Montserrat Volcano Observatory, Special Report 06

Montserrat Volcano Observatory, Montserrat, West Indies

Special Report 06
The Boxing Day Collapse

Eliza S Calder, Simon R Young, R Steve J Sparks, Jenni Barclay,
Barry Voight, Richard A Herd, Richard Luckett, Gill E Norton, Lutchman Pollard,
Lucy Ritchie, Richard E A Robertson and MVO Staff

Individual author affiliations.

Montserrat Volcano Observatory
Mongo Hill
Montserrat
West Indies
Tel: 1 664 491 5647
Fax: 1 664 491 2423

Any use of trade, product or firm names is for descriptive purposes only and does not imply endorsement by the Government of Montserrat

TABLE OF CONTENTS

Executive Summary

List of Figures

Precursor Activity

Evidence On Chronology

Morphological changes, vegetation and structural damage

Volume estimates

Dynamical Constraints on the Boxing Day Event

Interpretations of field data and observations

Frontispiece 1. Looking northeast over the devastated area, late-January 1998 (810x540 JPG, 103K -- 1080x715 JPG, 831K)

Executive Summary

A major collapse of the old volcanic edifice and fresh lava dome occurred at the Soufriere Hills volcano early on Boxing Day morning, 26 December 1997. This collapse represents the highest magnitude and perhaps most intense volcanic activity yet at the volcano. The major impact of these events was confined to the southwestern part of Montserrat, but important lessons must be learnt from the event; a similar event on the northern flanks of the volcano would probably have had more severe consequences.

The Boxing Day event was the culmination of a period of very rapid dome growth which had followed the explosive phase of September/October 1997. Dome growth began on 22 October at first within the explosion crater and then as a large lobe growing to the north. Subsequent extrusion (from c. 4 November) of a southerly lobe broke through and removed the southern wall of the explosion crater by collapse (on 4 November), and built up a large dome (with summit height of c. 1020 m a.m.s.l.) over the Galway's Wall. Seismic activity was low through much of the dome construction period, but a hybrid swarm began on Christmas Eve, which built up to continuous tremor in the few hours before the collapse.

The slope failure and dome collapse occurred at about 3 am local time on Boxing Day morning, and lasted only about 15 minutes. Darkness and cloud cover meant that visual observations were extremely limited. Seismic evidence provides information on the duration of the event and some timings of specific phenomena, but reconstruction of the event itself has been done chiefly by evaluation of the deposits, changes in dome and flank morphology and effects of transportation processes (physical damage).

The event comprises two main phases: initial generation of a debris avalanche from the Galway's Wall and Galway's Soufriere areas, which incorporated overlying dome talus material; and the consequent collapse of a destabilised portion of the lava dome. This resulted in the formation of two collapse scars; one eating into the base of the Galway's Wall and one with an amphitheatre-shaped scoop excavated into the dome behind it, typical of previous dome collapses.

The debris avalanche moved down the White River and locally spilled over the valley walls. Deposits are present through much of the White River valley, but much of the debris avalanche deposit has been blanketed by later pyroclastic flow deposits. A portion of the debris avalanche is thought to have reached the ocean although its volume is poorly constrained.

The succeeding dome collapse produced pyroclastic flows and ash-cloud surges. Much of the pyroclastic flow material was confined within the White River valley and a considerable volume of this material is also believed to have reached the sea.

A small tsunami was generated, probably by the steep flow front of the debris avalanche, possibly assisted by the pyroclastic flows as they entered the sea at the mouth of the White River.

Very intense pyroclastic surge activity also occurred during the collapse. Marginal surges occurred associated with the main flows but some surge units are thought to have been associated with one or more explosive 'blast' events sourced from the collapsing dome. Surges caused widespread devastation in the area south of Gingoes Ghaut and also impacted on areas further north.

A convective ash cloud was generated from the pyroclastic flows and surges, and reached about 47,000 feet (14.3 km) altitude. Over the following few days, this ash cloud drifted southwards across much of the eastern Caribbean. Fine ash deposits mantled much of southwestern Montserrat.

Dome growth recommenced very soon after these events. Talus from the growing dome started to spill into the lower depression within a few days. Seismicity has been at a generally low level through this re-growth period (late December 1997 to late January 1998), but dome rockfall and small pyroclastic flow signals have been recorded at moderate levels, indicative of dome growth rates comparable with those over the previous few months.

Frontispiece 2. Tree blow-down at Upper Galway's Estate (810x540 JPG, 120K -- 1075x715 JPG, 875K)

List of Figures
Clicking on the embedded figures will download largest figure version. See figure descriptions for more options.

Figure 1. Cross section and map of the Galway's Wall area prior to the Boxing Day collapse. Based on survey information from 23 November and 8 December and additional constraints from video and photographs.
Figure 2a. Photograph of the southern part of the lava dome and Galway's Wall area, taken from a helicopter above the upper part of the White River valley, 23 December 1997.
Figure 2b. Debris avalanche deposits and overlying coarse pyroclastic flow deposits from the Boxing Day collapse in the area of O'Garra's, at the mouth of the White River Valley.
Figure 3. Windy Hill (MBWH) seismic station, count of event types, 22 October to 31 December 1997.
Figure 4. Rate of occurrence and energy of hybrid earthquakes, RSAM for MBWH seismic station and tilt data from Long Ground for the period 22 to 27 December 1997.
Figure 5a. Partial copy of the Jack Boy Hill (MJHT) short period seismometer helicorder record from 25 December 1997.
Figure 5b. Partial copy of the Jack Boy Hill (MJHT) short period seismometer helicorder record from 26 December 1997.
Figure 6. Signal amplitude on MBWH, start time is 02:55 on 26 December 1997, amplitude scale is arbitrary.
Figure 7a. Plume distribution sketch based on GOES 8 satellite imagery for 04:45 on 26 December 1997.
Figure 7b. Plume distribution sketch based on GOES 8 satellite imagery for 10:45 on 26 December 1997.
Figure 7c. Plume distribution sketch based on GOES 8 satellite imagery for 15:45 on 26 December 1997.
Figure 7d. Plume distribution sketch based on GOES 8 satellite imagery for 04:15 on 27 December 1997.
Figure 7e. Ash distribution and SO₂ concentrations from TOMS satellite data, c. 09:00 on 26 December 1997.
Figure 7f. GOES 8 satellite image showing Montserrat plume, 08:15 on 26 December 1997.
Figure 8a. Map of deposits from the Boxing Day event, with approximate line of section for Fig. 8b.
Figure 8b. Schematic cross section showing deposit position and relationships for the Boxing Day event. Vertical scale is exaggerated.
Figure 9. Cross section and map of the Galway's Wall area after the Boxing Day collapse. Based on information from video and photographs.
Figure 10a. View of the lower collapse scar and new dome talus filling the scar. Photograph taken from above the White River valley, mid-January 1998.
Figure 10b. Destruction of buildings in the Trials area.
Figure 11. Map of damage zonation for the Boxing Day event.
Introduction
This report summarises the events leading up to and associated with the edifice failure in the Galway's Wall area, and the consequent collapse of the southern sector of the lava dome, during the early morning of 26 December 1997. The report includes an assessment of monitoring data relevant to this event and precursor activity. Ongoing fieldwork is evaluating the nature of the deposits; as this work is incomplete, the interpretations presented here are preliminary. However, enough information has been gathered to present a coherent scientific assessment of these important events.
There are many lessons to be learnt from this event, both for the evolving situation on Montserrat and for other similar volcanoes around the world. MVO is keen to maximise and share the knowledge gained from this volcano and acknowledges the high level of collaborative support which it has received throughout the current volcanic crisis.
The report summarises the work undertaken to date on the Boxing Day event by a number of staff at MVO. We are grateful to all those associated with MVO who provided comment on the events described here. We are also grateful to other external agencies for providing information relevant to these events, especially the Volcanic Ash Advisory Centre (NOAA Satellite Analysis Branch, Washington) and the National Climate Data Centre (NOAA). French colleagues at the Institute de Physique de la Globe, Paris and at the Commissariat a l'Energie Atomique, Bruyeres le Chatel assisted with modelling of tsunami hazards in 1996 which proved most useful.
Edifice instability at active volcanoes (McGuire et al., 1997) has been identified over two decades as a major volcanic hazard, and the events of Boxing Day highlight some of the problems in managing such hazards. Although the timing of this event could not be predicted, the possibilities for moderate-sized debris avalanches had been noted for over a year, with the Galway's Wall recognised to be particularly vulnerable to such a failure (MVO Special Report 2). The two-month build-up to the collapse was identified as a time of increasing danger to the southwestern part of Montserrat as highlighted in the MVO daily reports for the period, and the precursor hybrid earthquake swarm was interpreted as cause for heightened concern (e.g. MVO Evening Report, 25 December 1997).
All times given are local (GMT -4 hours). Supporting information is available from MVO. Work is ongoing and all findings described here are preliminary.
Precursor Activity
Dome Growth
Dome growth recommenced in the crater formed by the September and October explosions (MVO Special Report 5, 1998) immediately after the cessation of explosive activity on 21 October. This episode of dome growth occurred in two phases: growth of the 22 October lava lobe in the open northern sector (ceased by 2 November); and growth of a new, southern lobe (4 November lobe). This growth soon generated two collapses from the southern flank in association with hybrid earthquake activity. The first of these collapses (4 November) removed much of the southern side of the cold, pre-explosion dome complex (which grew mainly in April and May 1997). The second and larger collapse on 6 November primarily involved hotter material from the growing 4 November lobe. The flows associated with these collapses reached the sea at the mouth of the White River, further infilled the White River valley, and extended the delta. The collapses in early November together involved c. 18 million m³ of material.
Rockfalls and minor pyroclastic flows continued to be generated by the rapid extrusion of the southern lobe of the dome. Thus a thick talus apron accumulated at the base of the dome between the Galway's Wall and Soufriere area. By 8 December, the talus apron was 500 m wide and extended about 500 m beyond the base of the Galway's Wall (Figure 1) at an angle of repose of c. 33° The remnants of Galway's Wall were, by this stage, completely engulfed by the growing dome and new talus was accumulating in the area of the Soufriere. Although the lava dome/talus contact near the top of the talus apron was outwith the old English's Crater margin, the area of active intrusion into the base of the lobe was still within the crater (Figure 1). The southern part of the 4 November lobe and associated talus had a minimum volume of 26 million m³ at this time, and the growth rate for the early November to early December period was estimated at 11 m³/sec.

Figure 1: Cross section and map of the Galway's Wall area prior to the Boxing Day collapse. Based on survey information from 23 November and 8 December and additional constraints from video and photographs. (655x1035 GIF, 12K)

Subsequent to 8 December, growth continued to build up dense blocky material at the top of the talus slope, shedding further rockfalls and small pyroclastic flows over the Galway's Soufriere area and occasionally eastwards into the Tar River Valley. By 23 December, the southern lobe had reached an altitude of about 1020 m (± 10 m, estimated from scaling of video footage), the greatest height of the dome since the eruption began (Figure 2a). Using a simple geometrical reconstruction of additional dome growth between 8 and 26 December (including an estimate of additional talus), an estimated total volume for the pre-collapse dome is 110 million m³. This gives a growth rate for this period of c. 5 m³/sec.

Figure 2a: Photograph of the southern part of the lava dome and Galway's Wall area, taken from a helicopter above the upper part of the White River valley, 23 December 1997. (900x590 JPG, 120K -- 1195x780 JPG, 279K)

Commencing immediately after the explosions ceased in late October, seismic activity was dominated by hybrid earthquakes (Figure 3). A series of hybrid swarms started in late October and continued until 12 November. These swarms at times merged to tremor of similar frequency to that of the individual hybrids. On 6, 7 and 8 November, in particular, tremor was at a high level so that individual events did not trigger the seismic network. Throughout this period, tremor episodes and hybrid swarms were both associated with vigorous ash venting from the central cleft between the two lobes. On 4 and 6 November, the seismicity, whether hybrids or tremor, stopped soon after the major pyroclastic flows.

Figure 3: Windy Hill (MBWH) seismic station, count of event types, 22 October to 31 December 1997. (735x635 GIF, 21K -- 1385x1200 GIF, 53K)

Hybrid earthquakes recorded on 1 and 2 November had particularly high amplitudes at MBWH - the only broadband network station continuously recording since October 1996. The only hybrid events previously recorded with higher amplitude were those on 24 June, 1997. Some of the events on 1 and 2 November were reported by SRU to have triggered their stations on Dominica, Antigua and Nevis. The amplitudes of the hybrids then decreased through November.

From 12 November to end November, activity was dominated by rockfall signals although there continued to be large numbers of very low amplitude hybrids - many of which did not trigger the networks. These hybrids were scattered throughout the period and not grouped in swarms. Two short pyroclastic flow episodes occurred on 27 November and 1 December.

From the beginning of December until 24 December (Figure 3), there was a very low level of hybrid earthquake activity. The number of rockfall events remained constant at around 70 per day.

Through December, the frequency content of the rockfall signals changed slightly towards lower frequencies probably due to the attenuation of higher frequencies. In many cases it was difficult to distinguish between long period earthquakes and rockfall signals. The source of these signals was most likely debris falling from the dome and also the ash venting or degassing that can cause these falls to occur. In addition, a number of the events classified as long-period earthquakes were, in fact, more like short bursts of harmonic tremor. These lasted up to a few 10s of seconds and were nearly monochromatic at approximately 2 Hz, although there were often a few wavelengths at 1 Hz present at the start of an event.

Figure 4: Rate of occurrence and energy of hybrid earthquakes, RSAM for MBWH seismic station and tilt data from Long Ground for the period 22 to 27 December 1997. (760x625 GIF, 14K -- 1180x970 GIF, 28K)

There was a subtle increase above background in the number of hybrid earthquakes from 22 December (Figure 3). A distinct hybrid swarm then began at about 14:30 on 24 December (Figure 4). This swarm started out with events every 20 minutes or so, and slowly increased in intensity until approximately 20:00 on 25 December. The individual events also generally increased in amplitude as the swarm progressed (Figure 5a), but even the largest earthquakes were relatively small in amplitude, an order of magnitude smaller than those recorded in early November. At 20:00 the hybrids were occurring too frequently to trigger the network and from this time onwards the signal being recorded was effectively a tremor signal. The amplitude of this tremor peaked at about 23:00 on 25 December and then declined until 00:00 when individual events could again be discerned. These individual events merged back into tremor at about 01:30 on 26 December, which then built in amplitude up to the onset of the signal corresponding to the main collapse at 03:01 (Figure 5b).

Figure 5a: Partial copy of the Jack Boy Hill (MJHT) short period seismometer helicorder record from 25 December 1997. (1080x715 GIF, 64K -- 1620x1080 GIF, 138K)

Figure 5b: Partial copy of the Jack Boy Hill (MJHT) short period seismometer helicorder record from 26 December 1997. (870x615 JPG, 376K -- 1480x1050 JPG, 975K)

Other Monitoring

Other monitoring information gathered in the weeks between the end of explosive activity on 21 October and Boxing Day do not reveal any marked changes. Deformation measured by EDM (lines to Lees Yard reflector) and GPS (Hermitage to Harris Lookout) did not show any deviation from established trends in this period, although sampling frequency was insufficient to pick up short term (days) trends. The Long Ground electronic tiltmeter showed irregular fluctuations during the precursor hybrid swarm and a permanent positive offset on the X-axis of c. 3 ±rad (no offset on Y-axis) after Boxing Day (Figure 4). The orientation of the tiltmeter is such that this positive offset just on the X-axis is difficult to interpret as being related to inflation/deflation of the edifice itself. No quantitative gas flux data is available, although a vigorous gas plume was noted over the volcano on clear days during this period.

Evidence on Chronology

Seismic evidence for an eruptive chronology is taken mainly from the Windy Hill (MBWH) 1 Hz single component station, which is part of the digitally telemetered 'broadband' network. Additional information is taken from the 'short-period', analogue network; the St Patrick's (MSPT) station of this network stopped transmitting at 03:03.3 on 26 December when impacted by the collapse event.

There is no distinct onset of the collapse event (defined as the period of highest seismic amplitude), and a time of 03:01.0 is taken as the start. This period ended at 03:16.2, after which a period of lower but still high amplitude was recorded until seismicity dropped to background level at 03:32.1. In the course of this period, intervals of monochromatic tremor were recorded at 1.9 Hz, including 3 higher amplitude pulses. This gives a total duration for the main collapse event of 15.2 minutes and for the entire event of 31.1 minutes. The main collapse event has been divided into six pulses (Figure 6, Table 1).

Figure 6: Signal amplitude on MBWH, start time is 02:55 on 26 December 1997, amplitude scale is arbitrary. (800x625 GIF, 17K -- 1230x960 GIF, 37K)

Table 1. Individual pulse timings within main collapse signal.
Pulse No Start End Duration (secs) Comments
1 03:01.8 03:02.3 30 Relatively brief, low amplitude
2 03:02.8 03:04.6 108 Even build and decay around central peak, high amplitude at peak
3 03:04.6 03:06.0 84 Similar to pulse 2, lower amplitude
4 03:06.3 03:10.2 234 Complex, moderate amplitude, no strong peak
5 03:10.2 03:11.3 66 Short, moderate amplitude
6 03:11.3 03:14.2 174 Consistent high amplitude for most of duration

Signal frequency changed very little from the hybrids and tremor prior to the collapse, with dominant frequency below 2.8 Hz. The dominance of lower frequency signals is typical of the MBWH station, largely due to site effects. A little higher frequency energy is associated with the final two pulses of collapse activity, prior to a period of near-monochromatic tremor at 1.9 Hz.

A comparison of the seismic records indicates that the form of the collapse signal was similar to that produced by the 6 November collapse and pyroclastic flow activity, but with an amplitude of about 5 times greater.

The monochromatic tremor, best developed between 03:21 and 03:25, is punctuated by three high amplitude signals of the same dominant frequency (Figure 6). Each successive signal is smaller than the previous - these occur at 03:18.3, 03:23.8 and 03:26.7. The last one is followed by a rockfall signal. The nature of this tremor signal and the high amplitude peaks is very similar to that generated by ash venting and Vulcanian explosions respectively during the August to October period (MVO Special Reports 4 and 5, 1998).

Roaring was reported from a number of places on Montserrat and was described as similar to, but louder than, that heard during the ash-venting associated with the August to October explosions. The timing of these reports is consistent with the roaring being associated with the 'ash venting' signal described above. Police at the Salem checkpoint reported hearing two "explosions" and seeing flashes of light between 03:15 and 03:25. The exact timing of these reports is unknown, and there were separate reports of thunder and lightning also at this time. However, the police reports could coincide with the first two high amplitude 'explosion' signals.

GOES8 satellite data indicate that the eruption plume ascended to 14.3 km based on plume-top temperature estimates. These estimates were confirmed by radiosonde data for wind directions taken above Puerto Rico and Guadeloupe. Piarco FIC (Trinidad) reported ash at 12.2 km at 03:55 and a BWIA pilot reported seeing ash at 11 km at 06:05.

The plume was rapidly dispersed to the SSE (30 to 40 knots above 7.6 km) and SE (30 to 40 knots between 2.4 and 7.6 km) at upper levels, although a lower level (< 2.4 km) plume moved slowly (15 knots) westwards (Figures 7a-d, f). The high level ash plume passed over the central part of the eastern Caribbean and then out into the Atlantic. Ash was still visible over much of the region 24 hours after the event. TOMS data also show the ash plume, as well as SO₂ concentrations, which were highest at the leading edge of the plume (Figure 7e). The nature of the high level ash plume and distribution of ash on Montserrat suggests that it was generated from pyroclastic flow and surge activity rather than by a vertical explosion. The low level ash plume may have been associated with the ash venting phase of the event.

Direct observations were limited due to the occurrence of the event at night. Burning buildings in the Trials area were observed at about 03:45. At this time, the ash plume was seen from Garibaldi Hill and there was no evidence for a vertical column having occurred. The remaining low level ash plume was drifting southwest. Light ash fell at Garibaldi Hill but did not fall any further north.

Initial observations of the deposits were hampered by strong winds, burning vegetation and blowing ash. Heavy rain during the first weeks of 1998 caused some modification of the deposits, but also removed masking ash deposits so that observations were easier. A programme of fieldwork was undertaken during the month following Boxing Day.

Deposition occurred over a sector with a spread angle of about 70° centred on the village of Morris' (Figure 8a, b). A 2 km-wide delta built out to a maximum of 600 m from the pre-existing shore at the mouth of the White River valley. Although the delta has been widened significantly since the November collapses, a steep submarine slope 600 m from the shoreline constrains further extension of the fan out to sea.

Figure 8a: Map of deposits from the Boxing Day event, with approximate line of section for Figure 8b (below). (1095x587 GIF, 21K -- 1690x910 GIF, 39K)

Figure 8b: Schematic cross section showing deposit position and relationships for the Boxing Day event. Vertical scale is exaggerated. (1080x415 GIF, 11K -- 1515x585 GIF, 19K)

Five main depositional units have been identified: debris avalanche deposits; block-and-ash flow deposits; pyroclastic surge deposits; co-ignimbrite fallout; and a putative blast deposit. A significant amount of material is thought to have entered the ocean, although the volume and nature of the deposits offshore are yet to be ascertained.

Debris Avalanche Deposit

The central region of the delta and lower reaches of the White River valley are characterised by a hummocky orange-brown deposit of poorly-sorted, coarse blocky debris with an irregular bulbous front about 25 m high (Figure 2b). This is a debris avalanche deposit (DAD) resulting from slope failure of hydrothermally altered rocks in the Galway's Soufriere area, the lower outward flank of the Galway's Wall, and the overlying apron of fresh dome talus. Further outcrops of the distinctive hummocky debris avalanche are located higher up the White River valley, in a constriction at the top of the valley and in the Galway's Soufriere area. Much of the debris avalanche material can be seen to have a smoothed, heavily scoured upper surface with discontinuous remnants of pre-existing hydrothermally altered stratigraphy preserved within the deposit.

Figure 2b: Debris avalanche deposits and overlying coarse pyroclastic flow deposits from the Boxing Day collapse in the area of O'Garra's, at the mouth of the White River Valley. (540x360 JPG, 54K -- 1075x715 JPG, 892K)

The avalanche deposit is approximately 500 m wide and 25 to 70 m thick across the central axis of the delta, overlying pre-existing delta material, with the central lobe clearly feeding into the sea on the southern side of this axis (Figure 8a). The medial part of the White River valley is oriented to the southwest and bends to the south at about 1.7 km from the coast. Some of the avalanche material surmounted the side of the valley at this bend and was deposited on the plains above Morris' village. A small lip also spilled out on the western bank and deposited on the flank of Fergus Mountain.

The over-spill lobes of DAD above Morris' show internal stratigraphy, with segregation of relatively fresh, blocky dome talus (forming loose, matrix-depleted layers) beneath hydrothermally altered material (forming cohesive, matrix-rich layers). The lateral continuity of such stratigraphy is not clear.

Block and Ash Deposits

The main pyroclastic flow deposits are similar in nature to previous dome collapse flows produced by the Soufriere Hills volcano (Cole et al., 1998). They comprise dense to slightly vesicular (friable textured) blocks in a poorly sorted ash-rich matrix. Little internal organisation is seen in sections of these deposits but at the base there often exists a thin (1-3 cm) fines-depleted layer. During the Boxing Day event, the main pyroclastic flows were largely confined to the White River valley although some material did spill out over the prominent bend and travel down towards Morris' (Figure 8b). Little deposition occurred in this area, with only small piles of lag blocks stranded on the surface of the DAD. The flows produced marked erosion features over the whole area between several hundred metres east of the White River valley and Morris' village even though the gradient of the slope of this area is relatively low (c. 10°).

For the most part these deposits are ponded in the area immediately behind and locally upon the debris avalanche, filling the remainder of the White River valley (adjacent to the valley spill-out) to a maximum depth of 50 to 70 m.

On the delta, the DAD is partially overlain by relatively thin (50 to 70 cm), block and ash type pyroclastic flow deposits. These broad, flat-lying deposits are poorly sorted with blocks reaching a maximum size of about 1 m (blocks > 0.1 m forming c. 10% of the surface). These pyroclastic flow deposits grade laterally to thinner finer grained surge deposits towards the margins of the delta. These deposits are normally graded, with a coarse gravely sand base grading upwards to a medium to fine sand. Fluidisation pipes are observed locally. The fan surface is covered in deposits from these energetic flows and it is therefore assumed that a significant volume has now been lost to sea.

Surge Units

The surge deposits associated with the collapse cover an area on land of 9.1 km² over a sector of 70° around the south of the volcano (Figure 8a). They are quite variable in nature, and some at least are markedly different from previous surge deposits associated with pyroclastic flow emplacement at Soufriere Hills. The relationship between the various surge units is still under investigation and, although we are able to recognise the main events, we cannot as yet be confident as to which surge layers are associated with the different parent events.

Conventional ash-cloud surge deposits are found to the east of the White River valley, on the delta and in the Trials area. Surge clouds ran up and partially surmounted the South Soufriere Hills and spilled over the southwestern ridge near Fergus Mountain Estate. These deposits comprise a fine grained ash-rich, sandy deposit (6 to 10 cm) with an underlying thin (0.5 to 2 cm) fines-depleted coarse sand layer. It is considered that these units represent a single deposit from a marginal ash cloud surge derived from a parent block and ash flow. Similar, though slightly thicker and coarser-grained deposits are found between Gingoes and Germans ghauts, although it is currently unclear how these deposits grade laterally.

The surge deposits in the area between the White River valley and Germans Ghaut comprise several units of variable type. Although fine grained ash-rich units exist, the dominant facies, which is between 15-40 cm thick, is a coarse sand/gravel fines-depleted unit. Contacts between this and the underlying fine surge units are often erosive in nature. In some areas this deposit is seen to be overlain by a second fine grained surge deposit. The coarse surge deposits largely comprise sub-angular dense dome rock and crystals with little pumiceous or friable component. The deposits thin and become finer-grained both down-slope and laterally. This unit is generally massive, although in some places shows laminations. The origin of this unit is unclear. It may be associated with or be an extension of the blast deposit described below or may in itself represent a highly energetic surge pulse not derived from a parent block and ash flow but of a primary 'true' surge origin. Further investigation of field sections is required in order to clarify this.

Secondary Pyroclastic Flow Deposits

Small secondary pyroclastic flow deposits with abundant charcoaled wood occur in the deep ghauts that drain the area covered by the surge deposits (Figure 8a, b). These flow deposits are dominantly fine grained and are interpreted as contemporaneous drainage of fluidised fine surge deposits emplaced on gully slopes (i.e. condensed surges). One of these secondary pyroclastic flows drained down towards the eastern side of the Soufriere Hills, down Dry Ghaut from the saddle region between the South Soufriere Hills and Galway's Mountain. This thin, highly mobile flow is confined to the bottom of the ghaut (average width generally 2-4 m) and extends to within 300 m of the sea. The deposit is poorly sorted and 50 to 70 cm thick, consisting predominantly of fine ash-rich sand.

Blast Deposit

A deposit of enigmatic character is found on the southwestern flank of the volcano between Gingoes Ghaut and the White River, best seen in the vicinities of the upper and lower Galway's Estate (Figure 8a). The deposit comprises angular to sub-angular lithic clasts scattered on the surface, some up to 70 cm in maximum dimension. They now lie upon and possibly grade into coarse sand/gravel fines-depleted pyroclastic surge deposits, although co-ignimbrite ash which originally coated this deposit has been winnowed and washed around surface blocks. Coverage of blocks is 60 to 70 % of ground area (adjacent to the Upper Galway's Estate area), increasing upslope and decreasing downslope. The surface of the deposit is very subtly corrugated in the flow direction, suggesting a highly energetic emplacement mechanism. This deposit is distinctly different from thinly-spread 'normal' facies block and ash flows as it is locally only one clast thick and is completely fines depleted. This deposit grades distally into a gritty surface layer, extending about 3 km from the dome.

Dense, fresh, angular dome rock makes up most of the deposit, with small amounts of altered dome rock and sub-rounded, semi-vesicular, steely blue-grey dome rock which locally are more abundant. There is a marked lack of impact craters, bread crust-textured clast or any ballistic blocks. Increased deposit thickness occurs in the lee of obstructions and there are localised secondary flows (up to 50 cm thick) of this block-rich material, especially in very proximal localities. This deposit is almost certainly of a 'flow' nature as opposed to proximal ballistic fallout; however a highly energetic environment is needed to transport such large clasts. It is inferred that this may represent a lag deposit from an explosively generated blast surge which then grades downstream into coarse surge deposits. The exact relationship between this and the coarse 'primary' surge described above is still to be ascertained; they do not always appear to grade simply into each other.

Co-ignimbrite Fallout

Co-ignimbrite ash covers most of the southwestern part of Montserrat and drapes all the 26 December deposits, although heavy rains in early January altered the nature of this deposit. Typical preserved thickness across the area is a few mm to a few cm. Near the coast in the Trials area the co-ignimbrite ash fell as accretionary lapilli, caused by incorporation of steam generated by hot material entering the ocean. The accretionary lapilli are up to 8 mm in diameter and form a layer up to 4 cm thick.

Further upslope and in exposed positions, the ash is thinner, due probably to the ease of windblown re-mobilisation. The fine-grained, crystal-rich ash is typical of co-ignimbrite ash generated from pyroclastic flows sourced from dome collapse. The co-ignimbrite ash plume reached an altitude of c. 47,000 feet and light ash fall was reported from Guadeloupe, 60 km to the south-southwest (see Figure 7), St Vincent and Bequia.

Figures 7a-f - separate sheet

Temperatures

Temperatures determined from the various deposits several days after the eruption show values up to 293°C (Table 2). The debris avalanche deposit was mainly emplaced cold, although parts of the Galway's Soufriere and dome talus debris would have been warm at time of incorporation into the avalanche.

Table 2. Temperature measurements for deposits from the Boxing Day collapse.

Number Deposit type Location Depth of measurement (cm) Time after event (Days) Temperature °C)

1 Secondary pf Dry Ghaut 20 4 48

1 Secondary pf Dry Ghaut 25 4 138

1 Secondary pf Dry Ghaut 35 4 122

4 Surge WR delta 30 9 155

4 Surge WR delta 60 9 216

4 Surge WR delta 30 9 228

4 Surge WR delta 30 9 83

4 Surge WR delta 50 9 93

4 Fumarole WR delta 30 9 68

5 Secondary pf 50 10 190

7 Surge/pf over DAD 20 13 157

7 Surge/pf over DAD 25 13 103

7 Surge/pf over DAD 60 13 293

Morphological Changes, Vegetation and Structural Damage

The Galway's Soufriere and lower, outer parts of the Galway's Wall, and all of the dome and talus overstepping the wall, were removed by the Boxing Day event, leaving two collapse scars (Figure 9). The lower scar was a large arcuate step with its back wall c. 200 m south of where the crest of the Galway's Wall used to be and extending outwards c. 500 m. The upper scar was very similar in morphology to previous dome collapse scars producing a large spoon shaped amphitheatre. This scar filled rapidly with new dome lava after 26 December. After the collapse, the new valley floor in the area of the Galway's Soufriere was remarkably flat, covered with coarse block and ash flow deposits infilling the rough topography of the DAD.

Figure 9: Cross section and map of the Galway's Wall area after the Boxing Day collapse. Based on information from video and photographs. (505x780 GIF, 8K -- 700x1080 GIF, 13K)

The eastern edge of the Galway's Wall itself and the southern side of Chance's Peak were heavily eroded during the collapse although observations have again been hampered by limited visibility. In the central portion, the wall remnant is inferred to be little more than 650 m high and all of the outer faces of the wall have been heavily eroded. Within a few days after Boxing Day, a small notch had developed in the front face of Galway's step (the back wall of the lower scar), and new dome material was spilling down to form a new talus slope on the flat valley floor (Figure 10a).

Figure 10a: View of the lower collapse scar and new dome talus filling the scar. Photograph taken from above the White River valley, mid-January 1998. (540x340 JPG, 34K -- 1075x680 JPG, 190K)

All sides of the White River valley are heavily scoured other than in valley corners. The zones of vegetation and structural damage are illustrated in Figure 11. The upper surface of the DAD in the lower sections of the White River valley are smoothed and undulating with conspicuous striations and furrows. Some of the hummocky topography of the DAD has been infilled by blocky lag deposits of the overriding flows and surges. The land surrounding the White River valley shows extensive signs of strong erosion. The topsoil and vegetation has been stripped from most surfaces to expose strongly abraded bedrock. The eroded bedrock shows decimetre to metre-scale corrugations with the axes pointing downslope in the flow direction. Blackening of some of these surfaces is of unknown origin, but may represent creation of organic tar (see below).

Figure 11: Map of damage zonation for the Boxing Day event.(1020x656 GIF, 16K -- 1555x865 GIF, 28K)

Further from the valley to the northwest, where down-cutting erosion is not severe, soil is veneered by a blackened, polished and striated surface, caused by the incineration of organic matter to form a 3-5 mm matting of organic paste which smells like tar. This thin veneer is not present beneath the DAD, and is interpreted to have been formed beneath post-DAD pyroclastic surges which did not deposit material in this area. Extensive tracts of ground directly upstream from St. Patrick's are covered with this tar veneer and yet most of the buildings have been swept away. This implies the flows here had a strong directional force but carried little bedload at their base, as minimal erosion occurred on the ground surface. This area is referred to as a 'blast' zone.

Within the 'blast' zone, few man-made structures remain. The building remnants in the Morris' and St. Patrick's area are confined to structures below ground level. The only significant structure remaining above ground south of Gingoes Ghaut is the Mill Tower at Reid's Estate. Rebar supports for floors have all been bent in the direction of the flow downslope and water pipes have been broken and stretched in a similar orientation. All but the largest trees have been removed and those that remain are similarly bent in the direction of the flow. Following the storms early in January, large logs of carbonised and normal wood were washed ashore in the delta area; these include logs up to 5 m long and 0.5 m diameter. Tree blow-down is variable in the area above Upper Galway's Estate with a sharp contact between total removal of trees and true blow-down. Some pitting of ground floors in destroyed buildings can be seen and there are centimetre to decimetre-scale striations on the top surface of the tar deposit.

North of Gingoes Ghaut, most structures are standing, although bending or uprooting of trees in the flow direction and bending of rebar and fence posts can be seen all the way through Trials and Fairfield (Figure 10b). The outer limit of the impacted zone includes southern parts of Kinsale almost to Aymer's Ghaut, where charcoaled trees (on upslope side only) and many burned houses suggest impact by the edge of surge clouds. There is relatively little physical damage in this area apart from fire damage, although some screens and doors are blown in.

Figure 10b: Destruction of buildings in the Trials area. (545x360 JPG, 50K -- 1085x715 JPG, 837K)

All flow-direction indicators show a very consistent flow source at the head of the White River valley except in peripheral areas such as Trials and parts of South Soufriere Hills, where some local variations can be seen.

Volume Estimates

Volume estimates of the two scars formed during the Boxing Day collapse could not be made by the conventional MVO methods (Stasiuk et al., 1998) due to poor visibility in the weeks after Boxing Day. However, an estimate of scar volumes has been made using data generated from construction of cross sections and assuming relatively simple geometries. These suggest that the following inventory of material was shed on Boxing Day: 20 million m³ of hydrothermally altered rock from Galway's Soufriere area; 5 million m³ of Galway's Wall; 26 million m³ of 4 November dome lava; and 26 million m³ of dome talus. Assuming 80% of dense rock equivalent (DRE) for the Galway's Soufriere and Wall material (the same as is assumed for the deposits) and 97% and then 13% of DRE (to account for pore space in dome/talus and vesicularity respectively) for the dome and talus, this gives a DRE-equivalent volume for the collapsed material as 64 million m³.

A volume survey of the lower White River valley and new delta was undertaken on 4 January and of the upper part of the valley on 17 January. From these surveys, a well-constrained total volume for deposits from the Boxing Day collapse is 45.9 million m³. This does not include out-of-valley deposits, material lost to sea nor ash generated during the collapse. An estimate of the out-of-valley surge deposits yields a volume of 1.8 to 3.2 million m³, with an additional 0.05 million m³ of out of valley debris avalanche and pyroclastic flow deposits. Using an estimate of an additional 15% of DRE volume for ash deposits, this yields a total DRE volume for the collapse deposits, not including material lost to sea, as 44.5 million m³.

Given the errors involved in these estimates, the match between scar volume and deposit volume is good, and suggests that about 20 million m³ (25 million m³ if expanded into 'deposits') of material has been lost to sea. This would be consistent with the size of the tsunami wave generated during this event, assuming that most of the material entered the sea at the same time (see below). The magnitude of this event, at c. 44 million m³ of dome material, is thus 4 to 5 times the previous highest magnitude dome collapse events at the Soufriere Hills volcano during the current eruption, the 9 million m³ dome collapse of 21 September 1997 and the similar volume collapses of 4 and 6 November 1997.

On the night of the eruption there were reports of a wave inundating the Old Road Bay area. Investigations revealed that a wave had indeed come inshore near the jetty. The wave was estimated to have been about 1 m higher than the level of the road which lies approximately 2 m above water level and to have moved a maximum distance of 80 m inland. A variety of objects, including a small wooden boat, a roof to a shelter, and a stone table were displaced several metres inland and a large log was carried even further by the wave. There were also impact marks up to a metre high on the side of palm trees facing the sea. The grass was oriented in such a way as to indicate the retreat of the wave. An observer reported seeing the sea move out and then in, which is typical of a tsunami wave. Elsewhere around the western coast of Montserrat, there was no strong evidence for wave inundation higher than high-tide level.

The focusing of the wave at Old Road Bay can be attributed to the peculiarities of wave behaviour along a coastline and the abrupt change of coast direction at Old Road Bay. According to the models developed for the MVO by French colleagues (Caristan et al., 1997) a 25 million m³ landslide entering the ocean at 50 m/s should produce a source wave height of 5 m. These conditions are close to the estimates of conditions for the mouth of the White River soon after 3 am on Boxing Day. The coastal wave will decay at about the square root of the distance so should be about 1 m height at 10.7 km distance at Old Road Bay. This seems consistent with the observations. The wave moved inland here because the coast abruptly changes its direction and the wave moving parallel to the coast would have met the shore head-on. Also, the shallow offshore bathymetry and onshore topography in the area aided extended wave run-up.

Dynamical Constraints on the Boxing Day Event

This section provides some preliminary estimates of the energetics and dynamics of the Boxing Day event. The event consisted of a debris avalanche that originated from the Galway's Wall and Galway's Soufriere area, dome collapse that caused pyroclastic flows with accompanying ash cloud surge, and some sort of explosive surge/blast(s). The following constraints can be placed on the dynamics. The basic source of models for plumes and surges comes from Sparks et al. (1997) and Bursik and Woods (1996), including some new calculations carried out by Andy Woods for MVO.

Debris Avalanche

The debris avalanche surmounted the bend in the White River at 3 km from its source and climbed a height of 60 m. This climb height yields a minimum velocity of 35 m/s from conversion of kinetic to potential energy and neglecting frictional losses.

Pyroclastic Flow and Ash-Cloud Surge

The only control on pyroclastic flow velocity for this event is that flows jumped out of the valley and mantled DAD which had already been emplaced. This suggests that they had higher velocity than the DAD. Typical velocities of similar flows on Montserrat are 20 to 30 m/s, although velocities as high as 60 m/s have been recorded (Cole et al., 1998). The great degree of erosion seen on the flanks of the White River valley suggest that, although the Boxing Day flows are similar in nature to previous dome collapse flows, their velocity was markedly higher than typically generated during dome collapses. A velocity of 35 to 60 m/s seems reasonable.

Data from the GOES8 satellite constrain the eruption column height to be approximately 14.3 km. A satellite image at 03:15 (almost at the end of the main collapse event) indicates a wide source for the plume and it is inferred that this was a typical distributed plume formed by buoyant lofting from ash cloud surge derived from the pyroclastic flow. The total volume of pyroclastic flow material is c. 30 million m³ erupted over a 15 minute period, giving a flux of 30 to 35,000 m³/s. Empirical data collected over the past 2 years on Montserrat indicates that c. 15% of deposit volume for pyroclastic flows and surges feeds into the co-ignimbrite plume. The flux into the co-ignimbrite plume is thus 4,500 to 5,000 m³/s. This is in very good agreement with plume-rise models, which can be used to imply a flux of c. 4,000 m³/s.

There are some constraints on velocity of the ash-cloud surge. First, the surge climbed almost to the top of the South Soufriere Hills from the saddle between Galway's Mountain and the South Soufriere Hills (at which point part of the surge condensed to form the secondary pyroclastic flow in Dry Ghaut). The saddle is at 1.5 km from the dome source and the height climbed is about 200 m. Using the simple conversion of kinetic to potential energy gives a minimum velocity of 60 m/s. This direction is well off the axis of the main collapse direction so the main axis velocity is likely to have been significantly higher.

The St. Patrick's seismic station was destroyed at 03:03.3. This is 90 seconds after the onset of the first pulse in the collapse signal, giving a minimum velocity of 40 m/s. The head of a turbulent gravity current travels about 70% of the speed of the following current so this observation suggests speeds of 50 m/s. This is very much a minimum estimate as the first flow to reach St. Patrick's may have been smaller than the peak flow.

Surge/Blast

Calculations have been made by Andy Woods (Bristol University) for MVO which are particularly appropriate for modelling the surge/blast. Relevant documents are held at MVO and are based on the work published in Bursik and Woods (1996). The model is most appropriate for a flux of a dilute turbulent suspension from a source and so can be used for modelling the generation of a surge/blast by the disintegration and explosive expansion of the pressurised dome at source.

The surge would have been a supercritical flow (see Bursik and Woods, 1996). The closest model is for a radial flow down a cone with 10° slope (close to the average slope of the volcano from the dome to the sea) with a flux of 10⁸ kg/s and mean grain size of 1 mm for a particle density of 1,000 kg/m³. Note that this flux has been modified to take account of the 70° eruption sector. Preliminary sieve analyses of the samples from the surge deposit indicate a mean size between 300 and 150 microns. Although there are coarse facies in the deposit, the flow would still be dominated by fines. The surge is made predominantly of dome rock lithics and crystal ash, so the mean settling velocity is comparable to that used in the Woods model.

The model gives a velocity of about 100 m/s with a maximum of 110 m/s at 2 km distance and a velocity of 90 m/s at the end of the flow. The model run-out distance is 3.8 km and is determined by the distance at which the surge lifts off. The Boxing Day surge goes out further than this model predicts. This could be due to a number of factors including finer grain size in the surge, higher peak flux, differences between the simple model topography and the actual topography and less efficient entrainment than the model.

A critical question is whether such an event on the northern flanks would penetrate into the currently populated zone. Until new information (possibly provided by satellites) emerges the total run-out distance of the Boxing Day surge is unknown. The models of Woods imply a run-out of only 3.8 km, but these models are subject to significant uncertainties. The sensitivity tests in Woods and Bursik (1996) alone show that significantly finer grained flows or those that entrain less air could reach 5 km. This would still not be enough to reach Salem. However, run-out distance is a strong function of flux so that an event with a flux of 2.4 x 10⁸ kg/s (2.4 times the modelled flux) should be capable of reaching Salem.

Interpretations of Field Data and Observations

Despite the reconstruction of events being made difficult by the lack of visual observations and the problems in interpretation of the seismic evidence in terms of source processes, a good picture of the events of Boxing Day has emerged.

The single event on which there is a good time is the loss of the St Patrick's seismic station - however, field inspection gives no indication about the sequence of events at this location - much of the deposits in this area have been reworked but local remnants comprise a 15 cm layer of coarse sand surge ash and a thin layer of co-ignimbrite ash. It cannot be assumed that the loss of the station was simultaneous with the devastation of this area; a relatively minor event could have damaged the transmission electronics, which were outside, modestly protected under a balcony and open towards the volcano.

There is clear evidence that the debris avalanche was the initiating event in the collapse sequence, and there is no indication that there was deposition of pyroclastic flow or surge material or significant erosion prior to deposition from the debris avalanche. Constraints on the relative velocities for the debris avalanche and pyroclastic flows for this event (see above), supported by empirical data, would suggest that the debris avalanche started a minimum of 30 to 60 seconds prior to the first pyroclastic flow in order to have reached the end of the White River valley first.

The relative roles of dome talus instability and weak, altered rock within the Galway's Soufriere in prompting and propagating the debris avalanche is not clear. Loading of a crater wall area with a weak base clearly played an important part in the initiating landslide, but extrusion of a lava lobe so far over an inherently unstable talus slope must also have played some part in initiating the dome collapse (Figures 1, 9). This point has particular relevance to assessing the potential for this type of event elsewhere at the Soufriere Hills - weakened crater wall may not be the only prompt for debris avalanche events.

There are two lines of evidence to suggest that the seismic signal generated by the debris avalanche may have preceded the onset of collapse as defined seismically above. It is known that a significant amount of signal energy for rockfall and pyroclastic flow events is due to source effects at the dome rather than transport effects. The high level of 'dome' seismicity in the form of merged hybrid events could well mask any seismic energy generated by movement of the debris avalanche. The transport and emplacement mechanism for the debris avalanche would also tend to make it less seismically noisy. Station effects mean that frequency variations do not necessarily assist in finding a seismic signal for the debris avalanche; however, one would expect higher frequency signals to be enhanced during emplacement of a debris avalanche. There is no high frequency energy prior to or during the collapse event. It thus seems reasonable to suggest that the debris avalanche might have started prior to 03:01 but no further constraints can be given.

The onset of vigorous dome collapse and the start of the first coarse pyroclastic flows in the White River valley may then be taken as 03:01. The pulse-like nature of such dome collapse flows in the past indicates that pyroclastic flow generation would have continued for many minutes. Deep scouring of the sides of the White River valley and of the upper surface of the DAD is consistent with overrunning by energetic pyroclastic flows. Ash cloud surges would have been generated at this time, causing the blow-down and singeing of trees on the southeast side of the White River valley. It is unclear as to whether the surge damage to the north of Gingoes Ghaut occurred from marginal ash cloud surges from these events or whether they were derived in association with the subsequent 'blast' surges. Other ash cloud surge damage to the northwest of the valley was superseded by greater damage by later events. Secondary pyroclastic flows in the deeper ghauts were generated by condensed surge material.

A small tsunami was generated as the debris avalanche, perhaps synchronous with early pulses of pyroclastic flows, entered the ocean at the mouth of the White River valley. The tsunami wave was refracted around the coastline of Montserrat and achieved considerable run-up in Old Road Bay due to local focusing effects.

The nature of the surge deposits, the degree of damage, and the uniformity of flow direction directly from the dome suggest that the phenomena impacting last on the area between a few hundred metres northwest of the White River and Gingoes Ghaut are different to the conventional pyroclastic flows and ash cloud surges seen previously during the current eruption and focusing in this case on the White River valley. Our preliminary interpretation is that a single or series of laterally directed blasts occurred after or during the final phases of the main phase of dome collapse and pyroclastic flow emplacement. These blasts were followed by or came during the period of ash venting, and were probably driven by rapid decompression of inner dome material with relatively high pore pressure. Such surface blasts during dome collapse have been seen before at Montserrat on a smaller scale (February 1997), where vigorous rock-bursting occurred off the exposed scar. The blasts were relatively modest in intensity by world standards.

The lack of evidence for vertical explosive activity and highly gas-charged magma (lack of low density pumice) suggests that the main, vertical feeder conduit was not unroofed during the collapse event as has been interpreted to have happened during previous episodes of explosive activity at the Soufriere Hills volcano (MVO Special Reports 1, 1996; 4 and 5, 1998).

References

Caristan Y, Bouchez J, Heinrich P, Guiborg S, Roche R. 1997. Estimation des hauteurs de vagues aux Antilles generees par un effondrement aerien sur l'ile de Montserrat. Report of Commissariat a l'Energie Atomique, February 1997 (in French).

Cole PD, Calder ES, Druitt TH, Hoblitt RP, Lejeune AM, Robertson REA, Smith AL, Stasiuk MV, Sparks RSJ, Young SR and MVO staff. 1998. Pyroclastic flows generated by gravitational instability of the 1996-97 lava dome of the Soufriere Hills volcano, Montserrat. Submitted to Geophysical Research Letters.

McGuire WJ, Jones AP and Neuberg J (eds.). 1996. Volcano Instability on the Earth and Other Planets. Geological Society Special Publication 110. 389 p.

MVO Special Report 1. 1996. Report of the explosive event of 17/18 September 1996.

MVO Special Report 2. 1997. The Galway's Wall Crisis, November 1996 to March 1997.

MVO Special Report 4. 1998. Vulcanian explosions during August 1997. (In preparation).

MVO Special Report 5. 1998. Dome collapse and Vulcanian explosions during September and October 1997. (In preparation).

Sparks RSJ, Bursik MI, Carey SN, Gilbert JS, Glaze LS, Sigurdsson H, Woods AW. 1997. Volcanic Plumes. J Wiley and Sons, Chichester. 574 p.

Stasiuk MV, Shepherd JS, Barclay J, Calder ES, Cole PD, Herd RA, James M, Lejeune AM, Sparks RSJ, Toothill J, Watts R, Young SR and MVO Staff. 1998. Volume measurement methods for lava domes and related deposits developed for the Soufriere Hills Volcano, Montserrat. Submitted to Geophysical Research Letters.

Woods AW and Bursik M. 1996. The dynamics and thermodynamics of large ash flows. Bulletin of Volcanology, Vol. 58, 175-193.

Montserrat Volcano Observatory