alerting click here to receive free alerts when new articles cite this article service

Transcription

1 Geological Society, London, Memoirs Pyroclastic flow and explosive activity at Soufrière Hills Volcano, Montserrat, during a period of virtually no magma extrusion (March 1998 to November 1999) G. E. Norton, R. B. Watts, B. Voight, G. S. Mattioli, R. A. Herd, S. R. Young, G. E. Devine, W. P. Aspinnall, C. Bonadonna, B. J. Baptie, M. Edmonds, A. D. Jolly, S. C. Loughlin, R. Luckett and S. J. Sparks Geological Society, London, Memoirs 2002, v.21; p doi: /GSL.MEM alerting click here to receive free alerts when new articles cite this article service Permission click here to seek permission to re-use all or part of this article request Subscribe click here to subscribe to Geological Society, London, Memoirs or the Lyell Collection Notes The Geological Society of London 2013

2 Pyroclastic flow and explosive activity at Soufri~re Hills Volcano, Montserrat, during a period of virtually no magma extrusion (March 1998 to November 1999) G. E. NORTON 1'2, R. B. WATTS 3, B. VOGHT 4, G. S. MATTOL 5, R. A. HERD 1'2, S. R. YOUNG l, J. D. DEVNE 6, W. P. ASPNALL 7, C. BONADONNA 3, B. J. BAPTE 8, M. EDMONDS 9, C. L. HARFORD 3, A. D. JOLLY 1~ S. C. LOUGHLN 7, R. LUCKETT 11 & R. S. J. SPARKS 3 1 Montserrat Volcano Observatory, Mongo Hill, Montserrat, West ndies 2 British Geological Survey, Keyworth, Nottingham NG12 5GG, UK ( gen@bgs.ac.uk) 3 Department of Earth Sciences, University of Bristol, Bristol BS8 1R J, UK 4 Department of Geosciences, Pennsylvania State University, University Park, PA 16802, USA 5 Department of Geosciences, University of Arkansas, Fayetteville, Arkansas, USA 6 Department of Geological Sciences, Brown University, Providence R 02912, USA 7 Aspinall & Associates, 5 Woodside Close, Beaconsfield, Bucks HP9 1JQ, UK 8 British Geological Survey, Edinburgh EH9 3LA, UK 9 Department of Earth Sciences, Cambridge University, Cambridge CB2 3EN, UK l0 Geophysical nstitute, University of Alaska, Fairbanks, Alaska 99775, USA. 11 nternational Seismological Centre, Pipers Lane, Thatcham, Berkshire RG19 4NS, UK Abstract: Dome growth at Soufri+re Hills Volcano halted in early March After dome growth ceased, seismicity reduced significantly, but activity related to dome disintegration and degassing of magma at depth continued. A sustained episode of pyroclastic flows on 3 July 1998 marked the single largest collapse from March 1998 to November This led to a remarkable episode of dome collapses, low-energy explosions and ash-venting that resulted in the regular production of ash plumes, commonly reaching km above sea level (a.s.1), but sometimes up to 11 km a.s.l., and the development of a small block-and-ash cone around the explosion crater. During the period of this residual activity, higher levels of activity occurred approximately every five to six weeks. This periodicity was similar to the cycles observed during active dome growth during 1995 to 1998, and probably had a similar cause. The relatively high level of observed activity caused continued concern regarding volcanic hazards and their potential to impact upon the resident population. Vigorous magma extrusion resumed in November The activity of the intervening period is attributed to the continued cooling and degassing of the dome, conduit and deep magma body, the impact of rising volcanic gases in the volcanic edifice, and limited magma flow in the conduit. The eruption of Soufri6re Hills Volcano on Montserrat began on 18 July 1995 with phreatic explosions from Castle Peak, a prehistoric (c. 350 years Br') andesitic dome sited within the horseshoeshaped English's Crater (Fig. 1). After an initial phreatic phase, most of the following 32-month period involved activity related to the growth and collapse of a viscous lava dome (Young et al. 1998a). On 17 September 1996 a magmatic explosive eruption followed a 9-hour period of dome collapse involving c. 9.5 x 106 m 3 (dense rock equivalent; DRE) of the lava dome (Robertson et al. 1998; Calder et al. 2002). High magma production rates (7-10m 3 s -1) from May to October 1997 led to major gravitational collapses (e.g. Loughlin et al. 2002) and two series of Vulcanian explosions (Druitt et al. 2002). On 26 December 1997, catastrophic failure of the southwestern wall of English's Crater and flank of the volcano resulted in removal of 55 x 106 m 3 of the lava dome and talus (Calder et al. 2002; Sparks et al. 2002; Voight et al. 2002). This collapse, the largest of the eruption at the time of writing, was followed by rapid new growth of the dome. By early February 1998, the dome had restored the volume destroyed by the 26 December 1997 collapse and for the period 26 December 1997 to mid- February 1998 was estimated at a rate of 6m 3 s -1. Growth of the dome continued through the end of February 1998, but ceased on about 10 March 1998, at which time the total amount of magma erupted since 1995 was 300 x 106 m 3 (DRE), and the volume of the dome at this time was 113 x 106 m 3 (Fig. 2a). During magma extrusion from November 1995 to March 1998, cyclicity in the volcanic activity was noted on several timescales and in several datasets (Voight et al. 1998, 1999; Denlinger & Hoblitt 1999). Major dome collapses or periods of explosivity were thought to occur approximately every six to ten weeks (Voight et al. 1999). On a shorter timescale, cycles of between 8 and 14 hours were often noted in seismicity, tilt data and explosions or increased pyroclastic flow activity (Voight et al. 1998). This latter pattern persisted to early March 1998, although by this time the peaks in activity were barely recognizable. Major dome collapses during the period of magma extrusion were generally associated with intense precursory seismic activity. For example, the collapse on 26 December 1997 occurred after nearly 36 hours of a swarm of hybrid earthquakes (Sparks et al. 2002). The recognition of these short- and intermediate-scale patterns of activity enhanced the forecasting capability of the Montserrat Volcano Observatory (MVO). After the cessation of magma extrusion in March 1998, seismic activity decreased markedly, and the dome appeared to be stable. A sudden large dome collapse on 3 July 1998, without any recognized seismic or other precursory activity, restarted a series of hazardous dome collapses and explosions at a level thought to be unusual for the period following a major phase of dome-building extrusion. There was some evidence that magma may have been slowly ascending up the conduit, and surface deformation recorded by an array of six continuous global positioning system (CGPS) receivers showed ongoing inflation or uplift of the volcanic edifice during the period from February 1998 to November 1999 (Mattioli et al. 2000), suggesting that pressure in the magma chamber and conduit may have been increasing. This led to concerns about whether vigorous magma production at the surface would resume. n mid-november 1999, after 20 months of this residual activity, a second phase of lava extrusion commenced. This paper outlines the phenomena that were observed at Soufri6re Hills Volcano during the intervening period of virtually no magma extrusion from March 1998 to November There are few accounts in the literature that chronicle pauses in magma production or the ends of such dome-building eruptions (e.g. Nakada et al. 1999; Mori et al. 1996), and thus few cases with which to compare the activity under discussion. n general, the scientific literature contains many descriptions of the onsets of eruptions and precursory activity (e.g. Christiansen & Peterson 1981; Chouet et al. 1994), but much less information on dormancy of lava domes or on criteria for recognizing the end of lava dome eruptions. The observations made during the period of virtually no DRUTT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufridre Hills Volcano, Montserrat, from 1995 to Geological Society, London, Memoirs, 21, 467~1L /02/$15 9 The Geological Society of London

3 468 G.E. NORTON ET AL. 375ooone 380o0~ i 62~ ~5ooonE - 18~0000r~'~ 18~O000n~ -- N l HLL Montserrat (UK Overseas Territory) 2km Contour interval is 200 feet (melres = feet x 0,3048) -1855ooonN 1855ooo~ - r~ St Peters W.H. BRAMBLE \ ARPORT - 6~ 16~ - -le50oo0~ 1e50OOOr~- Hill PLYMOUTH!RE HLLS Galway'~Peak e45ooonn - B main road asoooo~ i i i i i i i 62~ q q Fig. 1. Map of Montserrat, showing main towns and locations mentioned in the text.

4 ACTVTY DURNG A PEROD OF VRTUALLY NO MAGMA EXTRUSON ::::::::: "::::::::: :~:::~ :~:::~...,-~,:.....! "~ ~ :~.:~:?...?..., o.. "~ ~ o '=,, ~ o = o "~''%. =, =,. = --~"/ t/ \_1_/\/\_ --'x- / l _ --1.~. ~ ~: ~Tel~l~l ~ "t" + " + + " = " = " k" " \ / ' k + Y,/\ -'~,,,./\ /,,t \ \ -,~[7'r 0 100m i,_ -/_ -/.~;\t -,/~ Exposed v ~,.," ~ \ -",~\ -",~\ / ~.~ crater walip. -- J-. -- '//- -/- -/- -/- - / - ~" x,,,. <.-~.~E~'scrate~ -/ / MARCH 1998 Dome Volume = 113x10~m ~ 13 JULY 1998 Extrusion Rate = zero Dome Volume = *,88 xl0"m ~ Extrusion Rate = zero ,.~ KEY December 1997 lobe 4 November 1997 lobe 22 October 1997 lobe 13 August 1997 lobe 2!~2~i2;~ii!+ 9. o. ". ~176 ~ ~ o o! 1.2ol = = ~ = (, = '~ "l" +..+ t. + = ~, + + +lepn1"& + ~_ +. + ar ~ Chanc s ~,'/,,-/,,, ~+ + ~.~ ** Peak',,,.,+.C,.~: - -,., / 2 -. J '~'" la~ ~ ~, + ~" ' - L / "" +, + +.~ "f \/_"" /_. -,/_. x.z_k-/_.. \, /" '- f" "! /~'/"/,//, -',t',/j \'~J \ t,f //--~ /-\ ; /''- J/--\ i ~ "~ ~/\/~J\ "~/\ / ~t, / ~[\\~'.,~ ~..~Eng~h,s Crate~ ~ "~ 17 May 1997 lobe 21 January 1997 lobe Tephra from explosions Persistent fumarole Measured height (m a.s..) +/- 2 m Rockfall pathway 26 JANUARY 1999 ' -'. ' " k... Dome Volume xl06m 3 Extrusion Rate = zero ~ JL- ~..al. Pyroclastic flow pathway Collapse scar rim r" "7 Blocky growth ""'-,,, Fracture Fig. 2. Map showing the main features of the dome complex as it appeared on (a) 10 March 1998, (b) 13 July 1998 and (e) 26 January 1999.

5 470 G. E. NORTON ET AL. magma extrusion at Soufri~re Hills Volcano have improved our understanding of the evolution of andesitic dome-building eruptions, particularly for hazards assessment and risk management purposes. Of particular importance are the observations that major collapses of the lava dome and Vulcanian explosions with fountaincollapse pyroclastic flows were possible without extrusion of new magma, and with no discernible precursory seismic activity. Monitoring methods Several methods were used to monitor the activity of Soufri~re Hills Volcano (Aspinall et al. 2002) during the period under discussion, including seismic, ground deformation, volcanological, and gasmonitoring techniques. This section summarizes the observations made during the period of dome growth from November 1995 to March 1998, and includes an overview of the data collected during the period of virtually no magma extrusion from March 1998 to November Discussion of these latter data is included in later more detailed sections. Se&mic monitoring The main monitoring method consisted of seismic data collection and analysis (Miller et al. 1998). Five main types of seismic signal were recognized at Soufri~re Hills Volcano: volcanotectonic earthquakes, long-period earthquakes, hybrid earthquakes, rockfall or pyroclastic flow signals, and explosion signals. Elevated levels of volcanic tremor were also measured periodically at Soufridre O (D Q o '~ 300 z E 20o Rockfalls a Volcanotectonic earthquake i... Volcanotectonic swarm on 22 May ~ ( earthquakes Eid of magma extrusion i i: :, 3 July[ 1998 collapse,,"!i i 9,,,,,,',,:, ;,, 9,,' ',,i.,,,,',,.,:, i, 2,,,,,,,,,,,, -,. ~. :.:,',,... 9.:....,.,, ;',. ;., ~ ~ :'; "., -..,,,'.,, ".;',,', :~ 0"..~ 200._o g...a r O > "6 50 r E z O 6 El. E t- t- ~D (D t i / 1 i ii l 9 EDM line length / / Number of dome collapses, / ash venting episodes / or explosions per week J 6 4 c6 9,,e ct} 9 0- E O~o 6 > 0-6 -n O Z O E "O e- "~ O r El_ co.._ c 1-Jan-98 3 July 1998 collapse 1 9 Ash venting on!... Diffusion tubes [ ' ~ August 1998 [ 9 COSPEC. 3.5 i 3.0,~ ', Re-start of dome extrusion 2.5 ll ~,l 9 1 i '' 9 "" 9,l ; 2.0,,D - i, :,, , 9 ', t.5! = Hi i i,? 9 9,,:',...~,,,,: ',,", ) \," mliqmmmmm..-" m~ 9 "i /, ~'.,t " "... '. ~1,~ '.',.",' ' r ~ o t-apt-98 1-Jul-98 1-Oct Jan-99 t-apt-99 1-Jul Oct-99 Date 4.0 "O w Z, -13 2o t-.j O-13. coo Fig. 3. Key monitoring data for 1 January 1998 to 30 November (a) The number of rockfalls and volcanotectonic earthquakes recorded each week by the MVO digital seismological network. (b) Changes in line length between a reflector site on the northern wall of English's Crater to an instrument site 3 km to the north of the volcano at Windy Hill (Fig. 1). Each point on the graph represents the average of ten direct measurements of slant distance done with a Leica TC1100 Total Station (see Jackson et al. (1998) for details of the technique used at the MVO). The number of dome collapses, explosions or ash-venting episodes per week is also shown. (c) SO2 flux as measured by COSPEC (x 10 6 kgday -L) and SO2 concentration from diffusion tubes (ppb).

6 ACTVTY DURNG A PEROD OF VRTUALLY NO MAGMA EXTRUSON 471 Hills Volcano. Volcanotectonic earthquakes had impulsive starts and then rapidly decreased in amplitude. They were predominantly high-frequency signals (>2 Hz). They were common at the start of the eruption and occurred sporadically during March 1998 to November 1999, often in swarms. Long-period earthquakes had a more emergent start and generally low frequency content (1-2 Hz), and were common during active dome growth. Hybrid earthquakes had impulsive starts but contained a significant amount of lowfrequency energy. These signals were often associated with periods of rapid dome growth, and hybrid swarms were recognized as precursors to major dome collapses during dome growth. Both long-period and hybrid earthquakes were rare during the period of virtually no magma extrusion. Rockfall or pyroclastic flow signals were the most common types of seismic signals. They had an emergent start and a gradual decrease in amplitude towards the end of the signal; they contained a mixture of frequencies. Pyroclastic flow signals were simply long-duration and high-amplitude rockfall signals. Explosions had quite characteristic signals. They often had a long-period component (1-2 Hz) at the start, followed by a highamplitude, mixed-frequency signal related to pyroclastic flows (Druitt et al. 2002). After the pyroclastic flow signal had died away, the long-period signal remained, but at a lower amplitude than the initial signal. This tremor was often related to ash-venting, and sometimes continued for many tens of minutes. The main types of seismicity recorded during the period of virtually no magma extrusion were volcanotectonic earthquakes, rockfalls, tremor and explosion signals (Fig. 3a, b). E 2o --,.. E 10 :~ 0.J O E F: lo 0 9 ~ -lo r -20 a 3 July 1998 collapse. 19 December 1998 exdosion -40 Deformation monitoring i Several different methods were used to measure the deformation of the flanks of the volcano, the main methods being global positioning system (GPS) geodesy, electronic distance measurement (EDM), and tilt measurements. During the period of virtually no magma extrusion, GPS data were downloaded daily from three Leica (MVO) and three Trimble (University of Puerto Rico) CGPS receivers stationed around the volcano. n addition, over 20 sites were occupied occasionally by temporary GPS surveys to provide a wider coverage of the deformation field of the volcano. EDM was used frequently in the early stages of the eruption to monitor the deformation of the flanks of the volcano, but three out of four networks were destroyed as a result of volcanic activity during 1997, and reinstallations of near-field reflectors were repeatedly destroyed. Electronic and dry-tilt measurements were taken periodically during the ongoing eruption. Electronic tilt was particularly successful during periods of rapid dome growth when cyclical activity with time periods of typically 8 to 14 hours was measured on the tiltmeters, showing repetitive inflation (with associated hybrid earthquake activity, and culminating in pyroclastic flows or explosions) and deflation (Voight et al. 1999; Denlinger & Hoblitt 1999). Tilt stations close to the summit of the volcano were destroyed in August 1997; stations at a greater distance away from the dome showed little or no variation. n general, large deformations of Soufribre Hills Volcano were confined to areas close to English's Crater, although measurable, volcanically induced deformation was recorded as far away as 8 km from the dome. n addition, a relatively large differential movement of + 10 cm was noted, increasing the distance between a site at Long Ground, 2 km to the ENE of the volcano, and a site at White's Yard, 3 km to the NE of the volcano (see Fig. 1), from June 1996 to December Since these sites are only 733m apart, this movement could not have been accommodated elastically. Ground inspection revealed a fracture that crossed the survey line, trending at 068 ~ There was insufficient evidence to determine whether this surface fracture was related to volcanic deformation or to localized mass movement disrupting the ground surface. Apart from minor, short-term variations, three certain trends were observed in the GPS data collected between March 1998 and November 1999: (i) eastward movement of stations to the east of :E i C 1-dan-g8 1-Jul-98 1-Jan-99 1-Jul-99 1-Dec-99 Date Fig. 4. Hermitage Estate CGPS coordinate changes, 1 January 1998 to 1 December 1999, showing (a) latitude, (b) longitude, and (c) vertical co-ordinate changes. All data are residuals plotted relative to a best-fit curve through the entire dataset since June the volcano; (ii) higher rates of movement for stations within 2 km of the dome; and (iii) significant vertical displacement at all GPS sites, implying inflation or uplift of the edifice relative to the Earth's centre of mass (Fig. 4). The rate of movement of an EDM reflector on the northern wall of English's Crater (Fig. 1) fell dramatically in March 1998 when dome growth stopped (see Fig. 3b). The rate of shortening of the line from the reflector to the instrument site at Windy Hill to the north dropped by an order of magnitude, from a peak of about 3 cm per month between late January and early March 1998 to less than 1 cm per month from early March to late May From late

7 472 G. E. NORTON ET AL. May 1998 to early September 1998 the line lengthened considerably, and then shortened again until the reflector site was destroyed in early October Volcanological monitoring Observations of the volcanic activity constituted a large proportion of the monitoring effort. Many different techniques were used, but the main parameters that were measured included the measurement of dome and deposit volumes (and consequently rates of extrusion), the geology and petrology of the new deposits, and measurements of ash cloud heights. These observations will be described in more detail below. Gas monitoring Two methods of monitoring gas emissions were used regularly throughout the eruption. The first method used a correlation spectrometer (COSPEC) to measure the daily output of SO2 from the volcano. SO2 fluxes during the period of dome growth fluctuated, but generally showed an increase in flux as the rate of magma extrusion increased (Young et al. 1998b). There was also evidence for shortterm fluctuations of SO2 flux in concert with short-term cycles in seismicity (Watson et al. 2000; Young et al. 2000). The maximum SO2 flux rate recorded during dome growth was 2.5 x 106 kg day -1, although the long-term average flux was c. 0.5 x 106kgday -1 (Young et al. 1998b). COSPEC data were not collected during periods of high magma extrusion rate from July 1997 to February 1998, or at various other times due to instrument malfunction or when precluded by inaccessibility due to volcanic activity. SO2 was also measured at ground level using diffusion tubes that were left at sites around the volcano for about two weeks (Norton 1997). Trends in these data closely tracked long-term averages of COSPEC data so that, when COSPEC measurements were not possible, diffusion tube results provided a useful proxy for SO2 output from the volcano. SO2 flux continued to vary during the period of virtually no magma extrusion (Fig. 3c), but generally remained comparable to gas fluxes recorded during the period of dome growth. This is somewhat surprising given the previously noted positive correlation between extrusion rates and gas fluxes, which would have suggested that gas fluxes in the period of virtually no magma extrusion should have been very low. The highest values of SO2 emission from March 1998 to November 1999 were usually associated with pyroclastic flow activity or periods of ash-venting (Fig. 3b). SO2 flux then tended to decrease after ash-venting episodes. One other method of measuring gas concentrations has been used occasionally on Montserrat: Fourier transform infrared spectroscopy (FTR; Oppenheimer et al. 2002). Using this method the ratios of different gases were measured, such as the ratio of hydrogen chloride to sulphur dioxide. During the period of virtually no magma extrusion, FTR was used during July through October 1998, and again in January 1999, and the results indicated that the ratio of sulphur to chlorine gases had increased by more than one order of magnitude in comparison with measurements first done in 1996 during magma extrusion (Oppenheimer et al. 2002). Chronology from March 1998 to November 1999 March to June 1998." cessation of dome growth and slow degradation of the dome On 1 March 1998, the summit of the lava dome was measured at 1011 m above sea level (a.s.1.) and was blocky with a small number of short spines. These spines grew rapidly during the ensuing week, and, on 9 March 1998, the most prominent spine was measured at a height of 1027m a.s.1. Seismic activity, which had been low since early February 1998, declined further in March, and volcanotectonic earthquakes and rockfall signals became the most common types of seismicity recorded. Hybrid earthquakes had been the dominant type of earthquake during dome growth (Aspinall et al. 1998; Miller et al. 1998). Theodolite measurements during clear conditions on 5 April 1998 indicated that the highest point on the dome was the top of the large spine that had grown rapidly before 10 March 1998 in the SW sector of the dome. This spine had attained a final elevation of 1031m a.s.1. (Fig. 2a). A survey of the dome complex conducted on 10 March 1998 indicated that the total dome volume was 113 x 106 m 3 (DRE). Visual observations of the dome complex from 10 March 1998 to 3 July 1998 indicated no major changes in the morphology, implying that magma extrusion had ceased. Slow degradation of the upper parts of the dome by occasional rockfalls from early March to late June 1998 formed deep gullies on the eastern and southwestern flanks. The average number of rockfalls decreased from 390 per week in February 1998 to 44 per week during March 1998 (Fig. 3a). Some of the larger rockfalls had runouts of as much as 1 km, and reached the base of the talus slope. Some small pyroclastic flows occurred down the eastern flank, the largest of which travelled down a narrow ravine and reached a runout distance of 1.8 km. These collapses each lasted a few minutes, generated minor amounts of convecting ash, and left fine-grained, thin deposits (<5 m thick) in a well developed chute extending from the upper flanks. Active fumaroles developed in a V-shaped cleft on the eastern side of the dome and in a gully between the western edge of the 26 December 1997 scar (Fig. 2a) and the area of new growth within the scar. Earthquake activity from early April to late June 1998 consisted principally of small amounts of volcanotectonic seismicity located at depths of km below the summit of the dome. Two exceptions to this were a small swarm of 12 high-amplitude hybrid earthquakes that occurred on 6 May 1998, and a volcanotectonic earthquake swarm on 4-5 June Ground deformation measurements (EDM) on the northern flank of the volcano indicated that the rate of deformation had slowed markedly between early March 1998 and late May 1998 (Fig. 3b). Major dome collapse of 3 July 1998 A sudden increase in activity occurred on 3 July 1998, with a major dome collapse to the east. No precursory seismic activity was recognized, although the average number of rockfall signals had increased slightly from 0-4 per day through most of June 1998 to 8-13 per day three days prior to the collapse. The seismic signal caused by the collapse started with a very high-amplitude, short-lived onset at 03:02 local time (all times are local time = UTC-4 hours); this was followed by approximately 2.5 hours of pyroclastic flow signals. Two high-amplitude pulses at 03:50 and 04:30 are evident in the seismic data, although neither was as large as the seismic signal at the initial failure. The main collapse period was followed by a period of heightened rockfall activity and an increase in the number of volcanotectonic earthquakes, which lasted until 14:07. The volcanotectonic earthquakes were probably caused by depressurization of the system during and after the collapse. The ash plume generated by the collapse rose to 9-14 km a.s.1. and the top of the cloud moved to the ENE. The lower-level plume moved to the NW. Most of the ash fell as accretionary lapilli (diameters as much as 2-4 mm in northwestern Montserrat) and the ash layer reached a thickness of 2 mm c. 6 km NW of the volcano. Pyroclastic flows were observed in the Tar River valley at 04:22 and had reached the sea 3km to the east by 04:50. The associated pyroclastic surge was well developed, scoured the south side of the Tar River valley and extended 700 m to the north of the valley (Fig. 5). A small explosion occurred later in the afternoon at 14:07.

8 m Downloaded from at University of Bristol Library on March 10, 2013 ACTVTY DURNG A PEROD OF VRTUALLY NO MAGMA EXTRUSON 473 Key Block and ash flow deposit Surge deposit r'f- 3July1998 " scar English's Crater -'White's 1 km 9 Yard Contour interval is 200 feet ~ ~ [metres = feet x 0,3048] _ } mN- Tar River Pyroclastic Fan 16~ m N - ;Peak Fig. 5. Map of deposits from the 3 July 1998 dome collapse. 385OoomE Ballistic blocks up to 1.5 m diameter, observed in a field of impact craters 1 km SE of the dome summit, suggested that this (or some component of the main collapse) might have been a small Vulcanian explosion. The collapse is estimated to have removed 15-19x 106m 3 (DRE) of the lava dome, leaving a canyon-like feature extending deep into the SE flank of the dome, with chutes leading down to the south edge of the Tar River valley (Fig. 2b). The pyroclastic fan at the mouth of the Tar River valley was extended 350 m to the north, and c. 6 x 106 m 3 (DRE) of material was added. A substantial volume of material must have been deposited offshore of the Tar River valley. After the collapse, strong fumarolic activity was observed along a clearly defined NW-trending linear fracture, m long, at the base of the new scar. This fracture was located in a general position that had been recognized since September 1997 as the site of vigorous fumarolic activity, and the rocks in this vicinity were probably hydrothermally altered. Significant pulse-like ash-venting continued for two to three weeks, and fumarolic activity was also evident on the southern and northern flanks of the dome. Watts et al. (2002) argue that the dome consists of a shell of loose blocky talus surrounding a main core of massive lava. The combination of hydrothermally altered rocks in the area of the 3 July 1998 collapse, and the presence of loose blocky talus, created a weakened area in the dome, which collapsed without precursory seismic activity. However, there was also intense rain prior to and during the collapse, and it is possible that rain infiltrated into the dome interior through the steep talus slope on the eastern flanks and contributed to the initiation of dome failure. This is not unprecedented, since rainfall-induced dome collapses have previously been recognized at Merapi Volcano, ndonesia (Ratdomopurbo & Poupinet 2000; Voight et al. 2000). Measurement of SO2 flux with a COSPEC in early July 1998 gave values of x 106kgday -1 between 5 July and 11 July, with one occasion when it rose to >4 x 106 kg day -1 on 13 July These levels represented the highest measured SO2 fluxes since the onset of the eruption. The flux rate declined more or less steadily throughout the rest of July to an average of 1 x 106kgday -J (Fig. 3c). Similarly, at a diffusion tube site 4.5 km west of the volcano, SO2 from March to June 1998 was generally below 25 ppb (Fig. 3c); this rose suddenly to over 200 ppb at the same site averaged over the two-week period from 29 June 1998 to 13 July This high flux suggests that high gas pressure may have existed within the dome and upper conduit prior to the collapse, and this gas pressure may have contributed to the failure. Steam generated by rainwater percolating into the dome interior may have further enhanced the gas pressures within the dome. Gas pressure acts at the boundaries of potential failure blocks within the dome, with an uplift force that reduces frictional resistance, and an outward push that increases destabilizing forces. The net effect of gas pressurization is thus to increase instability, leading to dome collapses, as discussed by Voight & Elsworth (2000). The northern EDM line from a reflector on the northern wall of English's Crater to Windy Hill lengthened by 8 cm between May and August 1998, although the data are sparse between May and July This may indicate a decrease of loading stresses on the wall of English's Crater following the 3 July 1998 collapse (Fig. 3b). n the weeks following the collapse, no fresh magma apparently reached the surface, despite the wholesale depressurization of the vent area due to the loss of dome above it. August to November 1998." small dome collapses Earthquake and rockfall activity remained at an elevated level immediately after the 3 July 1998 collapse, but then gradually declined through July, with the exception of a volcanotectonic earthquake swarm on 25 July The early part of August was quiet with few rockfalls or earthquakes. Active fumaroles on the southern side of the dome and associated rockfalls gradually undermined the dome and led to two episodes of pyroclastic flow activity down the southwestern flanks of the volcano during the second week of August. The flows travelled 1.8 km from the dome and each episode was followed by about an hour of continuous rockfall activity. A steep lava buttress overhanging the 3 July 1998 scar collapsed in mid-august and generated a series of pyroclastic flows, which reached the Tar River pyroclastic fan, 3 km east of the dome. An intense period of ash-venting, which lasted from 19 to 21 August, was correlated with nearly monochromatic volcanic tremor and high SO2 fluxes (>1 106kgday -1 measured by COSPEC; concentration of >300ppb measured by diffusion tubes). By late August, the level of activity had again declined, with low seismicity and a general reduction in levels of SO2 flux to an average of about 0.5 x 106 kgday -1 (Fig. 3c). Gradual degradation of the lava dome continued during September 1998, with three periods of pyroclastic flow generation caused by small-volume collapses to the east, west and north.

9 474 G. E. NORTON ET AL. A swarm of volcanotectonic earthquakes on 26 September 1998 was followed by periods of ash-venting and associated volcanic tremor, which continued almost daily until the end of the month. SO2 fluxes increased towards the middle of September 1998 in tandem with observations of enhanced ash-venting from the dome fumaroles (Fig. 3c). Measurements on the EDM line from the northern wall of English's Crater to Windy Hill made in September 1998 indicated that the line had again begun to shorten (Fig. 3b). The reflector was destroyed by pyroclastic flows in early October The CGPS site at Hermitage Estate (Fig. 1), located approximately 500 m from the dome, showed its most rapid horizontal acceleration to the NE from early September to the end of December 1998 (Fig. 4a, b). Small-volume dome collapses increased in frequency during October 1998, leading to further degradation of the dome. Five main collapses occurred on 13 October at 08:01 (to the NE and NW), 18 October at 09:16 (direction uncertain), 20 October at 22:41 (to the west), 26 October at 00:51 (to the east and south) and 31 October at 04:18 (to the south and west), each producing deposits up to 3 km from the dome. Ash plumes reached 2-8 km a.s.1, and deposited ash predominantly to the west and NW of the volcano, but light tephra fallout was also experienced in the north of the island. Volcanotectonic earthquake swarms located at 3-4 km depth were associated with the collapses on 13 and 18 October, but there was no anomalous volcanotectonic seismicity associated with the other three collapses. Volcanic tremor and associated ash-venting were recorded after each collapse and lasted for a few minutes to several hours. By late October 1998, collapse of dome lava down Gages valley had exposed the incandescent interior of the dome and excavated a deep gully, which extended to the southern side of the dome. A large fracture, oriented NE-SW, developed in the SW part of the dome. There was an apparent decrease in SO2 fluxes towards the end of October, and a single measurement on 2 November 1998 gave 0.7 x 106 kgday -1 (Fig. 3c). November 1998 began with a swarm of volcanotectonic earthquakes on 1 November, the largest of which were felt throughout the island. The frequency of small-volume dome collapses and the runout distances of the resulting pyroclastic flows increased in early November, reaching 3 km from the dome to the east on 3 November, and 5 km from the dome to the SW on 5 November. Volcanotectonic earthquakes again followed very rapidly from the collapse on 5 November. The largest collapse occurred at 06:07 on 12 November and probably involved a dome volume of 2.5 x 106 m 3 (DRE). This resulted in pyroclastic flows that reached the sea to the east, west and south. This collapse may have had an explosive component, as suggested by simultaneous generation ofpyroclastic flows in three different directions, possibly by fountain collapse. A convective ash plume rose to 8 km a.s.1. The collapse was followed by vigorous ash-venting, which caused continued tephra fallout to the NW of the volcano. As a consequence of this collapse and the previous collapses in October and November, a deep gorge was cut into the dome running approximately ENE-WSW. The gorge was about 150 m deep and 30m wide, and it split the dome into two peaks (Fig. 6). n the weeks following the large 12 November 1998 collapse, there were a few small pyroclastic flows and some periods of low-amplitude seismic tremor coupled with ash-venting. December 1998 to May 1999." minor explosions with ash-venting Small-volume (<106 m 3) dome collapses continued into December 1998, but then the style of activity changed to unequivocally explosive behaviour. At the onset of a small collapse at 10:34 on 19 December, powerful black jets of ash and rock burst from the east side of the dome, preceding the generation of a pyroclastic surge to the east. This was the first direct evidence of explosive activity, although it was unclear whether it preceded or followed the dome collapse. The resulting deposit was small in volume and almost entirely confined to a pre-existing incised channel within the Tar River valley. On 21 December, at the onset of a sudden, large seismic signal, dense black jets of ash and vigorously convecting ash clouds were observed coming from the main vent in the 3 July 1998 scar. Ballistic blocks were observed to heights of about 80m above the vent. Very vigorous ash-venting continued for over 30 minutes after the initial explosion, but no pyroclastic flows were generated. No pumiceous fallout occurred from these explosions. The expulsion of ballistics suggests that new magma may have risen to shallow levels in the conduit, plugged it, and then been explosively expelled. Periods of volcanic tremor occurred almost daily during December and were observed to coincide with episodes of ashventing of varying vigour. t is interesting to note that this period of explosive activity followed immediately after the period of most rapid horizontal displacement of the northeastern sector of the edifice as recorded at the Hermitage CGPS site (Fig. 4a, b). A spectrum of activity was observed throughout this period, from brief episodes of weak steam-venting, to extremely vigorous, pulse-like ash-venting commonly following dome collapses, to small explosions generating plumes up to 6 km a.s.1. After a volcanotectonic earthquake swarm on 4 to 5 December 1998, the number of volcanotectonic earthquakes declined towards the end of the month while the number of long-period signals increased. The hypocentres of the volcanotectonic earthquakes became more diffuse, and small clusters of earthquakes were located with epicentres 3 km to the NW, 1 km to the SE and 1 km to the NE of the vent, as well as directly below the volcano. A series of small explosions occurred during the first half of January These explosions often generated small-volume pyroclastic flows, possibly by fountain collapse, which reached up to 3 km to the east and south of the dome. The flow deposits were usually very thin (< 1 m thick throughout their length), fine-grained (<1 mm diameter grain size) and closely followed topographic features such as pre-existing incised channels in older deposits. Ash clouds rose to 6 km a.s.1., and weak, pulse-like ash-venting continued for 15 to 30 minutes after each explosion. Each explosion caused a pressure wave detected by an infrasonic sensor 2 km to the east of the dome, produced an audible sound within 6 km of the volcano and generated a seismic signal with distinct long-period energy. Brief periods of ash-venting, lasting minutes and correlating with seismic tremor, occurred throughout January. These became shorter and weaker towards the end of the month. During this period, the average SO2 flux was elevated, at >106 kgday -1. The dome had a volume of 77 x 106m 3 (DRE) at the end of January 1999, with its highest point (977 m a.s.l.) above the southern flank (Fig. 2c). The gorge from the 3 July 1998 and October- November 1998 collapses cut up to 100m deep into the pre-1995 floor of English's Crater and had removed a minimum of 5 x 106 m 3 of pre-1995 rock from this area. The part of the dome to the north of the gorge comprised three main buttresses above the northwestern, northern and northeastern flanks (Fig. 7), and contained two-thirds of the total dome volume in January Episodes of ash-venting or small explosions from within the gorge, accompanied by small pyroclastic flows with runouts of <2 km, occurred about once every 48 hours between 5 and 12 February. The explosion on 5 February was the largest in the February 1999 sequence, and the accompanying ash cloud rose rapidly to over 5 km a.s.1. Pyroclastic flows from this event travelled to the east down the Tar River valley. Although the COSPEC results showed a generally decreasing SO2 flux throughout most of January 1999, the levels remained elevated until February 1999 (Fig. 3c). There were increased levels of activity from 1 March to 31 May 1999, with 54 small explosions or ash-venting episodes during these three months. The largest explosions produced ash clouds up to approximately 9 km a.s.l., with fountain-collapse pyroclastic flows and associated lightning and tephra fallout. These occurred at a rate of a little less than one per day in late March, with the number of explosions increasing in the first week in April 1999 (Fig. 3b), and then decreasing again until mid-may The first week in April was the most intense period of activity during the time of virtually no magma extrusion. Ash-venting episodes throughout this period were at times very intense. n particular, a fumarole away from the

10 Downloaded from at University of Bristol Library on March 10, 2013 ACTVTY DURNG A PEROD OF VRTUALLY NO MAGMA EXTRUSON 475 Fig. 6. Photographs of Soufri+re Hills Volcano taken in late December 1998, showing the dome complex after the series of collapses culminating in the 12 November 1998 scar. (a) View from the west showing (1) Gages Mountain; (2) the dome; (3) the 12 November 1998 scar; (4) Chances Peak; and (5) the Gages pyroclastic fan. (b) View from the east showing (1) the 12 November 1998 and 3 July 1998 collapse scar; (2) an explosion crater with (3) a tuff ring being formed; (4) the northern part of the dome; (5) the southern part of the dome and (6) the southern wall of English's Crater (photographs by R. Robertson). main dome complex on the southern wall of English's Crater was so vigorous that a small cone of tephra had built around the vent in just a few days. Gradual degradation of the dome was caused by the explosions, and pyroclastic flows reached to the Tar River pyroclastic fan or to less than 2 km from the dome in other directions. Seismicity was dominated by small explosion signals and ashventing. Many of the explosion or ash-venting episodes had impulsive starts with a gradual decline in amplitude towards the end of the signal. The signals had a long-period component and each explosion generated a pressure wave. Volcanotectonic earthquakes (10-70 per week) and rockfall signals ( per week) continued throughout this period, with no seismic build-up before the ash-venting episodes and explosions. Fist-sized blocks were collected from fountain-collapse pyroclastic flows to the east of the volcano, from explosions on 1 March, 1, 8 and 14 April 1999; tephra fallout was also collected from these explosions. Some of this material was moderately vesiculated and petrological analysis indicated the presence of fresh green hornblende with no reaction rims, although the majority had been extensively oxidized. By comparison with magma erupted during active dome growth, this observation suggested that some of the magma may have ascended rapidly from the magma chamber (Devine et al. 1998a, b; Murphy et al. 1998), although it is possible that the samples with fresh hornblende may have been entrained into the pyroclastic flows or eruption columns from deposits resulting from the activity. No new dome extrusion was observed in the

11 Downloaded from at University of Bristol Library on March 10, G. E. NORTON ET AL. Fig. 7. Photograph of the northern flank of Soufri~re Hills Volcano in January 1999 showing (1) Perches Mountain; (2) the three main buttresses of the dome that constitute the northern section of the dome complex; (3) Chances Peak; and (4) the pyroclastic flow fan (photograph by R. Robertson). following weeks. The simplest explanation is that the explosions were evacuating the upper conduit and enabling it to be refilled from below, although the pressure driving the flow was not great enough to cause extrusion and dome building. M a y to November 1999." earthquake swarms, continued dome collapses and small explosions An intense volcanotectonic earthquake swarm occurred on 22 May 1999, the first since 5 December 1998, and also the most intense for at least a year. There were 121 volcanotectonic earthquakes in the swarm, which lasted from 07:00 to 17:00, with peaks in activity between 07:00 and 08:00, and between 13:00 and 15:00. A small ash cloud to 2 km a.s.1, was produced at 13:21 at the time of maximum intensity of earthquake activity. The volcanotectonic earthquakes were located consistently between 2.9 and 3.9km depth, either directly under the dome, or on a trend l km east-northeastwards away from the dome. Depths of previous swarms during the period of virtually no magma extrusion (mostly between October and December 1998) were also at about 3 kin, and predominantly under the dome, although several swarms had epicentral clusters in other areas; e.g. 1 km to the W N W or 1 km to the ESE of the dome. Focal mechanisms for volcanotectonic earthquakes throughout late 1998, and from the swarm on 22 May 1999, were oblique in nature (thrust) rather than strike-slip, thus implying an increase in pressure at the depth of origin. Towards the end of the swarm on 22 May 1999, the earthquakes became more strike-slip oriented, possibly implying a change in stress later in the day. These earthquakes showed some similarities to the seismicity from 1992 to 1995, prior to the start of the eruption in July 1995 (J. Shepherd, pers. comm.). The earthquake swarm was followed within 12 hours by a dome collapse at 02:43 on 23 May The ash cloud from the pyroclastic flows reached to a height of about 6 km a.s.1. Tephra fallout was predominantly dispersed to the west of the volcano, with accretionary lapilli to 3 mm diameter reported 4 km from the vent. New pyroclastic flow deposits were seen to the east, with a substantial blocky flow deposit in the main valley as far as the sea, 3 km from the dome. A new pyroclastic surge deposit had also reached to 1.5km from the dome and had spread 500m north of the main block-and-ash deposit. The nose of the block-and-ash flow deposit on the delta was pinkish in colour, with a few large blocks greater than 1 m diameter. A pyroclastic surge deposit was also observed on the southern flanks of the dome with a runout of less than 1 kin. SO2 fluxes after the dome collapse remained low ( x 106 k g d a y -x from 24 to 29 May 1999 compared to x 10 6 k g d a y -1 from 15 to 22 May 1999). A dome collapse occurred on 5 June 1999 at 17:45, with a plume up to 4 km a.s.1. Pyroclastic flows were seen entering the sea 3 km to the east, and flows were also seen travelling less than 2 km to the NE of the dome. Tephra-fallout deposits 7 km to the N W of the volcano reached l cm in thickness and included accretionary lapilli. Approximately 1.5 x 106 m 3 had collapsed from the northeastern part of the dome. A more substantial dome collapse with a volume of about 4.2 x 106 m 3 D R E occurred on 20 July This collapse generated pyroclastic flows that covered most of the pyroclastic fan at the end of the Tar River valley, as well as pyroclastic surges 1.5 km to the SE over a topographic barrier. This latter area had not been previously affected by pyroclastic surges during the period of active dome growth. The ash plume from the pyroclastic flows reached to 11 km a.s.l. Volcanic activity remained at an enhanced level for some time after 20 July 1999, including two small explosions and two volcanotectonic earthquake swarms before the end of July. Activity in August was generally at a lower level, with reduced gas emissions and occasional small dome collapses with associated pyroclastic flows. The flanks of the lava dome slowly became stabilized, as talus and loose blocks were shed from all areas. Volcanic activity levels increased again in early September 1999, with a substantial explosion on 3 September 1999 and subsequent dome collapses and enhanced rockfall activity in the following seven days. Volcanic activity reduced to a lower level by the beginning of October Minor ash-venting episodes and an increase in rockfall signals occurred towards the end of October At the end of October, a series of phreatic explosions was followed by an intense hybrid earthquake swarm on 3-8 November The hybrids were mostly located beneath the volcano at depths of km below English's Crater. Magmatic explosions on 8 and 9 November 1999 produced steamy ash clouds that rose up to 6-8 k m a.s.1. The tephra-fallout deposit contained small (<0.5 cm) pumice lapilli. These explosions were followed by periods of low-frequency tremor, but there was no associated domecollapse activity. On 27 November 1999, a new dome was observed inside the gorge in the centre of the dome, marking the start of the second phase of magma extrusion.

12 ACTVTY DURNG A PEROD OF VRTUALLY NO MAGMA EXTRUSON 477 Activity types The main types of volcanic activity that took place in the period March 1998 to November 1999 were pyroclastic flows, ash-venting episodes, and Vulcanian explosions. The following paragraphs summarize the essential features of these phenomena. t must be noted, however, that a continuum of types of activity occurred throughout this period, such that, for example, some dome-collapse pyroclastic flows may have had an explosive onset, and some ashventing episodes may have been vigorous enough to be termed explosions. Pyroclastic flows Pyroclastic flows during the period under review were formed by both dome collapse and fountain collapse. Cole et al. (1998) and Calder et al. (1999, 2002) distinguished two types of dome-collapse pyroclastic flows at Soufri6re Hills Volcano: (i) discrete collapses that involve the shedding of material as a single pulse; and (ii) sustained collapses that involve the incremental collapse of a significant portion of the dome. Dome-collapse pyroclastic flows during the period March 1998 to November 1999 were mainly single, discrete events. The flows often lasted only a few minutes and produced blocky (up to c. 5 m diameter boulders) deposits of fresh and hydrothermally altered dome material. Runout distances were generally <2 km, although the larger flows extended as far as 5 km from the dome (Fig. 8). The collapses were largely responsible for the excavation of deep scars and gullies in the dome. Vigorously convecting ash plumes, which rose to km a.s.l., were generated by these collapses. The 3 July 1998 collapse was exceptional during this period as it lasted for over 2.5 hours and involved a series of successive collapses. The seismic record for this collapse shows a sudden highamplitude onset followed by a decrease in energy, and then at least two further high-amplitude pulses. This implies a sudden failure of an outer part of the dome, followed by successive failures of the interior. With the onset of ash-venting and unequivocal low-magnitude explosions in December 1998, fountain-collapse pyroclastic flows started to occur. These flows had runouts of >2 km and generated relatively fine-grained deposits with very few blocks larger than 1 m diameter. Some blocks collected from the nose of the flows were semi-vesicular. The flows were not erosive and left only thin deposits along valley bottoms. Highly expanded pyroclastic surges were often associated with the flows, particularly close to the vent. Observations of the initiation of some of these flows indicated that they were generated by partial collapse of poorly developed eruption columns. n comparison with the Vulcanian explosions during the period of high extrusion rate in 1997 (Druitt et al. 2002), the explosions from December 1998 to October 1999 were relatively low-energy with fountain collapse probably occurring no more than 100 m above the vent, although few direct observations were made. Ash-venting Steam-venting continued throughout the period of virtually no magma extrusion. At times, vigorous emission of ash-laden steam occurred from various locations on the lava dome. This phenomenon is here termed ash-venting. Ash-venting episodes were discrete events that typically lasted about 30 minutes, but on rare occasions the episodes lasted several hours. The first unequivocal episode of ash-venting occurred on 19 August 1998, in the scar produced by the 3 July 1998 collapse. Jets of steam and ash were discharged by a particularly vigorous fumarole on the back wall of this scar. Ash-venting was intense for about 24 hours, but waned over the following days. Some of the fumaroles were temporarily buried by rockfalls within the scar, and venting declined to low levels by the end of August. Periods of vigorous ash-venting occurred again towards the end of September 1998, and were then correlated with low-amplitude seismic tremor and increased SO2 fluxes. The increased frequency of small-volume pyroclastic flows in October and November 1998 was accompanied by periods of vigorous ashventing, which often lasted for several hours. Key ~ Pyroclastic flow and surge deposits, April 1996 to March 1998 D Pyroclastic flow and surge deposits, March 1998 to November 1999 W, H. BRAMBLE ARPORT i i mE 16~ ~OooowN -ii m N PLYMOUTH N t HLLS ~" ) 1845OoomN Fig. 8. Map of the pyroclastic flow and surge deposits from Soufri~re Hills Volcano from April 1996 to March 1998 and from March 1998 to November km Contour interval is 200 feet [metres = feet x 0, ooomE 62~ a85000me i i i i

13 478 G. E. NORTON ET AL. Periods of volcanic tremor, which coincided with episodes of ash-venting, were recorded daily throughout December Spectrograms of the tremor associated with ash-venting were often sharply peaked at between 1 and 2Hz. These peaks generally remained constant over the duration of the tremor period, but occasionally the frequency of the dominant spectral peak changed as the amplitude of the tremor was also observed to change. This strong correlation with active ash-venting led to the interpretation that the vent itself may have been acting as the source for the tremor, or that the venting may have been generating some resonance in the body of the lava dome, near the vent orifice. Small-magnitude Vulcanian explosions Small-magnitude Vulcanian explosions, which generated buoyant plumes above the volcano, were unequivocally observed first in early December 1998, but the dome collapses in November 1998 may also have had explosive components. While these initially produced discrete blasts, which threw metre-sized ballistic blocks up to 80 m above the vent, they were unable to generate substantial convecting eruption columns with most ash clouds reaching no more than 6km a.s.1. By January 1999, the explosions were generating fountain-collapse pyroclastic flows, which reached the sea in the Tar River valley, 3 km east of the dome. The plumes initially rose rapidly to between 2 and 3 km a.s.1, and then convected slowly to between 3 and 6 kin a.s.1. The explosions originated from a crater located in the renmants of the 3 July 1998 scar, and caused detectable pressure waves. A blocky tuff ring formed around the developing explosion crater and gradually increased in width and depth through A boom was heard at the initiation of some explosions, and spectacular thunder and lightning were associated with some eruption plumes. The explosions were often accompanied by periods of intense ash-venting which were sometimes audible as a jet engine-like noise. The most active period of explosions was between the start of March 1999 and the end of May 1999, with 54 explosions or periods of ash-venting between 1 March and 31 May Most of the explosions reached between 1.5 and 6 km a.s.1., which is low relative to the Vulcanian explosions during the period of high extrusion rate in 1997 when heights of between 3 and 15 km a.s.1, were measured (Druitt et al. 2002). The explosions during the period of virtually no magma extrusion produced no pumiceous fallout, and the blocks collected from fountain-collapse pyroclastic flows were only semi-vesiculated. The fountain-collapse flows were only rarely distributed radially in several directions around the vent, with most being directed to the east or west by the gully in the dome, implying that fountain collapse occurred at a low elevation above the vent. Thus the explosions during this period must have occurred at low magma production rates, and with only limited vesiculation of the magma within the upper levels of the conduit. Discussion This paper documents a period of relative dormancy between two phases of magma extrusion. t is clear, however, that the volcano was not inactive. ndeed, apart from a four-month period from March to July 1998, the activity was never at a particularly low level: there were numerous, relatively small explosions, and pyroclastic flows impacted areas that had not been affected previously (Fig. 8). Throughout this period, critical issues were: (i) whether the activity that was being observed indicated continuing, but less energetic, ascent of magma; (ii) whether energetic magma extrusion at the surface would restart; and (iii) what precursors might be expected prior to the onset of any renewed dome growth. This type of residual volcanic activity between two phases of dome growth has not been commonly described in detail in the literature, and so it is important to make comparisons between the types of phenomena that were observed during this period of dormancy and those that have been described at the ends of other dome-building eruptions. Episodes of growth at dome-building eruptions tend to end with declining growth rates (over weeks to months), formation of spines, and reduction in seismic activity to near-background levels. The eruption of Mount Lamington ended with slow dome growth accompanied by slow extrusion of a series of spines (Taylor 1958). A sharp decrease in seismicity accompanied the cessation of dome growth at the end of the eruption of Mount Pinatubo (Mori et al. 1996). Low levels of earthquakes, gas flux and ground deformation all marked the end of the eruption of Mount Unzen (Nakada et al. 1999). ntense activity at the end of dome growth appears to be rare and, when it occurs, is most often in the form of minor phreatic explosions or dome collapses. For instance, small phreatic explosions occurred after cessation of dome growth at the 1976 eruption of St Augustine volcano (Swanson & Kienle 1988) and in 1992 at the end of the eruption of Pinatubo (Daag et al. 1996). Eruptions that ended with magmatic explosions are not common, but examples include Galeras (Stix et al. 1997), Lascar (Matthews et al. 1997) and razfi (Krushensky & Escalante 1967). At Galeras, cessation of dome growth in 1991 was followed by 7.5 months of low seismicity and low gas flux before sudden Vulcanian explosions began on 16 July 1992 (Stix et al. 1997). However, no further magma extrusion had occurred at Galeras at the time of writing. At Soufri6re Hills Volcano, in the three months before magma extrusion paused in March 1998, the rate of dome growth was still quite high, although growth rates appeared to have been gradually decreasing throughout February and early March 1998 (Watts et al. 2002). n March 1998, the number of seismic signals diminished to near-background levels, and the character of the seismicity changed: volcanotectonic earthquakes and rockfalls became dominant relative to the hybrid earthquakes that had been prominent throughout dome growth (Aspinall et al. 1998; Miller et al. 1998; Voight et al. 1999). Changes in dome morphology from March to July 1998 were minor and related only to minor disintegration of the dome. Hence it appeared that magma ascent had ceased. However, data from the CGPS sites, processed using absolute point positioning techniques and final orbit/clock products (see Mattioli et al. (1998) for discussion), showed a signal of uplift or inflation relative to the Earth's centre of mass (Mattioli et al. 2000). A major dome collapse on 3 July 1998 unsealed the cooling dome and conduit, and may have been a result of pressurization due to cooling and degassing of the stagnant magma column, possibly triggered by rainwater infiltration into the steep talus slope. After this collapse, SO2 flux was exceptionally high, with values up to >4.0 x 106kgday -1, and typically between 0.5 and 1.5 x 106 kg day-l, indicating a sudden increase in gas output relative to the previous three months. These levels were even higher than those measured during the period of active dome growth. Exposure of the vent area, and local development of steep slopes within the collapse crater, led to further dome collapses and ash-venting. Eight months after dome growth had ceased, mild Vulcanian explosions restarted, possibly in November 1998 and unequivocally in December These explosions continued, along with episodes of vigorous ash-venting and occasional dome collapses, until the renewal of vigorous magma extrusion in November The high gas fluxes and high ratio of sulphur to chlorine throughout this period (Oppenheimer et al. 2002) pointed to continued, on-going degassing of the dome, conduit and deeper magma body. There was some evidence that SO2 flux often peaked after an explosion or ash-venting, and that the flux decayed gradually thereafter (Fig. 3c). This suggests that the vent and conduit were opened during the explosion or venting episode, and then gradually resealed with time, possibly by slow ascent of new magma in the conduit. The general progression of the activity from dome collapses through ash-venting to small explosions from July 1998 to November 1999, and the overall decrease in SO2 flux through this period, may also indicate a longer-term gradual cooling, degassing and partial resealing of the conduit.

14 ACTVTY DURNG A PEROD OF VRTUALLY NO MAGMA EXTRUSON 479 There is some evidence for weak peaks in dome collapse or explosive activity occurring approximately every five to six weeks. This is most discernible in the frequency of explosions and dome collapses (Fig. 3b). This periodicity is similar to, although not as strongly defined as, the six to ten week cyclicity observed during active dome growth in (Voight et al. 1999), and may have been driven by similar repeated pressurization processes in the dome, conduit and deeper magma body. The key difference between the periodicity seen during active magma extrusion and that during dormancy is the lack of precursory hybrid earthquake activity in the latter. On the contrary, volcanotectonic earthquake activity sometimes followed major collapses, indicating a response to depressurization of the conduit. This lack of precursory activity made forecasting the timing of individual collapses and explosions more difficult, although the recognition of the weak cyclicity gave the MVO some ability to forecast periods of likely enhanced activity. The lack of precursory hybrid seismicity may reflect a lesser degree of pressurization than that which occurred in more active periods. Previously, it had been recognized that the development of hybrid seismicity responded to a pressure threshold (Voight et al. 1999), which may not have been surpassed in March 1998 to October There were some indications that magma may have been ascending slowly at various times during the period of dormancy, particularly after 3 July Petrological evidence indicated that fresh material might have been erupted from the vent during the explosive activity in March and April t was not clear, however, that the samples collected from the distal ends of pyroclastic flow deposits necessarily reflected magma newly erupted from depth. An alternative explanation is that they were reamed from the conduit wall or dome during the explosions, or even incorporated into the pyroclastic flows from earlier ( ) deposits during transport. Other evidence for slow ascent of magma includes the cumulative volume of material expelled in scores of explosive eruptions. For example, in the period of the most intense explosive activity from 1 March 1999 to 31 May 1999, given an average eruption column height of 3.5 km a.s.1, for each of 54 explosions or periods of ash-venting, the mass erupted per explosion can be calculated using the method and estimated parameters given in Druitt et al. (2002, equation 1). This requires an average eruptive volume of about 2000 m 3 (DRE) per explosion, thus giving an average magma production rate of 0.01 m 3 s -1 in this three-month period. t is also possible that slow extrusion of magma in the vent area may have occurred during periods of poor or limited visibility, and that any dome material was expelled from the vent during subsequent explosions. These lines of evidence suggest that magma may have been rising in the conduit throughout parts of the period of virtually no magma extrusion, although with less energy than from November 1995 to March 1998, such that the magma emerged only rarely, if ever, from the vent. Similarly, the volcanotectonic earthquake swarm on 22 May 1999 was similar to the seismicity prior to the onset of the ongoing eruption in 1995, and this may also have been an indication of magma movement at depth. We suggest that movements immediately after 3 July 1998 were very sluggish, due to the high viscosity of largely degassed magma, and also to the narrowing of the conduit from crystallization at the wall. Over the following months, the mean conduit viscosity gradually reduced as newer magma entered the vent, and ascent became slightly easier. Thus, by late 1998, activity became increasingly explosive as conduit volatile content could not be readily dissipated. This trend continued in 1999 as a gradual prelude to the more vigorous dome building that commenced in November However, the upward trend was not continuous, but included some minor peaks and valleys, with a noticeable peak occurring in April Possibly these variations reflected periodic repressurization of the magma chamber related to the multiple-week cyclicity. The above hypothesis can be considered in the light of conduitflow physics, as expressed simply by the expression: Q = {Trer4}/{8#L} where Q is the flow rate, P is the overpressure (total pressure minus magmastatic pressure) driving the conduit flow, r is the conduit radius, # is the bulk magma viscosity, and L is the length of the conduit. Over the period 26 December 1997 to mid-february 1998, the flow rate was approximately 6m 3 s -l. The chamber overpressure at depth L= 5000m was of the order of 10-15MPa during the period of vigorous flow (Voight et al. 1999), the outlet pressure was about 5MPa at the base of the 250-m-high dome, and the net overpressure driving conduit flow was therefore about 5-10MPa. By March 1998, flow had ceased, suggesting that chamber pressure had declined and was then counterbalanced by pressure due to the weight of the dome. The chamber overpressure was thus about 5 MPa, net overpressure was zero, and Q was zero. With the 3 July 1998 dome collapse, however, the dome counterweight was removed, such that the net overpressure driving conduit flow was again c. 5MPa. n a viscous system, the applied stress would have required flow to restart. f we assume that the rate Q = 6 m 3 s -1 in January 1998 was associated with an overpressure of 10MPa, then the flow in July driven by P=5MPa should have recommenced at the rate 3 m 3 s -1, all other factors remaining equal. However, no such large flow rate was observed, and this is consistent with the idea that the magma viscosity may have increased substantially during the stagnant interlude. ncreasing viscosity by a factor of 100 reduces Q to 0.03 m 3 s -1. Likewise, if conduit radius decreased by only 10% as a result of congealing at the conduit wall, the flow rate would be decreased by a further 34% or Q=0.02 m 3 s -1. These values concur with estimates of magma production rate during periods of increased explosive activity, for example, during March to May Finally, we note that various parameters such as overpressure and mean viscosity may have varied from time to time over the period of virtually no magma extrusion, so that some subperiods may have been more or less active than others. The volcanotectonic seismicity during the period of virtually no magma extrusion was most likely related to variations in the pressurization in the conduit at 3-4 km depth, coupled with localized structural adjustments to changes caused by the preceding magma movements and extrusion. CGPS data from early 1998 to late 1999, however, are best fitted by an inflating point-source at approximately 6 km depth, embedded in a simple elastic half-space (Mogi 1958). The derived depth is similar to that obtained by Mattioli et al. (1998) for GPS data from 1995 to mid-1997, although the dilatation is opposite in sign and about one-half the magnitude. A strong indication of vigorous renewed magma ascent was registered in early November 1999, when the first hybrid swarm since May 1998 occurred, to be followed by small Vulcanian explosions with pumiceous fallout. On 27 November 1999, a substantial new dome was observed to be forming over the vent reamed out by the numerous explosions since December The continuation of substantial activity at the surface, despite the apparent pause of magma extrusion, poses several problems for attempting to determine whether the eruption has ended, once a phase of magma extrusion ceases. The question is of vital importance to hazard management and government officials. f one defines an eruption purely in terms of the sustained arrival of magma at the Earth's surface, then the first modern eruption of Soufri6re Hills Volcano can be said to have ended in early March 1998, and a new one started in November However, pyroclastic flows and surges with accompanying ash clouds continued to occur and large collapses were experienced in July and November 1998, and June and July 1999, and Vulcanian explosions with ash plumes up to 9 km a.s.1, occurred frequently from December 1998 to November Thus, if one defines an eruption in terms of observable phenomena, then the volcano continued in eruption through the period being reported here. n this case, the results of monitoring data collected were also equivocal since, for example, COSPEC measurements showed gas fluxes were higher than during periods of magma extrusion, and the eastern flanks of the volcano continued to show evidence of ground deformation. The experience of volcanic activity at Soufri6re Hills Volcano demonstrates that hazardous activity can continue for extended periods even after magma has apparently ceased to erupt at the

15 480 G. E. NORTON ET AL. surface. With hindsight, it might be argued that the high level of observable activity, continued deformation and high gas fluxes, from March 1998 to November 1999, were an indication that the volcano was only in temporary repose, and that further vigorous magma extrusion was quite possible. ndeed, this was always recognized as a possibility in the probabilistic risk assessments (always examined for the 'next' six-month period), although the 'consensus' probability of a restart declined as time went on. n general it was suspected that the observed activity was more likely to have been due to cooling and degassing of the dome, conduit and deeper magma body, than any incipient renewal of eruptive activity. The need for vigilance in both monitoring and crisis management are clearly underscored. t seems necessary to continue dedicated, full-scale monitoring of this type of dome-building eruption for some years after magma extrusion has ceased. t also seems that observed activity has to be substantially diminished before an eruption can be deemed to be over. The visible evidence of ongoing eruptive activity, despite the apparent absence of new magma at the surface, posed several problems for crisis managers. The civilian population, who continued to observe dome collapses, ash-venting and explosions at the volcano similar to those seen during the height of the emergency, remained concerned about the activity. Even when there was no magma extrusion, dome-collapse activity still produced tephrafallout deposits rich in crystalline silica (10-24% in the <10#m fraction; Baxter et al. 1999), which was of great concern for people living on the island (Moore et al. 2002). Ash plumes continued to threaten aviation in the region and tephra fallout caused occasional problems for new agricultural developments in the north of the island, as well as adding to the volume of loose material on the flanks of the volcano available for lahar generation. Dome collapses and explosions remained very dangerous, with their potential to take lives and impact property on the flanks of the volcano. The Montserrat experience emphasizes the necessity to maintain close monitoring for a considerable time after such a dome-building eruption appears to have peaked and declined. Conclusions Volcanic activity at SoufriGre Hills Volcano continued at a high level during a period of virtually no magma extrusion from March 1998 to November Dome-collapse pyroclastic flows and Vulcanian explosions, often without apparent precursors, continued to endanger life and property on the flanks of the volcano, and provided serious hazards to aviation in the eastern Caribbean. There was no evidence for ascent of magma to the surface between March 1998 and July 1998, although CGPS results showed that there was gradual inflation of the volcanic edifice, probably as a result of pressurization in the magma chamber and conduit. The 3 July 1998 dome collapse involved dome lava from the dome-building phase from November 1995 to March 1998 and it may have been caused by gas pressurization front the buried conduit, possibly enhanced by rainwater infiltration and weakening of the dome rock by hydrothermal activity. Following this collapse, the vent was exposed, and sluggish ascent of volatile-depleted magma may have begun in a conduit narrowed by crystallization at the wall. Over the next six months, activity became increasingly explosive. The limited evidence suggests that, from July 1998, magma ascent may have occurred in the conduit, although with limited driving pressure such that any dome growth was not sufficiently voluminous to be observed. The main signals of renewed vigorous magma extrusion were the onset of hybrid earthquakes and pumiceous fallout from explosions in early November t is proposed that much of the activity throughout the period of virtually no magma extrusion was due to cooling and degassing of the dome, conduit and deeper magma body, leading to pressurization of the crystallizing magma. Slow magma ascent after July 1998 may have renewed the volatile and energy supply to higher conduit levels. Weak peaks in the intensity of the activity every five to six weeks were similar to the cyclicity observed during active dome growth. The cyclicity in both cases may have been linked to degassing and crystallization of the magma body and conduit. The activity at Soufri~re Hills Volcano during this period underscores the need to continue to monitor dome-building eruptions for several years after magma extrusion has apparently ceased. The authors would like to thank all colleagues at the Montserrat Volcano Observatory, particularly R. Robertson and all the local staff who have contributed to the continued detailed documentation of the ongoing eruption at Soufri~re Hills Volcano. The Government of Montserrat and the British Government (through DFD) supported the authors' work at the MVO. This paper is published by permission of the Director, British Geological Survey (NERC). Funding for UPRM equipment and GSM was provided by NASA grants NCC5-252 and NAG References ASPNALL, W. P., MLLER, A. D., LYNCH, L. L., LATCHMAN, J. L., STEWART, R. C., WHTE, R. A. & POWER, J. A Soufri~re Hills eruption, Montserrat : Volcanic earthquake locations and fault plane solutions. Geophysical Research Letters, 25, ASPNALL, W. P., LOUGHL1N, S. C., MCHAEL, F. V. ET AL The Montserrat Volcano Observatory: its evolution, organization, role and activities. n: DRUTT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufridre Hills Volcano, Montserrat,.[}om 1995 to Geological Society, London, Memoirs, 21, BAXTER, P. J., BONADONNA, C., DUPREE, R. ET AL Cristobalite in volcanic ash of the Soufri~re Hills Volcano, Montserrat: Hazards implications. Science, 283, CALDER, E. S., COLE, P. D., DADE, W. B. ET AL Mobility of pyroclastic flows and surges at the Soufriere Hills volcano, Montserrat. Geophysical Research Letters, 26, CALDER, E. S., LUCKETT, R., SPARKS, R. S. J. & VOGHT, B Mechanisms of lava dome instability and generation of rockfalls and pyroclastic flows at Soufri~re Hills Volcano, Montserrat. /n: DRUTT, T. H. & KOKELAAR, B. P. (eds) The Eruption of SoufriOre Hills Volcano, Montserrat, from 1995 to Geological Society, London, Memoirs, 21, CHOt)ET, B. A., PAGE, R. A., STEPHENS, C. D., LAHR, J. C. & POWER, J. A Precursory swarms of long-period events at Redoubt Volcano ( ), Alaska: Their origin and use as a forecasting tool. Journal of Volcanology and Geothermal Research, 62, CHmSTANSEN, R. L. & PETERSON, D. W Chronology of the 1980 eruptive activity. n: LPMAN, P. W. & MULLNEAUX, D. R. (eds) The 1980 Eruptions of Mount St Helens, Washington. United States Geological Survey, Professional Papers, 1250, COLE, P. D., CALDER, E. S., DRUTT, T. H., HOBLTT, R., ROBERTSON, R., SPARKS, R. S. J. & YOUNG, S. R Pyroclastic flows generated by gravitational instability of the lava dome of Soufri+re Hills Volcano, Montserrat. Geophysical Research Letters, 25, DAAG, A. S., DOLAN, M. T., LAGUERTA, E. P., MEEKER, G. P., NEWHALL, C. G., PALLSTER, J. S. & SOLDUM, R. U Growth of a postclimactic lava dome at Mount Pinatubo, July-October n: NEWHALL, C. G. & PUNONGBAYAN, R. S. (eds) Fire and Mud." Eruptions and Lahars of Mount Pinatubo, Philippines. University of Washington Press, DENLNGER, R. P. & HOBLTT, R. P Cyclic eruptive behavior of silicic volcanoes. Geology, 27, DEVNE, J. D., MURPHY, M. D., RUTHERFORD, M. J., SPARKS, R. S. J., CARROLL, M. R., YOUNG, S. R. & GARDNER, J. E. 1998a. Petrologic evidence for pre-eruptive pressure-temperature conditions, and recent reheating, of andesitic magma eruption at the SoufriGre Hills Volcano, Montserrat, W.. Geophysical Research Letters, 25, DEVNE, J. D., RUTHERFORD, M. J. & GARDNER, J. E. 1998b. Petrologic determination of magma ascent rates for the Soufri+re Hills Volcano andesite magma. Geophysical Research Letters, 25, DRUTT, T. H., YOUNG, S. R., BAPTE, B. ET AL Episodes of cyclic Vulcanian explosive activity with fountain collapse at SoufriGre Hills Volcano, Montserrat. n: DRUTT, T. H. & KOKELAAR, B. P. (eds) The

16 ACTVTY DURNG A PEROD OF VRTUALLY NO MAGMA EXTRUSON 481 Eruption of Soufridre Hills Volcano, Montserrat,,from 1995 to Geological Society, London, Memoirs, 21, JACKSON, P., SHEPHERD, J. B., ROBERTSON, R. E. A. & SKERRTT, G Ground deformation studies at Soufri6re Hills volcano, Montserrat : Electronic distance meter studies. Geophysical Research Letters, 25, KRUSHENSKY, R. D. & ESCALANTE, G Activity at razu and Poas Volcanoes, Costa Rica, November 1964-July Bulletin Volcanologique, 31, LOUGHLN, S. C., CALDER, E. S., CLARKE, A. B. ET AL Pyroclastic flows and surges generated by the 25 June 1997 dome collapse, Soufri6re Hills Volcano, Montserrat. n: DRUTT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufri&e Hills Volcano, Montserrat, from 1995 to Geological Society, London, Memoirs, 21, MATTHEWS, S. J., GARDEWEG, M. C. 8~; SPARKS, R. S. J The cyclic activity of Lascar volcano, Northern Chile: cycles of dome growth, dome subsidence, degassing and explosive eruptions. Bulletin of Volcanology, 59, MATTOL, G. S., DXON, T. H., FARNA, F., HOWELL, E., JANSMA, P. ~; SMTH, A. L GPS measurement of surface deformation around Soufri~re Hills volcano, Montserrat from October 1995 to July 1996, Geophysical Research Letters, 25, MATTOL, G. S., RODRGUEZ, L., SMTH, A. L., JANSMA, P. E. & HERD, R. A Changes in magmatic flux recorded by campaign and continuous GPS geodesy of Soufri6re Hills, Montserrat, BW (abstract). Geosciences Geological Society, London. MLLER, A. D., STEWART, R. C., WHTE, R. A. ETAL Seismicity associated with dome growth and collapse at the Soufri6re Hills Volcano, Montserrat. Geophysical Research Letters, 25, MOG, K Relations between the eruptions of various volcanoes and the deformations of the ground surfaces around them. Bulletin of the Earthquake Research nstitute, 36, MOORE, K. R., DUFFELL, H., NCHOLL, A. & SEARL, A Monitoring of airborne particulate matter during the eruption of Soufri~re Hills Volcano, Montserrat. n: DRUTT, T. H. & KOKELAAR, B. P. (eds) The Eruption of SoufriOre Hills Volcano, Montserrat..from 1995 to Geological Society, London, Memoirs, 21, MOR, J., WHTE, R. A., HARLOW, D. H. ETAL Volcanic earthquakes following the 1991 climactic eruption of Mount Pinatubo: Strong seismicity during a waning eruption. n: NEWHALL, C. G. & Punongbayan,, R. S. (eds) Fire and Mud. Eruptions and Lahars of Mount Pinatubo, Philippines. University of Washington Press, MURPHY, M. D., SPARKS, R. S. J., BARCLAY, J. ET AL The role of magma mixing in triggering the current eruption at the Soufri6re Hills Volcano, Montserrat, West ndies. Geophysical Research Letters, 25, NAKADA, S., SHMZU, H. 8~; OHTA, K Overview of the eruption at Unzen Volcano. Journal of Volcanology and Geothermal Research, 89, NORTON, G. E Summary of the environmental effects" of volcano emissions, SoufriOre Hills Volcano, Montserrat. MVO Open File Report, 97/15. OPPENHEMER, C., EDMONDS, M., FRANCS, P. 8z; BURTON, M Variation in HC1/SO2 gas ratios observed by Fourier transform spectroscopy at Soufri~re Hills Volcano, Montserrat. n: DRUTT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufridre Hills Volcano, Montserrat, from 1995 to Geological Society, London, Memoirs, 21, RATDOMOPURBO, A & POUPNET, G An overview of the seismicity of Merapi volcano (Java, ndonesia) Journal of Volcanology and Geothermal Research, 100, ROBERTSON, R., COLE, P., SPARKS, R. S. J. ETAL The explosive eruption of Soufri+re Hills Volcano, Montserrat, West ndies, 17 September, Geophysical Research Letters, 25, SPARKS, R. S. J., BARCLAY, J., CALDER, E. S. ET AL Generation of a debris avalanche and violent pyroclastic density current on 26 December (Boxing Day) 1997 at Soufribre Hills Volcano, Montserrat. n: DRUTT, T. H. & KOKELAAR, B. P. (eds) The Eruption of SoufriOre Hills' Volcano, Montserrat, from 1995 to Geological Society, London, Memoirs, 21, STX, J., TORRES, C. R., NARVAEZ, M. L., CORTES, J. G. P., RAGOSA, A. J., GOMEZ, M. D. & CASTONQUAY, R A model of Vulcanian eruptions at Galeras volcano, Columbia. Journal of Volcanology and Geothermal Research, 77, SWANSON, S. E. & KENLE, J The 1986 eruption of Mount St Augustine: Field test of a hazard evaluation. Journal of Geophysical Research, 93(B5), TAYLOR, G. A The 1951 Eruption of Mount Lamington, Papua. Department of National Development, Bureau of Mineral Resources, Geology and Geophysics. Bulletin No. 38. VOGHT, B. & ELSWORTH, D Stability and collapse of gas-pressurized lava domes. Geophysical Research Letters, 48, -4. VOGHT, B., HOBLTT, R. P., CLARKE, A., LOCKHART, A. B., MLLER, A. D., LYNCH, L. & MCMAHON, J Remarkable cyclic ground deformation monitored in real-time on Montserrat, and its use in eruption forecasting. Geophysical Research Letters, 25, VOGHT, B., SPARKS, R. S. J., MLLER, A. D. ET AL Magma flow instability and cyclic activity at Soufri6re Hills Volcano, Montserrat, B.W.. Science, 283, VOGHT, B., CONSTANTNE, E. K., S1SWOWDJOYO, S. 8,: TORLEY, R Historical eruptions of Merapi Volcano, Central Java, ndonesia, Journal of Volcanology and Geothermal Research, 100, VOGHT, B., KOMOROWSK, J.-C., NORTON, G. E. Er AL The 26 December (Boxing Day) 1997 sector collapse and debris avalanche at Soufri~re Hills Volcano, Montserrat. n: DRUTT, T. H. & KOKE-LAAR, B. P. (eds) The Eruption of Soufri~re Hills Volcano, Montserrat,.from 1995 to Geological Society, London, Memoirs, 21, WATSON,. M., OPPENHEMER, C., VOGHT, B. ET AL The relationship between degassing and ground deformation at Soufri6re Hills Volcano, Montserrat. Journal of Volcanology and Geothermal Research, 98, WATTS, R. B., HERD, R. A., SPARKS, R. S. J. 8~; YOUNG, S. R Growth patterns and emplacement of the andesitic lava dome at Soufri6re Hills Volcano, Montserrat. n: DRUTT, T. H. & KOKELAAR, B. P. (eds) The Eruption of Soufri&e Hills Volcano, Montserrat, from 1995 to Geological Society, London, Memoirs, 21, YOUNG, S. R., SPARKS, R. S. J., ASP1NALL, W. P., LYNCH, L. L., MLLER, A. D., ROBERTSON, R. E. A. & SHEPHERD, J. B. 1998a. Overview of the eruption of Soufri6re Hills Volcano, Montserrat, 18 July 1995 to December Geophysical Research Letters, 25, YOUNG, S. R, FRANCS, P. W., BARCLAY, J. ETAL. 1998b. Monitoring SO2 emission at the Soufri6re Hills Volcano: mplications for changes in eruptive conditions. Geophysical Research Letters, 25, YOUNG, S. R., DUFEELL, H. J., VO1GHT, B. ET AL Cyclic gas exhalations at Soufri6re Hills Volcano, Montserrat and implications for real-time hazard assessment. Proceedings of A VCE General Assembly, Bali, ndonesia, July 2000.