periods of heavy rainfall in November and December had a similar effect.

Plymouth Courthouse buried in ash and lahar deposits, March 2006.	Ruined buildings in central part of Plymouth which has been destroyed by pyroclastic flows and lahars.	View from St Georges Hill over centre of Plymouth in March 2006. The centre of town (top right near pier) is buried in up to 20m of lahar deposits.

Remains of the former Arrows Manstore buried in lahar deposits, Plymouth, January 2006. The store, like many others, has been re-established on the safe N side of the island.	Remains of Catholic Church Plymouth. The church was set on fire by a pyroclastic flow and then gradually buried in lahar deposits. January 2006.	West facade of Government Headquarters, Plymouth. The building is partially buried in lahar deposits. January 2006.

Huge boulders deposited by lahars in a residential area, Plymouth. January 2006.	Remains of house in Trials estate set on fire by Boxing Day 1997 lateral blast. Photo: January 2006.

Dome degradation continued into 1999, with a large collapse event throwing ash up to 12km on the 20th of July. At the end of November, renewed dome growth was observed.

By February 2000, small PFs were reaching the sea down Tar River valley as the new dome gained in stature. A first major collapse of the new dome occurred on 20 March 2000 during a period of heavy rainfall, with much of the dome making its way down Tar River valley. This was followed by vulcanian explosions and inevitable lahars. Dome growth resumed shortly afterwards.

Continued growth in 2001, resulted in the volume of the dome reaching a record 162 Million cubic meters by the end of July. On the 29th of July, during a period of heavy rainfall, 45 million cubic meters of the East flank of the dome collapsed over a period of 8 hours, resulting in near-continuous PFs down Tar River valley into the sea. A smaller collapse of the N side of the dome occurred on 4th October with several PFs again reaching the sea. On 14 October, outer parts of the dome on the SE gradually collapse, producing continuous low energy PFs down Tar River valley into the sea over a 6 hour period. On 28 December a small collapse of the NE of the dome produced PFs lasting for over an hour and flowing into the sea.

Piano coated in ash, Plymouth.	Pipettes coated in Ash, Montserrat Technical College, Plymouth.	Bottle coated in Ash, General Store, Church Road, Plymouth.

On January 12, 2002, an energetic flow down the Tar River valley was accompanied by a 2-3km ash column and an over 1km high steam column at the point of sea entry. In February, PFs continued to reach the sea and an extruded spine on the top of the dome was measured reaching up to 1080m (90m above the rest of the dome). Heightened PF activity was observed in the following months. On September 29, a minor dome collapse sent PFs into the sea near Spanish point. On the 8th December a further larger collapse (5 Million cubic meters) sent energetic PFs to the same area and filled much of White Ghaut.

Dome growth and PF activity continued into 2003, culminating in a major collapse event on 12/13 July. The collapse was the largest to date, involving 210 million cubic meters of dome material. Ash clouds reached a height of 15km and massive PFs reached the sea down Tar River valley resulting in secondary phreatomagmatic explosions which generated hot base surges that spread inland along the coast up to the Spanish Point area, burning vegetation in the area (Edmonds and Heard, 2005. Geology 33(4), p.245-248). Much of the Island was coated in a thick layer of ash. This was followed by 3 vulcanian explosions on 13, 14 and 15 July, throwing ash and lithics up to a height of 12km. After a period of intense ash venting on 1 August, dome growth ceased and activity remained low for the rest of the year.

The only notable event in 2004 was a collapse event on March 3rd. Ash clouds reached an altitude of 7km and PFs reached the sea down Tar River Valley.

View from Jack Boy Hill over eastern Montserrat. The location of Blackburne Airport is indicated. The dome is just off the top right of the picture.	Remains of Blackburne Airport viewed from runway which is covered in lahar deposits. January 2006.	Destroyed building in Spanish Point area south of Blackburne Airport. March 2006.

The early part of 2005 remained uneventful until increased seismic activity in May and June, accompanied by some ash venting, was followed by two explosive events on 28 June and 3 July, the first of which producing PFs into the sea at Tar River, and both producing about 6km high ash clouds. On August 5, signs of renewed dome growth were visible. Minor PF activity occurred in the following months.

The first major collapse of the new dome occurred on 20 May 2006. Most of the 90 million cubic meters of the dome collapsed over a period of under 3 hours. Seismic activity attributable to rockfalls and PFs increased in the early hours of the morning, and by 6:45 PFs were flowing down Tar River valley to the sea. At 7:40, a huge explosion threw an ash cloud to a height of 17km and large ballistics rained down on the flanks of the volcano. Massive PFs flowed to the sea generating a base surge which spread north along the coast and set vegetation alight as far as the Spanish Point area (as in July 2003). Vigorous ash venting continued for much of the day and rainfall caused massive mudflows down all sides of the volcano. By 23 May 2006, a new lava dome could be observed in the crater. On June 06, a small collapse of this dome occurred, once again sending PFs to the sea via Tar River valley. On the 29th August, PFs reached the sea and ash clouds reached an altitude of 10km. From 31 August to 10 September, several episodes of vigorous ash venting occurred from a vent located between the dome and Gages wall. Small PFs from N to E of dome were observed in following months as activity started to shift to the N.

On 08 Jan 2007, explosive activity accompanied by a 10km high ash cloud was reported. Importantly, a large PF was able to flow NW along the Belham River valley. This PF had a run-out of 5km, a new record for this side of the volcano. Numerous smaller PFs were observed to the N of the dome. During April 2007, the dome which had reached the huge volume of 208 million cubic meters ceased growing.

On 20 July 2008, a significant increase in volcano-tectonic tremors was detected at Soufriere Hills. Ash venting and small PFs were observed during the following days. At 23:27 on 28 July a notable eruptive event started with little warning, culminating in a series of explosions, the most powerful being recorded at 23:38. The resulting ash column reached an altitude of over 10km and PFs flowed down Tar River Valley, White River Valley and most notably into Plymouth and Lee´s Yard. Whilst the TRV flow mainly consisted of older material resulting from partial dome collapse, the other flows were pumice-rich suggesting that they resulted from partial collapse of the eruption column. Fist-sized pumice fragments also fell in inhabited areas NW of the volcano. The PF heading into Plymouth split into 2 lobes which reached the old police station and Pentecostal church, respectively. Several buildings (including the Pentecostal church) were set alight by the flows. This activity marked the beginning of a further phase of low-level lava extrusion focussed in the area of the explosion crater.

An explosive eruption occurred on 02 Dec 2008. This scattered large ballistics up to 1km from the dome and was accompanied by large pyroclastic flows on the W flank of the volcano and an ash cloud reaching over 10km in height. The PFs inundated the Lee's Yard area and flowed through Plymouth into the sea, setting fire to several buildings on the way. The PF deposits were mainly of older material, suggesting that they had resulted from partial collapse of the dome. Associated surges burnt vegetation to near the top of St. Georges Hill. Further explosions occurred on Dec 3, 4 and 5. Activity at the dome increased in mid-december with extrusion and intermittent ash venting from multiple vents occurring on the SW side of the dome. Several medium-sized PFs were observed. Signs of extrusion on N and W flanks were additionally reported a few days later and activity continued to increase until the end of the month with medium-sized PFs entering Tyers Ghaut, Tar River Valley and Plymouth. On 03 Jan 2009, following a period of increased seismic activity, two powerful explosions resulted in plume heights of over 10km. Both were associated with column collapses which caused pumiceous PFs that entered Plymouth. Activity dropped off rapidly after these events and no signs of extrusion or significant seismic activity have been reported in the following weeks. The dome nevertheless continues to pose a significant threat to surrounding areas since it is very large and may become unstable at very short notice (last update 01.03.09).

Nighttime view of Soufriere Hills lava dome with Tar River Valley in the foreground, March 2006.	Early morning view of pyroclastic flow deposits in Tar River Valley, March 2006.

Scientific Analysis of the Eruption of Soufriere Hills Volcano

Soufriere Hills has been the subject of significant multidisciplinary volcanological research in recent years, making it one of the most studied volcanoes in the world today.

Soufriere Hills is composed of at least 5 andesitic lava domes flanked by pyroclastic deposits (Murphy et al., 2000. J. Petrology 41(1), p21-42). The lava erupted during the current eruption has a similar composition to that of the 17C castle peak dome and to all other lavas erupted at the volcano in the past 18000 years. The magma chamber from which these lavas were derived is thought to lie at a depth of 5-7 km and contain a large body of highly crystalline andesitic magma at a temperature of approx. 860'C. The magma is andesitic with a silicate content of about 60%. It contains high levels of fine-grained crystals suspended in a "groundmass" consisting of even finer crystals (microlites with below 100 micrometer diamm.) in a high silica (77%) rhyolite glass. However, small basaltic inclusions are often found within the andesitic lava and are thought to be related to the trigger mechanism of the present and previous eruptions. It is considered that hotter mafic (basaltic) magma sometimes rises from greater depth into the bottom of the magma chamber, leading to heating and pressurization thereof, ultimately resulting in eruption. Seismic activity in 1992, 3 years before the start of the eruption, has been proposed to signify the onset of intrusion of basaltic magma into the magma chamber (Young et al., 1998. Geophys. Res. Lett. 25, p.3389-3392).

The gas plume from Soufriere Hills consists mainly of water vapour, yet also of significant amounts of sulphur dioxide (SO2) and halogens, particularly in the form of hydrochloric acid (HCl) and HBr. The level of each of these gases fluctuates and can provide indications as to the status of the volcano. The reported long term average for SO2 emission lies at about 500 Tonnes/day, although Rodriguez et al., 2008 (J. Volc. Geotherm. Res. 173, p.135-147) argue that levels are consistently underestimated several-fold. During periods of high-level extrusion, several thousand T of HCl may be degassed per day and levels of water vapour from 20000 T up to 1 Million T per day have been estimated (Young et al., 1998).

It is thought that ongoing high level sulphur dioxide emission from the volcano indicates that fresh lava continues to enter the magma chamber from a deeper source, since the andesitic magma body is thought incapable of producing such high levels of this gas. Levels of sulphur dioxide measured at the surface are influenced by both magma degassing at depth and in the conduit, but also by the permeability of the conduit and dome. Gradually increasing rates of sulphur dioxide emission during the early phase of the eruption from 1995-1998 broadly correlated to steadily increasing levels of extrusion and magma supply. Gases reach the surface within the extruded lava or through cracks and cavities in the volcanic edifice which can be formed by pressure or thermal cracking. From Mar. 98 to Nov. 99, no lava extrusion was reported and SO2 emission decreased. It is thought that dome permeability was reduced during this period as silica minerals such as quartz were hydrothermally deposited in cavities and fractures within the dome as it cooled. This would have trapped a significant amount of SO2 and can partially account for inflation of the volcano and the release of significant amounts of SO2 during partial dome collapse events in this period (Edmonds et al., 2003. J. Volc. Geotherm. Res. 124, p.23-43). In the 1999-2003 period, it appears that the dome was relatively permeable since overall levels of degassing were similar, but dome collapses were associated with less SO2 release and pyroclastic flows were observed to contain smaller surge components (possibly due to less SO2 being entrapped in the flow material).

Levels of HCl are correlated to the extrusion rate, since this gas is formed at shallow levels in the rising magma. Levels at Soufriere Hills are thought to be high due to influx of sea-water into the system. The HCl vapour was probably largely responsible for damage to vegetation in the vicinity of the crater during the early phase of the eruption. HBr is also emitted and is associated with depletion of the ozone layer.

The impact of volcanic gases on the inhabitants of Montserrat is generally minimal, since the plume tends to blow in a W to NW direction over uninhabited or sparsely inhabited areas and is most dense well above ground level. The levels of microlite silica crystals in the air are a greater concern, as long-term exposure to these can cause the severe lung condition silicosis or other respiratory disorders. All eruptions at Soufriere Hills tend to release microlite silicates which are often remobilized by the wind. It has been found that levels of microlites are particularly high in ash clouds resulting from pyroclastic flows since when lava entrained in PFs is ground and crushed during motion of the PF, the microlite-containing groundmass is more readily broken down than the larger crystals within it and thus makes a higher relative contribution to the fine ash in the plume rising from the PF (Horwell et al., 2001. J. Volc. Geotherm. Res. 109, p.247-262).

The eruption of Soufriere Hills initially involved a period of high lava output with rapid dome emplacement and relatively frequent small or moderate volume dome collapses. Activity in recent years (2003-2006) has been associated with slower dome growth and less frequent but higher volume collapses. It appears that gradual growth correlates with more voluminous domes. Interestingly, it also appears that extrusion rates of magma partially dictate dome morphology. At low extrusion rates, the magma is able to crystallize further before reaching the top of the conduit. This magma is more viscous and growth is exogenous (at the surface) and involves the extrusion of spines and whaleback features. At higher extrusion rates, the magma is less viscous and growth of the dome may be endogenous (causing gradual swelling without lava reaching the surface) or extrusion may occur at the surface without the formation of any prominent features (Watts et al., 2002. Geol. Soc. London Memoirs 21, p.115-152). Similar observations have been made at Unzen volcano (Kaneko et al., 2002. J. Volc. Geotherm. Res. 116, p.151-160).

It is also interesting to note the effusion is not constant even over shorter time-scales. For example, in the month before the 25.06.07 dome collapse, Soufriere Hills showed a cyclical inflation / deflation of 18-25 microradians over periods of 8-14 hours. During the inflation phase, an increase in hybrid earthquakes can be detected. The onset of deflation is accompanied by peak degassing, peak magma flow (resulting in increased rockfall activity) and a rapid decrease in hybrid events (Watson et al., 2000. J. Volc. Geotherm. Res. 98, p.117-126). Hence it appears that the magma is extruded in regular pulses. Various models have been proposed for the underlying mechanism. Green and Neuberg (J. Volc. Geotherm. Res. 153, p.51-63 (2006)) propose the following mechanism. The cycle starts as degassing causes increased microcrystallization and leads to the formation of a relatively dense plug at the top of the conduit. As pressure rises and the edifice inflates, fracturing of the plug starts to occur, causing hybrid earthquake signals. Due to the increasing pressure correlating to the inflation, magma flow (i.e. extrusion of the plug at the top of the conduit) gradually accelerates. Fracturing continually increases allowing gas release through the plug and a drop in the rate of inflation. Eventually the extent of fracturing allows more gas to escape than is being formed under the plug. This causes deflation of the edifice and as the pressure falls a drop in magma flow. The fractures then gradually seal and formation of the cycle starts again at the degassing and plug formation stage.

Ground deformations on a larger scale were observed using satellite-based radar interferometry and GPS during the 1998-2000 period. It was shown that when the dome was not growing, the western flank of the volcano gradually inflated by several cm/year (Wadge et al., 2006. J. Volc. Geotherm. Res. 152, p.157-173). Together with ongoing SO2 degassing, this indicated that magma infusion into the magma chamber continued even when no dome growth was occurring. Conversely, deflation could be detected after Nov. 99 when dome growth resumed, apparently allowing a net release of magma and gases from the chamber.

Detailed Analysis of Main Eruptive Events

Click on maps to enlarge. Eruptive events are shown superimposed on pre-eruption map. Detailed explanations follow below.

Montserrat Pre-1995	25 June 1997 Event	Destruction of Plymouth (3. Aug. 1997)

21 Sept. 1997 Event	26 Dec. 1997 Event	12-13 July 2003 Event

1. Fatal Pyroclastic Flow of 25 June 1997 [MAP]

In mid-May 1997 extrusion started to occur on the N flank of the dome from a subhorizontal shear lobe structure (Watts et al., 2002. Geol. Soc. London Memoirs 21, p.115-152). Dome growth assumed a clear cyclical inflation / deflation pattern with rockfall / PF activity largely during the deflationary phase (as explained in previous section). Cycle lengths ranged from 12-16 hours from the 5th to 14th of June with a measured amplitude of 16-18 microradians. Deflation phases were accompanied by increased rockfall and PF activity as new magma reached the surface of the dome.The small moat between the dome and the N wall of English Crater was quickly filled by rockfall debris, allowing dome material to spill into Tuitts Ghaut by May 19 and into Mosquito Ghaut as from June 5. Pyroclastic flows from partial dome collapses were observed in Mosquito Ghaut in mid-June with run-out distances of 1.6km and 4km on June 16 and 17, respectively. The 17 June event was preceeded by a particularly intense dome inflation. Weak cyclic activity resumed thereafter, rapidly increasing on the 22nd with cycles only taking 8-12 hours and involving a higher amplitude of up to 30microrad. Small pyroclastic flows accompanied each deflation cycle.

On the morning of the 25th of June, a deflation cycle commenced at around 6:10 in the morning. For the next hour, PF activity was almost continuous in Mosquito Ghaut with a maximum run-out distance of about 1.5km. Activity then dropped slightly as the dome entered its inflationary phase at around 9:00. Seismic activity was high on the 24th and 25th with almost continuous tremor marking the ends of the inflationary phases. This was again the case at about 12.45 on the 25th with continuous tremor accompanying a sharp deflation of the dome (less than 7 hours after the same point in the previous cycle). At 12:55, strong rockfall-associated seismic signals were detected, marking the onset of the fatal partial collapse of the lava dome.

The dome collapse resulted in PFs sweeping to the N down Mosquito Ghaut and into Paradise River which directed them NE towards Bramble and Bethel and thereafter N along Pea Ghaut to Farm and Trant's estates, and finally E along Farm River where they finally came to a standstill shortly before the airport. The flows spilled out of the river bed in various places, burying a number of settlements. Additionally, whilst the flows were channelled in Mosquito Ghaut, surges escaped from the Ghaut on the outside of several corners and headed N and NW until reaching the bases of Gun Hill and Windy Hill. Further, ash sedimenting from these surges flowed along the ground into Dyers / Belham River to the W, forming a secondary pyroclastic flow which extended for several km along the river bed (Druitt et al., 2002 Geol. Soc. Memoirs 21, p.263-279). The event lead to 19 fatalities, 7 in the Streatham area, 1 at Harris, and the remainder probably at Farms which was buried in flow deposits.

The pyroclastic flows can be divided into three main pulses based on seismic records, study of the deposits and eyewitness accounts. The event has been analysed in great detail in Loughlin et al., 2002 (Geol. Soc. London Memoirs 21, p.191-209). The first flow was confined to Mosquito Ghaut and Paradise River, reaching as far as the the beginning of Pea Ghaut near Bethel. The second pulse immediately followed the first, starting at about 13:00. This pulse was more powerful and of longer duration. The flow was again largely constrained by river valleys, and stopped just short of Bramble airport. However, associated surges were able to spill out of them, especially on the outside of rightward bends in the Mosquito Ghaut. These surges were reported to have reached the main road connecting the E and W sides of Montserrat. The third pulse was similarly powerful to the second, based on seismic records, and immediately followed it, starting at about 13:08.

The deposits from the earlier pulses laid the foundation for the destruction caused by the third pulse. These deposits raised the floor of the valleys and provided a relatively flat base on which the following flows could rapidly advance. As a consequence, the third flow was able to almost completely fill Mosquito Ghaut, and large surges were able to escape to the N and NW (notably in a rightward bend about 1.4 km from the crater rim), this time extending further than the previous surges and flowing through Streatham and as far as the base of Windy Hill. Further, the flow itself was able to escape from the river bed in several places, notably in the bend connecting Paradise River and Pea Ghaut (allowing it to flow into parts of Bethel) and then in several places along Pea Ghaut where the river bed was less deep than in the steeper-sided uphill sections and thus less able to restrain the flow. The settlements of Farm and Trant were almost completely buried as a result.

The third pulse reached speeds of over 20m/sec (compared to 15 and 16m/sec for pulses 1 and 2) in Mosquito Ghaut and Paradise River. This is thought to be not only due to movement over previous deposits but also due to the involvement of increasingly gas-rich material from deeper within the dome as the eruption progressed. The higher speed would have assisted the flow in escaping the river valleys since it would have increased the centrifugal force in the corners. Erosion was evident high on the outside of many bends that the flow had taken, and was almost continuous in the upper 2km of Mosquito Ghaut. This is partially a result of the massive blocks entrained in the flow, some of which had a diammeter as large as 5m even in the Bethel area. The flows were generally gas-rich and significant surges were formed almost immediately at the base of the dome. This is not surprising due to the large amount of recently extruded and thus gas-rich material involved. The convecting clouds rising from the flows rapidly reached an altitude of 10km and many areas downwind of the flows were shrouded in darkness for several minutes.

The dome collapse involved nearly 6 million cubic meters, of which about 14% is attributed to the 1st, and 43% to each of the 2nd and 3rd pulses. Run-out lengths were 4.8, 6.8 and 6.7km, respectively for the three pulses and the peak flow rate of the third flow in the upper reaches of Mosquito Ghaut has been estimated at about 70000 cubic meters / sec.

The victims died either by being engulfed in the pyroclastic flow itself, or by being caught in one of the surges that detached from it. The bodies of most victims caught in the actual pyroclastic flow (largely at Farms) were buried under several meter thick deposits and could not be recovered. Two bodies were found in pyroclastic flow deposits near Trants. These bodies had been dismembered as they were carried along by the flow. The victims in the Streatham area were unable to reach high ground (Windy Hill) quickly enough and were engulfed in the surge associated with the third flow pulse. Several bodies were found near the base of Windy Hill and these had severe skin burns and partially carbonized fingers. Hair and clothing had been largely burnt off (Loughlin et al., 2002. Geol. Soc. London Memoirs 21, p.211-230). The limbs were drawn close to the bodies, in the so-called "pugilistic attitude", as a result of muscle tissue coagulation which results from intense heat (Baxter et al., 1990. Bull. Volc. 52, p.532-544). This shows that even near the periphery of the surge, where these bodies were found, temperatures were extremely high. In the open, these victims had absolutely no chance of survival.

In general, humans caught even in the periphery in pyroclastic surges have little chance of survival unless they can reach a building which is able to restrict entry of the surge. The impact of pyroclastic flows and surges on humans is discussed in detail in Baxter et al.,1998 (Nat. Haz. 17, p.163-176). Pyroclastic surges are essentially hot currents of air and volcanic gases containing high particle (ash) concentrations. Even in surges with less extreme temperatures than the 25th of June one, humans will rapidly suffer skin burns, especially to unclothed areas, and will suffer airway obstruction and burning as a result of particle inhalation. Inhalation of large amounts of ash will cause lethal asphyxia even at low temperatures. Inhalation of hot ash, especially in combination with moisture, will additionally lead to rapid heat transfer to the lungs, causing severe damage. Several survivors from the periphery of the 03.06.91 surge at Unzen volcano developed laryngeal oedema within 25 min of exposure and all died later of acute respiratory distress syndrome. People caught in the most dynamic parts of surges will additionally be affected by the dynamic pressure exerted by the wave of heavy (particle-rich) air and may be mobilized by the flow. Fatality would be inevitable in such a case.

The impact of the June 25 event on buildings in the Streatham area has been assessed and compared to the impact of other events at Soufriere Hills (Baxter et al., 2005. Bull. Volcanol. 67, p.292-313). No obvious signs of damage from dynamic pressure were visible. However, most buildings along the main road and thus nearest to the dome were set on fire by the 400'C hot surge, leaving little but the gutted remains. When the surge reached flatter ground and then partially ascended Windy Hill it appears to have rapidly decelerated and cooled, which coincided with reduction in the level of impact on buildings. In a transitional area, windows and roofs were damaged and partially burnt on the side of the houses facing the volcano, whilst at the highest points on Windy Hill reached by the surge, the temperature was insufficient to set fire to wooden materials. In all areas affected by flows or surges it was noted that houses surrounded by significant deposits were almost all rapidly set alight. Most trees had been blown down in proximity to Mosquito Ghaut where the surge was clearly significantly more dynamic than at Streatham. The orientation of the tree-trunks provided a useful indicator of local flow direction.

It is noted that all the settlements hit by the event were in the officially evacuated exclusion zone. However, it is thought that about 80 people were in the area N, NE and E of the dome around lunchtime on the 25th of June. Whilst road blocks had been put in place to avoid unauthorized entry, those willing to avoid them could do so easily, and others, such as farmers who were supplying food for the shelters in the N of the Island were apparently often allowed to pass. The different motivations for entering the exclusion zone are well explained in Loughlin et al., 2002 (Geol. Soc. London Memoirs 21, p.211-230). Further, Kokelaar 2002 (Geol. Soc. London Memoirs 21, p.1-43) provides a detailed overview of the proportion of the population displaced at different times during the volcanic crisis. Critically, the economic situation of many people relocated to the N of the island was precarious and the temporary shelters that were provided were cramped (about 6000 people (over 80% of the population) had to be rehoused by mid-1997). Rental accommodation was scarce and too expensive for many. Consequently, a number of inhabitants were inofficially living in the exclusion zone or at least regularly visiting their houses there to seek respite from the cramped shelters. Others were tending property or retrieving belongings. Many were farming, either growing crops or tending livestock, especially those in the Streatham area. It should be noted that most of Montserrats farmland was in the exclusion zone. Whilst all had been made aware of the risks, previous flows had largely been topographically confined to river valleys, so an element of complacency had crept in. The fact that a small area of vegetation next to Mosquito Ghaut was already scorched by a surge on the 17th of June seemed to cause little alarm. Further, regular visitors to the zone had recognized the cyclical nature of the activity in the weeks preceeding the eruption and felt that they could predict it and thus avoid periods of heightened activity. Additionally, most thought they would be given enough warning to be able to reach higher ground, since it was expected that the extent of the flows would increase in little steps. However, the June 25 event affected a much larger area than previous events and some eyewitnesses did not even notice the approaching flows / surges until they had nearly reached them, since they were nearly silent. This gave many people insufficient time to run for higher ground or into buildings. A detailed summary of eyewitness accounts is found in Loughlin et al., 2002 (Geol. Soc. London Memoirs 21, p.211-230).

MVO Staff had noticed the significant increase in seismic activity on the morning preceeding the eruption and several had been deployed in the field to observe activity levels. Essential service personnel were advised to evacuate from Plymouth just 10 min before the onset of the eruption. Sirens were sounded in Plymouth and near the airport and a warning was broadcast on local radio. However, those deep in the exclusion zone couldnt hear either and others simply ignored them or reacted with little urgency.

Interestingly, many of the survivors suffered severe burns to their feet. These resulted from walking (often barefoot) over surge deposits. This was unavoidable in places, especially when survivors were forced to flee from burning houses. Even only cm-thick layers of ash deposited by the surges had temperatures of several hundred degrees shortly after emplacement. Deeper layers, especially in small ditches or near houses stayed hot for weeks and presented a hazard for rescue teams and scientists in the aftermath of the eruption. These layers are relatively fluid due to the fine-grained particles and one can sink quite deeply into them.

2. Destruction of Plymouth [MAP]

Little scientific attention has been attributed to the destruction of Plymouth since neither unusual, nor particularly powerful volcanic events were responsible. Nevertheless, a brief chronology of the destruction is warranted in view of the major loss of infrastructure involved. Plymouth, like many other settlements on Montserrat had been built on gently sloping prehistoric volcaniclastic deposits formed by PF and Lahar deposits (Roobol and Smith, 1998. Geophys. Res. Lett. 25, p.3393-3396). The vulnerability of the city had been noted in the Wadge and Isaacs report of 1987 (well before the eruption) which dealt with potential volcanic hazards on Montserrat. However, Plymouth had been estimated to be at risk once every 10000 years and the report received little attention. Consequently, core infrastructure was still being rebuilt in the city following the devastating caused by hurricane Hugo in 1989 when the eruption began (Kokelaar, 2002. Geol. Soc. London Memoirs 21, p.1-43).

Plymouth had remained relatively unscathed during the early stages of the eruption since it was shielded from the dome by the W flank of English crater and Gages Peak. However, ashfall regularly affected the city. On 21 August 1995, a dense cold pyroclastic density current descended into Plymouth following a phreatic explosion. This event ("Ash Monday") shrouded the city in darkness for several minutes and precipitated a first evacuation lasting for 2 weeks. The city was again evacuated in December 1995 as PF activity commenced on the E side of the volcano. The inhabitants returned again in January but were finally evacuated on 3-4 April 1996 when PF activity intensified (Kokelaar, 2002. Geol. Soc. London Memoirs 21, p.1-43). Plymouth Port continued to be used until 25 June 1997 as the city remained deserted but largely unscathed.

Between June 28 and 30, the rapidly growing dome overtopped Gages Wall to the extent that material started to spill into Fort Ghaut. PFs were observed descending the Ghaut every 8-12 hours (Cole et al., 2002. Geol. Soc. London Memoirs 21, p.231-262). These eventually reached to within about 400m of the mouth of Fort Ghaut which was located just south of Plymouth town center. The June flows were still largely constrained by the Ghaut, spilling out of it only in the uninhabited area near Gages Lower Soufriere. Material continued to spill into the upper reaches of the Ghaut during July as high level extrusion (5-10 cubic meters / sec) continued. Further buildings are damaged by the flows.

On the 3rd of August, a major collapse of the dome occurred involving about 8 million cubic meters of material. Much of this material formed energetic PFs which were channelled along Fort Ghaut into Plymouth. The small Ghaut was unable to constrain the flows which escaped from either side of it along much of its length . This resulted in destruction or setting alight of many buildings in Plymouth and its uphill suburbs as the flows progressed through town all the way to the harbour.

No PFs have reached the sea through Plymouth following this event, however the destruction of the city has continued due to inundation by lahars, occasional small PFs, and of course years of neglect, since the city has been uninhabited since April 1997. Major lahars resulted from the passage of Hurricane George on 20-21 Sept. 1998. Several hurricanes passed in 1999, resulting in progressive burial of lower-lying parts of Plymouth. The process continues unabated with the low-lying areas around the former path of Fort Ghaut being buried in over 10m of volcaniclastic deposits by 2006.

3. Vulcanian Eruption Series (1997)

Two series of Vulcanian explosions occurred at Soufriere Hills in 1997. The first 13 explosions were recorded from 4-12 August, following the 3. August dome collapse described in the section above. The remaining 75 explosions occurred between 22. Sept. and 21. October following the 21. Sept. dome collapse which is dealt with separately in the section below. The second series of explosions formed an about 300m wide crater in the remains of the dome at the location of the 1995 phreatic vent. Each series was triggered by a preceding dome collapse (Calder et al., 2002. Geol. Soc. London Memoir 21, p.173-190). Due to the regularity of the explosions, detailed video monitoring was possible allowing the exact sequence of events at the surface to be determined and compared with seismic and tilt data (Formenti et al., 2003. Bull. Volcanol. 65, p.587-605). Each explosion involved a high intensity phase lasting for about 10 minutes, followed by a 1-3 hour waning phase with low intensity and pulsatory discharge of ash and gases. The explosions commenced with jets of rocks and ash being dynamically propelled from the vent at speeds of 40 m/sec, rising to 140 m/sec within the following 10 seconds (speeds based on trajectories of large ballistics). The jets rapidly decelerated and reached heights of 300-650m above the vent before partially collapsing downwards, whilst larger rocks separated from the jets and landed in a wide radius around the dome. Blocks up to 1m in diameter were found nearly 2km from the vent (Druitt et al., 2002. Geol. Soc. London Memoir 21, p.281-306). The initial jets involved relatively degassed and thus dense material which had plugged the top of the conduct prior to the eruption. These were followed by eruption of highly pressurized gas-rich magma from further down the conduit as the fragmentation wave travelled down the conduit at speeds of between 30 and 70 m/sec. The climactic phase of the eruption ceased when the fragmentation wave reached magma which did not satisfy the conditions required for fragmentation (Spieler et al., 2004. Earth Planet Sci. Lett. 226(1-2), p.139-148).

Fountain collapse occurred primarily during the first 10-20 secs of the explosions, whilst buoyant plumes containing lighter material convected upwards, ultimately reaching heights of from 3-15 km. The most intense fountain collapse phase clearly coincided with the eruption of dense material at the beginning of the eruption. The resulting pyroclastic density currents spreading out radially in all directions at the base of the eruption column with an initial velocity of 30-60 m/sec. The pyroclastic surges reached distances of 1-2km with ash convecting upwards from them as they decelerated. In all but 2 explosions, pumiceous pyroclastic flows (of material from deeper in the conduit than the initial flows) emerged from under the surge clouds and flowed down several drainages at speeds of about 10 m/sec reaching distances of 3-6 km from the dome. It was estimated that each explosion involved 0.3 Million cubic meters of DRE (Dense Rock Equivalent), corresponding to the contents of the conduit up to a maximum depth of around 2km. About 65% of this material was estimated to have collapsed, forming pyroclastic flows or surges.

The regular occurrence of the explosions is explained by continuous input of magma into the conduit from the underlying magma chamber resulting in rapid reestablishment of pre-explosion conditions in the system, priming it for the following event. The radius of the conduit is estimated at 12-15m based on dimensions of extruded spines (Watts et al., 2002. Geol. Soc. London Memoir 21, p.115-152), although other authors use estimates of 20-30m. Following each explosion, magma gradually refilled the conduit within the next hours (initially rising a few cm per second). This magma eventually stagnated due to crystallization and subsequent degassing of magma near the top of the conduit. A burst in crystal nucleation occurring at pressures below 7 MPa which are encountered in the upper parts of the conduit is thought to be critical in this process (Clarke et al., 2007. J. Volc. Geotherm. Res. 161, p.261-274). The degassing magma formed a dense plug with an estimated thickness of 250-700m, and caused a build-up of pressure due to closed system degassing in the magma below. The onset of the next explosion relates to failure of the plug at the top of the conduit.

4. Major Dome Collapse of September 21, 1997 [MAP]

The dome collapse of Sept. 21 1997 occurred over a period of 20-30 min at around 4 in the morning and involved approx. 11 million cubic meters DRE (Calder et al., 2002. Geol. Soc. London Memoirs 21, p.173-190). The event occurred following a period of rapid dome growth and was preceded by hybrid swarms just like the previous collapse on June 25. The resulting PFs had the longest run-out distance recorded during the entire eruption, extending up to 6.9 km NE of the dome (Carn et al., 2004. J. Volc. Geotherm. Res. 131, p.241-264). The PFs were primarily channelled down Tuitt's Ghaut, but partially spilled over into nearby White's Ghaut at a distance of just over 1km from the dome. The flows continued along Tuitt's Ghaut until reaching the point where the Ghaut flowed into Paradise River near Bramble village. The river bed at this point had been completely filled by deposits from the 25. June event and was thus unable to contain the flows. This resulted in the flows spreading NE over a width of over 1km in places as they approached the coast. The remains of Bethel as well as the settlement of Spanish Point were completely buried in meter thick PF deposits. The flows inundated all of the areas of the coastal plane affected by the 25. June flows and further extending beyond these areas, especially to the N and NE. The main lobe spilled into the sea along about 500m of coastline from Spanish Point northwards, forming a small delta at the former site of Farm Bay. A further lobe proceded further northwards and partially flowed through the airport terminal building, setting it alight in the process. The flow in White's Ghaut also reached the coastline but was constrained by the river valley. Surges spilled out of each of the Ghauts and inundated much of the area between them and either side of White's Ghaut. Whilst the flow narrowly missed Tuitt's village, surge clouds having temperatures of at least 400'C set fire to all buildings therein and left a thin deposit of ash in the area (Baxter et al., 2005. Bull. Volc. 67, p.292-313) Whilst much property was destroyed, no fatalities were recorded since the whole area had been evacuated since Dec. 1995 and the deaths that occurred 3 months earlier had clearly put people off illegally entry into the area.

5. Boxing Day (26.12.1997) Lateral Blast [MAP]

On March 30, 1997 the first significant PF activity occurred to the S of the dome in the White River drainage. From September through November several PFs extended as far as the sea, with the largest events occurring on the 4th and 6th of November when sustained collapses mobilized 18 Million cubic meters of material, about a third of which reached the sea. The deposits from these events formed a small delta at the mouth of White River. Then, on 26 Dec. 1997 an extremely violent eruptive event lead to almost complete destruction of several villages on the SW flank of the volcano. The eruption is classified as a lateral blast (or blast-generated pyroclastic density current) and can be compared to similar larger events at Bezymiany (1956) and Mt St Helens (1980) volcanoes (Belousov et al., 2007. Bull. Volcanol. 69, p.701-740). The condition of dome and southern crater in the period preceding the blast is critical to explain the mechanism thereof. A year earlier, in Nov.1996, growth of the dome had switched from an exogenous (extrusion of lava at the surface) to endogenous style. The effect of the endogenous growth was a measurable swelling of the dome. An area of 300x800m of the dome near the S rim inflated by up to 100m during this endogenous growth phase. Growth subsequently switched to exogenous growth at other parts of the dome, before returning to the S sector in Nov. 1997. In the meantime, increasing instability of the hydrothermally weakened S crater wall (Galway's Wall) was noted, including displacement by about 1m, crack formation and detachment of slabs of rock from the steeper sections (Young et al., 2002. Geol. Soc. London Memoirs 21, p.349-363). A large volume of dome material was thus being supported by a critically weak crater wall.

Failure of the S crater wall occurred with little warning at 3:00 on Dec. 26. The resulting debris avalanche descended down White River valley as far as the coast, shortly followed by a series of extremely powerful (Pyroclastic Density Currents) PDCs which were directed SW and devastated a 70 degree wide area containing over 10 square km of land including the evacuated settlements of St. Patricks, Gingoes, Trials and Morris. Seismic data suggests that the dome fragmented in just over 15 min in three main pulses, the second of which was the largest. After this, three short vulcanian explosions were registered before activity waned (Sparks et al., 2002. Geol. Soc. London Memoirs 21, p.409-434). Each pulse mobilized huge volumes of material, since fragmentation occurred over a much wider surface area than during an eruption from a discrete vent. The crater wall was removed down to the level of Galway's Soufriere where the rocks were particularly weakened due to hydrothermal activity.

The collapse of the S crater wall would have exposed the side of the magma body in the dome. As with Mt St Helens and Bezymiany, the magma body was shallow and thus highly crystallized and partially degassed (thus making it more dense and prone to flowing down the flank of the volcano rather then convecting upwards). In all cases the magma body was at least partially restrained by an older volcanic edifice, which after removal allowed extremely rapid explosive fragmentation of the magma body in an at least partially lateral direction, after which much of the material flowed away from the "vent", initially partially channelled by the collapse scar. The lateral component is more important with respect to focussing the direction of the blast than for providing it with sustained velocity (i.e. a slightly inclined highly voluminous fountain may be sufficient as long as it transports large volumes of dense material onto an incline). Sustained velocity at distance from the vent is largely gravitational and a result of the density of the flow. The PDCs resulting from the first explosive pulse at Soufriere Hills may have had average frontal speeds of over 60m/sec and internal speeds of up to 80m/sec. Speeds of well over 100m/sec were probable near the vent and possible down the steep SW slope of the volcano during the second and most powerful pulse. Exact values cannot be determined and various analyses have given different values (see Belousov et al., 2007). The same uncertainty applies to estimates of the density of the currents with values of between 6 and 116kg /cubic meter being suggested. The latter value is only likely to have been reached in the St Patricks area.

The Soufriere Hills lateral blast deposits were unusual compared to those at St Helens or Bezymiany in that the deposits had only a negligible basal layer containing remnants of vegetation. The flows were extremely erosive (presumably as large dense rock fragments were mobilized by them) and removed the entire soil down to the bedrock in places. Nevertheless, little of this material is retained in the bottom layers of the deposit and it may have largely been carried into the sea. Otherwise the deposits show the usual density-grading with larger materials at the bottom and finer ones at the top. The layers were then coated by fine material from the plume. The eruption involved a debris avalanche with a volume of 50 Million cubic meters (mostly deposited on land), followed by the explosive mobilization of 30 Million cubic meters of dome material, of which 90% was deposited in the sea. The White River delta was extended up to 400m off the pre-eruption coastline and attained a width of 2km. Significant submarine deposits were detected extending over 3km offshore mostly in 3 pre-existing submarine valleys. It is assumed that much material was deposited even further offshore. In places it appeared that the PFs had eroded sediment from the sea-bed, resulting in areas with increased depth following the eruption.

The damage to buildings in the PDC-affected area has been illustrated in detail (Baxter et al., 2005. Bull. Volcanol. 67, p.292-313). The central axis of the PDCs extended through the settlement of St. Patricks. Here, all free-standing buildings (most of which were constructed with reinforced concrete) were razed to the ground, large boulders and trees had been dislodged and propelled through the air at high speed, acting as battering rams on anything that was able to withstand the overall dynamic pressure of the PDC (which was estimated at over 25kPa*). Moving out from the central axis to areas such as Reids Hill Estate or Gingoes, one observed houses which were still standing but had severe damage to external and internal walls. Roofs had been blown off and the entire contents of the houses destroyed. Damage due to large projectiles such as trees could still be observed in this area. It is noted that most projectiles were not overlain by any significant deposits, suggesting that the main blast was not followed by much further deposition in the area. Further from the PDC central axis, for example in Trials estate, windows had been blown out and the houses had been set on fire by the hot ash which was deposited therein. The PDC deposits were measured at around 300'C the following day and temperatures at this level and above are likely to have been encountered in much of the inundated area. At the outer edge of the area with PDC deposits, houses were mostly undamaged but trees were singed and PVC guttering (for example) was melted in places. Here the temperatures were presumably slightly lower due to increased mixing with air and also the PDC was less dense and dynamic. A PDCs density and speed can be correlated with increased heat transfer to objects in its path and also increased penetration of buildings outer shells, in turn correlating with an increased risk of fire.

The impact of blast-generated PDCs on buildings, had previously been observed during the 1902 eruption of Mount Pelee on the Caribbean island of Martinique which like Montserrat belongs to the Lesser Antilles Volcanic Arc (Fisher and Heiken 1982, J. Volc. Geotherm. Res. 13, p.339-371). This eruption cost over 30000 lives.

* Dynamic Pressure is measured in Pa (Pascals) and is determined according to the following equation: Dynamic pressure = 0.6 x V² (wherein 0.6 is a constant which represents the mass of air which is about 1 kg/cubic meter).

An 80km/h (22m/sec) wind exerts a dynamic pressure of 295 Pa (0.295 kPa), whereas a 500km/h (139m/sec) wind exerts a dynamic pressure of 11550Pa (11.5kPa) and a 750km/h (208m/sec) wind exerts 26kPa.

Taking a low estimate of 10kg /cubic meter density for the boxing day PDC and a speed of 80km/h one would get a dynamic pressure of 2.95kPa as the density is 10x higher than air and thus the constant becomes 6. Using a higher estimate of 100kg/ cubic meter one gets a dynamic pressure of 29.5 kPa which would be more than sufficient to account for the damage observed.

6. March 2000 / July 2001 Rainfall-Induced Dome Collapses

The dome collapse events of 20.03.2000 and 29-30.07.2001 (and possibly 03.07.1998) were unusual in that they were not preceded by significant seismic activity under the dome and were not associated with significant release of sulphur dioxide, suggesting that the dome was not highly pressurized at the time. Both events were however preceded by periods of unusually heavy rainfall. This correlation, which has been observed previously at other volcanoes, suggests that rainfall is capable of triggering dome collapse events.

Between Nov.1999 and Mar. 2000 a new lava dome with a volume of about 30 million cubic meters had been emplaced at Soufriere Hills. The collapse of this dome on 20.03.2000 is documented in detail in Carn et al., 2004 (J. Volc. Geotherm. Res. 31, p.241-264). Shortly after the onset of heavy rainfall, small PFs were observed in the TRV at relatively regular 5 min intervals. These flows were weak and not very convective, suggesting mobilization of relatively old degassed dome material. After about 2 hours (at about 18:00), the PFs became more vigorous, with several entering the sea. At 19:20 a very powerful PF entered the sea and the surge cloud reached 1.5 km off shore. This was followed by explosive ejection of incandescent material from the remnants of the dome, presumably as the conduit was exposed. The PF-generated and 10km high explosion-generated ash-clouds were the source of numerous lightning strikes. Large lahars swept down Belham River and carried boulders and other material onto the bridge. Most of the dome material was emplaced in the sea of the TRV delta, which was itself increased in size by 3 million cubic meters. The delta had deep channels in it, suggesting that the climactic PFs may have been unusually erosive, possibly due to the mobilization of large blocks from the dome. Regarding the triggering mechanism, Carn noted that for about 20min prior to the observation of the first weak PFs, seismic signals consistent with lahars in the TRV had been detected. It was thus suggested that initial erosion of the tallus at the base of the dome by lahars could have played a role in the collapse. Whatever the trigger mechanism, the collapse event increased in intensity over time as more and more gas-rich and hotter parts of the dome and eventually the conduit became exposed by sequential removal of the outer layers of the dome, resulting in the climactic explosion.

The 2001 event was essentially similar and removed a 45 million cubic meter dome which had began to grow directly after the Mar. 2000 collapse (Matthews et al., 2002. Geophys. Res. Lett. 29(13), 1644).

The mechanism by which rainfall-triggered dome collapses occur at Soufriere Hills has been modelled (Taron et al., 2007. J. Volc. Geotherm. Res. 160, p.195-209). It was concluded that during periods of heavy rainfall, water may penetrate to 100m and beyond into pre-existing fissures in the dome. The surface of the dome and top of the fissure need to be cooled sufficiently (e.g. by previous rainfall) for this to be possible. The weight (hydrostatic pressure) of the rainwater column forming in the fissure prevents degassing through the fissure, resulting in a small build-up of pressure underneath. Whilst the build-up of pressure is almost negligible compared to other pressures within the dome, it is thought that when a dome is already mechanically unstable due to oversteepening and internal fracturing, the small additional pressure may be sufficient to trigger the dome collapse (i.e. it acts as the proverbial straw that breaks the camel's back).

7. Major Dome Collapse of 12-13 July 2003 [MAP]

The July 2003 dome collapse was the largest historical event of this type at any volcano and involved mobilization of 210 million cubic meters of dome material over an 18 hour period, with 170 of these being mobilized in the most intense 160 minute period, and about 16 in the largest single event which lasted just over 2 minutes. The whole eruption has been documented in much detail (Herd et al., 2005. J. Volc. Geotherm. Res. 148, p.234-252), including the associated shock wave which damaged structures in Harris, 3km N of the dome, and the base surge and tsunami resulting from interaction of pyroclastic flows with sea-water.

From the 9th of July onwards small swarms of hybrid earthquakes were detected. About 9500 events from a relatively static source region at a depth of about 3km were detected. Each event lasted about 12 seconds and over time, albeit with some variations, the events increased in intensity and frequency until tremor became continuous at 12:00 on July 12. (Ottemöller 2008. J. Volc. Geotherm. Res. In Press). Tremor intensity gradually increased and merged into a constant tremor by the morning of July 12. The collapse started at around 13:30 with progressive degradation from the tallus at the base of the lava dome. This activity, which generated small PFs in the Tar River Valley (TRV), many of which reached the sea, gradually destabilized the overlying lava dome resulting in its progressive failure. At 22:30 significant explosive activity resulting from interaction of larger PFs with sea-water was first observed at the TRV delta. A further increase in activity was noted after midnight with 5 significant PFs occurring at 0:03, 0:22, 0:52, 1:03 and 1:25. Each of these flows had incandescent surge clouds which travelled out to sea for up to 2.5km. Ash convecting up from these clouds reached altitudes of up to 5km and intense lightning was associated with these, forcing observers to leave Jack Boy Hill at around 2:30. Seismic activity reached peak levels, far surpassing those measured during the rest of the eruption, at around 3:35. At this point a tremendous explosion occurred producing a powerful shock wave and an ash plume that eventually rose to a height of 15km. Massive PFs swept into the sea via TRV. This resulted in tsunamis reaching heights of several meters and running inland up to 15m a.s.l. in places (Herd et al., 2005) and a powerful hydrovolcanic explosion resulting in a base surge which flowed inland and along the coast as far as Spanish Point (Edmonds et al., 2006. J. Volc. Geotherm. Res. 153, p.313-330). Dense lithic fragments fell over the whole island and heavy ash-fall occurred, with 15cm of ash being deposited in Old Towne, several km to the NW. About 1000 Tonnes of sulphur dioxide were released. Following this climactic phase of the eruption, activity waned. Small collapses of material off the flanks of the collapse scar and ash venting occuring for several hours followed by a series of vulcanian explosions at 5:08, 13:10, 5:14 (on July 14) and 5:28 (on July 15). Each of these would have been notable in its own right with ash columns of over 10km in each case, yet they were dwarfed by the main explosive event. No fountain collapses were observed during these explosions. Between 21 and 28 July a small lava lobe (30000 cubic meters) was extruded before the volcano entered into a period of inactivity. The pressure wave associated with the main explosion and the phreatomagmatic (hydrovolcanic) event are now addressed in more detail.

The main explosive event at 3:35 generated a pressure wave which was detected 240km to the South on Martinique Island. No previous pressure waves from Soufriere Hills had been detected at this distance. Major structural damage to buildings at Harris, an evacuated settlement perched on top of a 250m high hill, 3km N of the dome, was noted after the eruption. Several cows were killed at the foot of the hill. Houses had collapsed, others had doors and windows blown out. Strangely, corrugated iron roofing, antennae and other structures were bent towards the volcano. A high pressure shock wave fronting a powerful blast of wind appears to have passed outwards from the vent following the main explosion. The upward sloping terrain infront of Harris probably served to locally focus and enhance the pressure wave. Shortly after its passage, air must have rushed back towards the dome into the low pressure area behind the shock wave. Whilst the rushing back air would have been less powerful than the preceding shock wave, it was sufficiently strong (estimated at 30-40m/sec) to obscure the damage from the shock wave and outward blowing wind by bending back roofing and antennae (that had initially been bent outwards) in the direction of the volcano. The effect of shock waves can be compared to the situation following nuclear explosions (Valentine 1998. J. Volc. Geotherm. Res. 87, p.117-140). As the shock wave passes a structure such as a house, the pressure in front of the wave is normal (atmospheric) but the pressure immediately behind it is extremely high. This extreme pressure difference from one side of the house to the other exerts a tremendous lateral force on the house, as well as an implosive force, since the inside of the house is at atmospheric pressure. After initial damage by the pressure wave, the high winds behind it will exert additional dynamic pressure on the structures. These winds would have affected narrow structures such as antennae which are relatively insensitive to pressure waves.

The hydrovolcanic explosion following entrance of pyroclastic flows into the sea was the first such event documented in recent times. It is probable that a similar event occurred in 1902 at Mt Pelee. Detailed analysis was possible based on study of its deposits and impact on vegetation (Edmonds et al., 2006. J. Volc. Geotherm. Res. 153, p.313-330). Deposits in TRV were generated by PF activity, and deposits over a small area on the N rim of the valley near Long Ground were from associated surges overspilling from the valley. However, a far larger area had been devastated, especially to the N, beyond Spanish Point on the coast and to areas up to an altitude of 300m inland as far as Paradise Ghaut. The deposits in this area were notably thicker near to the coast than near the volcano. The bottom and thickest layer of the deposits in most places was a single size-graded layer with larger clasts and burnt vegetation at the bottom and less dense material at the top, as would be expected from a single powerful and hot surge. This layer was overlain by finer layers from the less dynamic trailing parts of the surge and fallout from the convecting ash clouds above the PFs and surges. The thickness of deposits decreasing away from the sea, together with the fact that tree-trunks were singed and impacted by ballistics on the side facing the sea shows that the surge must have been generated at the shore. The extensive fragmentation of fine material and shape of larger clasts in the deposits are further consistent with known hydrovolcanic deposits.

The following mechanism was proposed by Edmonds. Massive volume PFs entered the sea at the climax of the eruption, displacing large amounts of sea-water (as evidenced by 1m high tsunamis which damaged fishing boats in Guadeloupe, 65km SE). Following the main PFs the displaced sea-water was able to rush back and vigorously mix with the PF deposits and ongoing PFs. This caused a wave of explosive fragmentation of the PF material and vaporization of sea-water resulting in the powerful hydrovolcanic explosion and associated base surge that devastated large areas of land. It is important to note that when the fragments of lava in the PF contact water, a violently frothing film of steam forms at the interface which may result in further fragmentation. This increases the interacting surface area and further mixes the magma and water until the whole water is heated sufficiently to vaporize and violently expand. If this process involves mixing on a large scale, a hydrovolcanic explosion as witnessed on July 13 may occur.

The volumes involved in this record-breaking dome collapse can be split into several components (Herd et al., 2005). About 164 million cubic meters of lava dome collapsed, accompanied by 46 million cubic meters of basement material. The massive PFs, which entrained huge boulders and were thus highly erosive, removed over 20 million cubic meters of older PF deposits from TRV and its delta. The erosive PFs left a 10-30m deep and 200m wide channel in the delta.

Most of the erupted / mobilized materials ended up in the sea offshore of the TRV delta (210-220 million cubic meters). Deposits from the surge associated with the hydrovolcanic explosion amount to about 1.5 million cubic meters and 10-15 million cubic meters can be accounted for by tephra deposits on land. In fact, the eruption was also associated with the heaviest tephra-fallout on populated areas ever recorded on Montserrat. The area E (Plymouth) and particularly NE of the volcano (e.g. Olde Towne) was most affected. Vegetation was severely damaged with total loss of crops and damage to trees. Car windscreens were smashed by large lithics mobilized by the vulcanian explosions and accumulation of ash led to the collapse of roofs. As much as 15cm of tephra (mainly ash) fell on Olde Towne with most of the material being deposited from plumes rising over the PFs and also from the main vulcanian eruption plume.

8. Major Dome Collapse of 20 May 2006

The 20 May dome collapse is the second largest dome collapse ever observed at any volcano, and is only surpassed by the 2003 event. The entire dome was mobilized during the event, together with a section of the crater wall. A total of 110 million cubic meters DRE were involved (Loughlin et al., 2007. Geophys. Res. Abstr. 9, 11090). Whilst the volcano was in a period of slightly elevated seismicity for the preceding 10 days, no obvious trigger event was evident. Heavy rainfall and a change in direction of dome growth both occurred in the hours preceeding the eruption and may have played a role in the timing of the collapse. Most of the dome collapsed over a period of under 3 hours. Seismic signals associated with rockfalls and small PFs were initially noted and by 6:45 PFs were first observed flowing down TRV to the sea. The eruption, just like most previous dome collapses, involved successive "peeling away" of layers of the dome (retrogressive collapse) exposing increasingly gas-rich and pressurized layers as the collapse progressed. The eruption rapidly intensified and during a peak period of 34 minutes with its climax around 7:40, huge PFs descended down the TRV into the sea resulting in a devastating base surge that swept back onto land and ignited vegetation along the coastline as far as Spanish Point (as in 2003). The peak in activity also involved 4 massive explosions in short succession which resulted in an ash cloud to a height of 17km and large ballistics rained down on the flanks of the volcano. Vigorous ash venting continued for much of the day and rainfall swept large amounts of ash that had been deposited on the flanks of the volcano into its drainages causing significant lahars.

------

Montserrat Island is only reachable by flights from Antigua. Flights land at the new airport, replacing Blackburne (Bramble) airport which has been destroyed by the eruption. Due to the wind-exposed location of the new airport, flights may not be possible for days during poor weather conditions. Most of the South of Montserrat is a volcanic exclusion zone which may not be entered without permission. The volcano can usually be safely viewed from the observation point on Jack Boy Hill at the edge of the exclusion zone, overlooking the remains of Blackburne airport on the East coast. The new volcano observatory on the hillside above Salem provides an alternative viewpoint to the West. During periods of small dome volume and low activity, it may also be possible to climb Garribaldi or St Georges Hill on the West side of the Island. Comfortable accommodation and friendly service can be found at the View Pointe Hotel when this is not closed due to the risk of PFs entering Belham valley. Various other accommodation is available for those on a lower budget.

The kind of hazards in the exclusion zone can be seen from the chronology of volcanic activity above. Collapse events can occur without visible precursors. Additional hazards in the exclusion zone include the risk of building collapses, but also free-running bulls and large pigs which have been reported to attack people.

For those interested in seeing moving pictures of Soufriere Hills eruptions, a series of informative / educational DVDs with unique footage of the eruption is obtainable from Montserrat-based filmer David Lea at www.priceofparadise.com. Updates on current activity can be found at the Montserrat Volcano Observatories Website www.mvo.ms.

Readers interested in the impact of Lahars and Pyroclastic Density Currents may also find the section on Chaiten volcano interesting.

Further Photos

Photovolcanica Full Index