GEOPHYSICAL RESEARCH LETTERS, VOL. 25, NO.18, PAGES 3425-3428, SEPTEMBER 15, 1998
Pyroclasticflows generatedby gravitational instability of the
1996-97 lava dome of Soufriere Hills Volcano, Montserrat
P.D. Cole1,E.S.Calder
2,T.H. Druitt3,R. Hoblitt
4, R. Robertson
5,R.S.J.Sparks
2, S.R.
Young
6
Abstract Numerouspyroclasticflows were producedduring
1996-97 by collapseof the growing andesiticlava dome at
Soufriere Hills Volcano, Montserrat. Measured deposit volumes
from
these
flows
range
from
0.2to9x 106m3.Flows
range
from
the derivationof the ashcloudsurgecomponent,
thebehaviourof
theflowsastheyenteredthe seaandtheextensiveerosioncaused
by thepassage
of theflows.Pyroclastic
flowsformedby fountain
collapseduringVulcanianexplosions
arenotconsidered
here.
discrete,
singlepulseeventsto sustained
largescaledomecollapse
events. Flows entered the sea on the eastern and southern coasts,
Overview of PyroclasticFlow Activity
depositinglarge fans of materialat the coast.Small runout
distance(<1 km) flows had averageflow front velocitiesin the
orderof 3-10 m/s while flow frontsof the larger runoutdistance
flows(up to 6.5 km) advanced
in theorderof 15-30m/s.Many
flows were locally highly erosive.Field relationsshow that
development
of the fine grainedashcloudsurgecomponentwas
enhanced
duringthe largersustained
events.Periodsof elevated
pyroclastic
flow productivity
andsustained
domecollapseevents
arelinkedto pulsesof highmagmaextrusionrates.
Pyroclastic
flowsat Montserrat
aredistinguished
as(i) discrete
events;where the material is shedas one pulse, which can be
categorised
in termsof runoutdistance
or (ii) assustained
events;
wherea significantportionof the domecollapsesincrementally
over a period of 20 mins - 4 hours. During 1996-97 discrete
events producedsmall runout distanceflows (<1 km) up to
severaltimesa day andlargerevents(up to 3km) on averagea
few times a month.Sustainedcollapsesoccurredlessfrequently,
duringtheseeventspyroclastic
flowsweregenerated
continuously
or semi-continuouslybuilding up to a distinct climax, during
which the largestflows were producedand thereafterwaning.
Introduction
Material collapsedincrementally,producingscoopshapedscars
The currenteruptionat SoufriereHills Volcano,Montserrat, into the dome. Pyroclasticflows producedduringthe sustained
beganon 18 July1995andby midNovember1995an andesitic collapseeventsgenerallyhadlongerrunoutdistances(3 - 6.5 km)
lava domehad startedto grow within English'sCrater (Fig 1). althoughsomeindividualpulsesmayhavebeensmall.
Rockfalls producedby gravitationallyunstableor actively
BetweenMarch and June 1996 all the pyroclasticflows in the
growingareasof thedomehavebeencharacteristic
of theactivity Tar River valley were moderate(< 3 km), discreteevents.On 12
sinceextrusion
firstbegan.The firstpyroclastic
flowsoccurred
in May 1996theflowsreachedtheseafor thefirsttime,2.7 km from
lateMarch
1996when
thesizeofthedome
(6.35x 106m3)was dome,buildinga smallfan at thecoast.BetweenlateJulyandmid
suchthat collapsingmaterialbeganto spill out of the open, September1996 a seriesof four sustaineddomecollapseevents
easternsideof the craterinto the upperreachesof the Tar River occurred,which producedpyroclasticflows, many of which
valley.Thesepyroclastic
flowscomprisea basalavalanche
of
travelled over 3km but which were still confined to the Tar River
blocks and ash and an overriding dilute ash cloud surge
valley.Duringthisperiodthe smallpyroclasticfan createdby the
May 1996 flows(Fig 1) wasenlargedso thatby earlySeptember
component
derivedby elutriation
fromtheunderlying
unit.In all
but the smallestflows, theseashcloudsurgeshave a significant
it extended 400 m from the shoreline and had a width of-
1 km
lateralcomponent
of motion.Upper,buoyantportions
of theash alongthecoast.A largesustained
collapseon 17 September
1996
cloudsurgedominated
by thermallydrivenconvection
produce with an eight hour period of semi-continuous
pyroclasticflow
loftingashplumesalongtheflowpathwhichrisea few hundred activitywasfollowedby an explosiveverticaleruption(Robertson
metres to several kilometres. Such flows produced by non-
et al. 1998). The pyroclasticflows producedwere the largestthat
explosive
collapse
of domesarecommonly
referred
to as'Merapi had occurred,spillingout of the lower part of the valley on both
type'flows.Thispaperdescribes
features
of thepyroclastic
flows its northernandsouthernmargins.Significantcollapsesdownthe
generated
by gravitational
domecollapseup to 25 December Tar River valley occurred again in mid December 1996 and
1997. Direct measurementsof velocities, temperatures, and
volumesweremade.In particularwe document
observations
on
January 1997.
Papernumber98GL01510.
During late March and early April, 1997 sustainedcollapses
occurredover the degradedGalway'sWall producingpyroclastic
flowsup to 4 km downtheWhite River valley.Particularlyrapid
growthlocalisedon the northernsectorof the domein May and
Juneproduced
flowswhichovertopped
thecraterwall to thenorth
first travellingdownTuittsandthenMosquitoGhaut.Flowsalso
travelled west down Fort Ghaut and into Plymouth during June
1997.A largesustained
collapseeventon 25 June1997 travelled
6.5 km down MosquitoGhautreachingwithin 50 m of the sea
near the airport. During this event a portionof the upper,ash
cloudsurgedetached
fromtheparentflow at a prominentbendin
theflow pathandtravelledwestwards
downtheBelhamvalleyto
0094-8534/98/98GL-01510505.00
Cork Hill (Fig 1).
i. Dept.Geology,University
of Luton,ParkSquare,
Luton,U.K.
2.Dept.Geology,University
of Bristol,Bristol,BS8 1RJ,U.K.
3.UMR6524CNRS,DepdesSciences
Terre,Clermont-Ferrand,
France.
4.US GS, Cascades
VolcanoObservatory,
Vancouver,WA 98661USA.
5.SeismicResearchUnit,, Portof Spain,Trinidad.
6.BritishGeolol•icalSurveyWest Mains Rd., EdinburghEH9 3LA, U.K.
Copyright1998by theAmericanGeophysical
Union.
3425
3426
COLE ET AL.: PYROCLASTIC FLOWS GENERATED BY GRAVITATIONAL
INSTABILITY
elevated
extrusion
rateintherange
4-10m3s
-1,bothpriortoand
WH
N
immediately after the collapses. The last four sustaineddome
collapses:25 June,21 September,and4 ! 6 November1997, have
Centre
Hills
involved
volumes of 4.2,
8, and 18 million
cubic metres
respectively.The maximum volume being shedin any one event
hasincreasedwith increasingaverageextrusionrates.
Valle
Observations
Hill
of the Flows
Initiation
lish's
y mo u t h
Wall
Crater
South
Soufriere
Hills
Figure 1. The distributionand extentof pyroclasticflow deposits
formed in 1996 - 1997 (up to 25 December 1997). The three
depositsshownin the Tar River Valley emplaced:3 April (solid),
12 May (mod stipple) and 17 September1996 (stipple). White
River: February/March 1997 (solid) andApril to November1997
(stipple). Mosquito Ghaut: 17 June 1997 (solid). Tuitts Ghaut:
June5 1997 (solid). Large stippledareato N andNE are the flows
depositsof 25 Juneand21 September1997. Fort Ghaut::17 June
(solid)and3 August1997 (stipple). The extentof flows includes
both the basalavalancheand the dilute surgeportions.
A further large sustainedcollapseevent producedpyroclastic
flows which travelled down Fort Ghaut, destroying much of
Plymouthon 3 August 1997. This eventwas followedby a phase
of explosiveactivity (3 - 10 August).Large domecollapseflows
occurreddown Tuitts Ghaut on 21 September1997, coveringa
wide area acrossthe NE flanks and impacting the airport. This
massunloadingof the dome initiated a phaseof 75 Vulcanian
explosionsover the next month.Renewedgrowth focusedon the
southernside of the dome generatingtwo major collapsesdown
the White River valley which occurredin early November 1997.
The small fan at the end of the White River valley was extended
duringthistime so thatit is currently1.5 km wide alongthe shore
andprotrudes500m into the sea.
Volumesof the flow depositshavebeenestimatedby surveying
the collapsescarsas well as the deposits.For discreteevents,such
as the 12 May 1996 flow, depositvolumeswere in the range o.f
The initiation of the pyroclasticflows was typically associated
with non-explosive, gravitational collapse of unstable dome
material, althoughthe processby which the cascadingrockfall
debris
transforms
over
a few
tens of metres
into
a coherent
pyroclastic flow is poorly understood. In a few cases the
generationof rockfallsand pyroclasticflows was associatedwith
small impulsive 'rockbursts' from active faces of the dome
especially during periods where growth was concentratedin a
1ocalisedarea. Minor, supe.
rficial explosionsalso occurredfrom
source areas immediately after the sheddingof material, these
subsequentlyfed further flows which were pulsatoryin nature.
Someflows were triggeredby long-period(LP) earthquakesand
associatedwith ash and gas ventingthat precededthe generation
of the flows by 10-20s. LP earthquakesare consideredto be an
inherent process in the generation of the rockfalls by
pressurizationfrom within the dome but they may also trigger
rockfallsand pyroclasticflows by shakingthe dome.
Flow Behaviour
The flow front of the ash cloud surges characteristically
developedwedgeshapedfrontswith a groundhuggingnoseand a
lobe andcleft type morphologyin the orderof a few metreshigh
(Fig 3). This morphologywas intermittentlydisruptedby material,
including large blocks, from within and behind the flow head
being ejectedforward. Velocity variationsbetweenthe body of
the ash cloud surge and the basal avalancheunit allowed brief
14
--
I
Sustained
collapses
•1o
E
-_-
{- 6
._o
Discrete flows
-i
o
lOO
__• PFdeposit
volume
'
•: 8o
•
•
Ash plume volume
Tar River fan
• 60
0.2 x 106m
3 denserockequivalent
(DRE)whilesustained"- 40
collapses
shed
between
1 - 9 x 106m
3 (DRE)material
fromthe
dome.Thetotalaccumulated
volume
ofdeposits
is 125x 106m
3 =E20
o
--_
--_
(DRE) 'whichis roughlyhalf of the total extrudedvolume(246
million) as of 25 December 1997. Weekly running average
extrusionrates are comparedwith cumulativepyroclasticflow
depositvolumes(Fig. 2). Although discreteeventshave occurred
regularly through the eruption, these only accountfor a small
componentof the volumeand were associatedwith periodsof low
go
15-Nov-95
;"'::::• ......... -5-25-May-96
4-Dec-96
15-Jun-97
25-Dec-97
Figure 2 .Weeklyrunningaveragedomeextrusionrate(solidline,
datapointsare givenwherevolumeestimatesare spacedat more
than 1 month)comparedwith cumulativepyroclasticflow deposit
volume, Tar River fan volume and estimated volume of lofted ash
toaverage
extrusion
rates(1-3m3s-1).
Sustained
domecollapseplumes.Periodsof discretepyroclasticflow activity (shortbars)
events however, were generally associated with periods of
andsustainedcollapseevents(longbars)are shown.
COLE ET AL.: PYROCLASTIC FLOWS GENERATED BY GRAVITATIONAL
INSTABILITY
3427
Lofting Ash Plumes
Vigorously convecting ash plumes were generatedfrom the
larger pyroclastic flows which frequently rose to heights of
severalthousandmetres.Those formed during sustainedcollapses
frequently reachedas high as -- 10 km. The lofting ash plumes
developprogressivelyalongthe entirelengthof the flow but tend
to form protuberanceswhich rise as discrete pulses. Vigorous
pulsesare producedabovebreaksin slopesuchas the baseof the
Galways Wall where dome materialcascadedinto the Soufriere
region and air entrainmentwas enhanced.Ascent velocities for
loftingashplumestypicallyshowcurvedvelocityprofiles(Calder
et a1.1997), 3 km runout, discrete flows in May 1996 produced
plumes which reacheda maximum of 10-12 m/s at heightsof
300-500
m.
Temperature Measurements
Direct temperaturemeasurementsof the ash cloud surgewere
obtainedfor three pyroclasticflows in the Tar River valley using
industrial temperaturepatchesfixed to posts 1.5 m above the
Figure3' Pyroclastic
flowat 13:17
hrson25June
1997in
SpanishPoint approx I km SSW of WH Bramble airport.
ground.Theseyieldedtemperatures
of 99-121øC, 99 - 149øC;
and200-250øC respectively.
Temperatures
withinanyoneflow
were consistentwithin the bracketsabove for severalpositionson
a profile runningtowardsthe valley axis. Temperaturevariations
glimpsesof the lower block-richcomponentas they accelerated
betweenflows may thusbe attributedto differing temperaturesof
over steeperportionsof the flow path. However for muchof the source material on the dome.
time the basal avalancheswere hidden by the dilute ash so that
one could not easily discern the exact behaviour of the coarse
debris. Although the basal avalanche component was strongly Erosion
topographicallycontrolledthey couldoccasionallybe seento have
A striking feature of the pyroclastic flows was their erosive
appreciable superelevationeffects at bends in flow paths. A capability. In the Tar River valley erosion has removed of all
distinctivefeatureof many flows was the highly variableratesof vegetation and topsoil as well as generally smoothing and
flow front advanceoften of stop-startnature,varyingfrom 0 to 10 rounding the topography, producing a 1 km wide denuded
m/s in 30 seconds for an individual
flow. Elsewhere,
such
behaviour has been considereda function of either topography
(Hoblitt, 1986, Levine and Keiffer, 1991) or non-linear mixing
processesat the flow head (Huppert et al. 1986): anotherpossible
cause at Montserrat is the pulsatory nature of flow generation.
During periodsof continuousor semi-continuous
pyroclasticflow
activity later pulses supply parts of flows initiated earlier,
occasionallydistincthigh velocity pulsescompletelyovertakeor
engulf earlier, slower moving parts. Initial estimatesof average
flow front velocities of the discrete flows are in the order of 3-10
m/s for < l km runout flows and 5 - 20 m/s for 3 km runout flows.
Locally,thefrontsof someof the largerflows'advance
asfastas
60 m/s.
Movement of Ash Cloud SurgesAcrossthe Sea
Good visual observationswere madeof the 12 May 1996 flow
as it travelledover the sea.The surgecomponentdetachedfrom
the basal avalanche, travelled across the water surface and
developedinto largelobeseachup to 50 m wide.The flow front
velocity
noticeably
accelerated
from8-15m/sasit approached
the
corridor. Eroded bedrock surfacesare conspicuousowing to the
orange/brown hydrothermally altered appearance of the
underlying prehistoric deposits. The side walls of ravines,
constrictionsand bendsin valley along which the flows passed
were severelyscouredand striated.Measurementsindicatethat in
placesover 5 m depth of material was removed. During the 17
September1996 collapse,a -50 m wide channelwhich could be
tracedthe whole lengthof the Tar River valley, was excavatedup
to 30 m deep into thick accumulations of pre-September
pyroclasticflow deposits.On a smallerscale,chutesare regularly
carved into the base of the dome by persistentrockfalls from a
localisedsourceregion. A spectacularillustrationof erosionby
pyroclasticflows wasthe incisionof a gully 60 m deepand 80 m
wide throughthe upper portionsof the southerncrater (Galways)
wall. The gully itself was excavatedin only a threehourperiodof
pyroclasticflow generationduringthe 31 March 1997 sustained
collapseevent.
Deposits
waterto a maximumvelocityof 25 m/s beforeit wanedandcame
to restapproximately200 m offshore.The densebasalportionof
The depositsof the pyroclastic flows are similar to those of
otherdocumenteddome-collapseeruptions(Nakada & Tositsugu,
1993, Boudonet al. 1993). They are predominantlycomposedof
the flow entered the sea but dark, sediment laden clouds within
dense, non-vesicular - microvesicular andesite (Robertson et al.
the water, obscuredobservations.Boiling water and explosive
sediment-richfountainsof the order-10 m high occurrednearthe
1998). Vesicular andesiteis subordinatebut has been presentin
increasing proportions since May 1997 and during explosive
phases.The developmentof the ashcloud surgefacies,the degree
of confinementof the entire flow and the slopeson which they
deposit vary with size of event. The talus apron immediately
below the dome comprisesamalgamatedfans of lobate deposits
marginof the new pyroclasticfan. Mushroomshapedloftingash
plumesformedabovethe distalend of the flow and thesewere
followed by columnsof dense steamrising several hundred
metres. Ash fallout from these clouds typically occurs as
accretionarylapilli - 2 mm in diameter.
with well defined
levee and channel
structures.
Individual
flow
3428
COLE ET AL.: PYROCLASTIC FLOWS GENERATED BY GRAVITATIONAL
depositsare typically 10-20 m wide, hundredsof metresin length
and are comprisedpredominantlyof coarse,block-rich material
with no distinct fine grained ash cloud surgedeposit,although
smalllofting ashplumeswere observedduringtheir emplacement.
Discrete pyroclasticflows of moderaterunout (2-3 km) were
stronglyconfinedby topography.Generallytwo distinctdeposit
facies can be identified: coarse,block-rich deposits,which are
confined solely to the valley axis and are interpretedas basal
avalanchedeposits;andthin, morewidespreadfine grainedblockpoorashcloudsurgedepositsThese ashcloudsurgedepositsare
generallylimited to the sidewallsof ravines,but spreadup to 100
m from the valley axis where flows sweptaroundbends.In the
lower reachesof valleys such as the White River valley, Tuitts
Ghautand early flows in MosquitoGhaut,manyof thesedeposits
havevirtually no laterallyextensiveashcloudsurgecomponent.
Pyroclasticflow depositswithin the Tar River valley formedby
the July - September1996 and December 1996/January1997
collapseeventshad a more widespreaddistributionsuchthat they
were not confinedto ravines. The flows depositedextensive,thin
(<50 cm) fine grainedashbut left only very limitedcoarsefacies
in the valley evenwithin the centralchannel.Theseeventseroded
pre-collapsedepositsout of the valley and carried most of the
coarsematerialdown to the fan. Likewise depositsfrom the large
collapsesdown Mosquito Ghaut (25 June 1997), Fort Ghaut (3
August 1997) and Tuitts Ghaut (21 Sept. 1997) all had a similar
widespreaddistributionof the ashcloudsurgecomponentandthe
volumesof material involved were suchthat they overflowedthe
Ghaut walls readily. During these events the ash cloud surge
componentof the flows producedwas more fully developedand
energetic. Vegetation knock-down directions oblique to the
directionof the valley axis and observationsof pyroclasticflows
indicate that the ash cloud surge detachesreadily at bends in
ravines.The 25 June1997 pyroclasticflow wasa typicalexample,
wherethe upper,finer graineddilute part of flow detached,spilled
out of the upper part of Mosquito Ghaut, and crossed the
watersheddrainingwestwardsinto the Belhamvalley.
The Fans
INSTABILITY
Discussion
The pyroclasticflows were all generatedfrom gravitational
collapse of the dome. Small flows are well confined by
topography whereas the larger flows produced during the
sustained collapse events involve more material and often
overspilled the valley walls. These flows also had better
developed ash cloud surge componentswhereas some of the
smallerdiscreteflows appearedto have had virtually none. The
development of the ash cloud surge by fragmentation and
elutriation
of ash from the basal avalanche
increases
with flow
volume,velocity and the excavationlevel of the collapsescarinto
the interior of the dome.
The runoutdistancesof theseflows can qualitativelybe seento
be proportionalto their volumes.However, volume discrepancies
between scar size and that of the depositsin the fan suggestas
much as 4 million cubic metresof material has passedover the
front of the Tar River fan into the deepwaterbeyond.
The ability of both small and largepyroclasticflows to erodeas
well as depositand thereby modify the topographyover which
they passhas consequences
for the routes taken by subsequent
flows. Evidencesuggestsa transitionexistsbetweenan erosional
regime and a depositionalregime which is largely dependanton
slopeand flow velocity.Large flows are erosivethe entire length
ofthevalleys
where
theaverage
slope
angleis 14-16
ø anddeposit
predominantly
whereslopeangles
in theorderof 4-6ø. Small
volume flows can also be highly erosivein their upper portions
but are capableof depositingcoarsematerial on slopesof up to
20ø.
Acknowledgements.Fundingfor this work from the UK Departmentfor
InternationalDevelopmentand NERC is acknowledged.This paper is
publishedby permissionof the Director BGS (NERC). We thank MVO
staff for their invaluablehelp, Peter Taylor for assistance
with diagrams
andtwo anonymous
reveiwersfor improvingthe manuscript.
References
Boudon,G., G. Camus,A. Gourgaud,and J. Lajoie, The 1984 nueeardente
depositsof Merapi volcano,CentralJava,Indonesia::stratigraphy,
texturalcharacteristics,
andtransportmechanisms,
Bull. Volcanol.,55,
Pyroclastic flows first reached the sea on 12 May 1996
327-342,1993.
depositinga small fan of deposits-100 m wide at the mouthof Calder,E.S., R.S.J. Sparks,and A.W. Woods,Dynamicsof co- ignimbrite
plumesgeneratedfrom the pyroclasticflows of Mount St. Helens
the Tar River valley. Since then, over 30 flows have reachedthe
(7 August 1980), Bull. Volcanol., 58, 432-440, 1997.
Tar River fan with growth mainly occurringin large increments Hoblitt,R.P. Observations
of the eruptionsof July 22 andAugust7, 1980,
during the sustainedcollapseevents(Fig 1). The Tar River fan
at Mount St Helens,Washington.U.S.G.S.Prof. Paper1335., 1986.
currently(December1997) protrudes600 m out into the seaandis Huppert,H., S. Turner,S.N. Carey,R.S.J.Sparks,andM.A. Hallworth.,A
laboratorysimulationof pyroclasticflows downslopes,J. Volcanol.
1200 m wide alongthe shoreits relativelystraightleadingedgeis
Geotherm Res., 30, 179-199, 1986.
thoughtto be constrainedby the submarineshelfof a paleo-fan.
Levine, A.H., andS.W. Kieffer., Hydraulicsof the August7, 1980,
pyroclasticflow, at Mount St Helens,Washington.Geology.,19,
Thevolume
ofthefan(December
1997)isestimated
at16x 106
m3 DRE,representing
almost
40%of thetotalcumulative
pyroclasticflow depositvolume.At the fan apex,the depositsare
60 m thickandslopegently(4-6ø) downto thefan margins.
Bathymetricsurveys(August1996) indicatethe angleof reposeof
thesubmarine
deposits
is typically
in therange12-17
ø, with
1121-1124, 1991.
Nakada,S. and F. Toshitsugu,.Preliminary
reporton the activity at Unzen
Volcano (Japan),November 1990-November1991: Dacite lava domes
andpyroclasticflows, J. Volcanol. and Geotherm.Res., 54,319333,1993
Robertson,R., P.D. Cole, R.S.J.Sparks,C. Harford,A.M. Lejeune,W.
maximum slopeson the leadingedge of the fan in the range 22McGuire,A. Miller, M. Murphy,G. Norton,N. Stevens,S.R. Young,
27ø. By April 1997thesurface
of thefancomprised
numerous The explosiveeruptionof SoufriereHills Volcano,Montserrat,17
September,1996, GeophysResearchletters.This issue.
overlapping lobate depositsand boulder trains which radiate
outwardsin broad arcs from the mouth of the valley. On the
shoreline, where the flows entered the sea the coarse blocks show
P.D. Cole, Dept of Geology, Universityof Luton,Luton,UK
E.S.Calder,andR.S.J.Sparks,Bristol, University,Bristol,UK.
red oxidised margins and distinctive prismatic radial joints
indicativeof quenching.
T.H.Druitt, Universite Blaise Pascal, Clermont Ferrand, France.
R Hoblitt, USGS, CVO, Vancouver, WA 98661, USA
A fan was built at the end of the White River (4.5 km from the
R. Robertson,SRU, U.W.I, Portof Spain,Trinidad.
S. Young,BritishGeologicalSurvey,Edinburgh,UK.
dome) by pyroclastic flows associated with the Vulcanian
explosionsin October1997. This wasenlargedby domecollapses (Received:November 13, 1997, Revised:February23, 1998, Accepted:
in November
1997withtheaddition
of 6 x 106m3 of material.
April 14, 1998.)