Abstract
The 1995–present eruption of Soufrière Hills Volcano on Montserrat has produced over a cubic kilometre of andesitic magma, creating a series of lava domes that were successively destroyed, with much of their mass deposited in the sea. There have been five phases of lava extrusion to form these lava domes: November 1995–March 1998; November 1999–July 2003; August 2005–April 2007; July 2008–January 2009; and October 2009–February 2010. It has been one of the most intensively studied volcanoes in the world during this time, and there are long instrumental and observational datasets. From these have sprung major new insights concerning: the cyclicity of magma transport; low-frequency earthquakes associated with conduit magma flow; the dynamics of lateral blasts and Vulcanian explosions; the role that basalt–andesite magma mingling in the mid-crust has in powering the eruption; identification using seismic tomography of the uppermost magma reservoir at a depth of 5.5 > 7.5 km; and many others. Parallel to the research effort, there has been a consistent programme of quantitative risk assessment since 1997 that has both pioneered new methods and provided a solid evidential source for the civil authority to use in mitigating the risks to the people of Montserrat.
At the time of writing (January 2013), the Soufrière Hills Volcano (SHV) is in its 17th year of eruption, placing it with a select group of very long-lived eruptions from silicic volcanoes. Within the Caribbean plate, only two other silicic volcanoes have had longer-lived historical extrusive eruptions: Santiaguito, Guatemala (1922–present: Harris et al. 2003); and Arenal, Costa Rica (1968–present: Wadge et al. 2006). As with both of these volcanoes, SHV has not extruded lava continuously for the whole eruption, but has erupted intermittently. In fact, it has been extruding lava for a cumulative total of only 8.5 years out of 17. So in what sense can we consider this a single eruption rather than a series of shorter ones? The longest gaps between extrusive phases have been about 2–3 years (July 2003–August 2005; February 2010–present). However, during the pauses between extrusion at SHV, there have been many indications that the volcanic system remains active. These have included inflation of the island's surface, the release of volatiles in quantities similar to those seen during lava extrusion, swarms of low-frequency seismicity, explosions and the production of ash. This suggests that there are processes within the system responsible for producing alternating extrusive and non-extrusive states. Melnik & Sparks (2005) showed theoretically how non-linear response to relatively minor perturbations of the magmatic system can result in such behaviour.
Recognition of the long-term continuity of the eruption process has direct consequences for the tasks of monitoring and estimating the risks posed by the volcano. It means that the Montserrat Volcano Observatory (MVO) must monitor the volcano continuously, even during periods of relative quiescence and apparent inactivity. The probabilities of a restart of lava release, and the enhanced hazards they may pose, need to be evaluated regularly. The Scientific Advisory Committee (SAC) has played the main role in quantitative risk assessment for this period and its reports (published every 6 months or 1 year) represent the longest such series of detailed assessments at any active volcano.
SHV continues to surprise. A great variety of eruptive styles, some unexpected, were a characteristic of Phase 1 of the eruption from 1995 to 1998 (described in detail in Geophysical Research Letters Volume 25; the Geological Society, London, Memoir 21; Druitt & Kokelaar 2002). As the system became increasingly vigorous, the number and the scale of phenomena increased: initial lava extrusion; the first pyroclastic density currents (PDCs); lethal PDCs; cyclic behaviour of the conduit flow and dome extrusions; collapses of the dome triggered by lava intrusion or by intense rainfall; vertical explosions triggered by dome collapse; repetitive explosions with pumice flows generated by collapse of the eruption column; formation of surge-generated secondary pyroclastic flows; structural collapse of the dome flank with accompanying lateral blast; and the generation of lahars from the remobilization of proximal deposits by intense rainfall. Whilst the 2000–2010 period has not seen the same range of new phenomena, there have been several firsts: wholesale collapses of the dome (e.g. 12 July 2003, 20 May 2006: Herd et al. 2005); the occurrence of offshore-generated hydrovolcanic explosions with on-land and offshore deposits (Edmonds & Herd 2005); the highest Vulcanian plume (17 km for the main plume (Loughlin et al. 2010) with a central overshoot to over 20 km (Prata et al. 2007)) on 20 May 2006; the largest Vulcanian explosions producing a pyroclastic flow with a 6.5 km runout threatening a populated area near the Belham Valley (8 January 2010: Cole et al. 2014b); the generation of tsunamis related to major dome collapse (Mattioli et al. 2007); and the substantial economic damage to the fruit crops of neighbouring Guadeloupe caused by the 11 February 2010 ash plume (Komorowski et al. 2011). The latest phases of extrusive activity (4 and 5) have been as vigorous as any.
What is less surprising is the continued high level of scientific interest in SHV as a place to learn more about silicic volcanism. The presence of a well-run volcano observatory, the relative ease of access, the spectacular variety of phenomena and the emergence of new techniques to be tested have all played a role in maintaining a vibrant scientific effort. Major international research programmes like SEA-CALIPSO benefited from the years of prior experience with the volcano (Voight et al. 2010c). However, there have been some retreats. It has proved impossible, for safety reasons as the dome has grown ever higher, to continue the near-field measurements of tilt that provided such a wonderful source of data in 1996–1997 (Voight et al. 1998, 1999; Sparks & Young 2002). Similarly, we don't have any borehole instruments measuring strain, electromagnetic or hydrological signals sited on the volcano itself. Also, the ability to take petrological and geochemical samples has been often limited by safety concerns.
Early in the eruption, it was realized that SHV represented a complex, non-linear system, but with some regularity in behaviour that promised the possibility of some degree of predictability (e.g. Melnik & Sparks 1999; Voight et al. 1999). There is continued evidence for a variety of behavioural states possible within the system, some semi-deterministic and quasi-periodic (e.g. the repetitive cycles of seismicity, deformation and Vulcanian explosions seen in 1997). In particular, at least three timescales of quasi-periodic fluctuation have been recognized with timescales of many hours, many weeks and a few years (Sparks & Young 2002; Loughlin et al. 2010; Mattioli et al. 2010; Wadge et al. 2010; Odbert et al. 2014b).
Underpinning our understanding of the volcano's behaviour is the hypothesis of an andesitic magma body being intruded by a basaltic magma releasing heat and volatiles, and generating a magmatic overpressure that powers the eruption. Slowly, from a variety of geodetic measurements supported by insights gained from seismic tomography and petrology, a picture of the geometry and dynamics of the andesite magma transport is being resolved, although that picture is still quite fuzzy at depth (Barclay et al. 1998; Mattioli et al. 2010; Voight et al. 2010b; Paulatto et al. 2012; Odbert et al. 2014a). The shallower components of the transport system are now thought to be represented by a dyke with a cylindrical conduit above (Costa et al. 2007; Hautmann et al. 2009; Chardot et al. 2010; Linde et al. 2010). Multi-week and sub-daily cycles of lava extrusion, first recognized in 1997 (Voight et al. 1998), are now associated with these two conduit components, respectively (Costa et al. 2007, 2012, 2013; Lensky et al. 2008). Improved temporal resolution of the measurements of lava and gas flux, together with seismicity and deformation coupled with models of these conduits, offer the prospect of improving the understanding of the shallower processes.
The setting of Soufrière Hills
SHV is a Peléean lava dome complex constructed largely of hornblende andesite (Rea 1974). It sits at the volcanic front of the northern Lesser Antilles island arc and is smaller (in elevation and volume) than the volcanoes generally found in the central islands of the arc. The subducting North American plate is at a depth of about 120 km below Montserrat (Syracuse & Abers 2006), and the vector of convergence between the two plates is ENE at 20 mm a−1. However, regional global positioning system (GPS) measurements show that the northern Lesser Antilles crust is moving in a slightly more northerly direction in the global reference frame relative to the plate model (Fig. 1.1). This can be accommodated by north–south extension of this part of the Caribbean plate (Lopez et al. 2006). There is contemporary evidence for such extension from seismicity and geological evidence from roughly east–west normal faulting (Feuillet et al. 2002, 2010; Kenedi et al. 2010). In particular, the WNW-trending Belham Valley Fault that crosses Montserrat adjacent to SHV forms part of a large transtensional fault system running NNW between Montserrat and Basse Terre, Guadeloupe – the Montserrat–Bouillante Fault System (Feuillet et al. 2002). In 1843, a major, interplate thrust fault earthquake on the subduction zone (Bernard & Lambert 1988) occurred east of Montserrat. Feuillet et al. (2011) argued that this had a magnitude as great as 8.5 and a rupture length of about 300 km (Fig. 1.1). Notable, smaller intraplate earthquakes occurred at the southern end of the Montserrat–Bouillante Fault System in 1851 and 1897 (Feuillet et al. 2011) (Fig. 1.1), although the 1897 event may have been located further north than shown in Figure 1.1. Feuillet et al. (2011, fig. 10) argued that increased Coulomb stress loading from the 1843 event could have initiated slip on the Montserrat–Bouillante Fault System in 1897. From 1897 to 1899, SHV showed the first unequivocal signs of volcanic unrest (Perret 1939), followed by three further episodes of volcano-tectonic (VT) seismicity, deformation and changes to the hydrothermal system in 1933–1935 (Powell 1938), 1966–1967 (Shepherd et al. 1971) and 1992–1995. This apparent 30 year cycle of unrest probably represents repeated attempts by magma to reach the surface through the brittle crust. Burgisser & Bergantz (2011) proposed a general model of remobilization of a mushy andesitic magma reservoir by a basal mafic magma intrusion leading to convective mixing, and, perhaps, the eruption of silicic magma. Their model, when parameterized for SHV conditions, suggests a period of a few years between initial intrusion and the convective overturn, leading to the buoyant rise of andesitic magma.
Montserrat's position on the NE part of the Caribbean plate means that it experiences intraplate NNW extension (inset panel). The last major thrust earthquake on the subduction zone east of Montserrat was in 1843 (the orange ellipse is a possible rupture trace) and may have triggered subsequent smaller earthquakes in Guadeloupe in 1851 and 1897 at the end of the Montserrat–Bouillante Fault System (shown as the red line in the main panel). In the orange panel: the four volcano-seismic crises leading to the eruption that began in 1897 are shown in white, following the three seismic events discussed (in black). The previous (Castle Peak) eruption occurred in about 1600.
40Ar/39Ar ages of the oldest rocks from Montserrat (Harford et al. 2002) suggest that magmatism began on the island around the last major change in Caribbean plate motion, approximately 2.5 Ma. Drilling in 2012 through about 130 m of marine sediment, around 60 km SW of Montserrat (IODP Expedition 340 Scientists 2012), revealed tephra layers from much of the approximately 5 Ma record there, but their provenance is not yet certain. The southerly migration of the locus of volcanism (Silver Hills, c. 2 Ma; Centre Hills, c. 1 Ma; SHV, <0.25 Ma) occurred at about 5 mm a−1, compatible with the predicted rate of Caribbean plate extension in this region of about 6 mm a−1 (Lopez et al. 2006). Silver Hills and Centre Hills are surrounded by distinct submarine shelves about 5 km wide and as deep as 60–100 m (mainly 20–60 m) below sea level (Fig. 1.2) (Le Friant et al. 2004). This must represent marine erosion produced during lower sea-level stands in the late Pleistocene (Le Friant et al. 2004). SHV, on the other hand, has no such shelf, suggesting a relatively youthful history and ongoing activity that prevented such a feature forming. South Soufrière Hills is anomalous as it is the only known basaltic centre on the island and seems to have developed over a short time period (c. 140–127 ka: Harford et al. 2002; Le Friant et al. 2009; Cassidy et al. 2014). Like other basaltic centres on the major volcanic islands of the Lesser Antilles (e.g. M. Caraibes, Basse Terre, Guadeloupe), it has a peripheral location compared to the main andesitic massifs, suggesting an independent conduit through the crust.
A chronological summary of the eruption
This summary of the main stages of the eruption is largely taken from both the MVO reports and published papers. Extrusive phases (Table 1.1) are treated as beginning when lava first appeared, times and dates are given in Coordinated Universal Time (UTC), and the locations mentioned are shown in Figure 1.2.
Map of Montserrat with the topography as it was in late 2009. Contours are at 100 m intervals, the main roads in red. The outline of the largely buried English's Crater is shown by the dashed black line with GW marking Gages Wall and FW marking Farrells Wall. The location of MVO and the four CALIPSO boreholes (GERD, OLVN, AIRS and TRNT) are marked. A shaded relief digital elevation model (DEM) image of Montserrat and its submarine shelf is inset.
Characteristics of the phases of extrusion and pauses
1995–1998
Phase 1 (15 November 1995–10 March 1998)
This period of the eruption is covered thoroughly by Geological Society, London, Memoir 21 (Druitt & Kokelaar 2002). Following 3 years of elevated local seismicity (Shepherd et al. 2003; Odbert et al. 2014b), the eruption began on 18 July 2005 with a phreatic phase that became magmatic on 15 November 2005, following extrusion of a solid plug in late September. The first PDCs were produced in March 1996. The rate of lava extrusion, dome growth and explosivity accelerated during 1996 (Sparks et al. 1998), and the lava dome began to overtop the previously enclosing English's Crater, spilling rockfalls and PDCs into the valleys radially around the volcano (Cole et al. 1998, 2002). A large Vulcanian explosion on 17 September 1996 destroyed about 35% of the dome (Robertson et al. 1998; Calder et al. 2002), and set a pattern that was to be repeated many times. The morphology of the dome varied considerably and the production of spines, whalebacks, shear lobes and ‘pancake’ lobes was shown to be partly dependent on the rate of lava extrusion (Watts et al. 2002). On 25 June 1997, part of the lava dome on the northern flank collapsed, sending PDCs and surges over 6 km to the NE as far as Bramble airport and killing 19 people (Loughlin et al. 2002). By early August 1997, PDCs entered the long-evacuated town of Plymouth, destroying the centre and removing all doubt about the long-term viability of the island's former capital. Between August and October of 1997 there were two series of Vulcanian explosions following a sub-daily cyclic pattern, 88 in total (Druitt et al. 2002b). Major dome collapses occurred on 21 September and early in November, consistent with a multi-week (6–7 week) cyclic pattern (Voight et al. 1998), and, on 26 December (Boxing Day) 1997, the long-forecast collapse of the hydrothermally altered Galway's Wall on the SW side of English's Crater occurred (Sparks et al. 2002; Young et al. 2002). This produced a large debris avalanche (Voight et al. 2002), followed by a southwards-directed lateral blast as the lava dome core was exposed. Lava extrusion stopped on about 10 March 1998, completing what is now known as Phase 1 of lava extrusion.
There followed an 18 month pause in lava extrusion. There was initially little surface activity, but this increased considerably later following a major collapse of the southern part of the dome on 3 July 1998. During this pause in extrusion, higher levels of activity were observed every 5–6 weeks (Norton et al. 2002), a period similar to the cycles recognized in Phase 1. Seismic swarms, ash venting and small explosions, described in detail by Norton et al. (2002), culminated in pumice-bearing explosions on 8–9 November 1999 following a 5 day hybrid earthquake swarm. The resumption of lava extrusion on 27 November 1999 marked the start of Phase 2.
1999–2010
Phase 2 (27 November 1999–28 July 2003)
This was the longest period of near-continuous lava extrusion (3.5 years) in the eruption to date. However, within it there were two probable cases of short breaks in extrusion of about 68 days (3 March 2001–10 May 2001) and about 48 days (1 June–19 July 2002). Phase 2 was characterized by three large dome-collapse events, sending much of the lava dome into the sea to the east via Tar River Valley. Young et al. (2003) showed that sub-daily cyclic behaviour involving magma flux and gas flux occurred during this phase almost from its inception, with the gas flux peak being delayed by several tens of minutes from the peak of swarm seismicity.
The partial collapse of the dome that occurred on 3 July 1998 during the pause in extrusion (Norton et al. 2002) left a deeply exposed core to the dome within which new lava began to extrude in late November 1999. This new dome grew rapidly in early 2000. However, on 20 March 2000, virtually the whole of the new volume added (c. 28×106 m3 dense rock equivalent) collapsed down the Tar River Valley over a 5 h period, producing many PDCs and a large ash column 9 km high (Carn et al. 2004). Vigorous lava extrusion at rates of 3–5 m3 s−1 began almost immediately. The dome grew in height until, by mid-2000, its summit exceeded 1000 m above sea level (asl). When the dome has a summit elevation above about 960 m asl it is usually sufficiently large to shed PDCs in any direction (Wadge 2009) (Fig. 1.3). From mid-2000 to mid-2001, dome-collapse PDCs went to the east and to the south. Following the hiatus in extrusion from March to May 2001, dome growth continued to the south until another large-volume (45×106 m3) collapse occurred to the east on 29 July 2001. Again, most of the collapse material ended in the sea (Le Friant et al. 2010) (Fig. 1.4). The case for a rainfall trigger for this collapse event was made by Matthews et al. (2002) (Table 1.2); this topic is discussed later.
Changing elevation of the summit of the lava dome throughout the eruption (black line), together with the direction of dome growth coloured by quadrant. (a) Phase 1, (b) Phase 2 and (c) Phases 3, 4 and 5. Note the different timescales on each plot. The dome has to achieve an elevation of about 960 m asl (black dashed line) before it can overtop English's Crater rim and send dome-collapse pyroclastic flows in any direction other than east.
Looking west up the Tar River Valley on 31 July 2001, 2 days after the 29 July 2001 dome collapse removed about 45×106 m3 of the dome, mainly into the sea. Note how the PDCs created during the collapse blanketed the surrounding area but also eroded the valley floor and the delta that had been growing there since 1996. The erosive channel is about 150 m wide at the sea. Image ©BGS (NERC)/Government of Montserrat.
Major dome collapses (>10×106×m3)
Extrusion resumed to the east about 5 days after the 29 July 2001 event. In November 2001, extrusion switched to the west and rapidly rebuilt the dome up to about 1000 m asl by the end of the year. Following the June–July 2002 hiatus, lava began to extrude to the north for the first time since 1997, and the period up to the end of 2002 produced a number of substantial PDCs with runouts of 4–6 km to the NE. In April 2003, the rate of extrusion began to fall as lava was directed to the NE and the dome reached an elevation of over 1100 m asl (Figs 1.3 & 1.5).
The Tar River Delta and Valley looking to the west on 21 February 2003. The large dome (summit c. 1050 m asl) has an apparent conical profile. Since the previous large collapse on 29 July 2001 (Fig. 1.4), 19 months of deposition from PDCs has filled much of the Tar River Valley and expanded the delta. Image ©BGS (NERC)/Government of Montserrat.
Phase 2 ended soon after the collapse of the dome on 12 July 2003. This was the first time that a major collapse seemingly was able to end magma flow to the surface. A swarm of 9500 hybrid earthquakes began on 9 July and merged into tremor by 12 July, indicative of magma pressurizing within the upper conduit. Ottemoller (2008) identified five ‘epochs’ within this series and changes to the interevent periods. There was also a period of heavy (c. 30 mm in 140 min) rainfall immediately preceding the collapse. The collapse began gradually and lasted for a total of about 18 h, with relatively small PDCs to the east building in size and removing much of the eastern talus deposit over a 10 h period (Herd et al. 2005). Much larger PDCs followed, eating into the dome core, and most of this was removed in a climactic period of less than 3 h. The collapse was captured by the CALIPSO borehole strainmeters (Voight et al. 2006) which indicated two major collapse pulses. The largest PDCs caused tsunamis (Mattioli et al. 2007) and a series of hydrovolcanic explosions (Edmonds & Herd 2005) that sent surges inland from the shore (Fig. 1.6). The submarine deposits from these flows formed two major lobes and a distal turbidite deposit (Trofimovs et al. 2006). The collapse event ended with a Vulcanian explosion, and four further explosions at progressively longer intervals from 13 to 15 July 2003 (Edmonds et al. 2006; Voight et al. 2010b) were followed in the sequence by a pulse of sulphur dioxide on 16 July (Edmonds & Herd 2007). Magma appears to have risen in the conduit after this, but only just breached the surface to form a very small (tens of metres in diameter) extrusion between 21 and 29 July 2003 (Fig. 1.7). The collapse removed about 210×106 m3 of the dome, talus and pre-existing Castle Peak dome, easily the largest such event of the eruption (Herd et al. 2005) (Fig. 1.8). Ostensibly, this was the largest historical dome collapse worldwide. The resulting amphitheatre largely followed the form of the old English's Crater but with a spectacular cross-section through the remaining Phase 2 talus on the northern side (Fig. 1.9). Also, a north–south-orientated ridge of dome core was preserved to the east of the conduit vent (Fig. 1.9). The ash fallout from the 12 July eruption produced substantial deposits in the populated NW coastal areas. These deposits caused difficulties in removal, damage to infrastructure, impact on health and major expense (Edmonds et al. 2006; Baxter et al. 2014) (Fig. 1.10).
The sugar mill at White's Yard, NE flank of the volcano, on 21 July 2003. A surge deposit about 0.2 m thick was left here during the 12–13 July 2003 collapse as a result of hydrovolcanic explosions when large PDCs entered the sea at the height of the event (Edmonds & Herd 2005). The on-land surge direction was from right to left (to the NW). Image ©BGS (NERC)/Government of Montserrat.
The crater area viewed from the SSE in September 2003 following the 12–13 July 2003 dome-collapse event. A vertical, white plume of gas rises from the vent area. Immediately above the conduit is a small extrusion of lava, a jagged triangular feature just right of the centre of the image, that formed unobserved during the last week of July 2003 at an altitude of about 650 m asl. In front of this is a tephra rampart that probably formed during one of the Vulcanian explosions following the collapse, between 13 and 15 July. In the right central part of the image is a ridge of old dome core rock that extends through the fume to the north wall of English's Crater and survived the 12–13 July event. Image ©BGS (NERC)/Government of Montserrat.
Images taken from Windy Hill looking south to the volcano before (left) and after (right) the dome collapse of 12–13 July 2003. The dashed black line shows the skyline after the collapse. The summit of the dome was about 1100 m asl before and about 850 m asl after the collapse. The post-collapse summit known as the NW Buttress on the right was itself destroyed during the explosion of 4 March 2004. Image ©BGS(NERC)/Government of Montserrat.
View NW across English's Crater from Perche's Mountain on 5 November 2005. On the left is the 3 month-old growing lava dome of Phase 3 with white fume around it. Grey talus from the dome is just starting to spill out from the ridge of old dome core (brown/black colours) that survived the 12–13 July 2003 collapse (Fig. 1.8). The finely runnelled buff-coloured slopes to the right are comprised of talus from the Phase 2 (1999–2003) dome. Almost all of these deposits were removed in the wholesale dome collapse of 20 May 2006. ©D. Macfarlane.
Ash deposits at Flemmings on 13 July 2003 following the dome-collapse event. Because of unfavourable winds, several centimetres of ash were deposited over much of northern Montserrat, which took a considerable effort to remove. Image ©BGS(NERC)/Government of Montserrat.
A pause in lava extrusion of 2 years followed the collapse of July 2003. Activity in the rest of 2003 and early 2004 was limited to occasional short periods of ash venting from the conduit, seven hybrid and long-period earthquake swarms (from 27 September 2003 to 21 February 2004), and pulses in sulphur dioxide emission. On 4 March 2004, and preceded by about 20 h of low-level tremor, there was an explosion that produced a 7 km eruption column and a PDC to the sea in the Tar River Valley (Green & Neuberg 2005; Linde et al. 2010). Also, one of the remnants of the Phase 2 dome core, the ‘NW Buttress’, was destroyed. In the next 2 weeks, there were small explosions, PDCs and vigorous ash venting, but no indication of new lava. Tremor signals decayed through April, ending in May 2004. The next ‘event’ was somewhat cryptic. In mid-October 2004 there was an unusually large (13 000 t/day) pulse of sulphur dioxide emission roughly coincident with a period of increased extensional deformation and elevated numbers of VT earthquakes. This period also appeared to mark the initiation of unusually elevated levels of spring flow from Centre Hills springs. Another episode of elevated sulphur dioxide emission coupled with VT earthquakes occurred on 6–7 March 2005. On 15 April 2005, a strong steam and non-juvenile ash-venting episode began on a NW-orientated fracture within the crater. This was later recognized as the start of a 3.5 month-long series of phreatic episodes precursory to the beginning of Phase 3 (Cole et al. 2014a). On 13 June 2005, a series of new vents opened on a NNE-trending fracture across the crater floor. Some of the ash produced in these events formed curvilinear tephra ramparts. On 28 June 2005, a Vulcanian explosion with a 7 km-column and minor PDCs occurred. This was followed by four more explosions on 3, 9, 18 and 27 July 2005, the most vigorous (on 18 July) reaching 10 km in altitude. Juvenile material was recognized in the last of these.
Phase 3 (1 August 2005–20 April 2007)
A newly extruded lava dome was first recognized (in a period of cloud cover) on 8 August, and growth is assumed to have begun on 1 August 2005 (Loughlin et al. 2010). The rate of growth of the dome was low (<1 m3 s−1) and the dome grew above the long-term conduit, buttressed on its eastern side by the north–south ridge of the remnant Phase 2 dome. The NW-trending fracture and gas vent of 15 April 2005 also appeared to play a role in initial asymmetric growth to the NW. By late October, the rate of growth was about 1 m3 s−1 as the ridge was buried. The transition from endogenous to exogenous growth (cf. Hale & Wadge 2008) occurred in November 2005 (Fig. 1.9). In late December, the extrusion rate increased by about fourfold (Ryan et al. 2010) and the dome grew rapidly. On 7 February 2006, following a series of VT earthquakes, extrusion seems to have stopped until, on 9 February, a swarm of long-period earthquakes marked the opening of a NNE-trending intracrater fissure with vigorous ash and steam emission at a northern vent (Fig. 1.11). By 11 February, the extrusion of lava was at a very high rate (>15 m3 s−1: Loughlin et al. 2010) which was maintained for the next 2 weeks, and produced a pancake-shaped lobe almost reaching the rim of Gages Wall to the west (Fig. 1.12). Thus, the start of Phase 3 involved high-level fractures facilitating magma transport, perhaps enhanced by precursory phreatic activity.
The lava dome on 10 February 2006 seen from the ENE. Across the summit of the dome there is vigorous ash venting along a NNE–SSW line, including orange ash. Lava extrusion at a high rate following this formed a ‘pancake lobe’ on the west (far) side of the dome. This marked the early part of a multi-week cycle of activity that lasted until March (Loughlin et al. 2010; Odbert et al. 2014b). Image copyright BGS(NERC)/Government of Montserrat.
View from the SE over Galway's Mountain on 14 February 2006 shows the steep-sided dome with a small spine, behind which (to the NW) is the flat-topped pancake lobe, with an overhanging margin, which began growing a few days before (Fig. 1.11). The dome is still entirely confined by the walls of English's Crater. Image copyright BGS(NERC)/Government of Montserrat.
By March 2006, dome growth was directed to the SE. In late March, there was an increase in growth rate (Wadge et al. 2008a, b), and, by mid-May, the dome had a stubby lava spine with a diameter of about 50 m (wider than the average 30 m) reaching just over 1000 m asl (Fig. 1.13). The occurrence of low-frequency earthquakes increased on 18 May 2006, and the direction of dome growth switched on 20 May just prior to another major collapse. This event appears to have been triggered by a high flux pulse of magma coupled with heavy rainfall (Loughlin et al. 2010). The collapse lasted about 3 h, but the most intense phase took only 30 min. The collapse removed the whole of the Phase 3 dome and the Phase 2 remnants, a total volume of about 97×106 m3. Like the 12 July 2003 event, high-energy PDCs entering the sea at Tar River (Trofimovs et al. 2011) produced a small tsunami, and hydrovolcanic surges moved inland to the NW (Fig. 1.14). Ash from two explosions at the end of the event sent fragments of up to 6 cm in diameter as far as Olveston, but little pumice. About 0.2 Tg (1 Tg=1×1012 g) of sulphur dioxide emitted into the atmosphere, together with the 17–20 km high ash plume, was measured by satellite, the largest such event in the eruption (Carn & Prata 2010). Some people in Old Towne experienced breathing difficulties in the hour after the last explosion, possibly due to a gravity flow of gas along the Belham Valley. Whilst there are many similarities between this collapse event and that of 12 July 2003, in the case of 20 May 2006, the collapse was much briefer, more intense, and the dome mass involved was much more recently emplaced and, hence, probably hotter and more gas-rich.
A surveying party on Galway's Mountain looking south to the dome on 18 May 2006. An unusually wide spine (c. 50 m in diameter) had been emplaced on the SW side of the dome, with vigorous summit ash venting. Two days after this, the whole dome and part of the substrate collapsed on 20 May 2006. Image ©BGS (NERC) /Government of Montserrat.
View south along the east coast of Montserrat during the dome-collapse event of 20 May 2006 (just prior to 08:15 ECT and near the climax of the collapse). The ash clouds are from pyroclastic density currents entering the sea off Tar River Valley. Image courtesy of R. Bellers.
Lava extrusion resumed in less than 8 h of the collapse at a high flux (c. 10 m3 s−1). The removal of the residual Phase 2 core lava in the 20 May collapse might have made this new dome even more prone to collapse down the Tar River Valley, but this proved not to be so and there were no major collapses of the dome to the east in the succeeding 7 years. By mid-June 2006, the first evidence of the transition from endogenous to exogenous dome growth was observed in the new dome, followed by a succession of shear lobes and minor collapses. By the end of August 2006, two ‘roaring’ vents producing high-pressure gas had formed; one to the north and one to the west of the new dome. The western one came to be known as the Gages Wall vent (Fig. 1.15). By September 2006, the dome had a volume of about 80×106 m3. Dome growth continued to the NE and then to the SW, and the northern Farrell's Wall of English's Crater was overtopped by small PDCs in early December, for the first time in Phase 3.
The NW rim of English's Crater on 9 November 2006. The lava dome is to the left, largely shrouded in steam and ash. To the right is Chance's Peak. Just right of centre is a column of ash from the newly established Gage's Wall vent just inside the crater rim. New deposits from short fountain-collapse flows can been seen extending into Gage's Valley. Image ©BGS (NERC)/Government of Montserrat.
A vigorous pulse of lava extrusion began on 24 December 2006 on the NW side of the dome, rapidly expanded the dome there, and produced ash venting and explosions to 4 km. Small PDCs began to descend Tyers Ghaut. On 8 January 2007, an explosion and collapse on the NW sector of the dome (De Angelis et al. 2007) sent a PDC into the Belham Valley (Fig. 1.16), the largest in that sector to that date, with a runout of 5 km, as far as Cork Hill (Fig. 1.17). Dome growth continued at diminishing rates for the first 4 months of 2007, building up a NE-facing lobe. Extrusion became increasingly difficult to detect in April, but a date of 20 April 2007 is taken as the end of Phase 3, with the dome having a volume of about 173×106 m3 and a summit elevation of about 1047 m asl (Ryan et al. 2010).
On 8 January 2007, a PDC originating on the NW side of the dome entered Tyers Ghaut and extended down the Belham Valley. This view shows Tyers Ghaut that same day looking SE towards the lava dome. Whilst the Ghaut contains several metres of block-and-ash deposits, the valley margins are draped by much thinner but more extensive surge deposits. Image ©BGS (NERC)/Government of Montserrat.
Looking west down the Belham Valley on 8 January 2007. The pale block-and-ash flow deposits formed that morning have a sharp margin. Contrast this with the much wider and more diffuse proximal surge deposits (Fig. 1.16). The deposits extend as far as Cork Hill about 5 km from the volcano. Image ©BGS (NERC)/Government of Montserrat.
Phase 3 was different from the earlier phases of the eruption in the relative paucity of explosions relative to Phase 1, and only one major collapse event compared to several in Phase 2. However, the normalized curves of magma output for phases 1 and 3 are similar, with pronounced acceleration at the start; deceleration at the end of Phase 3 was more pronounced than in Phase 1 (Ryan et al. 2010). Loughlin et al. (2010) showed that cyclic behaviour, including 2–6 week pulses of lava flux (equivalent to the c. 6 week cycles of Phase 1) and approximately 2 week cycles of seismicity, played a variable role in Phase 3.
During the first year of the pause following Phase 3, there was little surface activity, but the Gages Wall vent was a prominent feature, producing gas, occasionally at high pressure, and was associated with long-period earthquake swarms (e.g. 28 January 2008). An increase in VT earthquake activity between 27 April and 5 May 2008 heralded a period from 5 May to 27 July 2008 that involved ash venting and several small explosions up to 3 km high from the Gages Wall vent (Cole et al. 2014a), which by this time was clearly the main vent for gas release.
Phase 4 (29 July 2008–3 January 2009)
The increasing tempo of surface activity culminated in a major seismic crisis leading to an explosion on 29 July 2008. From 21 to 26 July, a series of VT earthquakes was replaced by a 2 day-long swarm of hybrid and long-period events starting on 27 July. After an 8 h lull in seismicity, the explosion on 29 July produced a 12 km-high ash column rich in pumice, generating pumice flows to the east and west (Chardot et al. 2010; Komorowski et al. 2010). The source of the explosion was shown to be an enlarged Gages Wall vent, the first clearly off-conduit vent of the eruption (International Charter on Space and Major Disasters 2008).
When the Gages Wall vent was first observed directly after the explosion on 14 August, its enlarged upper rim (c. 50–100 m across) was filled with lava that was spilling westwards into the upper Gages Valley (Fig. 1.18). On 25 August, after heavy rainfall, a block-and-ash flow was generated from this lava, travelling about 2 km. Any further extrusion of lava was at a very low rate and probably stopped altogether in early September. It was also observed that the western side of the lava dome had become extensively fractured, probably during the July seismic crisis, with high heat flow from those fractures seen in thermal imagery. The 3 months from September to early December 2008 showed no surface volcanic activity, although there was considerable erosion of the upper talus slopes on the eastern side of the dome, particularly during the passage of Hurricane Omar in mid-October 2008.
Lava spills from the crater formed during the 29 July 2008 explosion at the Gages Wall vent into the Gages Valley, imaged here on 14 August 2008. To the right is the dark mass of Chances Peak, to the upper left is Gages Wall with the lava dome above. The buff-coloured rocks in the foreground are on the southern side of Gages Mountain. Image ©MVO/Government of Montserrat.
An explosion on 3 December 2008 followed a small dome collapse and PDC, and marked the start of a nearly 1 month-long period of increased seismicity and, later, of rapid dome growth (Komorowski et al. 2010) (Fig. 1.19). The explosions fired ballistic blocks up to 2 km away. An increase in seismicity before this event was detected from 28 November onwards, but only interpreted as precursory after the event. By 10 December, a new lobe of lava, shedding PDCs into the Gages Valley from a vent above the Gages Wall vent, built rapidly at an extrusion rate of >10 m3 s−1. This was the first time since late 2001 that the dome had produced major PDCs outside the crater to the west (Figs 1.3 & 1.19). The Gages Wall vent appears to be blocked from this time onwards. By 20 December 2008, PDCs were sent into both White River to the south and Tyers Ghaut to the north. On 24 December 2008, a PDC reached the sea at Plymouth, and a 3 km-long PDC reached Lee's via Tyers Ghaut on 2 January. Two explosions on 3 January 2009 lofted ash to 11 km (Fig. 1.20). In contrast to those of 3 December 2008, these explosions were pumiceous (Komorowski et al. 2010), and they mark the end of Phase 4 after the extrusion of about 39×106 m3 of lava. Strains generated by the 29 July 2008 and 3 January 2009 explosions were similar and interpreted as evacuations of the cylindrical conduit, but the 3 December explosion was distinctive, interpreted as including the precursory pressurization of a feeder dyke (Chardot et al. 2010).
View of Plymouth taken on 17 December 2008 from the NW, with the Gages Valley and Gages Mountain in the background, and St Georges Hill to the left. The dome and southern part of the volcano are obscured by ash-bearing plume. The palest surfaces are block-and-ash deposits and marginal surge deposits from PDCs produced on 5, 10 and 17 December 2008. Below them are the deposits from the PDC on 3 December 2008, whose surge extended almost to the top of St Georges Hill, and almost to Richmond Hill in the lower left of the image. Image courtesy of A. Finizola, J. B. de Chaballier & J.-C. Komorowski.
Eruption plume of the 3 January 2009 Vulcanian explosion seen from the NW (Old Towne). Image courtesy of E. Tomme.
Phase 4 with its short duration, high degree of explosivity, and, towards the end, very strong sub-daily cyclicity and a high average extrusion rate had a very different pattern to those of phases 1–3. Activity precursory to Phase 5 began with a series of ash-venting episodes preceded by a small swarm of 24 VT earthquakes in 1 h on 5 October 2009. From 5 to 7 October 2009, there was a series of 13 mildly explosive episodes, some with roaring sounds, and with plume heights of up to 6 km, and minor PDCs, mainly to the south (Cole et al. 2014a).
Phase 5 (9 October 2009–11 February 2010)
Lava was confirmed to be extruding at the southern end of the dome on 9 October 2009. Extrusion was mainly focused on the SW sector of the dome summit for October (Fig. 1.21) and much of November 2009, after a period of extrusion to the north between 11 and 14 October 2009 (Stinton et al. 2014a). PDCs reached the sea to the south via White River and also entered Gingoes Ghaut for the first time in the eruption (on 4 November 2009) as the upper reaches of the White River drainage filled with deposits (Fig. 1.22). The great height and central location of the vent meant that sometimes PDCs were able to descend several valleys at the same time. By 10 November 2009, growth moved to the west and PDCs entered the Gages Valley and from there into Spring Ghaut.
Oblique, true-colour image of Montserrat taken from the International Space Station on 11 October 2009 (north is to the bottom right). An eruption plume rises above the volcano and is then advected to the west. Lava first began extruding 2 days earlier, marking the start of Phase 5. New (paler) block-and-ash flow deposits from these first few days of activity can be seen extending to the south via White River to the delta off the south coast. Image ©NASA.
A PDC created by dome collapse heads southwestwards down Gingoes Ghaut towards the sea, 4 November 2009. Note the recent infill by PDC deposits of the White River in the foreground. Image ©MVO/Government of Montserrat.
Following a decrease in activity, on 19 November 2009, a series of hybrid and VT earthquakes marked a renewed vigour and a shift of the locus of extrusion to the NE. At this time and throughout much of Phase 5, sub-daily cycles of seismicity and surface activity were present. At the peak of these cycles, ash venting and plumes (up to 5 km) with accompanying PDCs were typical. This behaviour was similar to that during the cyclicity of October 1997, but less explosive. By early December 2009, extrusion shifted to the NW and Tyers Ghaut was filled with the deposits from many short PDCs. This pattern of valley-filling by block-and-ash deposits rotating around the volcano was well captured at the time by satellite radar (Wadge et al. 2011). A large Vulcanian explosion on 8 January sent a plume to about 8 km in altitude and produced fountain-collapse PDCs in several valleys. By volume (about 2.7×106 m3 DRE (dense rock equivalent)) this explosion was the largest of the eruption to date, yet it was relatively pumice-poor. The PDC that entered the Belham Valley from this event had a runout of 6 km, the longest yet in that drainage, and just short of the populated area (Fig. 1.23). There were two more Vulcanian explosions on 10 and 11 January, one of which generated fine-grained pumice lapilli fallout. Accompanying these explosions, the lava extrusion was more directed to the west, and PDCs reached the sea at Kinsale via Spring Ghaut for the first time on 18 January 2010. Sub-daily cycles were weaker in January 2010. Another Vulcanian explosion occurred on 5 February, again sending PDCs into Spring Ghaut and out to sea, and another on 8 February 2010.
A PDC advances down the Belham Valley observed from MVO on 8 January 2010. The runout distance of this flow was 6 km, the furthest to date in the Belham Valley, and the Vulcanian explosion responsible was the most voluminous to date, with multiple flows produced by fountain collapse. Image ©MVO/Government of Montserrat.
Phase 5 was terminated (as was Phase 2 in July 2003) by a large dome collapse, on 11 February 2010, the largest to occur on the northern side of the volcano (Stinton et al. 2014b). This produced a horseshoe-shaped amphitheatre facing north, and involved about 40×106–50×106 m3 of dome and talus (Fig. 1.24). Unlike the collapses eastwards into the Tar River Valley, the base level for this collapse was over 100 m higher, at about 800 m asl, the elevation of the English's Crater walls on the north side. The event took about 2 h, with initial dome-collapse PDCs reaching the NE plains. The collapse became more vigorous, and high-energy surges destroyed much of the previously abandoned villages of Harris and Streatham (Figs 1.24 & 1.25). The deposits covered most of the NE slopes to depths of 2–10 m and extended the shoreline there by up to 650 m (Figs 1.26 & 1.27) (Stinton et al. 2014b). Two, terminal, Vulcanian explosions resulted in tephra plumes reaching 15 km in height (Fig. 1.28). These plumes dispersed slowly over the SE Caribbean, and deposited ash on Guadeloupe, Dominica and St Lucia, damaging fruit crops and causing aircraft disruption (Komorowski et al. 2011).
Remnants of a building in Harris village following the high-energy pyroclastic currents associated with the 11 February 2010 dome-collapse event. The 11 February collapse scar is in the background. Image courtesy of B. Voight.
Giant block of dome lava in dome-collapse PDC deposits of 11 February 2010, at the headwaters of Farm River near Bugby Hole Estate. The block moved in the flow, overtopped the ‘Harris Ridge’ near Windy Hill and descended downslope towards the river. Image courtesy of J.-C. Komorowski.
False-colour satellite images (red represents vegetation-covered surfaces; and grey represents volcanic deposits) of Montserrat illustrating the new deposits produced during Phases 4 and 5 of the eruption. The left–hand panel image was taken on 6 July 2007 by Envisat 3 months after the end of Phase 3. The right-hand panel image was taken on 23 February 2010 by ASTER, just 12 days after the huge collapse to the north of 11 February 2010. Envisat image ©European Space Agency; ASTER image is ©JAXA.
Aerial view from Trants Bay SW towards the volcano, looking over the new deposits and shoreline extended by approximately 650 m after the major pyroclastic events of 11 February 2010. Image courtesy of R. Syers.
The 15 km-high tephra column produced during the lava dome collapse of 11 February 2010 imaged from a commercial aircraft and looking to the NE. A separate small tephra plume can be seen (arrowed) rising from a PDC travelling down the Belham Valley. Image courtesy of M.-J. Pekala.
Prior to the collapse, the dome had reached its highest elevation to date, 1150 m asl, and was reduced to about 1080 m asl after the event. In addition to the very clear sub-daily cycles demonstrated in Phase 5, there is also evidence (Odbert et al. 2014b) for three multi-week cycles (9 October–20 November 2009; 20 November 2009–8 January 2010; 8 January–11 February 2010). Phase 5 was like a longer version of the December–January part of Phase 4, and strengthens the case for a change in the behavioural regime between Phase 3 and Phase 4.
After Phase 5 ended, the surface activity was very muted, with ash venting on 25 June and 2 July 2010. Hot magmatic gases maintained semi-permanent incandescent fumaroles throughout 2010–2012, particularly on the northern headwall of the 11 February 2010 collapse crater, although the hottest was at the base of the scar near the location of the buried English's Crater rim. This may be equivalent to the ‘northern’ vent of early 2006.
Summary of eruption phases and trends
There have been five phases of lava extrusion with intervening pauses. As discussed already, Phases 1–3 were much longer-lived than Phases 4–5, although the latter were as vigorous as anything seen during the whole eruption. Observed surface activity during the pauses varied considerably. Typically, there was very little activity in the months immediately after extrusion ended but then increasing levels of activity, precursory to the resumption of lava extrusion (Cole et al. 2014a) (Table 1.3). This suggests that magma makes it way to progressively higher levels, perhaps involving the water table and initiating seismic swarms, ash venting and explosions (e.g. July–November 1995; July 2008–November 1999; April–July 2005; May–July 2008). However, Phase 5 had only 3 days of precursory mildly explosive activity. The ash venting and explosions produced tephra comprising variable proportions of lithic and juvenile components (Cole et al. 2014a).
Precursory activity prior to extrusive phases
Each phase has involved addition of lava to a lava dome via a ‘stable’ cylindrical conduit. The location (16.7105°N, 62.1761°W) and ‘top’ of the conduit (c. 650 m asl) for magma extrusion has remained constant throughout the eruption. The one debatable exception is the extrusion of lava from the Gages Wall vent in the first part of Phase 4, although this could be interpreted as a westerly-inclined, upwards branch from the main conduit. Whilst the conduit position appears constant, the domes themselves are not. Although the morphology of each new dome is generally similar to predecessors (Wadge et al. 2009), repeated collapses have removed all of the pre-existing Castle Peak dome and all of the domes emplaced from 1995 to May 2006. Phases 4 and 5 added lava to the western half of the dome (Fig. 1.29).
Stacked profiles of the dome taken at different times from images acquired by the automated camera on Windy Hill, approximately 4 km NNW of the dome. The blue profiles are from the Phase 2 (1999–2003) dome, the red profiles from the August 2005–May 2006 Phase 3 dome, and the green profiles from the May 2006–April 2007 Phase 3 dome. The purple profile is taken from an AVTIS-3 acquired radar image in May 2011. The thick dashed lines represent the approximate positions of the boundary between core lava and talus on different domes. Based on Wadge et al. (2009).
Much of the material lost from the dome has been transported to the sea (Le Friant et al. 2010) in collapses, and by PDCs and lahars. Lahars are triggered by heavy rain, most often late in the rainy season from June to November after rain rates of >10 mm/day, irrespective of active dome growth (Barclay et al. 2007). However, Alexander et al. (2010) showed that during the 20 May 2006 dome-collapse event, synchronous tephra fall and heavy rainfall produced extreme flash flood conditions. In the Belham River Valley between 2001 and 2004, the average aggradation rate in the valley bottom was about 0.4 m a−1 and the shoreline at Old Road Bay had advanced by over 120 m due to prograde deposition (Barclay et al. 2007).
Lava volume and flux
About 1063×106 m3 of andesite lava was erupted from 1995 to 2010 at an eruption-average rate of 2.4 m3 s−1. The DRE volume estimates are added cumulatively in Figure 1.30. The volume estimate data for Phase 5 from Stinton et al. (2014a) have been added to the data used by Wadge et al. (2010). The volumes produced during phases 1–3 were each about 300×106 m3, those in phases 4 (c. 39×106 m3) and 5 (c. 74×106 m3) were much less (Fig. 1.31). The similar volumes of phases 1–3 suggest that the magma reservoir behaved in the same way each time, limited either by volume or magmatic overpressure. Phase-averaged extrusion rates are given in Table 1.2, and vary between 2.9 and 6.8 m3 s−1. The near-constant output rate (smoothed over months) for most of these phases further suggests that magmatic decompression played a major role (Voight et al. 2010). To first order, the theoretical flux of magma released by an evacuating reservoir of compressible andesite magma, supplied from depth by volatile-saturated basaltic magma, can be maintained at a constant rate of about 5 m3 s−1 for a system like SHV (Woods & Huppert 2003).
Cumulative DRE volume of lava emitted (106 m3), colour-coded by extrusive phase: orange, Phase 1; green, Phase 2; blue, Phase 3; grey; Phase 4; purple, Phase 5.
Cumulative volume of lava extruded through the five phases of lava extrusion from 1995 to 2010 (extended from Wadge et al. 2010) (black line). Cumulative emission of sulphur dioxide (dashed red line) throughout the eruption showing its continued emission during pauses in lava extrusion. The dashed blue curve is the, smoothed, northwards motion of the MVO1 cGPS receiver, showing the first-order ‘sawtooth’ behaviour of surface deformation associated with the extrusive phases. The five extrusive phases are represented as the pale brown panels, numbered 1–5. Periods of ash venting and explosions precursory to extrusion are shown as pale blue panels. Numbers at the top of the plot are years.
Volatiles
The emission rate of sulphur dioxide has been measured daily, when possible, for most of the eruption (Christopher et al. 2010). These measurements have shown an average emission rate up to 2010 of about 574 t/day (Fig. 1.31). Continued degassing during pauses in lava extrusion and the low content of sulphur in the andesite magma means there must be near-continuous volatile release from basaltic magma (Edmonds et al. 2001, 2014). The long-term permeability of the system to volatile transport must also be maintained (Edmonds et al. 2003b), but the physical components contributing to this permeability may vary (e.g. fractures, bubble rise). Spikes in volatile release during explosions and major collapses, usually measured from space (Carn & Prata 2010), are consistent with higher-level storage and/or rapid transport (Edmonds & Herd 2007). The pattern of SO2 emission has showed distinct variation, but the peaks and troughs are not simply related to the phases of lava extrusion and pauses (Christopher et al. 2010, fig. 1). Second-order modulation of the steady-state emission pattern hints at other deep processes (Christopher et al. 2011), and short-term fluctuations may be driven by upper conduit dynamics (e.g. Watson et al. 2000; Young et al. 2003). The Fourier transform infrared spectroscopy (FTIR)-measured HCl/SO2 ratio generally rises by an order of magnitude when andesite magma is extruding (Christopher et al. 2010), and this may give an index of the rate of lava effusion (Edmonds et al. 2002), but has no forecast capability.
Deformation
The first-order pattern of long-term surface deformation measured by continuous GPS on Montserrat has proved to be quite simple (Fig. 1.31). When lava is extruding (at high rates of c. 5 m3 s−1) the surface deflates, and then inflates in times of non-extrusion (e.g. Odbert et al. 2014a). Modelling the crust as an elastic body shows that the source of the deformation is centred below the volcano at mid-crustal depths (with a mean of c. 13 km: Hautmann et al. 2010; Mattioli et al. 2010). However, measurements closest to the volcano are the least compatible with this view, suggesting that there is also another ‘near-field’ source of deformation. Independent borehole strainmeter measurements show a non-axisymmetrical character to this near-field deformation, inferred to be due to a dyke (Chardot et al. 2010; Linde et al. 2010). This interpretation is supported by modelling of the multi-week tiltmeter data from 1997 (Hautmann et al. 2009). The orientation of the feeding dyke may have varied during the course of the eruption (Chardot et al. 2010), although this seems unlikely. The strainmeters record transient strain events associated with magma transport during periods of both extrusion and non-extrusion. The sub-daily cycles of near-field deformation that were measured by tiltmeters in late 1996 and 1997 (Voight et al. 1998) have not been measured subsequently, but seismicity with a similar range of cyclicity, and previously correlated with proximal tilt (Voight et al. 1999), has been recorded in the other phases (e.g. Young et al. 2003; Stewart et al. 2009) and was clearly evident in Phase 5 (Odbert et al. 2014b).
Seismicity
The classification scheme used by most authors for the majority of volcano-seismic events at SHV is as follows (Miller et al. 1998): volcano-tectonic (VT) earthquakes, due to brittle faulting, generally impulsive P and S waves, and significant energy above 5 Hz; long-period (LP) earthquakes with nearly monochromatic energy at approximately 1–2.5 Hz and emergent waveforms; hybrid earthquakes with impulsive high-frequency onsets and long-period codas with typical spectral bandwidths of 0.5–4 Hz, often in swarms with repetitive signatures; tremor, a continuous seismic disturbance lasting several minutes or more, often due to coalescence of closely spaced hybrid events, with the spectra of the tremor identical to that of preceding hybrid events; explosion signals, starting with a long-period 1–2 Hz signal, followed commonly by a high-amplitude, high-frequency signal denoting ballistics and fountain-collapse PDCs; rockfall signals, emergent phaseless high-frequency signals that correlate with observed rockfalls and PDCs. Hybrid swarms have been the most common type of dome-growth related seismicity. Neuberg et al. (1998) emphasized that, although low-frequency events can be subdivided into the long-period and hybrid families, the two event types are similar and represent end members of a continuum, with common source processes.
There has been an overall reduction in the amount of seismicity recorded throughout the eruption. This is particularly so for hybrid events (associated with magma ascent and precursory to numerous dome collapses: Luckett et al. 2008) and VT events, which tend to appear in swarms prior to the renewal of extrusion. Fault-plane solutions for some of these VT events show how the stress field responsible can change abruptly (by 90°) at various times in the eruption (Roman et al. 2008; Miller et al. 2010), as can shear wave anisotropy (Roman et al. 2011). Low-frequency (mainly hybrid) and rockfall events tend to occur during lava extrusion. The low-frequency earthquakes are the best studied (Neuberg et al. 2000, 2006; Green & Neuberg 2006) and reflect dynamic processes within the conduit caused by magma flow (e.g. Collier & Neuberg 2006). As discussed later, we know that these events are formed at a limited depth range within the magma conduit itself, produce different families of events with nearly identical signatures that indicate nearly the same source locations and are sometimes intimately linked to the sub-daily cyclic processes of conduit pressurization.
Petrology
SHV has erupted hornblende–andesite lava of relatively restricted bulk composition, typically containing mafic enclaves, throughout the eruption. The lava is highly porphyritic, typically with 30–46% phenocrysts set in a groundmass ranging from glassy to microcrystalline with a rhyolite composition. Bulk lava SiO2 content has a restricted range, mostly between 56 and 62%. The proportion of mafic enclaves (with SiO2 content in the range 49–56%) in the lava increased from about 1% in Phase 1 to 3–12% in subsequent phases (Barclay et al. 2010).
Studies of the Phase 1 rocks (mainly found in Geophysical Research Letters Volume 25, 1998; Murphy et al. 2000; and in Journal of Petrology Volume 44, (8), 2003) included phase equilibria studies (Barclay et al. 1998; Couch et al. 2003; Rutherford & Devine 2003), mineral chemistry (Murphy et al. 2000), major, trace element and isotope geochemistry (Zellmer et al. 2003a), and melt inclusion studies (Edmonds et al. 2001). These studies established that the magma phenocryst assemblage and water content of melt inclusions are consistent with crystallization in a shallow magma chamber at depths constrained to be at least 5 km at temperatures in the range 820–870 °C. The SHV magma, however, is texturally complex and contains many disequilibrium features, including the break-up of mafic enclaves, indicative of the mingling of andesite magma with more mafic magma (Murphy et al. 2000; Humphreys et al. 2009a, 2010; Plail et al. 2014), convective mixing (Couch et al. 2001) and assimilation of older consolidated andesite (Harford & Sparks 2001; Genareau & Clarke 2010). Hornblende crystals commonly show reaction rims, which indicate the effects of both heating in the chamber and decompression (Devine et al. 1998; Buckley et al. 2006). Comparison of glass, melt inclusion and spectroscopic gas data has shown that gas is sourced from both the rhyolitic interstitial melt within the andesite magma and the quenching of more mafic magma against the andesite (Edmonds et al. 2011). The issue of long-term petrological and geochemical change over the course of the eruption is reported by Christopher et al. (2014) in this Memoir. They conclude that, although there was no progressive change of bulk composition with time, there was evidence that the Fe content of the mafic enclaves and the andesite host did change systematically. Also, the lavas of Phases 1 and 2 show evidence of assimilation of plutonic residue from earlier basaltic intrusion that is lacking in the lavas from later phases.
Some petrological and geochemical constraints have emerged on the deeper parts of the SHV system. Ridolfi et al. (2010) analysed amphiboles in SHV products and applied an Al-geobarometer. The main hornblende phenocrysts give pressures consistent with previous petrological estimates of a magma reservoir at about 5 km depth, but they also find pargasite crystals consistent with crystallization from more mafic magma at higher temperatures (946–994 °C) and pressures consistent with estimated depths of 12–23.5 km. Similar high temperatures were inferred for the mafic magma by Humphreys et al. (2009a, b). A problem with the Ridolfi et al. (2010) study is that they used amphiboles from mafic enclaves. Since the amphiboles were formed by quenching in the enclaves, it is hard to see how they could have formed at high pressure. Kiddle et al. (2010) have described plutonic xenoliths on Montserrat, some of which show cumulate and crescumulate textures, and are probably derived from the middle and lower crust. Zellmer et al. (2003a) showed that the SHV andesite is depleted in MREE (middle rare earth elements) relative to LREE (light rare earth elements) and HREE (heavy rare earth elements), a diagnostic feature of amphibole fractionation. Amphibole-rich cumulates observed by Kiddle et al. (2010) are probably examples of the complementary rocks formed by fractionation to generate the andesites.
Studies of trace element zoning in plagioclase phenocrysts were used by Zellmer et al. (2003b) to infer that these crystals have a residence time in the shallow upper-crustal chamber of no more than a few centuries. An apparent paradox is that the andesites all show radioactive equilibrium for U and Th isotopes (Zellmer et al. 2003b), which either means that, unusually for arc magmas, there was no enrichment of U during magma genesis in the mantle, or that the crustal residence time of the andesite magma is very long (>105 years).
Petrological methods have also been used to constrain eruption processes. Amphibole rims related to decompression can be applied to infer magma ascent speeds (Devine et al. 1998; Rutherford & Devine 2003) and the results are in broad agreement with similar studies at Mount St Helens, with ascent speeds of less than a few millimetres per second producing well-developed reaction rims. The compositions of microlites of plagioclase have been used to infer the pressures in explosively erupted clasts derived from magma in the conduit just prior to Vulcanian eruptions (Clarke et al. 2007). Secondary ion mass spectrometry depth profiling of plagioclase crystals has also demonstrated an increase in conduit temperatures during Vulcanian explosions (Genareau et al. 2009). The main findings are that the magma that is erupted in Vulcanian explosions extends down to about 2 km depth and that a highly vesicular region of magma is overlain by a dense, low-porosity plug (Burgisser et al. 2010; Giachetti et al. 2010). Devine & Rutherford (2014) argue in this Memoir that the tephra from the 11 February 2010 explosion is consistent with heating of the source andesite magma by about 10 °C prior to the eruption.
Much of the interpretation of the SHV andesite petrology depends on understanding and constraining the volatile content, in particular in relation to the pressure and inferred depth range of the magmas. These volatiles are discussed further below.
Monitoring, observations and experiments
MVO
The establishment of the Montserrat Volcano Observatory (MVO) was described thoroughly by Aspinall et al. (2002). Since 1997, it has been a statutory body of the Government of Montserrat and was initially managed by the British Geological Survey (BGS). The MVO was given a permanent home at Flemmings in 2002. In 2008, management of the MVO passed from the BGS to a partnership of the Seismic Research Centre (SRC) of the University of the West Indies and the Institut de Physique du Globe de Paris (IPGP). The staff currently comprises a Director and four scientists, together with between seven and nine technicians and administrative staff.
The monitoring backbone of the MVO was established between 1995 and 2000: local networks of seismometers, continuous GPS (cGPS) receivers, spectroscopic measurements of sulphur dioxide and geodetic measurement of the gross shape of the lava dome (Mattioli et al. 1998; Miller et al. 1998; Neuberg et al. 1998; Sparks et al. 1998; White et al. 1998; Young et al. 1998; Gardner & White 2002). It was improved technically over the next 10 years. The current network of 10 three-component broadband seismometers was upgraded in 2005 (Luckett et al. 2007). The original cGPS network, installed by G. Mattioli (University of Puerto Rico and then University of Arkansas) in collaboration with the MVO, was augmented in 2003 with Caribbean Andesite Lava Island Precision Seismo-geodetic Observatory (CALIPSO) stations, and was eventually incorporated into an MVO-operated network in 2009. In 2002, the COSPEC method was replaced by an automated network of scanning differential absorption spectrometers (DOAS) (Edmonds et al. 2003a). The dome continued to be surveyed by a combination of ground-based photogrammetry with occasional helicopter-based photogrammetry, and ground-based and helicopter-based (June 2010) LiDAR (Light Detection And Ranging) survey. Telemetered data from fixed cameras provided continuous coverage at 1 min intervals for some periods (Fig. 1.29). A number of minor monitoring programmes were also pursued, including: environmental monitoring (e.g. ground-level sulphur dioxide via diffusion tubes); EDM (Electronic Distance Meter) readings (a small network on the northern side of the volcano was re-established in 2004 and replaced after destruction in 2011); thermal camera imagery from the MVO and elsewhere; and carbon dioxide flux from soil profiles.
The data collected at the MVO between 1997 and 2008 is joint copyright BGS(NERC)/Government of Montserrat, and so is held at both the BGS(NERC) and the MVO.
The lure of SHV as a test bed for observational technologies brought a number of new developments. A series of sea-floor bathymetric imaging, sampling and drilling surveys from 1998 to 2010 revealed the extent of offshore deposits, and the complexity of depositional processes (e.g. Trofimovs et al. 2006; Le Friant et al. 2009, 2010). Infrasound arrays deployed at St George's Hill, Lees Yard and the Waterworks showed how the timing and directionality of explosive signals could be interpreted (Ripepe et al. 2010). Integration of infrasound with thermal camera data has allowed the dynamics of Vulcanian explosions and PDC emplacement to be measured (Delle Donne et al. 2014). Gravity surveys in the north of the island revealed signals that were interpreted in terms of groundwater dynamics (Hautmann et al. 2010). Radar was used in several experiments: acquired directly on Montserrat from satellite using a receiving station at the MVO in 2000; recording PDC deposits during Phase 5 using high-resolution satellite radars (Wadge et al. 2011); and measuring topography and dome growth using a custom-built ground-based mm-wave instrument, the All-weather Volcano Topography Imaging Sensor (AVTIS), from 2004 to 2011 (Wadge et al. 2008a, b, 2014).
Three internationally funded experiments had SHV as their focus during 2000–2010.
MULTIMO
This EU-funded, 2001–2004 project recognized that volcanic systems were often chaotic, and that multiple time series of geophysical data can be analysed in an integrated manner using geostatistical and other techniques (Jaquet et al. 2006a). It showed how new forecasting tools might be developed with optimum data (Jaquet et al. 2006b), and was reported in a special issue (issue 1–2) of Volume 153 of the Journal of Volcanology and Geothermal Research.
CALIPSO
At the heart of the NSF/NERC-funded CALIPSO project was the deployment of Sacks–Evertson dilatometers (strainmeters) in boreholes of sufficient depth (c. 200 m) to preserve the great sensitivity of the instruments (Fig. 1.32). Four such boreholes were drilled (Trants, Geralds, Olveston and Air Studios), and accompanied by tiltmeters, seismometers and cGPS receivers (Fig. 1.33). The strainmeter signals produced interesting insights into the crustal strain-fields associated with specific events: the collapse of 12 July 2003 (Voight et al. 2006), and the associated tsunami (Mattioli et al. 2007) and Vulcanian explosions (Voight et al. 2010a); the explosion of 4 March 2004 (Linde et al. 2010), and the explosions and month-long deformation of Phase 4 in 2008–2009 (Chardot et al. 2010; Hautmann et al. 2013). The addition of four cGPS receivers substantially strengthened the GPS network, enabling higher precision deformation analyses and improved insights (Mattioli et al. 2010; Voight et al. 2010b).
The CALIPSO Olveston site (OLVN) in March 2003, during strainmeter installation. The instrument installed here was designed for high temperatures by Selwyn Sacks (left). Technicians Brian Schleigle and the late Nelson McWhorter (in the background) assist. The site is now covered by a protective housing. Image courtesy of B. Voight.
Portable wireline drill rig at the Geralds (GRLD) site in north Montserrat, February 2003. The strainmeter and seismometer instruments were installed by wireline and cemented in place at a nominal depth of 200 m. Image courtesy of B. Voight.
SEA-CALIPSO
This was a seismic tomography experiment undertaken in December 2007 (Voight & Sparks 2010; Voight et al. 2010c, 2014), and funded by NSF, NERC, Discovery Channel TV, BGS and the FCO. It used temporary arrays of about 240 seismometers on land and 10 more on the ocean floor, together with shipborne airguns and streamers (Fig. 1.34). The results improved our knowledge of: the shallow tectonic structure; the local depth to the Moho (c. 30 km); the presence of denser, intrusive bodies in the upper crust below the three volcanic centres of Montserrat; and the existence of a low-velocity zone beneath SHV at depths of between about 5 and, at least, 7.5 km, probably representing a magma body that feeds the current eruption (Fig. 1.35). The early results from SEA-CALIPSO were published in Paulatto et al. (2009) and in a special collection of the Geophysical Research Letters in 2010 (e.g. Paulatto et al. 2010; Shalev et al. 2010), with the results of improved tomographical modelling (180 000 ray paths) and thermal modelling appearing subsequently (Fig. 1.35) (Paulatto et al. 2012).
SEA-CALIPSO experiment in December 2007, with airgun shots bubbling off the stern of the RRS James Cook, sending energy towards seismometer arrays on Montserrat and sea-bottom instruments. Image courtesy of S. Sparks.
Final seismic velocity model derived from SEA-CALIPSO experiments (after Paulatto et al. 2012). (a)–(c) West–east sections through the three major volcanic centres. The high-seismic-velocity cores of the volcanoes are marked with white dashed lines representing the 0.25 km s−1 seismic velocity anomaly contour with respect to the average seismic velocity of the island. (d) South–north section. (e) & (f) Horizontal sections at 2 and 7 km depth below sea level, respectively. The coastline and the 200 m-depth contour are marked with thick black lines. The white circles bound the area over which the reference model for seismic velocity anomalies was calculated. Lighter areas have no ray coverage.
Risk assessment and management
SAC
In 2003, the Risk Assessment Panel (RAP) method of assessing hazards and risks at SHV (Aspinall et al. 2002) was put on a more formal footing and with a wider remit by the UK Foreign and Commonwealth Office (FCO) as the SAC on Montserrat Volcanic Activity. The code of practice used was that devised for the provision of expert scientific advice throughout the UK government. Since then, the SAC has met at the MVO about every 6 months to make a quantitative risk assessment, and also electronically during times of crisis. Given the longevity of this eruption and the continuing close management needs, this model of an observatory team handling day-to-day matters (MVO) and an external team (SAC) running a bi-annual assessment of risk has worked well (Sparks & Aspinall 2004). The meetings themselves acted as a useful focus for MVO staff to report on activity, and to discuss data and their interpretation. Relative to both FCO and MVO scientific staff, SAC members effectively became one of the important conduits of long-term memory of the volcano's behaviour and management.
Risk assessments
The risk assessment process at SHV has evolved over 14 years and is reviewed by Wadge & Aspinall (2014). Following briefings on recent behaviour and any new insights into hazardous processes (e.g. PDC computer simulations), the members of the SAC and MVO staff are elicited as to the probability bounds of scenarios of future activity using the Classical Model of Cooke (1991). The window of future behaviour assessed is typically 1 year, reassessed every 6 months–1 year. The risks are expressed in two ways: in terms of the risk of death from volcanic activity to individuals; and the impact on Montserrat society as a whole. The performance of the risk assessment in terms of what actually happened is generally considerably better than a random guess at the outcome (Wadge & Aspinall 2014). The geographical proximity of people to hazards is usually calculated in terms of the zonation of the Hazard Level System (see below) and knowledge of the spatial distribution of the population via census. The zonation itself has been determined partly by the results of stochastic modelling of the PDC hazard in the Belham Valley (Fig. 1.36). Separately, the SAC is also asked to assess risks associated with specific work-related scenarios (e.g. sand mining). The SAC has sought to convey the risks in a variety of ways to improve wide understanding: as simple odds; on a logarithmic scale with descriptive classes (e.g. ‘very low’); as decimal increments above the average risk of accidental death; and graphically as risk ladders with comparator values for ‘everyday’ activities. A report is issued by the SAC after each meeting that becomes public on the MVO website (www.mvo.ms). An analysis of the risk assessment process from the perspective of management was produced by Wadge et al. (2008a) for the FCO.
Limits of the approximately 20×106 m3 PDC simulations on the northern side of the lower Belham Valley undertaken for the SAC15 assessment (November 2010: SAC 2011). The blue, red and green lines are the results for the Pyroflow, Titan2D and PFz code simulations. The yellow dashed line is the average position of the three results, and the pink area is the envelope of uncertainty defined by the three limits. The black lines are the Hazard Level boundary lines for zones A and B at that time.
Hazard level system
This is a common tool in volcanology that was used throughout much of the eruption since Phase 1 (although earlier termed the ‘Alert Level’). It consists of a map of the island and the immediate offshore region with a number of designated ‘Hazard Zones’. There is also a set of Hazard Levels (from 1 to 5), with a brief synopsis of volcanic activity associated with each level. At any given level there is a prescribed set of conditions and actions for each zone. The system has had several major overhauls, the last in 2008 (Fig. 1.37). Because it is difficult to foresee all future circumstances, there is a tendency to overcomplicate such systems (Aspinall et al. 2002). However, the current Hazard Level System has kept the definition of activity broad enough to maintain flexibility. The MVO makes the decision to change from one level to another, and this is ratified by the National Disaster Preparedness and Advisory Committee (NDPRAC).
Hazard Level System for SHV in November 2011. Map shows the zone boundaries; the lower tables show the hazard levels and actions on access to be taken. An up-to-date version of this is available from www.mvo.ms.
New scientific understanding, new problems
SHV never fails to throw up new problems and puzzles, but seems to yield answers much less readily. The following synopses highlight some of the advances that have been made and new questions posed in volcanology as a result of work carried out on Montserrat, many of which are reported in depth elsewhere in this Memoir.
Volcanic seismicity
By the time a seismic network was first established at SHV on 28 July 1995, a diffuse field of VT seismicity was revealed at depths down to approximately 7 km below sea level underneath the volcano, but also to the NE and separately under St George's Hill to the NW (Aspinall et al. 1998). By 1996, the VT seismicity had shrunk to a much smaller source volume below the lava dome, extending between about 1 and 4 km below sea level, and 1.5 km in diameter, and has remained so since. Aspinall et al. (1998) interpreted focal mechanism solutions from these events in 1995–1997 in terms of magma-induced radial stress impinging variably on a faulted, tectonic stress field. Temporal variations in the focal mechanism solutions of VT events (Roman et al. 2008; Miller et al. 2010) and regional seismic anisotropy measurements (Roman et al. 2011) at the eruption onset and throughout the eruption were interpreted in terms of the pressurization and depressurization of a NNE-orientated dyke conduit associated with the phases of lava extrusion leading to local, volcanic stress field reorientation.
Seismicity remains the mainstay of monitoring at the MVO because of its robust operational character and its sensitivity to short-term changes in the volcanic system. Since the start of the eruption, the general trend has been for fewer and weaker swarms of both VT and hybrid events (Luckett et al. 2008). Occasional major swarms of hybrid events, such as that leading to the collapse of 12–13 July 2003 (Ottemoller 2008), have shown distinct event characteristics that varied systematically, corresponding to sub-daily epochs of behaviour. Low-frequency (mainly hybrid) seismicity has been associated with magma transport within the shallow conduit, but the location (c. 400–600 m below sea level: Neuberg et al. 2006; c. 100–300 m bsl, De Angelis & Henton 2011) and character of the events has been determined precisely by careful modelling of their spectral and waveform character (Neuberg et al. 2000, 2006). Green & Neuberg (2006) showed how swarms of such events are comprised of several ‘families’, each of which has very similar waveforms indicating repeated excitation of the same mechanism at the same place in the conduit. Magma fracture and healing were advocated by Neuberg et al. (2006) as the source mechanism for such events at depths determined dynamically by pressure and rheology (De Angelis & Henton 2011), although the fossilized evidence of such fractures is not common in the dome rocks at SHV. The seismicity can be used to constrain two-dimensional (2D) conduit flow models involving a brittle failure mechanism within the magma (Collier & Neuberg 2006; Thomas & Neuberg 2012). Rockfall seismicity has a more obvious source mechanism. Despite the difficulties of non-uniform energy exchange at the surface, Jolly et al. (2002) were able to show that it is possible to determine the location and migration of events generated during PDC emplacement using the signal amplitudes from different stations in the network. This is complementary with the ability of infrasound arrays and thermal cameras to track flow advance (Ripepe et al. 2010). Passive seismic interferometry, or noise analysis, was used by Baptie (2010) to show that for major perturbations to the volcano – in this case the 12 July 2003 collapse – the seismic velocity character of the volcano changed measurably at the event, although it did not appear to have value as a warning measure.
The role of groundwater
We know very little about how groundwater interacts with the volcano. During the initial phreatic stage of the eruption (July–November 1995), large, sudden expulsions of water from cracks around the old Castle Peak dome were observed to produce lahars. Subsequently, there have also been a few reports of occasional ‘rainless lahars’ forming in the uppermost parts of some catchments. This suggests that pressurized groundwater in fractures can be forced to the surface. The precursory period of ash venting and explosions before Phase 3 (15 April–30 July 2005) had fine ash considered to be of phreatic origin (Cole et al. 2014a). It is unclear how much access the groundwater has to the conduit when there is no magma there. The above episodes imply that in some cases groundwater must be driven off from around the upper conduit before an uninterrupted magma flow re-establishes itself. There are some components of an active hydrothermal system associated with SHV, but seemingly little evidence of direct access to the magma conduit by this system (Boudon et al. 1998). The contribution of meteoric water to the volcanic plume has yet to be tested quantitatively.
Further from the volcano, the water supply for the island comes from springs in the Centre Hills. Groundwater flow to these springs increased considerably during October 2004 in conjunction with other MVO-measured signals (SAC 2005, fig. 1), perhaps in response to an adjustment to the volcanic system at depth. This suggests that stress relaxation in the rock mass underlying the Centre Hills may have occurred following a magmatically induced strain event and this, in turn, increased the permeability of the rock mass. Gravimeter measurements made in the western Centre Hills from 2006 to 2009 detected temporal changes in mass that may correspond to changes in the water table there (Hautmann et al. 2010, 2014).
Explosive variety and mechanisms
Explosions have been very helpful scientifically in that they have provided (via tephra fallout and ballistics) a ready way to sample the contents of the dome or magma conduit, and to constrain the geometry and dynamic conditions therein. Observations have also led to advances in understanding mechanisms of explosive eruptions through model development. For instance, numerical simulations by Clarke et al. (2002a, b) resolved highly unsteady vent exit conditions such as velocity, pressure and mass flux, and the temporal and spatial dispersal of pyroclasts during the first few minutes, using one gas phase and three solid phases representing clasts of different size. The models mimicked the observed explosions and revealed that volatile depletion in the conduit was an important factor in the dynamic behaviour of the explosions and related density currents.
SHV currently does not seem to be capable of producing Plinian intensity explosions. It does produce relatively short-lived (seconds to a few minutes) Vulcanian explosions. The majority of these explosions eject volumes of <106 m3, and the two largest are the 3.2×106 and 2.7×106 m3 DRE volume explosions of 17 September 1996 and 8 January 2010, respectively (Table 1.4). The Vulcanian explosion plumes are mainly limited to altitudes up to about 15 km and, as a result, stay within the troposphere. However, the plume derived from the 20 May 2006 explosion during dome collapse did reach the stratosphere (17–20 km altitude) (Prata et al. 2007). The remarkable series of Vulcanian explosions from September to October 1997 (Druitt et al. 2002b) were shown to have had repose intervals that followed a log-logistic distribution, and were interpreted as reflecting the competition between rheological stiffening and gas escape (Connor et al. 2003). In contrast, the series of five explosions that followed the 12 July 2003 collapse showed that the repose periods between explosions increased as the fourth power of their rank in the sequence (Edmonds & Herd 2007), more consistent with a decaying system not being recharged. A decrease in recharge flux rate could explain the delay in meeting the explosion criterion as well as the lack of significant explosions after 15 July 2003 (Voight et al. 2010b). Jaquet et al. (2006a) interpreted the time-series data of the 1997 explosions in terms of an effective memory of earlier events lasting about 60 h, commensurate with the ascent of magma through the last few kilometres to the surface.
Significant explosions
Burgisser et al. (2011) showed how the variety within these explosions could be modelled as variants on a tripartite conduit layering: a dense, stiff cap; a transition zone of heterogeneous vesicularity; and a lower, homogeneous, low-porosity zone. The volume density of plagioclase microlites is inferred to show a sharp transition at about 700 m depth, consistent with the base of the stiff cap or plug (Clarke et al. 2007). Giachetti et al. (2010) showed that detailed textural analysis of vesicles could be used to recognize populations of bubbles formed by explosions, and to understand nucleation mechanisms and decompression rates. The formation of the dense cap is predicted by the model of Michaut et al. (2009), in which the magmatic permeability displays hysteresis with respect to gas volume fraction. This leads to a dense, low-permeability cap of collapsed foam above a highly vesicular magma. But what prevents such explosions continuing for much longer and evacuating the crustal reservoir at SHV? Mason et al. (2006) argued that the permeability of SHV magma is sufficiently high that, during the foam-generating stage of the explosive drawdown in the conduit, so much gas is released that the overpressure is lost and the explosion ends, to be replaced by ash venting, gas release and seismic tremor. Another possibility is that fragmentation ceases at the interface between the dyke and the cylindrical conduit after about 1×106 m3 of magma has been removed from the conduit above this point.
Phases 4 and 5 were more explosive than Phases 2 and 3, particularly at the start and end of lava extrusion. This is probably due to higher levels of shallow pressurization being achieved in the upper conduit and dome. Komorowski et al. (2010) invoked a reduction in gas permeability due to silica sealing to help explain this during Phase 4. However, the initial conditions for these large, isolated explosions (as opposed to the cyclic series of smaller explosions in 1997) are clearly quite variable and currently unpredictable. Also, the proportion of conduit-derived pumice involved is highly variable, suggesting that some events comprise largely dome rocks (e.g. 8 January 2010) (Cole et al. 2014b). Strainmeter data, coupled with seismic, gravity and infrasound records, show quite distinctly different signatures for the 29 July 2008 and 3 December 2008 explosions (Chardot et al. 2010; Gottsmann et al. 2011). In general, the strainmeter data show that magma reservoir and dyke pressurization preceded some explosions, such as those of 4 March 2004 and 3 December 2008. However, in many others, the pressurization transients seem to be confined to the top 1.5–2 km of a putative cylindrical conduit, as on 29 July 2008 (e.g. Chardot et al. 2010; Linde et al. 2010).
Volatile budget
SHV is a volatile-rich volcano in which mafic magma releases sulphur dioxide and carbon dioxide as it mingles with andesite magma at mid-crustal pressures of 200–300 MPa (Edmonds et al. 2014). Decompressing andesite magma rising to the surface releases hydrogen chloride and water at shallower depths (<2 km). The eruption-long measurements of sulphur dioxide emission rate are roughly constant (Fig. 1.31), indicating the near steady state of this process, which is equivalent to the complete degassing of a mafic magma containing 2000–3000 ppm sulphur (Christopher et al. 2010) and supplied continuously from depth at a rate of approximately 1 m3 s−1, which is just under half the eruption-long extrusion rate of andesite magma (2.3 m3 s−1). Using non-dispersive infrared spectrometry and a MultiGas sensor, Edmonds et al. (2011) were able to measure CO2, SO2 and H2S concentrations directly for the first time in July 2008 during the Phase 3–4 pause. These concentrations are consistent with the mafic magma having a CO2 content of 2000–5000 ppm and a water content of up to 8%, and the rhyolitic melt component of the andesite having a CO2 content of 2000 ppm and a water content of 4–6%. Whilst the SO2 and CO2 fluxes are good measures of the deep magma mingling process, the HCl/SO2 ratio is a potential proxy for the andesite eruption rate (Christopher et al. 2010). Melt inclusion studies of H2O and Cl, and zoning of Cl in hornblende (Humphreys et al. 2009b), are consistent with the accumulation of a CO2-rich vapour in the andesite after transfer from basaltic magma. This fluid is then decoupled from the magma and released as CO2 at the surface.
The pre-eruptive andesite magma is shown to coexist with an exsolved vapour that comprises 3–9.5 wt% of the bulk magma (Edmonds et al. 2014). The presence of ubiquitous small bubbles of exsolved vapour in the rising magma means that there is little or no need to nucleate new bubbles during conduit flow, which many of the existing conduit flow models have done.
The measured temporal variability of gas flux is a function of the permeability of the system as a whole, the dynamics of explosions and dome-collapse events, and any unsteady component to the deep supply of mafic magma. Christopher et al. (2011) noted that there is a second-order variability in the sulphur dioxide emission, with a periodicity of a few years that clearly does not correspond to the extrusive phases and may be related to the dynamics of deep magma supply. Higher in the system, major collapse of lava domes can produce large SO2 loading events in the atmosphere, the gas for which must be largely stored within the pores and cracks of the dome rocks themselves (Carn & Prata 2010). The rate of gas transport through the whole system is a key issue. Studies of diffusive timescales in glasses (Humphreys et al. 2010) reinforce earlier estimates of the time to transport post-mingling magma to the surface as about 1 month, with gas presumably faster.
In addition to the daily spectroscopic measurement of sulphur dioxide flux in the gas plume (Edmonds et al. 2003a), a series of ground-level diffusion tube measurements of sulphur dioxide accumulation over a period of about 1 month have been made throughout the eruption (Murrell et al. 2014). These data show a general pattern of low sulphur dioxide values at ground level during extrusive phases, and higher values during pauses in extrusion. This suggests that the plume behaves differently in the boundary layer depending on the presence or not of a tall, hot lava dome. When the dome is high and hot, the plume rises and is advected offshore before it sinks to sea level. The recorded levels of gas pose little problem in the currently inhabited areas.
Dome instability
Peléean lava domes are inherently unstable because they grow by central extrusion and intrusion of lava close to the point of gravity-induced failure on steep slopes, producing innumerable rockfalls that form angular talus deposits downslope. This tends to an equilibrium process that maintains a central core of lava of near-cylindrical shape surrounded by a talus with slopes at the angle of rest (33°–42°). From a hazard perspective, it is the size of the hot and volatile-rich core and its potential instability that matters, not the overall (core+talus) volume of the dome (Wadge et al. 2009). Above a volume of about 103 m3, individual rockfalls develop increased mobility due to gas release and fragmentation, and PDCs form (Calder et al. 2002). The source masses of PDCs and their behaviour immediately preceding collapse are poorly observed, but toppling and sliding of joint-bounded dome rocks are presumed to be dominant first motions. Larger collapse events, involving masses with volumes in the range 105–108 m3, probably involve other motive forces, including intrusive push, rainfall/steam explosions and cyclic gas pressurization. The frequently observed sequence of cessation of surface extrusion, hybrid earthquake swarm and a new lava lobe appearing in a different sector implies a new intrusive pathway being taken in the upper dome (Watts et al. 2002). The intrusive forces may lead to the near-surface rocks being pushed beyond the point of collapse. This, in turn, may lead to the rapid exposure of the newly intruding lava and a much more vigorous PDC developing, perhaps evolving to a lateral explosion. Loughlin et al. (2010) described the start of the 20 May 2006 collapse in these terms. Hale et al. (2009) raised the possibility of intrusion into the talus in Phase 3, although observational evidence for this mechanism and, indeed, the mechanics of dome collapse generally are currently lacking.
The correlation of heavy, sustained rainfall and major collapse of the dome was first proposed for the 29 July 2001 event by Matthews et al. (2002). A crack filling with rainwater and flashing to steam was tested as a potential model for this and required a minimum rain rate of 15 mm h−1 for 2–3 h (Matthews & Barclay 2004), whilst Elsworth et al. (2004) and Taron et al. (2007) invoked a role for internal magmatic gas pressurization in fissures capped by a rain-saturated carapace (cf. Voight & Elsworth 2000). Herd et al. (2005) considered that rain-induced erosion of the talus was an important trigger of dome collapse. A wider study of the lags between 229 rainfall events and seismic activity from 2001 to 2003 (Matthews et al. 2009) suggested almost immediate surface response to produce rockfall activity, but a lag of 12–24 h for deeper seismicity and larger failures. The major (>107 m3) collapses of 3 July 1998, 20 March 2000 and 29 July 2001 do appear to have had intense rainfall triggers, and the 12 July 2003 and 20 May 2006 events had rainfall events preceding them but at lesser rain rates. The only large collapse since 2006, on 11 February 2010, was not preceded by heavy rain.
Submarine deposition
Between 1995 and 2009 at least 65% of the output of SHV had been deposited in the sea (Le Friant et al. 2010). Multiple research cruises offshore Montserrat in 1998, 2002, 2005, 2009 and 2010 allowed the quantitative effect of submarine deposition, and particularly the deposits of the major dome collapses, to be measured using seismic survey, swath bathymetry and coring. For example, the largest collapse, of 12 July 2003, produced two 10–60 m-thick lobes up to 6 km off the eastern shore. This comprised 60% of the submarine mass and the remainder constituted a 40 km-long turbidite deposit (Trofimovs et al. 2008). The higher energy of the very rapid 20 May 2006 collapse event resulted in longer submarine runout than the large volume July 2003 event (Trofimovs et al. 2011). The cumulative effect of the collapse events and deposition can be identified stratigraphically (Trofimovs et al. 2006, 2011) to the east and SE of the island.
Comparing these new deposits with those from previous eruptions helps put the current eruption in perspective. It seems clear that there have been complex multiple-event eruptions like the current one, but they are relatively infrequent – the last one being around 20 ka from the land evidence (Smith et al. 2007), or around 12–14 ka from offshore evidence (Trofimovs et al. 2013). There are several debris avalanche deposits identifiable in echo-sounder profiles, and in bathymetry as ‘hummocky terrain’ in the sea around Soufrière Hills (Le Friant et al. 2004). Some are from identifiable flank collapse scars, others not. The volumes of the largest such older deposits (20–30 km3) are comparable with those from the larger islands to the south. The preponderance of these deposits to the south and east of the island indicates control of deposition and, perhaps, initiation by the Montserrat–Bouillante Fault System (Lebas et al. 2011). Three-dimensional (3D) seismic data illuminates subtle deposit fabrics that allow some of the emplacement dynamic features of these deposits to be identified (Crutchley et al. 2013). English's Crater is probably the source for one of these deposits to the east and may represent an event about 2 ka ago (Boudon et al. 2007). The balance between subaerially derived mass and reworked submarine sediment has been shown to be important in understanding the dynamics of individual debris avalanche events, including their ability to produce tsunamis (Watt et al. 2012). Dating of tephra in the top 6 m of the CAR-MON2 core 55 km SW of Montserrat reveals ages up to about 246 ka (Le Friant et al. 2008). This is equivalent to most of the history of SHV, and is represented in the core by about 15 eruptive periods. Basaltic tephra, presumed to represent the activity of South Soufriere Hills Volcano is restricted to the period 147–127 ka by land and offshore 40Ar/39Ar dates (Le Friant et al. 2008; Cassidy et al. 2014). Pumice-rich horizons possibly associated with Plinian eruptions are found in some deposits up to 77 ka, but not since.
PDC hazards
A taxonomy of PDCs at SHV now includes dome-collapse, fountain-collapse, surge-derived, lateral blast and hydrovolcanic types (e.g. Calder et al. 2002; Druitt et al. 2002a, b; Sparks et al. 2002). Only the hydrovolcanic flows, described by Edmonds & Herd (2005) and Edmonds et al. (2006) for the 12 July 2003 eruption, had not been recognized during Phase 1. Such surges generally produced fine-grained deposits and were the result of a landwards-directed portion of a steam explosion entraining fine particles (c. 0.1 mm) and originating from high-flux PDCs (with heights of c. 200 m) entering the sea (Dufek et al. 2007). A similar flow occurred during the 20 May 2006 collapse event. Detailed 3D numerical modelling of the 1997 Boxing Day lateral blast by Esposti Ongaro et al. (2008) and constrained by the earlier detailed fieldwork defines three phases: a burst phase lasting about 5 s; an asymmetrical gravitational collapse phase; and a topography-influenced PDC phase. The extreme damage inflicted on St Patrick's village was consistent with a subsonic flow but also high dynamic pressure in the PDCs and the impacts of large missiles. The event on 11 February 2010 that ended Phase 5 was of a similar volume (40×106–50×106 m3) to that of the 1997 Boxing Day event (Stinton et al. 2014b), and also produced a highly destructive PDC in the Harris and Streatham villages, although observations of severe damage in Harris village imply a topographical influence on the flow path. Similarly powerful PDCs, largely directed seawards, evidently occurred during previous large collapses of the dome to the east, as on 12 July 2003 and 20 May 2006, when the hydrovolcanic explosions occurred.
Following heightened concern on two occasions, the SAC assessed the likelihood of large PDCs descending the Belham Valley in 2007 and 2010. The problem has two linked components: the mass (volume) from the dome capable of being emplaced in the upper reaches of the valley; and the subsequent flow mobility and deposition. Dome growth tends to be directed to one particular sector (e.g. NW to enter the Belham Valley) for weeks, typically via shear lobes, and this provides the source of most PDCs. For collapses to the north, the English's Crater rim, at about 800 m asl, appears to form an effective base level to the collapse volume (Wadge 2009), as in the case of the 11 February 2010 event. Three numerical codes have been used to simulate dome-collapse pyroclastic flows: Titan2D (Widiwijayanti et al. 2004, 2006; MVO 2010; SAC 2011; C. Widiwijayanti, D. Hidayat & B. Voight unpublished modelling); Pyroflow (Wadge et al. 1998; Wadge 2009); and PFz (Widiwijayanti et al. 2008). The conclusion from these simulations was that a flow volume of about 20×106 m3 represents an upper limit to the volume of dome mass that can enter the Belham River Valley in one collapse event, and that flows with volumes of >3×106 m3 could reach close to the populated zones of the valley with a >7 km runout (SAC 2011). The success of this modelling effort was highlighted by the fidelity to which the 2007 simulation replicated the inundated area of the 11 February 2010 event (Fig. 1.38). The lateral blast hazard was assessed using the PDAC code (Esposti Ongaro et al. 2008).
Comparison of the extent of the PDC deposits of 11 February 2010 (top) and a computer simulation (Titan2D model with 20×106 m3 collapsing to the north) of a PDC and deposits produced by B. Voight, C. Widiwijayanti & D. Hidayat for the crisis response considered by SAC8 in March 2007 (bottom). In the top map, the image is from the ASTER satellite taken after the 11 February 2010 event and the pale blue lines show the outline of the deposits from that event. The black dashed lines (1–3) are the ‘watersheds’ (topographical divides) that affected the directions (arrowed lines) of pyroclastic currents on the Farrell's plain, Gages Mountain and Gages Valley, respectively. They were determined from post-11 February 2010 topography measured by radar interferometry using TerraSAR-X data (Wadge et al. 2011).
Cyclicity
One of the main discoveries of the eruption is evidence for cyclic behaviour of the volcano on several timescales (Odbert et al. 2014b). On the longest timescales, geochronological and stratigraphic data (Harford et al. 2002; Le Friant et al. 2008) indicate that the volcano is highly active over relatively short periods of, perhaps, no more than a few thousand years interspersed with periods of tens of thousands of years of low or no activity. While the data are not good enough to show whether these marked fluctuations in activity are really cyclic, they are consistent with: the petrological data that suggest short residence time of the andesitic magma in the upper crust (Zellmer et al. 2003b); thermal models (of Mont Pelée, Martinique) which indicate that magma fluxes into the upper crust have to be similar to the time-averaged flux out of the volcano in order to form magma chambers (Annen et al. 2008); and radioactive equilibrium of U and Th, implying long residence times of andesite magma in the deeper crust (Zellmer et al. 2003b). There is also the cyclicity of geophysical unrest at approximately 30 year intervals in 1896–1897, 1933–1937 and 1966–1967 that preceded the onset of the current eruption in 1995. Neither of these longer-term fluctuations are well understood, although processes that govern magma generation and magma ascent through the mantle and crust are probably involved.
Multi-year
The eruption itself has revealed three different timescales of cyclicity. The longest is represented in the five phases of dome growth that have alternated with protracted periods of no dome growth accompanied by inflation, continued gas emissions, and, towards their end, by increased seismicity and explosive eruptions. This activity, precursory to the resumption of lava extrusion (Table 1.3), must represent the relatively slow rise of magma up the conduit, usually over periods of approximately 3 months (phases 1, 3 and 4) (Cole et al. 2014a). The volume of magma involved must be quite modest because there is negligible GPS-measured deformation associated with it (Fig. 1.31). In contrast, Phase 5 had a much shorter period of precursory activity (3 days). This could mean that the magma rose much faster than usual, perhaps with the conduit containing a core of still-mobile residual magma. Alternatively, it could be that Phase 4 and Phase 5 should be regarded as comprising a single phase with two major non-extrusive interludes during which magma remained at high levels, and did not have to re-establish a flow path between Phase 4 and Phase 5, as had been the case following Phases 1, 2 and 3.
Multi-week
An intermediate timescale involves multi-week (6–7 week) cycles manifested in systematic variations of ground deformation, seismicity and eruptive activity. The recognition of these cycles in 1997 was noted as one of the scientific highlights of Phase 1 (Sparks & Young 2002). In 1997, the onset of cycles was marked by abrupt changes in tilt patterns, intense seismic swarms and major volcanic events, such as large dome collapses and explosive activity (Voight et al. 1998, 1999). The rates of deformation, seismicity and eruptive activity declined over the period of a cycle. The regularity of the cycles enabled the MVO to provide some forewarning of large events. Although these cycles were most prominent in 1997, they were also recognized subsequently, such as in 1999–2000 (Young et al. 2003), in 2008 (Stewart et al. 2009) and in Phase 5 (Odbert et al. 2014b), in which, for the first time, a cGPS-measured deflationary signal associated with them was also detected. Loughlin et al. (2010) also described a series of cycles from Phase 3, although several of theirs tended to be shorter and less voluminous than those from Phase 1 and Phase 5.
Sub-daily
Cyclicity is recognized over sub-daily periods mainly within the range 3–30 h, with a median of about 11 h, and has been a persistent feature during periods of dome growth (Odbert et al. 2014b). When the tiltmeters located at Chances Peak were in operation during 1997, the cycles consisted of an oscillating pattern of inflation and deflation, accompanied by strongly correlated seismicity in which hybrid earthquakes characterized the inflations, while rockfall signals commonly occurred during deflationary periods (Voight et al. 1999). During periods of Vulcanian activity, the explosions occurred at the peak of the deformation cycle, and gas emissions were also tuned to the cycles (Voight et al. 1998, fig. 5; Watson et al. 2000). The period of the sub-daily cycles sometimes increased over the course of a multi-week cycle, with an abrupt reversal at the end. Green & Neuberg (2006) showed that low-frequency earthquake swarms began at the inflection in the approximately sinusoidal inflation pattern of the cycles, and that the swarms consisted of a small number of repetitive and near-identical earthquake families located at a depth of 1–2 km. Other, more subtle, cycles may exist. Odbert & Wadge (2009), using wavelet analysis, found evidence for an approximately 3 day frequency modulation in the tiltmeter record from 1997, a cycle that was also recognized in seismic data (Jaquet et al. 2006a, b) and speculated as being due to some form of ‘memory’ of earlier cycles within the magma. The sub-daily cycles have been apparent for much of the eruption, but the tilt data from 1997 has meant that most of the research on their causation has focused on analysis of the 1997 data.
The quest to explain the cyclic patterns has led to major advances in concepts on volcanic processes, and the understanding of flows of magmas through magma chamber and conduit systems. The origins of the cyclicity have been related to coupling of rheological changes in ascending and degassing magma, phase separation between gas and magma, and pressure variations in deformable magma chambers and conduits with variable geometry. The models (e.g. Denlinger & Hoblitt 1999; Melnik & Sparks 1999, 2002, 2005; Barmin et al. 2002; Costa et al. 2007, 2012) are characteristically highly non-linear, with kinetic and dynamic controls related to crystallization and degassing. Rheological stiffening of the magma in the uppermost parts of the conduit plays a key role in governing shorter cycles, shallow ground deformation and earthquake location (Sparks 1997; Melnik & Sparks 1999; Voight et al. 1999). Various models of pressurization/depressurization within the upper conduit have been invoked to explain the sub-daily cyclicity and the 1997 tiltmeter record used to constrain them: volatile-dependent viscosity (Wylie et al. 1999); shear traction (Green et al. 2006); extra-conduit pressurized fluids (Widiwijayanti et al. 2005); gas pressure build-up due to gas diffusion from supersaturated magma in a confined volume (Lensky et al. 2008); and polymer stick–slip (Costa et al. 2012). Evidence from the Vulcanian explosions (Clarke et al. 2007) suggests that rapid formation of a degassed, dense and strong plug overlying gas-rich bubbly magma in which pressure builds up is a key aspect of these cycles. Michaut et al. (2009) showed how the compaction with gas escape in ascending bubbly magma can lead to the formation of dense degassed plugs.
The multi-week cycle has been modelled in terms of the deformation of a dyke (Costa et al. 2007; Hautmann et al. 2009) containing a rheologically stiffening and degassing magma. The cycle involves the pressure rise and relaxation of the elastic-walled dyke during charge of magma from below and its discharge through the conduit above. Costa et al. (2013) coupled the elastic dyke model with a stick–slip model of conduit flow (Costa et al. 2012) to simulate the combined interaction of the sub-daily and multi-week cyclic behaviour.
So far, an explanation of the longer-term cycles of dome growth alternating with quiescence has been elusive. Heuristic linear models in which deformation and lava extrusion volumes are coupled have been developed (Elsworth et al. 2008; Foroozan et al. 2011). Fluxes between a deep chamber, a shallow chamber and the surface are constrained by deformation models, and extrusion volume and volume flux data. Such models describe the first-order phenomena well, but provide limited insight into the underlying dynamics and mechanics. Although the observed behaviour qualitatively could be explained by the models of cyclicity from pressurized magma chambers developed by Melnik & Sparks (2005) and Melnik & Costa (2014), patterns of extrusion rate v. deflation do not fully agree with the models. It appears that some important factors are still missing from these dynamical models.
The elusive dyke
The phreatic stage of the eruption in 1995 involved numerous fractures crossing the Castle Peak dome, and extrusive and degassing fractures also played a strong role during phases 3 and 4. Yet, there are no dykes exposed in the walls of English's Crater, and the longevity of the cylindrical (evident from spine shapes) conduit that fed successive lava domes has led to a view of the shallow feeder from the reservoir as an axisymmetric pipe. However, Costa et al. (2007) made a strong case for an elastic-walled dyke capable of acting as a high-level temporary reservoir of magma to be the mechanism driving the multi-week cyclicity. This model dyke had approximate dimensions of 4 km high, 300–500 m long and 3–6 m wide, with the shear modulus of the host rocks increasing downwards so that it could expand much more at shallow depths, and was connected to a 30 m-diameter cylinder above a depth of about 1 km. Support for such a dyke comes from a variety of sources. Mattioli et al. (1998) had argued for a shallow NW-trending dyke source above a Mogi source, based on early GPS data. Hautmann et al. (2009) argued that the tangential components of the 1997 tiltmeter record were non-zero and thus the source was non-axisymmetric. Fault-plane solution analysis of VT earthquakes (Roman et al. 2011) suggested a NE-trending dyke, although the predicted 90° rotation in stress axes in moving from dyke-filled to dyke-empty appears too simplistic. However, Miller et al. (2010) examined 1995–1996 VT data, and interpreted them to indicate a dyke of NNE trend, which altered the stress distribution to promote localized fault slip, and caused localized dilation with changes in pore-fluid pressures, to either strengthen or weaken the rock mass depending on the polarity of the local strain.
Strainmeter transients for the 3 March 2004 explosion during the Phase 3–4 pause (Linde et al. 2010) are compatible with a WNW-trending dyke, those for the 3 December 2008 explosion suggested an ENE-trending dyke (Chardot et al. 2010), whilst Hautmann et al. (2009) argued for one trending NNW in 1997. These results are enigmatic. The WNW-trending Belham Valley Fault appears to cross the island, and elements of this system may underlie SHV and form part of the Montserrat–Bouillante Fault System. There is a left-lateral, transtensional sense of motion that offsets the Centre Hills and Soufrière Hills upper-crustal seismic anomalies (Shalev et al. 2010) that seems consistent with regional north–south extension (Feuillet et al. 2010). Miller et al. (2010) argued for involvement of this structure in the opening phases of the eruption when VT earthquakes were prevalent, as did Feuillet et al. (2011), who invoked a NNE-dipping normal fault with left-lateral motion on a plane about 5 km below the volcano. However, the details of local (c. 1 km scale) dyke- and fault-control in the vicinity of SHV are, as yet, unclear.
Better constraints on magma reservoirs and conduits
Just as the simple view of the upper magma conduit at SHV evolved to incorporate a dyke beneath the cylinder feeding magma to the surface, so the concept of a simple magma reservoir at >5 km depth has changed. SEA-CALIPSO receiver function analysis and seismic tomography has revealed the crustal thickness beneath Montserrat (30 km: Sevilla et al. 2010), and the upper-crustal structure in general terms (Paulatto et al. 2009, 2010). Integrated seismic tomography and thermal modelling by Paulatto et al. (2012) suggested the existence of a magma reservoir with a volume of approximately 13 km3, and over 30% melt fraction between depths of 5.5 and 7.5 km beneath SHV. cGPS-constrained model results from Phases 1–3 deformation data are consistent with a pressure source(s) in the range 5–19 km. The restricted geographical sampling of the deformation field on Montserrat does not permit tight constraints on the source depth to be applied (Elsworth et al. 2014). As a result, it has been possible to argue for an idealized single Mogi point source (Mattioli et al. 2010), a prolate ellipsoid source (an idealization to represent a more complex, vertically elongated, source or series of sources) (Voight et al. 2010a; Hautmann et al. 2010) or twin Mogi sources (at 6 and 17 km) (Elsworth et al. 2008).
Two clear conclusions can be made from joint consideration of the deformation, the magma budget and the sulphur dioxide degassing. First, the deformation signal is very much smaller than we would expect from the purely elastic behaviour of magma reservoir wall rocks. This leads to the conclusion that the magma in the reservoir is highly compressed and decompresses during extrusive phases, with only limited accompanying elastic response (Voight et al. 2010b). Secondly, the near-continuous gas flux implies a near-constant background mafic magma supply throughout the eruption. This has led to linear magma supply models of storage and release (Elsworth et al. 2008, 2014; Foroozan et al. 2011) in which the deep supply is at a rate of about 1 m3 s−1 and the andesite reservoir loses about 0.5 km3 of magma over the course of the eruption to date. Thus, of the total erupted volume of approximately 1 km3, half derives from depth and half from the reheated andesite within the crustal reservoir. If there is a dual reservoir, then most of the mass is derived from the deeper reservoir. However, dynamic models of magma flow in dual reservoir systems (Melnik & Costa 2014) point to the possibility of the flux also being controlled by the shallow reservoir at different times during the extrusive phase, depending on the ease of connectivity between the two reservoirs.
Figure 1.39 summarizes schematically the main plumbing elements of the upper part of the magma supply system inferred beneath the SHV from the observations just described above and elsewhere. Below the dome, the uppermost conduit is thought to be of cylindrical form, based on the occurrence of approximately 30 m-diameter spines of solidified lava. This conduit extends from about 650 m above sea level to about 500 m below sea level (bsl), at which limit the low-frequency earthquakes are found. Chardot et al. (2010) and Voight et al. (2010b) suggested an approximately 2 km-long cylindrical conduit, based on estimated explosion volumes and a lack of dyke-like signature to some measured strain events. A 30 m-diameter conduit 1.5 km long has a volume of around 1×106 m3, which is similar to the volume of the average pumice-rich Vulcanian explosion deposits recorded in Table 1.4. This suggests a physical limit to magma drawdown during explosions. The shape of the junction between the cylinder and underlying dyke is not known. Its location may be due to a change in the host rocks at this depth or the limit of explosive reaming. The change in geometry must act as a bottleneck to magma flow, and this could partly control the sub-daily cycle of activity (e.g. Costa et al. 2012). From roughly 500–5500 m bsl, the dyke is capable of storing a magma volume of the order of 10 Mm3 by elastic displacement of the wall rocks, preferentially at shallower depths. This storage and discharge of dyke magma is thought to be involved in the multi-week cycle of activity. The record of contraction on three dilatometers over 4 weeks at the end of Phase 4 was analysed by Hautmann et al. (2013) to be consistent with decompression of two linked magma reservoirs at 5.5 and 11.3 km bsl, and dilation of the shallow dyke conduit by about 0.8 m.
The upper part of this dyke, from about 500 to 4000 m bsl, is also associated with VT earthquakes, although the hypocentres of the earthquakes do not give any clear evidence of the orientation of the dyke. The upper part of the dyke corresponds to a depth interval of positive seismic velocity anomalies beneath SHV, presumed to be due to shallow intrusions (e.g. crystallized dykes, sills and filled conduits). Seismic tomography picks up the first evidence of a magma reservoir at about 5500 m bsl. Between 4000 and 5500 m bsl is a low-seismicity region that could correspond to rocks with elevated temperatures and increased ductility above the magma body. This interval is where the seismic velocity gradient seen in rocks above falls sharply. The vertical extent of the shallow magma reservoir is not known and the connection to a deeper magma reservoir is unconstrained but is shown in Figure 1.39 as a dyke. Modelling suggests that multi-annual magma flux dynamics can be modulated by the ease of connectivity between the two reservoirs, a feature that does not appear to hold true for the passage of sulphur dioxide from the zone of andesite–basalt mingling in the deeper reservoir.
Schematic cross-section through the upper part of the SHV conduit/reservoir system (depth relative to sea level in km). Orange, lava dome; red, cylindrical conduit; pink, dyke; purple, andesite magma; blue, basaltic magma. The approximate relative volumes of the upper and lower reservoirs are shown. Crosses, volcano-tectonic (VT) earthquakes; asterisks, low-frequency earthquakes. The sense of geothermal groundwater flow is shown on one side only by the green arrowed line. The two inset schematic plots show how the sub-daily cyclicity is controlled by processes in the cylindrical conduit, and the multi-week cyclicity by storage and release of magma in the upper dyke.
Acknowledgments
There must now be hundreds of scientists and students who have worked on SHV since the eruption started, too many to list, although their hard work and ideas are clear from the large number of published papers sampled here. This work could not have happened in the way it has without the MVO and the support given to it from the British Geological Survey, the Seismic Research Centre of the University of the West Indies, the Institut de Physique du Globe de Paris, the US Geological Survey, and the Government of Montserrat. In addition, many key contributions have been made to the MVO and the wider science effort by many international research groups. Much of the work would not have happened at all without the financial support of research funding bodies. In particular, we are grateful to NERC in the UK, NSF in the USA, and also, on the government side, DFID and the FCO for long-term financial support. We thank T. Druitt for reviewing an earlier draft. Published with the permission of the Executive Director of the British Geological Survey (Natural Environment Research Council).
- © The Geological Society of London 2014