Montserrat Volcano Observatory, Montserrat, West Indies

Scientific Report 45
08 December 1996

This report is for the 15-day period up to 24:00 8 December.


The level of activity at the Soufriere Hills volcano reached its most concerning level of the current eruption during this period. The south part of the crater wall, at Galway's, became increasingly unstable, and large cracks appeared across the crater wall adjacent to Chances Peak. A major volcano-tectonic earthquake swarm, with over 1700 detected earthquakes of up to magnitude 3, occurred between 30 November and 8 December. The earthquakes were located at shallow depths beneath the crater. The maximum size of the earthquakes increased with time, and several were felt by workers at the crater rim. Growth of the October 1 dome slowed, and perhaps stopped completely. The alert level was increased from Amber to Orange early on 28 November and the volcanic ash map was re-assessed twice.

Visual Observations

Visibility at the start and end of the period under review was excellent, and enabled good observations of both the dome and the crater walls.

Two detailed surveys were completed (23/24 November and 1 December). Maps of the growing dome were constructed from kinematic GPS surveys and photographic and theodolite data.

Growth during this period was predominantly in the northern and south western sectors of the October 1 dome, resulting in the flattening of the top of the dome into a broader plateau with no change in the maximum vertical height. The volume of the October 1 dome was 4.47 million cubic metres (Dense Rock Equivalent) on 24 November and 4.65 million cubic metres (DRE) by 1 December. This is equivalent to a mean extrusion rate of 47,500 m3/day; significantly less than the mean extrusion rate of 86,250 m3/day reported for the period 7 to 24 November. The active north-eastern face of the dome remained in equilibrium with limited amounts of growth balanced by the removal of material during rockfalls in two erosive chutes. 180,000m3 (DRE) of material was added to the developing talus fan associated with the new dome. Some small pyroclastic flows were also generated during this period; the largest of which (27 November) reached a maximum runout 800m beyond the high point of the dome.

Although no quantitative surveys of the dome have been completed since this date due to cloud cover, detailed visual observations suggest that dome growth was extremely limited for the latter part of the reporting period, concurrent with the marked decrease in dome rockfall activity recorded by the seismic networks. Night observations on 29 November indicated that exogenous growth was no longer very vigorous with continual incandescence limited to the centre of the summit plateau, associated with the growth of a small (<2m) spine.

Overnight on 27 November, heavy rainfall prompted significant generation of lahars due to remobilisation of old pyroclastic flow material in the Tar River valley and landslide material in the Whites River. Collapses from the October 1 dome caused small pyroclastic flows at the same time.

Galways Wall

Major landslides from the Galway's Wall, part of the southern crater wall, were first noticed in mid-October. By 26 November, the wall was showing signs of considerable weakening with a series of fractures visible on the wall surface, and talus debris from small rock avalanches were observed both on the inner (towards the new dome) and outer portion of the wall.

The wall is largely composed of interbedded lithified talus and block and ash deposits from prehistoric eruptions, cross cut by some small (c. 2m wide) laterally impersistant sills and one anastamosing dyke (c. 3m wide). The base of the wall largely consists of relatively thinly bedded tuffs. These beds dip towards the centre of English's Crater at approximately 15 degrees towards the NE in the western portion of the wall and 5 degrees towards the east in the eastern portion.

On the morning of 27 November, a large rockfall on the Galways Wall was observed by field teams. This produced a scar in the central portion of the wall and was calculated to have removed approximately 150,000 m3 of material. The talus from this avalanche travelled as far as the break in slope, approximately 300m from the crest of the wall. During subsequent helicopter flights vertical and wall-parallel fractures were noted in the wall; on 29 November a series of NW-SE fractures dipping steeply eastwards were also seen. On 1 December two vertical fractures, trending c. 060 degrees, were seen on the eastern col of Chances Peak near the intersection with Galways Wall. These large, through-going fractures were more than 50 cm wide and extended for at least a few metres below the surface of the wall.

Rockfall activity from Galway's Wall (both on the inner and outer wall) accelerated in the later part of the period, strongly association with the intense volcano-tectonic earthquake swarm of 30 November to 8 December. At least two more avalanches of a similar magnitude to that of 27 November have now occurred, along with many smaller events. Coarser rockfalls are associated with avalanching from the central portion of the wall. At the end of the reporting period material from the extreme eastern and western portions of the wall was confined to regolith, which accounts for approximately 50% of the material shed from the outer wall. The finer brown regolith tend to have a longer run out, now reaching 600m from the crest of the wall. Some of this thick regolith layer may be associated with the effects of hydrothermal alteration. Talus from the central portion has become increasingly blocky (blocks as large as 5m reaching the break in slope at the base of the wall), consistent with the increasing penetration of the fracture network within the wall. New fractures were observed within the Galways wall on a daily basis.


Review Of Recent Seismicity

Shallow volcano-tectonic swarm activity began on 21 October. In the 8 days before then, there was diffuse VT activity at depths down to 4 km, located slightly north of the crater. From 21 October until 20 November, there were 11 swarms which lasted between 5 hours and 3 days (Table 1). The gaps between the swarms varied from 20 hours to 2 days. The swarm characteristics changed with the swarm on 1-2 November. Before then, swarms were small, of short duration and with short intervals between them; both less than a day. After 1 November, the swarms had many more events, were much longer and had longer inter-event intervals.

Table 1: Number of VT earthquakes that triggered the short-period seismic network

Swarm   Start          End           Duration  No. of         Earthquakes/hour
                                     (hours)   Earthquakes    Mean Peak
1       08:33 21/10    13:10 21/10    4.6       32            7.0  14
2       20:47 23/10    03:30 24/10    6.7       26            4.0  7
3       00:08 25/10    06:44 25/10    6.5       18            2.8  5
4       03:11 26/10    21:18 26/10    18.1      80            4.4  8
5       17:22 27/10    01:13 28/10    7.8       25            3.2  4
6       14:46 30/10    23:35 30/10    8.8       25            2.8  9
7       00:27 01/11    20:31 02/11    44.1      413           9.4  38
8       19:24 03/11    19:11 06/11    48.9      388           7.9  14
9       17:50 09/11    18:09 11/11    48.3      200           4.1  9
10      01:28 14/11    21:51 14/11    20.4      41            2.0  5
11      06:27 19/11    23:51 20/11    41.4      53            1.3  4
12      18:54 25/11    07:41 28/11    60.8      109           1.8  6
13      06:45 30/11    07:46 08/12    193.0     1671          8.7  21

With the exception of a short period on 2 November, all the VT earthquakes in these swarms were located at shallow depths (0-2 km) beneath the crater. The signals showed no clear S arrivals and had distinct long-period codas. These can be interpreted as either surface waves, consistent with a very shallow source, or as long-period earthquakes triggered by the VT earthquakes. The latter interpretation would indicate that the VTs are located near the conduit or a body of magma. Throughout the swarm activity, there has been no detectable migration of the locations in either depth or horizontal position.

The swarm on 1-2 November was the most intense period of VT earthquake activity since dome growth began in November 1995. It was also the only swarm during the current activity which had hypocentres deeper than 3 km. The swarm was, at first, similar in intensity to the previous swarms, with up to ten events per hour. Deeper VT activity started almost 24 hours later at 00:06 on 2 November. This only lasted about four hours, but was very intense, with up to 38 earthquakes per hour. The depths of these events were mainly between 2 and 4 km and the signals showed clear S phases. Almost immediately after the deep activity there was an increase in the amount of the shallow VT activity, with the rate increasing to up to 20 per hour.

The magnitudes of these swarm earthquakes reached a maximum of about 2.5 during this period. The size of the largest earthquakes gradually increased following the activity on 1 and 2 November.

Rockfall activity from the dome had been at a fairly low level since the pyroclastic flows on 17 and 18 September. A period of significant rockfall activity started on 21 November.

The definite correlation of observed Galway's Wall landslides with seismic signals has enabled retrospective identification of previous large landslides. An analysis of rockfall signals has shown that rock avalanches from the Galway's Wall have been occurring since at least as early as 24 October. The largest of these by far occurred on 4 November. The debris from this avalanche had been spotted on 6 November during a helicopter observation flight.

Seismicity During The Reporting Period

Table 2: Earthquake types: 23 November to 7 December 1996
These earthquake counts are of events that triggered the short-period seismic network event recording system between 0000 and 2400 each day.

Date      VT        LP        Hybrid    RF

23 NOV 96 0 1 0 14 24 NOV 96 1 0 0 19 25 NOV 96 2 0 0 12 26 NOV 96 54 0 0 2 27 NOV 96 45 0 0 4 28 NOV 96 6 3 0 8 29 NOV 96 0 0 0 10 30 NOV 96 64 1 0 9 01 DEC 96 178 0 0 4 02 DEC 96 181 0 0 2 03 DEC 96 256 0 0 3 04 DEC 96 279 0 0 1 05 DEC 96 314 0 0 3 06 DEC 96 159 0 0 3 07 DEC 96 146 0 0 9 08 DEC 96 41 1 0 11

The seismicity during this two-week period was dominated by the largest volcano-tectonic earthquake swarm recorded since the start of the current crisis in July 1995. All these earthquakes were in similar locations to those in the previous swarms, at shallow depths beneath the crater. The swarm lasted almost 8 days, beginning at 06:50 on 30 November, and ending abruptly at 05:30 on 8 December. In that time, a total of 1671 earthquakes were recorded by the short-period seismic network. This represents a mean rate of about 10 events per hour. The rate of earthquakes increased gradually to a high of 21 events per hour on 5 December. It quickly dropped back down and fluctuated between 10 and 15 events per hour until the end of the swarm.

Figure 1 (also below) shows data on this swarm activity from the broad-band seismic network. This has the advantage that signals are not clipped and so measurements on the sizes of the earthquakes can be made. The figure shows data between 23 November (day 328) and 8 December (day 344). Note that the times are in UTC, four hours ahead of Montserrat time. The hourly count of shallow VT events is shown at the bottom of the plot. Above this is the amplitude of individual events at the seismic station at Gages, the station closest to the activity. This can be assumed to be related to the magnitude if the earthquake hypocentres are in approximately the same place, and if there is no significant radiation pattern effect. The plot of amplitudes shows that the size of the largest events has been slowly increasing during this swarm. In the later part of the period, many of the events were felt strongly by MVO staff on Chances Peak and at an observation post on the ridge to the south-west of Galways Wall. Some of the larger events were reported as felt by residents of the closest occupied area, which is Weekes, on the western side St Georges Hill. It is estimated that the magnitude of the largest events is about 3.0. The amplitudes are plotted on a logarithmic scale at the top of Figure 1. This is equivalent to plotting them as magnitude. It is clear that there is a bimodal pattern to the magnitudes, with small earthquakes (M about 1) dominating and few earthquakes of intermediate size. During the increase of activity up until 5 December, the number of small earthquakes increased with time, while the number of large ones remained almost constant. This means that the b-value of the earthquake distribution increased, although these earthquakes clearly do not follow a classical magnitude-frequency relationship.

Figure 1. shows data on this swarm activity from the broad-band seismic network. (click on image for download)

The rockfall activity on the dome, which started on 21 November, continued and peaked in the early morning of 24 November, with the highest amount of rockfall activity since the start of growth of the October 1 dome. Rockfall activity declined after that and had returned to a low level by 26 November.

It is now clear that, since mid-November, there is a strong inverse correlation between dome rockfalls and the shallow VT swarm activity. Dome rockfalls are not completely absent during VT swarms, but some of these may be due to the strong shaking making parts of the dome unstable.

Between 01:39 and 03:00 on 28 November strong, continuous seismic signals were recorded at most stations in the seismic network. The signals were caused by pyroclastic flows in the Tar River Valley, and debris flows in the Tar River, White's River and Fort Ghaut. Further debris-flow activity occurred between 05:00 and 06:00 the same morning.

Signals from landslides on the Galway's Wall continued to be recorded during this period. There has been a gradual increase in the number and size of these landslides. There is also a correlation between times of landslide activity and swarm VT activity. The strong shaking caused by VTs has been observed to trigger landslides on Galway's Wall. It cannot however be the only cause; no landslide activity was recorded during the intense seismicity on 1 and 2 November.

Interpretation Of The Seismicity

Our interpretation of the recent seismic activity is that the VT earthquakes are caused by pressurised magma at shallow depths. The working hypothesis is that the magma outlet becomes periodically blocked, leading to a slowing of dome growth (and a reduction in rockfalls). Until the "blockage" clears, the magma pressure is high, causing rock fracturing around the magma body. One possibility is that the lava is highly non-Newtonian and thus requires a certain pressure to yield and flow.

The correlation between Galway's Wall landslides and the VT swarms suggests that magma pressurisation is increasing the stress on the base of the wall.

The occurrence of deeper VTs at the start of this phase of activity (mid-October) and when the level of activity increased (November 2), suggests that the current phase is a response to some deeper volcanic activity, possibly fresh injection of magma.

Ground Deformation

EDM measurements were made with the MVO total station on the eastern triangle on 23, 28 and 30 November and 4 December. Until 30 November, the measurements showed the lines to Castle Peak shortened by about 6 mm/day, continuing the recent trend. The last measurement showed 3.8 cm shortening over a 4 day period, a significant increase in the rate.

The southern triangle was remeasured on 2 December, following replacement of the EDM reflector on Chances Peak. A line shortening of 2.6 cm since 19 October was recorded, a rate higher than the previous trend on this line. The line from O'Garras to Chances Peak was remeasured on 3 December, and showed no change since the previous day.

Two EDM reflectors were placed in temporary positions on the northern end of Galway's Wall, next to Chances Peak on 2 December. However it was not possible to get reflections from them from O'Garros, probably because of poor atmospheric conditions. Baseline measurements were made across the two large cracks on Chances Peak on 4 December. These were made between nails that had been hammered into trees on either side of the cracks. Both extensional and shear displacement will be measured.

GPS surveys were done on 23, 25, 26 and 27 November and on 5 and 7 December covering the three main networks. A network extending to the north of the island was also occupied. All line lengths and station heights are within the 95% confidence level of their long-term means. An experiment on 1 December showed a 1-sigma error of 3.5 mm on a 1.5km baseline; this is much better than the manufacturer's quoted value of 6.5mm for such a line. The equipment can therefore detect strains significantly below 10^-5 ppm.

Gas Measurements

The COSPEC instrument is currently out of service, and so no sulphur dioxide flux measurements were made during this period.

Hazard Assessment

The alert level was changed from Amber to Orange early in the morning of 28 November, because of the increased instability of Galway's Wall and the fears that catastrophic collapse of the wall might cause a lateral blast. A new risk map was issued on 26 November to account for the increased risk to St Patricks and surrounding areas, which were moved from zones B & C to A. On 3 December, complete closure of zones A to D were recommended as a temporary measure, because the scientific team thought that a larger collapse was possible, involving more of the southern part of the crater wall, and potentially causing major dome instability and pyroclastic flows in any direction. This change was formalised on 5 December with a temporary revision of the risk map, which made all of the south of the island, from Foxes Bay across to Spanish Point, zone A/B.

Tsunami Hazard Assessment

On 29 November, scientists from MVO became concerned that the collapse of the Galway's Wall and a portion of the dome towards the SW might lead to the sudden entry of a mass flow into the Caribbean sea. Such an event might generate tsunami-type waves that could travel to Guadeloupe (60 km away) and potentially other islands. Scientists from the French nuclear agency (CEA) responded on 3 December with initial analytical solutions to the problem of tsunami wave propagation based on data provided by the MVO and on analytical models developed by Slingerland and Voight (1979: in Engineering Sites, Vol 2 of Rockslides and Avalanches, ed. B. Voight, Elsevier, New York).

After a fact-finding trip by french scientists and discussions with MVO scientists and Dr Barry Voight, further detailed bathymetry, volume estimates, and hypotheses were used by the team of Drs Caristan, Bouchez, and Heinrich in France to produce a numerical tri-dimensional simulation (Caristan et al., 7/12/96, map and report) of the propagation of waves along the coast of Montserrat, and their effect on the coast of Guadeloupe in the case of a landslide of the new dome and Castle Peak dome into the Atlantic Ocean to the east.

In the worst-case scenario, a landslide of 80 x 10^6 m3 (includes Castle Peak dome, new dome complex and Tar River pyroclastic delta) would generate a wave that would propagate around the coast of Montserrat. The possiblility of such a volume of material reaching the sea all at once is thought to be very low by MVO scientists. Little Bay would be reached within 9 min by waves less than 1 m in height. In the Plymouth area wave height could be as high as 2 m whereas in Old Road Bay directly at the coast the wave amplitude would be less than 1 m. Run-up height would be a maximum of double the wave height. Waves reaching Guadeloupe after about 10 minutes would have an amplitude between 0.5 and 1 m and a run up height of up to 3 m. The wave height would be less than 10 cm on the coast of Martinique about 250 km southeast from Montserrat.

A similar numerical simulation was carried out for a collapse of Galway's Wall and part of the dome (25 x 106 m3) into the Carribean Sea on the SW (worst case scenario). Final results are not available yet, but suggest an epicentral wave height of 5 m that would gradually attenuate while propagating to the north. Waves reaching Guadeloupe would be smaller than for the Tar River collapse.

Whether collapse will occur suddenly or gradually remains uncertain. In the event of a sudden collapse, the volumes of material reaching the sea are likely to be much less than those used in these worst-case scenario simulations.

Staff changes
Chandradath Ramsingh, SRU
Dr Jean-Cristophe Komorowski, Institut de Physique du Globe de Paris


Dr Barry Voight, Penn State University
Michel Feuillard, Guadeloupe Volcano Observatory
Dr Herve Traineau, BRGM
Olivier Sedan, BRGM
M. Simeon, BRGM
Dr Georges Boudon, Institut de Physique du Globe de Paris
Christian Antenor, Guadeloupe Volcano Observatory

Montserrat Volcano Observatory