Direct observations of a three million cubic meter rock-slope collapse with almost immediate initiation of ensuing debris flows

Geomorphology351(2020)106933

ContentslistsavailableatScienceDirect

Geomorphology

j o ur na l h o me p a g e :w w w . e l s e v i e r . c o m / l o c a t e / g e o m o r p h

Direct

observations

of

a

three

million

cubic

meter

rock-slope

collapse

with

almost

immediate

initiation

of

ensuing

debris

flows

Fabian

Walter

,

Florian

Amann

_,

_Andrew

_Kos

_,

_Robert

_Kenner

_,

_Marcia

_Phillips

_,

Antoine

de

Preux

_,

_Matthias

_Huss

_,

_Christian

_Tognacca

_,

_John

_Clinton

_,

_Tobias

_Diehl

_,

Yves

Bonanomi

a_{LaboratoryofHydraulics,HydrologyandGlaciology(VAW),ETHZurich,Zurich,Switzerland} b_{ChairofEngineeringGeologyandHydrogeology,RWTHAachenUniversity,Germany} c_{TerrasenseSwitzerlandLtd,BuchsSG,Switzerland}

d_{WSLInstituteforSnowandAvalancheResearchSLF,Davos,Switzerland} e_{MartiAG,Bern,Switzerland}

f_{DepartmentofGeosciences,UniversityofFribourg,Fribourg,Switzerland} g_{BeffaTognaccaGmbh,Switzerland}

h_{SwissSeismologicalService,ETHZurich,Zurich,Switzerland} i_{BonanomiAG,Igis,Switzerland}

a

r

t

i

c

l

e

i

n

f

o

Received26June2019

Receivedinrevisedform29October2019 Accepted30October2019

Availableonline2November2019 Keywords: Rockavalanche Debrisflows Permafrost Environmentalseismology

a

b

s

t

r

a

c

t

Catastrophiccollapseoflargerockslopesranksasoneofthemosthazardousnaturalphenomenain mountainlandscapes.Thecascadeofevents,fromrock-slopefailure,torockavalancheandthe near-immediatereleaseofdebrisflowshasnotpreviouslybeendescribedfromdirectobservations.Wereport onthe2017,3.0×106_m3_{failureonPizzoCengaloinSwitzerland,whichledtohumancasualtiesand} significantdamagetoinfrastructure.Basedonremotesensingandfieldinvestigations,wefindachange incriticalslopestabilitypriortofailureforwhichpermafrostmayhaveplayedadestabilizingrole.The resultingrockavalanchetraveledfor3.2kmandremovedoveronemillionm3_{ofglaciericeanddebris} depositsfromapreviousrockavalanchein2011.Whereasthisentrainmentdidnotleadtoanunusually largerunoutdistance,itfavoreddebrisflowactivityfromthe2017rockavalanchedeposits:thefirst debrisflowoccurredwithadelayof30sfollowedbytendebrisflowswithin9.5handtwoadditional eventstwodayslater,notablyintheabsenceofrainfall.Wehypothesizethatentrainmentandimpact loadingofsaturatedsedimentsexplaintheinitialmobilityofthe2017rockavalanchedepositsleadingto anear-immediateinitiationofdebrisflows.Thisexplainswhyanearlierrockavalancheatthesamesite in2011wasnotdirectlyfollowedbydebrisflowsandunderlinestheimportanceofconsideringsediment saturationinarockavalanche’srunoutpathforAlpinehazardassessments.

©2019TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/4.0/).

1. Introduction

Thefailureoflargerockslopesresultsinrockavalanches,which canhavealongrun-out(>1km)andchargethelandscapewith vastquantitiesoffragmentedrockdebris.Asaresult,theriskfor remobilizingtheunconsolidateddepositmaterialasdebrisflows, triggeredbyprimaryfactorssuchasrainfallorearthquakesor sec-ondaryfactorssuchasantecedentrainfallorsnowmelt(Huang andFan,2013)dramaticallyincreases,andremainshighforsome timeafterrockslopefailure(i.e.daystomonth,dependingon

seis-∗ Correspondingauthor.

E-mailaddress:walter@vaw.baug.ethz.ch(F.Walter).

micityandthetypeandquantityofwatersupplytoenhancethe debrismobility)(Bergeretal.,2011;Baeretal.,2017,Franketal., 2019).

Afewcasestudiessuggestadirecttransitionofrockavalanches intolargeanddestructivedebrisflows(e.g.Hauser,2002,Plafker andEricksen,1978,Petrakov etal.,2008;Huggel,2009).Hauser (2002) analysed the 1987 6.0×106_m3 _{Cerro Rabicano rock}

avalanche (Chile). Entrainment of a significant volume of path materialconsistingoffluvialsedimentsand5–10mofsnowcover mayhavecaused adirect transitionfroma rockavalancheinto adebrisflow.Asimilartransitionisreportedfortheearthquake triggered, 1970 rock-ice-avalanche at the north peak of Neva-dosHuascaran,PeruvianAndes(PlafkerandEricksen,1978).The transition from the ice-rock avalanche with an initial volume https://doi.org/10.1016/j.geomorph.2019.106933

ofapproximately50–100×106_m3 _{(theicevolumewas}

approxi-mately5.0×106_m3_{)intoadebrisflowwaslikelyassociatedwith}

theentrainmentofdebrisandasignificantamountofsnowand firn.Theaccumulatedsnowdepthatthe2.4 km-longtransition zonewhentherockmassmovedacrossGlacier511wasestimated tobeatleast28mandthetotalentrained,equivalentwatervolume was10–20×106_m3_{(PlafkerandEricksen,1978).}

Largerockslopesfailwhentheyreachacriticalstress thresh-old(Kosetal.,2016).Recognisingthechangeincriticality,well inadvanceoffailurehowever,remainsachallenge.Eventhough theprocesscascadefromslopefailuretorockavalancheanddebris flowisarecognisedphenomenon,theunderlyingmechanisms con-trollingthetransitionremainpoorlyunderstood.Frompublished accounts,rockavalanchedepositswereeitherremobilizedasdebris flowslongaftertheinitialrockavalancheevent,oreventshadto bereconstructedviascientificback-analysis(e.g.,Marietan,1925; Evansand Clague,1994,Kotlyakovet al.(2004),Haeberlietal., 2004;Huggeletal.,2005;Petrakovetal.,2008;Huggel,2009;Kääb etal.,2018).

Herewefocusonthesequenceofeventsbefore,duringandafter themassive2017rockslopecollapseonPizzoCengalointheSwiss Alps.Thecollapsewasobservedanddocumentedinunprecedented detail,resultinginanewpictureofcascadinghazardinteractions forlargerockslopeinstabilities.Incontrasttoothercasestudies, wheredebrisflowsweretriggeredbyearthquakes,rainfallorsnow meltlongafterarockavalanche(e.g.Bergeretal.,2011;Baeretal., 2017)ordirectlytransitionedintoadebrisflow(e.g.Plafkerand Ericksen,1978,Hauser,2002),thePizzoCengalorockavalanche wasdepositedandfollowedbyafirstdebrisflowwithadelayof 30sandinabsenceofanyprecipitation.Understandingthe fac-torsinfluencingtheobservedcascadesandthemechanismsbehind processinteractionsiscriticalfordevelopingfuturehazard man-agementstrategies.

2. Studysiteandhistoryofrockslopefailure

PizzoCengalo(WGS84:46◦1749.090N9◦3637.688E,3369m asl) is situated in the Val Bregaglia region within Eastern Switzerland’sCantonGrisons.Itlocatesatthevalleyheadofthe BondascaValley,some6kmup-valleyfromthevillageofBondo (Figs.1and2).NearPizzoCengalo,theBondascaValleyandits lateralcanyonshostanumberofsmallglaciers(Figs.1and2b). Atloweraltitudes,thevalleyhasbeensparselydevelopedwith anaccessroad,vacationhomes,stables,popularhikingtrails, a seasonallyoccupiedmountainhut(“ScioraHut”)andparkingareas. PizzoCengalo’sNEfacehasbeensubjecttovaryingsizesofrock slopefailures.Systematicobservationoftherockslope instabil-itycommencedin 2012,inresponsetoa1.5×106_m3 _rockwall

collapse,whichoccurredinDecember2011(Fig.2c).The2011 col-lapseevolvedintoa rockavalanche, traveling2.7km downthe Bondascavalley,andchargedthelandscapewithaconsiderable depositofrockdebris(Fig.2c).Photographsshowthattheback scarpofthe2011failurewaslargelycoveredwithathicklayerof bluepermafrostice(Fig.S3a),whichmayhavecontributedtothe catastrophicfailure.Duringsummer2012,heavyrainfall remobi-lizedpartsofthe2011rockavalanchedeposit,andfourdebrisflows weretriggered,whichreachedthevillageofBondo,locatedata dis-tanceof6kmatthevalleybottom(Baeretal.,2017).Thedebrisflow hazardpromptedtheconstructionofaretentionbasininBondoand theinstallationofanearlywarningsystematPrä(Fig.1).

On23August2017,another,significantlylargerrockvolume of3.00×106 _±0.02m3 _{collapsedatthePizzo CengaloNEface,}

(Figs.1and2d)asdeterminedfromrepeatedterrestriallaserscans (TLS,seeSections3andTextS2)andtime-lapsphotographs,which alsodocumentaprecursory0.10-0.15×106_m3_{rockfalleventon21}

August2017(personalcommunicationM.Keiser,CantonGrisons). Themaincollapse on23 August2017resultedina3.2kmlong rock avalanche, which overran the deposits of the2011 event andreachedthehikingtrailtotheScioraHut(Fig.2d).Therock avalancheclaimedthelivesofeighthikerswhopassedtheaffected trailsectionatthetimeoffailure.

3. Slopestabilitypriortothe2017event

Displacementmonitoringwithterrestrialradarinterferometry (TRI,seeSectionS3inthesupplementarymaterialfordetails)and TLSbetween2012and2017promptedanearlywarningissued sev-eralweekspriortothe2017catastrophicfailure.TRIdatashowed thatannualdisplacementsvariedbetween40mmyr−1and60mm yr−1duringthe2012–2015period(Fig.4).BetweenAugust2015 andJuly2016annualdisplacementsnotablyincreasedbyafactor 1.5,andwerethenfollowedbyadramatic,factor2.5–3.0increase betweenJuly2016andJuly2017.TRImeasurementsinsummer 2013indicatenoslopedisplacementsduringthesummerseason (Fig.4b–c).Thus,themeasuredannualdisplacementsaccumulated solely betweenthe beginning offall and end of spring. During this period,increasedfluid pressureassociated withsnowmelt infiltrationcannotexplaintheobserveddisplacements,assnow accumulatesonlyatafewlocationsinthesteeprockwall. Addi-tionally,snowmeltonNorth-facingrockslopestendstoforma basalicelayerunderthesnowpackinspring,largelypreventing theinfiltrationofmeltwater(Phillipsetal.(2016)).

Afterthe2017failure,first-handobservationsviatelescope con-firmed small patches of ice in thebackscarp, which appeared coveredorinterspersedwithdebrisanddust.Inadditiontothese observationsofice,arecentlydevelopedpermafrostdistribution mapconfirmsthepresenceofpermafrostintheNEslopeofPizzo Cengalo(Kenneretal.,2019).Permafrostisknowntoplayan impor-tantroleinrockslopestability(Allenetal.,2013;Krautblatteretal., 2013;Delineetal.,2015;Phillipsetal.,2016).Icesegregationor freezethawcyclescaninduceicegrowthinfractures(ice-wedging) whichenhance crackopening(Matsuoka,2001; Draebingetal., 2017)followedbyprogressiverockbridgefailure(Phillipsetal., 2016).The fact that at least in 2013displacements were only observedbetweenthebeginningoffallandendofspringbutnot duringthesummerperiodsuggeststhatice-wedgingisan impor-tantpreparatoryfactor.

Theincreaseddisplacementratesafter2015suggestanincrease inthedegreeofcriticality(i.e.,agreaterproportionof discontinu-itiesintherockmassareclosetotheirstrengthlimit)resulting fromrockbridgefailurethatoccurredinthepreviousyears.This isconsistentwiththeTLSdataobtainedbetweenlatespringand earlyfall2016,whichincontrastto2013alsoshowedsignificant displacementsduringthesummerperiod.Theincreasedcriticality impliedanincreasedsensitivityoftherockslopetostresschanges induced,forexample,bychangesinwaterpressureassociatedwith summerrainfallevents.

ThefinaltriggerfortherockslopefailureinAugust2017cannot beassociatedwithasingleevent.Post-failureanalysishowever showsthatatthetimeoffailure,thewaterlevelsindaylighting structuresatthelateralreleasesurfacewereapproximately60m abovethebasalfailureplane(Fig.3).

Theorientationofgeologicalstructuresisakeycontrolling fac-torforthePizzoCengalorockslopeinstability.Structuresinthe granodioriteincludethreepenetrative,brittlesetsof discontinu-ities(Fig.3 andS1in theSupplementaryinformation).Priorto the2017catastrophic failure, kinematicanalysisof the geolog-icalstructures(Supplementaryinformation)indicatesaprimary topplingmechanismalongslope-paralleldiscontinuities(Set3in Fig.3andS1).Thosediscontinuitiesaffecttherockslope

stabil-F.Walter,F.Amann,A.Kosetal./Geomorphology351(2020)106933 3

Fig.1. OverviewmapandtopographiccrosssectionshowingtheBondascaValley,PizzoCengaloandthevillageofBondo.Limitsofthe2011and2017rockavalanche deposits,andlimitsofthedebrisflowsarealsoshown.NotethedebrisflowretentionbasininBondoandanearlywarningsystemfordebrisflowsinstalledatPrä.

ityabove2800masl.Belowthiselevation,thelowerdipangleof thesediscontinuitiesrenderstopplingunlikely.Thisissupported bydisplacementmapsobtainedfromradarinterferometric mea-surements, which indicate nodisplacements below 2800masl (Fig.4).Duetotoppling,thesteepslope-parallelstructures(Set 2)arekinematicallyunconstrainedandslidingcanoccurasa sec-ondarymechanism.Set1(Fig.3c)isofparticularimportancefor thedeep-seatedrockslope instabilitybecauseit strikesnormal totheslopeand,duetoitswidespreadpersistence(e.g.>50m), formsmajorlateralreleaseplanes(Fig.3).Thebasalfailureplane oftheslopeinstabilityisnotassociatedwithpre-existing geologi-calstructures,butratherexhibitsastep-shapedsurfaceindicating progressiverockbridgefailure(Fig.3)(Gischigetal.,2011a,2011;

Phillipsetal.,2016).Failureofrockbridgeswasfrequentlyheard asalow-frequencyrumblingduringsummerfieldtripsbetween 2015and2017.Thequalitativeintensityoftheaudiblesoundswas notablystrongerin2016and2017,whichanecdotallycorrelates withthedramaticincreaseindisplacementratesmeasuredwith TRIandTLS.

4. Rockavalanchedynamics

Thecatastrophiccollapseon23August2017wascapturedon video(videoS1;Supplementaryinformation).Thefootageshows howthefailedrockmassimpactedasmallglacierlocatedatthefoot oftheslope(Fig.2b).Theimpactcausedtheejectionofawhitejet

Fig.2.a)DetailsofBondascaValleyheadandterrainchangesinresponsetothe2011and2017PizzoCengalorockavalanches.a)SitesituationwithinSwitzerlandandthe BondascaValley.b)GlaciercoverageandthestreamnetworkmappedonanorthophotowhichshowsPizzoCengaloandtheupperBondascaValleybeforethe2011and 2017rockavalanches.Mapdata:swissimage©2019swisstopo(5704000000).c)&d)Changesinsurfaceelevationcausedbytherockslopefailuresin2011and2017.Map: SwissMapRaster©2019swisstopo(5704000000).Redcirclesina),b),c)andd):mountainhutCapannadiSciora.Theyellowtriangleswitharedoutlineinb),c)andd)

F.Walter,F.Amann,A.Kosetal./Geomorphology351(2020)106933 5

Fig.3.a)ThePizzoCengaloNEfacebeforethe2017rockfalleventandb)6hafterthe2017failure.The2011and2017releaseareasareoutlinedinredandorange,c)Zoom tothefailurezonewiththelateralreleaseplane(red)correspondingtoSet1structuresandSet3structures(blue)responsiblefortheprimarytopplingmechanicsandSet 2structures(orange)thatformslopeparallelreleaseplanes.d)Photograph48hafterthemainrockfallevent.ThewatertablecanbeinferredfromwaterflowfromSet3 structuresatthelateralreleaseplane.

indicatethepositionfromwhichterrestriallaserandradarmeasurementswerecarriedout.Thenumberedareaswithblackoutlinesind)definetheindividualdeposition orerosionareas.Thewhitearrowinsector1ashowsthepositionoftheblowholeslinkedtotheoccurrenceofgeyser-likewaterfountains(FigureS3b).Erosion/deposition distributionsaredifferencesoftheFederalOfficeofTopographyswisstopo(www.swisstopo.admin.ch)digitalelevationmodelsregisteredon18July2012-26September 2011(Panelc)and25August2017-30August2015(Paneld).

Fig.4. Resultsofrockslopedisplacementmonitoringacquiredwithaportableradarinterferometer.Displacementsinpanelsa,b,d–gareshownasinterferograms.Eachcolor cycle(e.g.,fringe)representsan8.72mmdisplacementandisdefinedbythewavelengthoftheradar’scentralfrequency(e.g.,17.2GHz).c,aphase-unwrappeddisplacement map,scaledto±3.0mm.Theinterferogramsanddisplacementmapwereprojectedintoathree-dimensionalphotogrammetricmodelforvisualization.AtpointsP1andP2 displacementdatawereextractedandplottedinpanelh.h)yearlydisplacementratesattwoselectedpointsextractedfrominterferogramsbetween2012–2017(P1and P2).Thedateofcatastrophicfailureisshownbythedashedline.

oficeand/orwaterparticleswithasub-verticaltrajectorydirected towardstheW-faceoftheBügeleisenridgeline(Fig.3a).Itwasthen redirectedalongaNEtrajectorydowntheBondascavalley(e.g.in thedirectionoftherockavalancherun-outFig.2).Theicejetwas airborneforapproximately8–9sandthefall-outoftheparticles occurredalongadistanceofapproximately450–550mmainlyin area1a(Fig.2d).Ourfirst-orderapproximationfromvideoS1 sug-gestsavelocityofupto70ms−1forthelargerandfasterparticleson

theleadingedgeoftheicejet.Avolumeof0.6±0.1×106_m3_glacier

ice wasremoved by an undetermined combination of melting, impactanderosion(forthevolumeestimationseeSupplementary information).

Thecascadeofeventswasrecordedonnearbyseismicstations, andwasequivalenttoanearthquakeofMI3.0(Fig.5)asreported bytheSwissSeismologicalService.Atdistancesexceeding100km, seismicstationsrecordedabroadbandsignalbetween0.005and

F.Walter,F.Amann,A.Kosetal./Geomorphology351(2020)106933 7

Fig.5.Spectrogram(top)derivedfromtheground-velocityseismogram(bottom)ofthe2017rockavalancheeventrecordedattheclosetseismicstationCH.VDL(46.48317, 9.44956;FigureS4).Afirstpeakinamplitudeoccursabout15–30safterthefirstonsetofthesignal,followedbyalargersecondpeakatabout40–60s.Withadelayof30s, therockavalanchesignalisfollowedbyasustainedhigh-frequencydebrisflowtremorwithinitiallyelevatedenergyabove20Hzforatleast200–300s.

severalHz(Fig.S5),allowingforadetailedreconstructionofthe rockavalanche run-outdynamics.Seismicenergy above1Hz is associated withinter-particle collisionsand particleinteraction withobstaclesintherun-outpath,whereasthelowerfrequencies arecausedbytheforce,whichtherockavalanche’sbulkmotion exertsontheEarth’ssurface(EkströmandStark,2013;Allstadt, 2013;Allstadtetal.,2018,andreferencestherein;seedetailsin TextS8).

FollowingAllstadt(2013),weinvertthelowfrequency seismo-grams(TextS8)fortheunderlyingforcehistory(Fig.6).Uponinitial release,therockmassacceleratesdownwardandeastwardandthe elasticreaction(“rebound”)ofPizzoCengaloisopposite,i.e.upward andwestward(PhaseAinFig.6).Thislikelyoverlappedwiththe

north-slopingglacierimpactpushingthegroundsouthward.The massthenacceleratesnorthwardacrosstheglacier,whichresults inafurthersouthwarddirectedreboundforce.Withinafewtens ofseconds,themassimpactstheBügeleisenridgelinetotheeastof theglacier.Theridgelineexertsawestwardforceontherockmass forcingitonamorenorthwardtrajectory.Thisimpliesaneastward reactionforceexertedbytherockmassontheridgeline(Phase B).Duringthefinalfrictionalslowdown,therockmasspushesthe groundprimarilynorth-northwestanddownward(PhaseC).

Fromtherockavalanche’sforcehistory(Fig.6)wecalculatethe centerofmasstrajectory(Fig.7;seeTextS8fordetails).The tra-jectorybulgessomewhattoofareastwardthanexpectedfromthe valleyshape.Wethereforeinterpretthetrajectorywithcare

confin-Fig.6.Forcehistoryofthe2017PizzoCengalorockavalanche.Dashedlines rep-resentuncertaintyestimates.Greenlinesarethelow-frequency(0.006–0.1Hz) seismogramsfromthenearestbroadbandstationVDL(FigureS4).Bluesolidline rep-resentsthecalculatedrockavalanchespeedsalongitstrajectorywithdashedlines indicatinguncertainties.ThegreyseismogramistheVDLrecordfilteredbetween1 and5Hz.Differenttrajectoryphasesdiscussedinthemaintextarelabeled(A:mass detachmentandglacierimpact,B:impactwithBügeleisenridgeline,C:final slow-down).Verticalgreybarsdenotethemanuallypickedonsetandendoftheforce history.

ingaquantitativediscussiontothoseaspectsthatcanbeexplained withtopographicfeaturesalongthetrajectory.

Thefirstofthetwoavalanche’sspeedmaxima(Figs.6Cand7)is reachedasthecenterofmassleavesthesteepglacierportionand entersflatterterrain(Fig.2).Thesecondmaximum(Figs.6Cand7) locatesjustbehindthesteepterrainstepinthecentralrunoutarea (Figs.1and2).Therockavalanche’shigh-frequencyseismogram exhibitstwoburstsseparatedby20–30s(Figs.5and6)andthe firstburstisassociatedwiththeglacierimpactbeforethefirstspeed maximumwhereasthesecondhigh-frequencyburstresultsfrom theimpactbehindthedownstreamterrainstep(Figs.6Cand7).

The timingof high-frequency energy transmission and rock avalanchedynamicsisconsistentwithpreviousstudies:Adelay

Fig.7.Rockavalanchetrajectoryfromtheforcehistorycalculationwithcenterof massspeedcolorcoded.Thetwogreendotsindicatespeedmaximaandthered circledenotesthemountainhutCapannadiSciora.Ascalculatedrockavalanche speeddoesnotvanishattheend,thecalculatedtrajectorywastruncated.

betweenthehigh-frequency seismicsignal aftertherock mass detachmentmanifestingitselfinthestartoftheforcehistorycan be explained with progressive fragmentation of the rock mass (Allstadt,2013;Hibertetal.,2014].Similarly,acorrespondenceof highrockmassmomentumandhigh-frequencyseismicsignalshas beenobservedforotherevents(Hibertetal.,2017).However,the generationofthefirsthigh-frequencyseismicburstofthePiz Cen-galoeventislikelyalsorelatedtothecollisionwiththeBügeleisen ridgeline,constitutingatopographicalbarrier(Hibertetal.,2014). Weusethetemporalseparationofthehigh-frequencybursts (20–30 s) and the distance between the two ground impact pointsalongthetrajectory(about1200m),toestimatetherock avalanche’sspeedbetweenthefirstandsecondimpactpointsto 40–60ms−1.Thisrangeofvaluesisconsistentwiththeaverage speed(Fig.6),calculatedasfollows:Weintegratethespeedtime seriesbetweenthetwohigh-frequencyburstsanddivideitbythe timedifference,whichgives44-64ms−1.About60safterthefirst impacttherockavalanchecametoastopwithmostofthedebris depositedinArea3(Fig.2).

Inourinversionofthelow-frequencyseismicsignaltherock avalanche’sforcehistoryisonlyconstrainedbyseismicdata.As aconsequence,thoseportionsoftheaccelerationanddeceleration phaseswhichhappenslowlyenoughtosuppressameasurable seis-micsignal(especiallyduringthefinalslowdown),arenotincluded intheinvertedforcehistoryandcalculatedtrajectory.Thisexplains thenon-vanishingfinal rock avalanche speed of26m/s (Fig.6) calculated with our inverted force history. This problem could bemitigated withtheinversion algorithmofEkströmand Stark (2013),whichrequiresstationarityoftherockmass’bulk momen-tum.Moreover,suchconstrainedinversionsavoidspuriousforce histories for band-limited seismic signals (Hibert et al., 2015). ForthePizCengaloeventwefollowtheunconstrainedinversion approachofAllstadt(2013)inordertoelucidatedynamicdetailsof therockavalancheandthereforemaximizethebandwidthofthe invertedlow-frequencyseismicsignals(TextS8).

F.Walter,F.Amann,A.Kosetal./Geomorphology351(2020)106933 9

5. Fromrockavalanchetoremobilizationofrockdebris

Within8.5days,aseriesof15debrisflowsformedfromthe depositsofthe2017rockavalancheasdocumentedbyeye wit-ness observationsand seismicdatafrom stationVDL (personal communication M.Keiser, CantonGrisons, Fig.5 and S4;Table S1). Theeventsstarted nearlyimmediatelyafter the2017rock avalanche. Theexact number of individualeventsis somewhat subjectivebecauseitissometimesnotclearwhethersubsequent surgesbelongedtothesamedebrisflow.Infact,thefirstdebrisflow wasovertakenbytheseconddebrisflow.

Bycomparingdigitalelevationmodels(DEMs) weestimated thevolumesofthesingledebrisflowevents.Asonlyfour rele-vantDEMsareavailable(30.08.2015,25.08.2017,30.08.2017and 05.09.2017;TextS4)onlycumulativevolumesofsuccessivedebris flowscanbecalculated.Volumeestimatesforsingleeventsrelied atleastpartiallyonindirectevidence.Forexample,thegeometry ofthechannelcarvedbyadebrisflowwasusedinitsflowvolume calculation.

Alarge debrisflow withaninitialvolume ofapproximately 0.545±0.05×106_m3_{wasinitiatedwithadelayofapproximately}

30s(Fig.5;seeTextS9)afterthedepositionoftherockavalanche. This event was followed by ten debris flows within 9.5h and anothertwo debrisflowstwodaysafterrockavalanche deposi-tion(TableS1).Notably,remobilizationofthedebriswithinthese twodaysoccurredintheabsenceofrainfall.Aweekaftertherock avalanche,twodebrisflowsoccurredduringprecipitationon31 August2017.

Directobservationsshowthatthefirstdebrisflowwasahighly viscous,slowmovingphenomenon.Theinitialvelocityofthefirst debrisflow wasapproximately 8m/s and canbe reconstructed fromtheinitiationtimederivedfromtheseismicdataanalysisand thetriggeringtimeoftheearlywarningsystemin“Prä”(Fig.1) at09:35:36UTCon23August2017(localtimeisUTC+2h).The debrisflowwasfirsttimeseenandphotographedataround09:43 closetotheexitoftheBondascavalley,about950mdownstream oftheearlywarningsystem.Theseobservationssuggestan aver-agevelocityof2m/s.Eyewitnessesreportedthatthedebrisflow wasmovingbelowwalkingspeedwhenitarrivedattheoldbridge at09:48(averagevelocityaround0.4m/s)anditwasovertakenby thefollowingdebrisflow(TableS1).Thisfirstdebrisflowdestroyed elevenstablesandvacationhomesintheBondascavalley(Fig.1and TableS1intheSupplementaryinformation).

Approximatelyonehouraftertherockavalanche, helicopter-borne observations of a distinct channel in the rock avalanche depositionindicatethatthefirstdebrisfloworiginatedfromArea3 (Fig.2d).Thisfirsteventhasatypicaldebrisflowseismogram show-ingasustainedtremoraftertherockavalanchesignalwithhigh frequenciesupto50Hz(Fig.5)(Allstadtetal.,2018).Inthe spec-trogram(Figs.5andS7)thesignalmostlyresidesbelow10Hzand towardsitsbeginningseismicitybetween15and50Hziselevated for3–4minutes.Aspectralgapabove10Hz(“delay”inFigs.5and S7)clearlyseparatestherockavalancheseismogramfromthis ini-tiationsignal(TextS9).Thisshowsthatthefirstdebrisflowformed withadelayofaround30saftertherockavalanchewasdeposited (Fig.5,TableS1).Observationsfromthehelicoptershow12 addi-tionaldebrisflowson23and25Augustthatreachedtheretention basininBondo(TableS1andFig.1).Incontrasttothefirstdebris flowtheyhadalowerviscosityandwereinitiatedfromArea1a (Fig.2d).

Thetotalvolumeofthetendebrisflowsfollowingthefirstmajor oneon23August2017was2.20–2.3×105_m3_{.Thefirsteightdebris}

flowsdidnotcausedamageinthevillageofBondo,butsuccessively filledthedebrisflowretentionbasin.Subsequentdebrisflows over-toppedtheretentionbasin anddamaged thecantonal roadand housesinthevillageofBondo.Thetotalvolumeofdebrisflows

reachingBondoon25August2017was7.0-8.5×104_m3_.The

trans-porteddebrisagainovertoppedtheretentionbasinand areasof Bondowereoverrun,causingfurtherdamage.Noneofthedebris flowscausedhumancasualties.Finally,on31Augustheavy precip-itationtriggeredtwodebrisflowswithtotalvolumesofbetween 2.6and2.85×105_m3_.

6. Discussion

Ourobservationsrevealthefollowingdetailsaboutthe2017 rockavalanche:

1Detachment:Rockbridgefailureleadtoenhanceddisplacement

ratesandacriticalstabilitystateinthePizzoCengalo’sNErock face(Figs.3and4).Icewedgingmayhavecontributedtorock bridgefailure.

2Dynamics:Therockavalanche wasconstrained byridgelines,

slopebreaksandothertopographicalfeatures(Fig.6),iteroded ca.0.6×106_m3_{ofglacialice(Fig.S2)andproduceddepositsthat}

coveredthoseofthe2011rockavalancheandotherpre-existing valleyfillssuchasmorainesandalluvialfans(Fig.2).

3Immediatedebrisflows:Withinhalfaminutetoseveraldays

afterthe2017rockavalanche,aseriesof15debrisflowswere mobilizedindepositsaboveandbelowtheterrainstepalongthe rockavalanche’srunoutpath(Fig.2,TableS1),mostofthemin theabsenceofprecipitation.

Wenowaddressthekeyquestionaboutwhatcausedthehigh mobility of the 2017 rock avalanche deposits leading to near-immediateformationofthefirstdebrisflowinArea3(Fig.2),a terraininclinedonlymoderatelyataround13◦(Fig.1).

Theinitiationofdebrisflowsrequiresrelativelyhighwater con-tent(e.g.JohnsonandSitar,1990).Apuzzlingquestioninthiscase is:Wheredidthewatercomefromtomobilizethedebrisflows in theabsence of precipitation?The water contentin therock slopeisvolumetricallynegligible,andtheporespaceswithinthe disintegratedrockavalanchematerialareessentiallydry. Further-more,volumeexpansionastherockmassdisintegratesprohibits porepressureincreaseduringthedynamicflow(HungrandEvans, 2004).

Therockavalancheeroded6.0×105_m3_{ofglacierice.Muchof}

thismasslikelymeltedimmediatelyuponrockavalancheimpact andduringsubsequenttransportwithintherockavalanche.This issupportedbyaccountsofrescueteamssearchingfortheeight missinghikers,hoursaftertheevent,whodidnotreportanyicein thedeposits(personalcommunicationM.Keiser,CantonGrisons). However,theobservedice/waterjet,formedbytheinitialimpact ontheglacier(supplementalvideo),showsthatnotalloftheglacier icewasentrainedatthemovingfrontoftherockavalanche: Com-paringtheaveragerockavalanchespeedcalculatedfromthelow frequencyseismicity(44–64ms−1,Section4)withthespeedof thefastesticejetparticles(70ms−1,Section4)suggeststhatthe majorityofjetice,whichmovedmuchslowerthanitsfront,was strewnacrosstheuppersurfaceoftherockavalanchewhereit sub-sequentlymeltedandcontributedtoahigherwatercontentinthe depositedrockavalanchedebris,thuscontributingtoitsmobility (Area3inFig.2).

Apart from thelost glacier ice, Fig. 2 shows a distinct pat-tern of material erosion and deposition during the 2017 rock avalanche:Abovethetopographicstep,debriswasdepositedin Area 1betweenthetoeof thePizzoCengaloand theBondasca morainetotheEast(Fig.2).BetweenAreas1and 3,in Area2, erosionandthusentrainmentof sedimentsoccurredduringthe rockavalancherun-out(Fig.2).Wefocusonareasshowing depo-sition in 2011 and erosion in 2017 tospecifically quantify the

amountofentrainedmaterialfromthe2011depositinthe2017 rockavalanche.DEMdifferences(Fig.2)showthatthicknessofthe 2011depositslocallyexceeded10mandatleast6.65×105_m3_of

thesedepositswereerodedandentrainedbythe2017event. Inadditiontotheentrainmentof6.65×105_m3_{ofthe2011rock}

avalanchedeposit,DEMdifferencingyieldssome2.0×105_m3 _of

othersedimentsthathavebeenentrainedalongtherun-outpath. Theseestimatesaresubjecttolargeuncertaintiesbecausethe addi-tionalsedimentswerecoveredbyvegetationbeforethe2017rock avalanche.Ourestimateofthetotalentrainmentvolumeranges between8.0and 9.0×105_m3 _{and is considereda lowerbound}

becausesedimentreplacementcannotbeexcluded.Ouranalysis showsthat entrainmentoccurredtosomeextentin Area1 but mainlyinArea2(Fig.2d).Consequently,depositsinArea3,where thefirst debris flow was initiated, included much of the 2011 deposits.

Quantitativestudies(e.g.Hungr andEvans, 2004;Mangeney etal., 2007, Crostaet al.2009, Aaronet al.,2019)suggest that entrainment of saturated sediments along the rock avalanche run-outpathcanenhancetherockavalanchemobility.Therock avalanchemobilityiscommonlyexpressedbytheindexH/L(i.e. theratiooffallheighttohorizontaltraveldistance).Interestingly, theH/Lratioforthe2017PizzoCengalorockavalanche,withafall heightofH=1.7kmandahorizontaltraveldistanceofL=3.3kmto themostdistalpointofArea3is0.53.ThisH/Lratioiswithinthe rangeofrockavalanchesofsimilarvolumesinnon-glaciatedareas, withnegligiblesediment entrainment(Hungrand Evans, 2004; Sosioetal.,2012),andclosetoavalueof0.6fordry,fragmented rock,assuggestedbyHeim(1932).Thus,theH/Lratioindicatesthat entrainmentdidnotsignificantlyenhancethemobilityoftherock avalanchethatcametoahaltinArea3,30sbeforethefirstdebris flowwasinitiated,asshownbytheseismicfrequencysignature (TextS9,Fig.S7).Wepropose,however,thatthewatercontentof theentrainedsedimentswasacriticalfactorinthedeposit mobil-itycontributingtotheinitiationofdebrisflowsintheabsenceof precipitation.

Thesurfacehydrologypriortothe2017event(Fig.2)shows thatmeltwaterfromtheBondascaglacierdrainedthroughthe pre-existingsediments(includingthe2011deposits)whichwerelikely saturatedatthetimeofthe2017rockavalanche.Onceentrained in the2017 rockavalanche, these sediments provided a major sourceofwater. Unliketheentrainmentofglacierice, whichat leastpartiallyoccurredbehindtherockavalanchefront,muchof thesaturatedsedimentswereentrainedinArea2anddepositedin Area3,wherethefirstdebrisflowinitiated.

Thefirst-hand accountsof climbers who coincidentally wit-nessed theevent, reportedthat in Area 1a(Fig.2), geyser-like waterfountains(blowouts)occurred25and40minaftertherock avalanche(Fig.S3).Theclimbersreportedtwoexplosiveandclearly audiblewaterfountainswithanestimatedheightofafewmeters. Theremnantsoftheblowoutswerelocatedlaterinthefieldbythe authors(locationshownbyanarrowinFig.2d).Hourslater res-cueteamsreportedsaturatedmudholesadjacenttosoliddebris depositsalongthemarginsofArea3(Fig.2),whichpersistedfor manydaysaftertheevent.Theseobservationsnotonlycorroborate thepresenceofwaterinthe2017deposits,butalsosuggestthatthis waterwaspressurizedandmayhavediffusedtotheearthsurface whereitformedmudholes.Wehypothesizethatdynamic load-inguponimpact(HutchinsonandBahndari,1971;BovisandDagg, 1992;HungranEvans,2004)inducedtheexcessporepressures, whichpersistedhourstodaysasaresultoflowpressure dissipa-tionrates(e.g.MajorandIverson,1999,andreferencestherein).The resultingmobilityofthedebrismaterialisalsosupportedbydirect observationsoftheinitiationofdebrisflowsfromArea1a(Table S1):Shallow-seatedmechanicalinstabilitywithinthedebrisledto aprogressivevolumeincreaseofmovingdebrisandeventuallyto

theformationofdebrisflows.Further,alltwelvedebrisflowevents fromArea1ainitiatedfromdebristhatcoveredthepre-existing maindrainagepathoftheBondascaglacieroutflowstream(Fig.2d). Themobilityofthedebrismaterialwasthusadditionallyenhanced byacontinuoussub-surfacewatersupply.

Ourhypothesisofentrainmentanddynamicloadingofwater saturatedsedimentsexplainswhyincontrasttothe2017event, the2011rockavalanchewasnotdirectlyfollowedbydebrisflows. Instead,remobilizationoftherockavalanchedepositdidnotoccur until six tonine monthsafter the event,in response toheavy summerrainfallin 2012(Baeretal., 2017).Aerialphotographs takenon26September2011beforethe2011rockavalancheshow thatthedistributionofthesedimentsintheBondascavalleywas significantlysmallerand agreaterareaofbedrocksurfaceswas exposed(Fig.2b).Theimpactofthefailedrockmasserodedonly ∼0.1×106_m3 _{of glacierice (TextS6) andsince the2011event}

occurredinwinter,runofffromtheBondascaglacierwasnegligible. Sedimentsalongtherun-outpathandinthedepositswere there-forelikely unsaturatedin 2011.Thediffering mobility between the2011and2017depositsisinlinewithlaboratoryexperiments ofwater-saturatedgranularflows(Iversonetal.,2011):Forflows overridingwetsedimentsthereexistsapositivefeedbackbetween entrainmentandmomentum.Forflowsoverridingdrysediments this feedback is negative.Consequently, entrained dry material tendstoreducetherun-outdistancewhereasentrainmentof sat-urateddebriscanenhancethemobilityofthedebris(Abele,1997; Crostaetal.,2009).

Fromthiscomparisonwerecognizethattwopre-existing fac-torsplayedacriticalroleinthe2017hazardcascadethatranged fromrockwallfailureoverrockavalanchepropagationto ensu-ing debrisflows: 1)The landscapewas covered withsufficient sedimentsfromthe2011rockavalanche,and2)thepre-existing sedimentsweresaturatedwithsummermeltwaterfromthe Bon-dascaglacierandbyintenserainfallduringthunderstormsinJuly andAugust2017.

7. Conclusions

Thisinvestigationpresentsauniqueopportunitytostudya cas-cadeofAlpinemassmovementprocesses.Favoredbypermafrost conditions,a3.0×106_m3_{rockmasscatastrophicallyfailedatthe}

PizzoCengaloNErockfaceproducingamajorrockavalancheinfall 2017.Inversionofseismicbroadbanddatayieldsaforcehistory, whichdescribesdynamicdetailsoftherockavalanche’scenterof masstrajectorythatagreeswellwithlocaltopographicfeatureslike aconstraininglateralridgelineandtopographicsteps.

Ourobservationssuggestthatentrainmentanddynamic load-ingofwatersaturatedsedimentsmayhavebeendominantfactors in the hazard cascade. They did not increase the reach of the 2017 rock avalanche compared to typical non-entraining rock avalanches.However,entrainmentofwatersaturatedsediments and the hypothesized dynamic loading may explain the high depositmobilitythatcauseddebrisflowsnearlyimmediatelyafter therock avalanchedepositionand in absenceof any precipita-tion.Thisisincontrasttothe2011rockavalanche.In2011less sedimentscovered therockavalancherun-out pathwhich ren-dersentrainmentlesslikely.Inaddition,thesedimentswerelikely notsaturatedsincetheeventoccurredinwintertimewithonly minimalwaterrun-offfromtheBondascaglacier.Saturated sedi-mententrainmentwasthuslikelythepivotalfactorindebrisflow productionfollowingthe2017rockavalanche.

Thestudyelucidatestheimportanceofinvestigatingconditions alongarockavalancherunoutpathwhenforecastingorassessing hazardousAlpinemassmovements.Focusingonconditionsinthe releasezone,only,whenestimatingtiming,volumeandrunout

dis-F.Walter,F.Amann,A.Kosetal./Geomorphology351(2020)106933 11

tancesofrockcollapsesandensuingdebrisflowsisinsufficientto fullygraspthepotentialhazardchaindownslope,aswasshownin the2017PizzoCengaloevent.

DeclarationofCompetingInterest

Theauthorsdeclarethattheyhavenoknowncompeting finan-cialinterestsorpersonalrelationshipsthatcouldhaveappearedto influencetheworkreportedinthispaper.

Acknowledgments

Data are available online at the institutional repository of RWTH Aachen University (http://dx.doi.org/10.18154/RWTH-2018-228255)andtheSwissSeismologicalService(http://arclink. ethz.ch/webinterface/). The salary of FW was paid by the Swiss National Science Foundation via Grants PP00P2157551 andPP00P2 183719).Terrestrial measurementswerefundedby ArgeAlps.WethankAmandineSergeantforfruitfuldiscussionson theseismicinversion.

AppendixA. Supplementarydata

Supplementarymaterial relatedto thisarticle canbe found, intheonline version,atdoi:https://doi.org/10.1016/j.geomorph. 2019.106933.

References

Aaron,J.,McDougall,S.,Jordan,P.,2019.Dynamicanalysisofthe2012Johnsons Land-inglandslideatKootenayLake,BC:theimportanceofundrainedflowpotential. Can.Geotech.J.(ja).

Abele,G.,1997.Rockslidemovementsupportedbythemobilizationof groundwater-saturatedvalleyfloorsediments.ZeitschriftfurGeomorphologie41(1),1–20.

Allen,S.,Huggel,C.,2013.Extremelywarmtemperaturesasapotentialcauseof recenthighmountainrockfall.Glob.Planet.Change107,59–69.

Allstadt,K.,2013.ExtractingsourcecharacteristicsanddynamicsoftheAugust2010 MountMeagerlandslidefrombroadbandseismograms.J.Geophys.Res.Earth Surf.118(3),1472–1490.

Allstadt,K.E.,Matoza,R.S.,Lockhart,A.,Moran,S.C.,Caplan-Auerbach,J.,Haney,M., Thelen,W.A.,Malone,S.D.,2018.Seismicandacousticsignaturesofsurficial massmovementsatvolcanoes.J.Volcanol.Geotherm.Res.

Baer,P.,Huggel,C.H.,McArdell,B.W.,Frank,F.,2017.Changingdebrisflowactivity aftersuddensedimentinput:acasestudyfromtheSwissAlps.Geol.Today33 (6),November–December2017.

Berger,C.,McArdell,B.W.,Schlunegger,F.,2011.Sedimenttransferpatternsatthe Illgrabencatchment,Switzerland:implicationsforthetimescalesofdebrisflow activities.Geomorphology125(421–432),2011.

Bovis,M.J.,Dagg,R.,1992.Debrisflowtriggeringbyimpulsiveloading:mechanical modellingandcasestudies.Can.Geotech.J.29,345–352.

Crosta,G.B.,Imposimato,S.,Roddeman,D.,2009.Numericalmodellingof entrain-ment/depositioninrockanddebris-avalanches.Eng.Geol.109(1–-2),135–145.

Deline,P.,Gruber,S.,Delaloye,R.,Fischer,L.,Geertsema,M.,Giardino,M.,etal.,2015.

Icelossandslopestabilityinhigh-mountainregions.In:SnowandIce-related Hazards,RisksandDisasters.AcademicPress,pp.521–561.

Draebing,D.,Krautblatter,M.,Hoffmann,T.,2017.Thermo-cryogeniccontrolsof fracturekinematicsinpermafrostrockwalls.Geophys.Res.Lett.44,3535–3544.

Ekström,G.,Stark,C.P.,2013.Simplescalingofcatastrophiclandslidedynamics. Science339(6126),1416–1419.

Evans,S.G.,Clague,J.J.,1994.Recentclimaticchangeandcatastrophicgeomorphic processesinmountainenvironments.Geomorphol.Nat.Hazards,107–128.

Frank,F.,Huggel,C.,McArdell,B.W.,Vieli,A.,2019.Landslidesandincreased debris-flowactivity:asystematiccomparisonofsixcatchmentsinSwitzerland.Earth Surf.ProcessesLandforms44(3),699–712.

Gischig,V.S.,Moore,J.R.,Evans,K.F.,Amann,F.,Loew,S.,2011a.Thermomechanical forcingofdeeprockslopedeformation:1.Conceptualstudyofasimplifiedslope. J.Geophys.Res.EarthSurf.116(F4),F04010.

Gischig,V.,Amann,F.,Moore,J.R.,Loew,S.,Eisenbeiss,H.,Stempfhuber,W.,2011b.

Compositerockslopekinematicsatthecurrentrandainstability,Switzerland, basedonremotesensingandnumericalmodeling.Eng.Geol.118(1–2),37–53.

Haeberli,W.,Huggel,C.,Kääb,A.,Zgraggen-Oswald,S.,Polkvoj,A.,Galushkin,I.,etal., 2004.TheKolka-Karmadonrock/iceslideof20September2002:an extraordi-naryeventofhistoricaldimensionsinNorthOssetia,RussianCaucasus.J.Glaciol. 50(171),533–546.

Hauser,A.,2002.RockavalancheandresultingdebrisflowinEsteroParraguirreand RioColorado,regionmetropolitana,Chile.Geol.Soc.Am.XV,135–148.

Heim,A.,1932.BergsturzundMenschenleben(LandslidesandHumanLives). Trans-latedbyN.SkermerBitechPress,Vancouver,1932.

Hibert,C.,Ekström,G.,Stark,C.P.,2014.DynamicsoftheBinghamCanyonMine landslidesfromseismicsignalanalysis.Geophys.Res.Lett.41(13),4535–4541.

Hibert,C.,Stark,C.P.,Ekström,G.,2015.DynamicsoftheOso-Steelheadlandslide frombroadbandseismicanalysis.Nat.HazardsEarthSyst.Sci.15(6),1265–1273.

Hibert,C.,Ekström,G.,Stark,C.P.,2017. Therelationship betweenbulk-mass momentumandshort-periodseismicradiationincatastrophiclandslides.J. Geophys.Res.EarthSurf.122(5),1201–1215.

Huang,R.,Fan,X.,2013.Thelandslidestory.Nat.Geosci.6,325–326,http://dx.doi. org/10.1038/ngeo1806,2013.

Huggel,C.,2009.Recentextremeslopefailuresinglacialenvironments:effectsof thermalperturbation.Quat.Sci.Rev.28(11–12),1119–1130.

Huggel,C.,Zgraggen-Oswald,S.,Haeberli,W.,Kääb,A.,Polkvoj,A.,Galushkin,I., Evans,S.G.,2005.The2002rock/iceavalancheatKolka/Karmadon,Russian Caucasus:assessmentofextraordinaryavalancheformationandmobility,and applicationofQuickBirdsatelliteimagery.Nat.HazardsEarthSyst.Sci.5(2), 173–187.

Hungr,O.,Evans,S.G.,2004.Entrainmentofdebrisinrockavalanches:ananalysisof longrun-outmechanism.GSABull.116(9/10),1240–1252,September/October, 2004.

Hutchinson,J.N.,Bahndari,R.K.,1971.Undrainedloading,afundamentalmechanism ofmudflowandothermassmovements.Geotechnique21(4),353–358.

Iverson,R.M.,Reid,M.E.,Logan,M.,LaHusen,R.G.,Godt,J.W.,Griswold,J.P.,2011.

Positivefeedbackandmomentumgrowthduringdebris-flowentrainmentof wetbedsediment.Nat.Geosci.4(2),116.

Johnson,K.A.,Sitar,N.,1990.Hydrologicconditionsleadingtodebris-flowinitiation. Can.Geotech.J.27(6),789–801.

Kääb,A.,Leinss,S.,Gilbert,A.,Bühler,Y.,Gascoin,S.,Evans,S.G.,etal.,2018.Massive collapseoftwoglaciersinwesternTibetin2016aftersurge-likeinstability.Nat. Geosci.,1.

Kenner,R.,Noetzli,J.,Hoelzle,M.,Raetzo,H.,Phillips,M.,2019.Distinguishing ice-richandice-poorpermafrosttomapgroundtemperaturesandgroundice occurrenceintheSwissAlps.TheCryosphere13,1925–1941,http://dx.doi.org/ 10.5194/tc-13-1925-2019.

Kos,A.,Amann,F.,Strozzi,T.,vonRuette,J.,Delaloye,R.,Springman,S.,2016. Contem-poraryglacierretreattriggersarapidlandslideresponse,GreatAletschGlacier, Switzerland.Geophys.Res.Lett.43,http://dx.doi.org/10.1002/2016GL071708, 12‘466–12‘474.

Kotlyakov,V.M.,Rototaeva,O.V.,Nosenko,G.A.,2004.TheSeptember2002Kolka glaciercatastropheinNorthOssetia,RussianFederation:evidenceandanalysis. MountainRes.Dev.24(1),78–83.

Krautblatter,M.,Funk,D.,Günzel,F.K.,2013.Whypermafrostrocksbecome unsta-ble:arock–ice-mechanicalmodelintimeandspace.EarthSurf.Process.Landf. 38(8),876–887.

Major, J.M., Iverson, R.M., 1999. Debris-flow deposition: effectsof pore-fluid pressureandfrictionconcentratedatflowmargins.GSABull.111(No.10), 1424–1434,October,1999.

Mangeney,A.,Tsimring,L.S.,Volfson,D.,Aranson,I.S.,Bouchut,F.,2007.Avalanche mobilityinducedbythepresenceofanerodiblebedandassociatedentrainment. Geophys.Res.Lett.34(22),http://dx.doi.org/10.1029/2007GL031348. Marietan,I.,1925.LeséboulementsdelaCimedel’EstdesDentsduMidien1926et

leBois-Noir.JournalBull.delaMurithienne44,67–93,1925.

Matsuoka,N.,2001.Directobservationoffrostwedginginalpinebedrock.EarthSurf. Process.Landf.26(6),601–614,http://dx.doi.org/10.1016/j.coldregions.2016. 02.010.

Petrakov,D.A.,Chernomorets,S.S.,Evans,S.G.,Tutubalina,O.V.,2008.Catastrophic glacialmulti-phasemassmovements:aspecialtypeofglacialhazard.Adv. Geosci.14,211–218.

Phillips,M.,Wolter,A.,Lüthi,R.,Amann,F.,Kenner,R.,Bühler,Y.,2016.Rockslope failureinarecentlydeglaciatedpermafrostrockwallatPizKesch(EasternSwiss Alps).EarthSurf.Process.Landf.,http://dx.doi.org/10.1002/esp.3992,07/2016. Plafker,G.,Ericksen,G.E.,1978.NevadosHuascaranavalanches,Peru.In

Develop-mentsinGeotechnicalEngineering,Vol.14.Elsevier,pp.277–314.

Sosio,R.,Croat,G.B.,Chen,J.H.,Hungr,O.,2012.Moellingrockavalanchepropagation onglacier.Quart.Sci.Rev.47,23–40,2012.