<%BANNER%>
PRIVATE ITEM
Digitization of this item is currently in progress.
Probabilistic approach to modeling lava flow inundation: a lava flow hazard assessment for a nuclear facility in Armenia
CITATION DOWNLOADS PDF VIEWER
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/AA00009671/00001
 Material Information
Title: Probabilistic approach to modeling lava flow inundation: a lava flow hazard assessment for a nuclear facility in Armenia
Series Title: Journal of Applied Volcanology
Physical Description: Book
Language: English
Creator: Connor, Laura J.
Connor, Charles B.
Meliksetian, Khackatur
Savov, Ivan
Publisher: BioMed Central
Publication Date: 2012
 Notes
Abstract: Abstract Probabilistic modeling of lava flow hazard is a two-stage process. The first step is an estimation of the possible locations of future eruptive vents followed by an estimation of probable areas of inundation by lava flows issuing from these vents. We present a methodology using this two-stage approach to estimate lava flow hazard at a nuclear power plant site near Aragats, a Quaternary volcano in Armenia. Keywords: lava flow simulation, modeling code, probabilistic hazard assessment, spatial density, Monte Carlo method, Armenia
General Note: Additional material under Downloads Tab: Additional file 1: PERL script that estimates spatial density. This code depends on inputs generated by the SAMSE bandwidth estimation routine from the ‘ks’ library package as part of the ‘R’ programming package. This PERL script is an ASCII (text) file that can be viewed with any text editor. It is run from the command line: perl. Additional file 2: PERL script that simulates volume-limited lava flows from vents on and around the Shamiram Plateau. This lava flow script depends on the output spatial density grid file generated by the above mentioned spatial density script (additional file 1). It is an ASCII file that can be viewed with any text editor. It is run from the command line: perl. Additional file 3: Perl script that simulates effusion rate-limited lava flows from vents located on the flanks of Aragats. This lava flow script depends on the output spatial density grid file generated by the above mentioned spatial density script (additional file 1). It is an ASCII file that can be viewed with any text editor. It is run from the command line: perl.
 Record Information
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution.
Resource Identifier: doi - 10.1186/2191-5040-1-3
System ID: AA00009671:00001

Downloads

This item is only available as the following downloads:

( PL )

( PL )

( PL )

( PDF )


Full Text

PAGE 1

METHODOLOGY OpenAccessProbabilisticapproachtomodelinglavaflow inundation:alavaflowhazardassessmentfora nuclearfacilityinArmeniaLauraJConnor1*,CharlesBConnor1,KhachaturMeliksetian2andIvanSavov3AbstractProbabilisticmodelingoflavaflowhazardisatwo-stageprocess.Thefirststepisanestimationofthepossible locationsoffutureeruptiveventsfollowedbyanestimationofprobableareasofinundationbylavaflowsissuing fromthesevents.Wepresentamethodologyusingthistwo-stageapproachtoestimatelavaflowhazardata nuclearpowerplantsitenearAragats,aQuaternaryvolcanoinArmenia. Keywords: lavaflowsimulation,modelingcode,probabilistichazardassessment,spatialdensity,MonteCarlo method,ArmeniaBackgroundVolcanichazardassessmentsareoftenconductedforspecificsites,suchasnuclearfacilities,dams,portsandsimilarcriticalfacilitiesthatmustbelocatedinareasofvery lowgeologicrisk(Volentiketal2009;Connoretal 2009).Thesehazardassessmentsconsiderthehazardand riskposedbyspecificvolcanicphenomena,suchaslava flows,tephrafallout,orpyroclasticdensitycurrents (IAEA2011;Hilletal2009).Althoughsitehazardscould beconsideredintermsofthecumulativeeffectsofthese variousvolcanicphenomena,abetterapproachisto assessthehazardandriskofeachphenomenonseparately,astheyhavevaryingcharacteristicsandimpacts. Here,wedevelopamethodologyforsite-specifichazard assessmentforlavaflows.Lavaflowsareconsideredtobe beyondthedesignbasisofnuclearfacilities,meaningthat thepotentialfortheoccurrenceoflavaflowsabovesome levelofacceptablelikelihoodwouldexcludethesitefrom developmentofnuclearfacilitiesbecausesafecontrolor shutdownofthefacilityundercircumstancesoflavaflow inundationcannotbeassured(IAEA2011). Thispaperdescribesacomputermodelusedtoestimatetheconditionalprobabilitythatalavaflowwill inundateadesignatedsitearea,giventhataneffusive eruptionoriginatesfromaventwithinthevolcanic systemofinterest.Therearetwoessentialfeaturesofthe analysis.First,thelocationofthelavaflowsourceis sampledfromaspatialdensitymodelofnew,potentially eruptivevents.Second,themodelsimulatestheeffusion oflavafromthisventbasedonfieldmeasurementsof thicknessesandvolumesofpreviouslyeruptedlavaflows withinanareaencompassingthesiteofinterest.The simulatedlavaflowsfollowthetopography,represented byadigitalelevationmodel(DEM).Inputdatathatare neededtodevelopaprobabilitymodelincludethespatial distributionofpasteruptivevents,thedistributionof pastlavaflowswithinanareasurroundingthesite,and measurablelavaflowfeaturesincludingthickness,length, volume,andarea,forpreviouslyeruptedlavaflows.Thus, themodeldependsonmappablefeaturesfoundinthe sitearea.Giventheseinputdata,MonteCarlosimulationsgeneratemanypossibleventlocationsandmany possiblelavaflows,fromwhichtheconditionalprobabilityofsiteinundationbylavaflow,giventheopeningofa newvent,isestimated.Anexamplebasedonanuclear powerplantsiteinArmeniademonstratesthestrengths ofthistypeofanalysis(Figure1).SpatialdensityestimationSite-specificlavaflowhazardassessmentsrequirethat thehazardoflavainundationbeestimatedlongbefore lavabeginstoeruptfromanyspecificvent.Inmany eruptions,lavaseruptfrom newlyformedvents,hence, *Correspondence:Iconnor@usf.edu1UniversityofSouthFlorida,4202E.FowlerAve,Tampa,FL33620,USA FulllistofauthorinformationisavailableattheendofthearticleConnor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 2012Connoretal;licenseeSpringer.ThisisanOpenAccessarticledistributedunderthetermsoftheCreativeCommonsAttribution License(http://creativecommons.org/licenses/by/2.0),whichpermitsunrestricteduse,distribution,andreproductioninanymedium, providedtheoriginalworkisproperlycited.

PAGE 2

thepotentialspatialdistributionofnewventsmustbe estimatedaspartoftheanalysis.Thisisparticularly importantbecausethetopographyaroundvolcanoesis oftencomplexandcharacterizedbysteepslopes.Small variationsinventlocationmaycauselavatoflowina completelydifferentdirectiondowntheflanksofthe volcano.Thus,probabilisticmodelsoflavaflowinundationarequitesensitivetomodelsofventlocation. Furthermore,manyvolcanicsystemsaredistributed. Examplesincludemonogeneticvolcanicfields( e.g.the Michoacn-Guanajuatovolcanicfield,Mexico),distributedcompositevolcanoeswhichlackacentralcrater( e.g Kirishimavolcano,Japan),andvolcanoeswithsignificant flankactivity( e.g.Mt.Etna,Italy).Spatialdensityestimatesarealsoneededtoforecastpotentialventlocationswithinsuchdistributedvolcanicsystems(Cappello etal2011). Inaddition,lociofactivitymaywaxandwanewith time,suchthatpastventpatternsmaynotaccurately forecastfutureventlocations(ConditandConnor 1996).Thus,itisimportanttodetermineiftemporal patternsarepresentinthedistributionofpastevents,so thatanappropriatetimeintervalcanbeselectedforthe analysis(i.e. ,useonlythoseventsthatrepresentlikely futurepatternsofactivity,notolderventsthatmay representpastpatterns). Kerneldensityestimationi sanon-parametricmethod forestimatingthespatialdensityoffuturevolcanicevents basedonthethelocationsofpastvolcanicevents(ConnorandConnor2009;Kiyosugietal2010;Bebbington andCronin2010).Twoimportantpartsofthespatial densityestimatearethe kernelfunction andits bandwidth ,orsmoothingparameter.Thekernelfunctionisa probabilitydensityfunctionthatdefinestheprobability offutureventformationatlocationswithinaregionof interest.Thekernelfunctioncanbeanypositivefunction thatintegratestoone.Spatialdensityestimatesusingkernelfunctionsareexplicitlydatadriven.Abasicadvantage ofthisapproachisthatthespatialdensityestimatewill beconsistentwithknowndata,thatis,thespatialdistributionofpastvolcanicevents.Apotentialdisadvantage ofthesekernelfunctionsisthattheyarenotinherently Figure1 LocationofstudyareainArmenia .Thestudyarea,outlinedbyaredboxonthelocationmap,islocatedinSWArmenia.Themore detailedviewshowsthearealextentandlocationofeffusion-limited(lightercolored)andvolume-limited(darkercolored)lavaflowslocated aroundAragatsvolcano.DetailsofeachoftheselavaflowscanbefoundinTable1.Thedashedredboxidentifiestheboundariesofthelava flowsimulationarea.TheShamiramPlateauisanelevatedregion(withinthecentralportionofthelavaflowsimulationarea)comprisinglava flowsfromShamiram,Atomakhumb,Dashtakar,Blrashark,andKarmratarvolcanoes.TheANPPsite(blackbox)islocatedontheShamiram Plateau.PhotoshowstheANPPsiteandAtomakhumbvolcano. Connor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 Page2of19

PAGE 3

sensitivetogeologicboundaries.Ifageologicboundaryis knownitispossibletomodifythedensityestimatewith dataderivedfromfieldobservationsandmapping.Connoretal(2000)andMartinetal(2004)discussvarious methodsofweightingdensityestimatesinlightofgeologicalorgeophysicalinformation,inamannersimilarto Ward(1994).Adifficultywithsuchweightingisthesubjectivityinvolvedinrecastinggeologicobservationsas densityfunctions. Atwo-dimensionalradially-symmetricGaussiankernel forestimatingspatialdensityisgivenbySilverman (1978);Diggle(1985);Silverman(1986);Wandand Jones(1995): ( s )= 1 2 h2NNi =1exp 1 2 di h 2 (1) Thelocalspatialdensityestimate, ( s ) ,isbasedon N totalevents,anddependsonthedistance, di,toeach eventlocationfromthepointofthespatialdensityestimate, s ,andthesmoothingbandwidth, h .Therateof changeinspatialdensitywithdistancefromevents dependsonthesizeofthebandwidth,which,inthe caseofaGaussiankernelfunction,isequivalenttothe varianceofthekernel.Inthisexample,thekernelis radiallysymmetric,thatis, h isconstantinalldirections. Nearlyallkernelestimatorsusedingeologichazard assessmentshavebeenofthistype(Woo1996;Stock andSmith2002;ConnorandHill1995;Conditand Connor1996).Thebandwidthisselectedusingsome criterion,oftenvisualsmoothnessoftheresultingspatial densityplots,andthespatialdensityfunctioniscalculatedusingthisbandwidth.Atwo-dimensionalelliptical kernelwithabandwidththatvariesinmagnitudeand directionisgivenbyWandandJones(1995), ( s )= 1 2 N | H |Ni =1exp 1 2 bTb b where, b = H-1/2x (2) Equation1isasimplificat ionofthismoregeneral case,wherebytheamountofsmoothingbythebandwidth, h ,variesconsistentlyinboththeN-SandE-W directions.Thebandwidth, H ,ontheotherhand,isa2 2elementmatrixthatspecifiestwodistinctsmoothing patterns,oneinaN-Strendingdirectionandanotherin anE-Wtrendingdirection.Thisbandwidthmatrixis bothpositiveanddefinite,importantbecausethematrix musthaveasquareroot.| H| isthedeterminantofthis matrixand H-1/2istheinverseofitssquareroot. x isa 1distancematrix( i.e .the x -distanceand y -distance from s toanevent), b isthecrossproductof x and H-1/2, and bTisitstransform.Theresultingspatialdensityat eachpointlocation, s,isusuallydistributedonagridthat islargeenoughtocovertheentireregionofinterest. Bandwidthselectionisakeyfeatureofkerneldensity estimation(StockandSmith2002;Connoretal2000; Molinaetal2001;Abrahamson2006;Jaquetetal2008; ConnorandConnor2009),andisparticularlyrelevantto lavaflowhazardstudies.Bandwidthsthatarenarrow focusdensitynearthelocationsofpastevents.Conversely,alargebandwidthmayover-smooththedensityestimate,resultinginunreason ablylowdensityestimates nearclustersofpastevents,andoverestimatedensityfar frompastevents.Thisdependenceonbandwidthcan createambiguityintheinterpretationofspatialdensityif bandwidthsarearbitrarilyse lected.Afurtherdifficulty withellipticalkernelsisthatallelementsofthebandwidthmatrixmustbeestimated,thatisthemagnitude anddirectionofsmoothingintwodirections.Several methodshavebeendevelopedforestimatinganoptimal bandwidthmatrixbasedonthelocationsoftheevent data(WandandJones1995),andhavebeensummarized byDuong(2007).Hereweut ilizeamodifiedasymptotic meanintegratedsquarederr or(AMISE)method,developedbyDuongandHazelton(2003),calledtheSAMSE pilotbandwidthselector,tooptimallyestimatethe smoothingbandwidthforourGaussiankernelfunction. ThesebandwidthestimatorsarefoundinthefreelyavailableRStatisticalPackage(Hornik2009;Duong2007). BivariatebandwidthselectorsliketheSAMSEmethod areextremelyusefulbecause,althoughtheyaremathematicallycomplex,theyfindoptimalbandwidthsusing theactualdatalocations,removingsubjectivityfromthe process.Thebandwidthsel ectorsusedinthishazard assessmentprovideglobale stimatesofdensity,inthe sensethatonebandwidthorbandwidthmatrixisusedto describevariationacrosstheentireregion. Giventhatspatialdensityestimatesarebasedonthe distributionofpastvolcanicevents,existingvolcanic ventswithinaregionandtimeperiodofinterestfirst needtobeidentifiedandlo cated.Thiscompilationis thenusedasthebasisforestimatingtheprobabilityof theopeningofnewventswithinaregion.Ourlavaflow hazardassessmentmethodisconcernedwiththelikelihoodoftheopeningofnewventsthateruptlavaflows. Suchventsmayformwhenmagmafirstreachesthesurface,forminganewvolcano,ormayformduringan extendedepisodeofactivity,wherebymultipleventsmay formwhileaneruptiveepisodecontinuesoversomeperiodoftime,generallymonthstoyears(LuhrandSimkin 1993),andthelocusofactivityshiftsasnewdikesare injectedintotheshallowestpa rtofthecrust.Therefore, forthepurposesofthisstudy,aneventisdefinedasthe openingofanewventatanewlocationduringanewConnor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 Page3of19

PAGE 4

episodeofvolcanicactivity.Multipleventsformedduring asingleepisodeofvolcanismarenotsimulated.NumericalSimulationofLavaflowsOnland,alavaflowisadynamicoutpouringofmolten rockthatoccursduringaneffusivevolcaniceruption whenhot,volatile-poor, relativelydegassedmagma reachesthesurface(KilburnandLuongo1993).These lavaflowsaremassivevolcanicphenomenathatinundate areasathightemperature( > 800C),destroyingstructures,evenwholetowns,byentombingthemwithin metersofrock.Thehighlydestructivenatureoflava flowsdemandsparticularattentionwhencriticalfacilities arelocatedwithintheirpotentialreach. Theareainundatedbylavaflowsdependsontheeruptionrate,thetotalvolumeerupted,magmarheological properties,whichinturnareafunctionofcomposition andtemperature,andtheslopeofthefinaltopographic surface(DragoniandTallar ico1994;Griffiths2000; CostaandMacedonio2005).Previousstudieshavemodeledthephysicsoflavaflow susingtheNavier-Stokes equationsandsimplifiedequationsofstate(Dragoni 1989;DelNegroetal2005;MiyamotoandSasaki1997). Otherstudieshaveconcentratedoncharacterizingthe geometryoflavaflows,andstudyingtheirdevelopment duringeffusivevolcaniceruptions(Walker1973;Kilburn andLopes1988;StasiukandJaupart1997;Harrisand Rowland2009).Thesemorphologicalstudiesaremirroredbymodelsthatconcentrateonthearealextentof lavaflows,ratherthantheirflowdynamics.Thesemodels generallyabstractthehighlycomplexrheologicalpropertiesoflavaflowsusinggeometrictermsand/orsimplified coolingmodels(Barcaetal1994;Wadgeetal1994; HarrisandRowland2001;Rowlandetal2005). Anewlavaflowsimulationcode,writteninPERL,was createdtoassessthepotentialforsiteinundationbylava flows,similar,inprinciple,toareal-extentmodels.This lavaflowsimulationtoolisusedtoassesstheprobability ofsiteinundationratherthanattemptingtomodelthe complexreal-timephysicalpropertiesoflavaflows.Since theprimaryphysicalinformationavailableforlavaflows istheirthickness,area,leng thandvolume,thismodelis guidedbythesemeasurableparametersandnotdirectly concernedwithlavaflowrates,theirfluid-dynamicproperties,ortheirchemicalmakeupandcomposition.The purposeofthemodelistodeterminetheconditional probabilitythatflowinundationofasitewilloccur,given aneffusiveeruptionataparticularlocationestimated usingthespatialdensitymodeldiscussedpreviously. Atotalvolumeoflavatobeeruptedissetatthestart ofeachmodelrun.Themodelassumesthateachcell inundatedbylavaretainsoraccumulatesaresidual amountoflava.Theresidualmustberetainedinacell beforethatcellwillpassanylavatoadjacentcells.This residualcorrespondstothemodalthicknessofthelava flow.Lavamayaccumulateinanycelltoamountsgreater thanthisresidualvalueifthetopographyallowspooling oflava.Asflowthicknessvariesbetweenlavaflows,the residualvaluechosenfortheflowmodelalsovariesfrom simulationtosimulation.Here,ourterm residual correspondstotheterm adherence ,usedincodesdeveloped byWadgeetal(1994)andBarcaetal(1994).Inourcase, residuallavadoesnotdependontemperatureorunderlyingtopography,butrather,isusedtomaintainamodal lavaflowthickness.Lavafl owthicknesses,measured withinthesitearea,arefittoastatisticaldistribution whichissampledstochasticallyinordertochoosearesidual( i.e .modalthickness)valueforeachrealization.Lava flowsimulationrequiresadigitalelevationmodel(DEM) oftheregionofinterest.OnesourceoftopographicDEM dataistheShuttleRADARTopographyMission(SRTM) database.The90-metergridspacingofSRTMdatalimits theresolutionofthelavaflow.Topographicdetailssmallerthan90mcaninfluenceflowpath,butthesecannot beaccountedforusinga90 -mDEM.Amoredetailed DEMcouldprovideenhancedflowdetail,butadecrease inDEMgridspacingincreasesthetotalnumberofgrid cells,thusincreasingcomputationtimeastheflowhasto passthroughanincreasingnumberofgridcells.Abalanceneedstobemaintainedbetweencapturingimportantflowdetailoverthetopographyandlimitingthe overalltimerequiredtocalculatethefullextentofthe flow.Criticalconsiderationsforgridspacingarethe topographyofthesiteareaandthevolumesandflow ratesoflocallavaflows.Lavaflowseruptedathighrate orhighviscositywouldquicklyoverwhelmsurrounding topography,sointhesecasesacoarse90-mDEMmaybe sufficientforflowmodeling.Forlowflowratesorlow viscosities,lavaflowswouldmeanderaroundsmaller topographicfeatureswhichwouldbeunresolvedina coarse90-mDEM.Therefore,inthesecasesahigher resolutionDEMwouldbenecessarytoachievecredible modelresults.Inourstudy,a90-mDEMwasconsidered adequateduetotheunavailabilityofinformationregardinglavaflowratesintheareaandassumedhigherflow ratesbasedonflowgeometriesmeasuredinthefield. Also,theboundariesoftheplateauonwhichtheANPP siteislocatedwasdeterminedtobeadequatelyresolved bya90-mDEM. Asimplealgorithmisusedtodistributethelavafroma sourcecelltoeachofitsadjacentcellsoncetheresidual oflavahasaccumulated.Adjacentcellsaredefinedas thosecellsdirectlynorth,south,eastandwestofasource cell.Foreaseofcalculation,volumesarechangedto thicknesses.Cellsthatreceivelavaareaddedtoalistof active cellstotrackrelevantpropertiesregardingcell state,including:locationwithintheDEM,currentlava thickness,andinitialelevation.ActivecellshaveoneConnor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 Page4of19

PAGE 5

parent cell,fromwhichtheyreceivelava,andupto3 neighbor cellswhichreceivetheirexcesslava.Acell becomesaneighboronlyifitseffectiveelevation( i.e .lava thickness+originalelevation)islessthanitsparent s effectiveelevation.Ifanactivecellhasneighbors,thenits excesslavaisdistributedproportionallytoeachneighbor basedontheeffectiveelevationdifferencebetweenthe activecellandeachofitsneighbors.Lavadistribution canbesummarizedwiththefollowingequation: Ln= XaDn/ T (3) where Lnreferstothelavathicknessinmetersreceived byaneighbor, Xaistheexcesslavathicknessanactivecell hastogiveaway. Dnisthedifferenceintheeffectiveelevationbetweenanactivecellandaneighboringcell, Dn= EaEn,where Eareferstotheeffectiveelevationoftheactive celland Enreferstotheeffectiveelevationofanadjacent neighbor.Theeffectiveelevationisdefinedasthethicknessoflavainacellplusitsoriginalelevationfromthe DEM. T ,isthetotalelevationdifferencebetweenanactive cellandallofitsadjacentneighbors,1 -N T =Nn =1Dn. Iterationscontinueuntilthetotalflowvolumeis depleted.Someexamplelavaflowssimulatedinthis fashionareshowninFigure2.LavaflowhazardattheArmeniannuclearpowerplant siteLavaflowsareacommonfeatureoftheArmenianlandscape.SomemappedflowsarehighlightedinFigure2.A groupof18volcaniccenterscompriseanareaknownas theShamiramPlateau(thisareaislocatedwithinthered boxinFigure1).TheArmeniannuclearpowerplant (ANPP)sitelieswithinthiscomparativelydensevolcanic clusteratthesouthernmarginoftheShamiramPlateau. Ourlavaflowhazardassessmentisdesignedtoassessthe conditionalprobabilitythatlavaflowsreachtheboundary ofthesitearea,givenaneffusiveeruptionontheShamiramPlateau.Inaddition,large-volumelavaflowsare foundontheflanksofAragatsvolcano,a70-km-diameter basalt-trachyandesitetotrachydacitevolcanolocated immediatelynorthoftheShamiramPlateau. ThemappedlavaflowsontheShamiramPlateaucan bedividedintotwoagegroups,pre-ignimbritelava flowsthatrangeinagefromapproximately0.91-1.1Ma, andpost-ignimbritelavaflowsthatcovertheignimbrites ofAragatsvolcano.TheyoungestfeaturesofAragats Volcanoarelargevolumelavaflowsfromtwocinder cones,Tirinkatar(0.45Ma)andAshtarak(0.53Ma).All oftheseagedeterminationsarebasedonK-Ardatingby Chernyshevetal(2002).Theyoungestsmall-volume lavaflowsoftheShamiramPlateauaretheDashtakar groupofcindercones,basedonboreholeevidenceindicatingthattheDashtakarflowsoverlayoneofthese ignimbritesofAragats. LavaflowsoftheShamiramPlateauaretypicalof monogeneticfields,beingofcomparativelylowvolume, generally < 0.03km3,andshorttotallength,generally < 5km.Basedonloggingdatafromfourboreholesand includingtheentireareaoftheShamiramPlateauand estimatedthicknessofthelavapile,thetotalvolumeof lavaflowsmakinguptheplateauis~11 24km3.Given thesevalues,hundredsofindividuallavaflowscomprise theentireplateau.Thus,thereisapossibilitythatlava flowswillinundatethesiteinthefuture,associatedwith theeruptionofmonogeneticvolcanoesontheShamiramPlateau,shouldsucheruptionsoccur. MappedlavaflowsoftheShamiramPlateauare volume-limitedflows(KilburnandLopes1988;Stasiuk andJaupart1997;HarrisandRowland,2009),trachyandesitetotrachydaciteincomposition.Lengthsrange from1.4km,fromShamiramvolcano,to2.49kmfrom Blrasharkvolcano;volumesrangefrom310-3km3, fromKarmratarvolcano,to2.310-2km3fromAtomakhumbvolcano(Table1). Volume-limitedflowsoccurwhensmallbatchesof magmareachthesurfaceanderuptforabriefperiodof time,forminglavaflowsassociatedwithindividual monogeneticcenters.Theseeruptionsoftenoccurin pulsesanderuptingventsmaymigrateashortdistance, generally < 1km,duringtheeruption.Eachpulseof activityintheformationofthemonogeneticcentermay produceanewindividuallavaflow,hence,constructinga flowfieldovertime.Thelongestlavaflowsinthesefields aregenerallythoseassociatedwiththeearlystagesofthe eruption,wheneruptionrat esaregreatest(Kilburnand Lopes,1988).WithintheShamiramPlateauarea,individualmonogeneticcentershaveone( e.g.Shamiramvolcano)tomany( e.g.Blrasharkvolcano)individuallava flows. LongerlavaflowsarealsofoundonAragatsvolcano, especiallyhigheronitsflanks(Table1).Thesesummit lavascompriseathicksequenceoftrachyandesitesand trachydaciteshavingatotalvolume > 500km3.Themost recentlavaflowsfromtheflanksofAragatsinclude Tirinkatar,whichisseparatedintotwoindividualtrachybasaltflowsTirinkatar-1andTirinkatar-2,andtheAshtaraklavaflow.Tirinkatar-1andAshtarakeachhave volumes~0.5km3.Thelargestvolumeflanklavaflows arepartofthetrachydaciticCakhkasarlavaflowofPokr Bogutluvolcano,withatotalvolume~18km3,onthe sameorderasthelargesthistoricaleruptionsoflava flowsworldwide(ThordarsonandSelf1993).Theselargervolumelavaflowsareeffusionrate-limited,sincetheConnor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 Page5of19

PAGE 6

lengthofthelavaflowiscontrolledbytheeffusionrateat thevent.ThelengthsoftheAshtarakandTirinkatar-1 lavaflowsexceed20km.Basedoncomparisonwith observedhistoricaleruption s,theireffusionrateswere likelyontheorderof100m3s-1(Walker,1973;Malin 1980;KilburnandLopes,1988;HarrisandRowland, 2009).Thus,whilevolume-limitedflowseruptonthe ShamiramPlateauintheimmediatevicinityofthesite, effusionrate-limitedflowseruptathigherelevationson theflanksofAragatsvolcano.Whileitisconceivablethat theselargervolumeflows mayreachthesitebecauseof theirgreatpotentiallength,thiseventislesslikely becausetheiroccurrenceissoinfrequent.AnotherdeterrentisthefactthattheShamiramplateauactsasatopographicbarriertotheselonger,largerflowsreachingthe ANPPsite.Eachclassoflavaflows,smallervolume-limited Figure2 SomesimulatedlavaflowsontheShamiramPlateau .Exampleoutputfromthelavaflowsim ulationcode.Lavaflows(colored regions)areeruptedfromvents(blackdots)thatarerandomlysampledfromaspatialdensitymodelofventsontheShamiramPlateau.FlowpathfollowstheDEM.Thesiteareaisconsideredtobeinundatedifthelavaflowintersectsthewhiterectangle.Inthisexample,twooftheten lavaflowsintersectthesiteandoneventfallswiththesiteboundaries. Connor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 Page6of19

PAGE 7

flowsandlargereffusionrate-limitedflows,isconsidered separatelywhenassessinglavaflowhazardattheANPP site.ResultsandDiscussionUsingspatialdensityestimationLocatingthesourceregionoferuptinglavaiscriticalin determiningtheareainundatedbyalavaflow.Probable sourceregionsareestimatedusingaspatialdensity model,whichinturndependsonageologicalmapidentifyingthelocationsofpasteruptivevents.Inthiscontext,volcanicventsaredefinedastheapproximate locationswheremagmahasormayhavereachedthesurfaceanderuptedinthepast.Aprimarydifficultyinusing adatasetofthedistributionofvolcanicventsisdeterminationofindependenceofevents.Instatisticalparlance, independenteventsaredrawnfromthesamestatistical distribution,buttheoccurrenceofoneeventdoesnot influencetheprobabilityofoccurrenceofanotherevent. Weareinterestedinconstructingaspatialdensitymodel onlyusingindependentevents .Unfortunately,itisdifficulttodeterminefrommappingandstratigraphicanalysisifventsformedduringthesameeruptiveepisodeor occurredasindependenteventsduringdifferentvolcanic eruptions.Someoftheseareeasilyrecognized( e.g .boccasthatarelocatedadjacenttoscoriacones).Inother cases,itisuncertainifindividualvolcanoesshouldbe consideredtobeindependentevents,orwereinreality partofthesameevent.Becauseofthisuncertainty,alternativedatasetsareusefulwhenestimatingthespatial density.Here,weuseonedatasettomaximizethe potentialnumberofvolcanicevents:allmappedventsare includedinthedatasetasindependentevents.Analternativedatasetcouldconsidervolcaniceventstobecomprisedofgroupsofvolcanicventsthatarecloselyspaced andnoteasilydistinguishedstratigraphically. Inordertoapplythespatialdensityestimate,itis assumedthat18mappedvolcaniccentersrepresentthe potentialdistributionoffuturevolcanicventsonthe ShamiramPlateau.Someolderventsarenodoubtburiedbysubsequentvolcanicactivity.Itisalsopossible thatolderventsareburiedinsedimentoftheYerevan basin,southoftheANPPsite. Usingadatasetthatincludes18volcanicevents mappedontheShamiramPlateau(Table2),theSAMSE selectoryieldsthefollowingoptimalbandwidthmatrix Table1SizeestimatesoflavaflowsVolcano (source) Area (km2) Thickness (m) Volume (km3) Length (km) Composition Arich16.380.1309.48TB1,BTA1Atomakhumb 3.9 6 0.023 3.43 BA1,BTA Barcradir(Bartsradir)32.990.29612.10TB,BTA Bazmaberd13.1140.1846.34BA,BTA Blrashark 1.6 6 0.010 2.49 TA1,TD1Blrashark 2.5 7 0.018 3.13 TA,TD Bolorsar 2.2 6 0.013 2.72 BTA,TA Dashtakar 2.1 10 0.021 4.44 BA,BTA Dashtakar 1.6 6 0.009 3.66 BA,BTA Karmratar 0.7 4 0.003 3.61 TA MetsMantash 8.9 9 0.080 8.47 TB,BTA Shamiram 1.0 4 0.004 1.41 TA Siserasar 0.8 11 0.009 1.72 TA Tirinkatar-2 13.3 4 0.053 6.54 BTA,BA Topqar(Topkar) 2.9 9 0.026 3.07 BTA,TA Ashtarak 84 6 0.50 26.50 BA,BTA Irind 66 55 3.65 20.53 Dacite Paros 109 8 0.87 33.36 TB,BTA Tirinkatar-1 75 7 0.53 26.36 BTA,BA PokrBogutlu 165 110 18.18 27.92 TD (Cakhkasar)1Note:TB(trachybasalt),BTA(basalt-trachyandesite), BA(basaltic-andesite),TA(trachyandesite),TD(trachydacite) ThevolcanicrocknomenclaturefollowstheoneofLeBasetal(1986) SizeestimatesforsomelavaflowsassociatedwithmonogeneticventsoftheShamiramPlateauandelsewhereontheflanksofAragatsvolcano.Theinput parametersforthelavaflowsimulationswerebasedontheobservedcharacteristicsofthesmaller-volumeflows.Volcanoeslocatedwithintheareaofthe ShamiramPlateauappearinitalicfont.Sizeestimatesforthe5largestlavaflowsontheflanksofAragatsvolcanoarelistedlast.Connor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 Page7of19

PAGE 8

andcorrespondingsquarerootmatrix: H = 0.84 0.01 0.012.1 b H = 0.92 0.005 0.0051.5 b (4) Inequation4,theupperleftandlowerrightdiagonal elementsrepresentsmoothingintheE-WandN S directions,respectively.The H indicatesanactual E-Wsmoothingdistanceof920mandaN-Ssmoothing distanceof1500m.AN-Sellipticityisreflectedinthe overallshapeofthebandwidth(Figure3).Theresulting spatialdensitymapiscontouredinFigure4.Agrid-basedflowregimeTheSRTMdatabasefromCGI AR-CSI(theConsultative GrouponInternationalAgriculturalResearch-Consortium forSpatialInformation)isusedasamodeloftopographic variationontheShamiramPlateauandadjacentareas. Thisconsortium(Jarvisetal,2008)hasimprovedthequalityofSRTMdigitaltopographicdatabyfurtherprocessing version2(releasedbyNASAin2005)usinghole-filling algorithmsandauxiliaryDEMstofillvoidsandprovide continuoustopographicalsurfaces.Forthelavaflowsimulation,thesedataareconvertedtoaUTMZone38Nprojection,usingtheUSGSprogram,PROJ4,andre-sampled ata100100mgridspacing,usingthemappingprogram GMT.Inthemodel,lavaisdistributedfromone100m2gridcelltoitsadjacentgridcells. Theregionthatwaschosenforthelavaflowmodelis identifiedinFigure1(red-dashedbox).Withinthisarea anewventlocationisrandomlyselectedbasedona spatialdensitymodelof18eventsclusteredwithinand aroundtheShamiramPlateau(Figure4).Themodel simulatesaflowoflavafromthisnewventlocation ontothesurroundingtopography.Thetotalvolumeof lavatobeeruptedisspecifiedattheonsetofamodel run.LavaisaddedincrementallytotheDEMsurfaceat theventlocationuntilthetotalspecifiedlavaflow volumeisreached.Ateachiteration,105m3isaddedto thegridcellatthelocationofthevent(source)andis distributedoveradjoininggridcells.Giventhatagrid cellis100m2,thiscorrespondstoaddingatotaldepth of10mtotheventcellateachiteration. Thelavaflowsimulationisnotintendedtomimicthe fluid-dynamicsoflavaflows,sotheseiterationsareonly looselyassociatedwithtimesteps.Forexample,volumelimitedlavaflowsoftheShamiramPlateauaregenerally < 5kminlength,withvolumesontheorderof0.3-2.3 10-2km3.Thesevolumesandlengthsagreewellwithlavas fromcompilationsbyMalin(1980)andPinkertonandWilson(1994).Forsuchlavaflows,effusionratesof10-100 m3s-1areexpected(HarrisandRowland,2009).Using theseempiricalrelations,aniterationaddingavolumeof Table2VolcanicventsmappedontheShamiramPlateauEasting Northing 425507 4449732 425649 4449144 425992 4449400 425053 4449362 428682 4452894 429363 4452946 429504 4452711 429931 4452251 427322 4449676 427383 4449840 427835 4450008 428332 4444255 427386 4454344 427538 4453062 430618 4442102 427623 4452343 426857 4451520 425285 4454652Thelocationof18volcaniceventsusedinthespatialdensityanalysisof futurevolcanismontheShamiramPlateau,unitsareUTMmeters.Thesevent locationsareusedtodetermineacloser-to-optimaldata-drivenbandwidth. Figure3 Shapeofthekerneldensityfunction .Shapeofthe kerneldensityfunctionaroundasinglevolcanodeterminedusinga datasetof18volcaniccentersandtheSAMSEbandwidth estimationalgorithm,contouredatthe50th,84th,90thpercentiles. Note:theN-Selongationofthekernelfunctionreflectstheoverall patternofvolcanismontheShamiramPlateau. Connor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 Page8of19

PAGE 9

105m3oflavacorrespondstoanelapsedtimeof103104s. Lavaisdistributedtoadjacentcellsonlyateachiteration, sothiseffusionratecorrespondstoflow-frontvelocityon theorderof0.01-0.1ms1,inreasonableagreementwith observationsofvolume-limitedflow-frontvelocities.ParameterestimationforMonteCarlosimulationManysimulationsarerequiredtoestimatetheprobability ofsiteinundationbylava.Lavaflowpathsaresignificantly affectedbythelargevariabilityinpossiblelavaflow volumes,lavaflowlengths,andcomplextopography.A computingclusterisusedtoexecutethislargenumberof simulationsinatimelymanner.Basedonthevolumesof somelavaflowsmeasuredwithinandsurroundingthe ShamiramPlateau(Table1),therangeofflowvolumesfor thesimulatedflowswasdeterminedtobelog-normally distributed,withalog(mean)of7.2(107.2m3)andalog (standarddeviation)of0.5.Basedontheseobservations, Figure4 ModelforspatialdensityontheShamiramPlateau.Thespatialdensitymodelofthepotentialforvolcanismisshownforanarea aboutasite(ANPP),basedon18mappedvolcaniccenters(whitecircles,seeTable2).TheSAMSEestimatorisusedtogenerateanoptimal smoothingbandwidthbasedontheclusteringbehaviorofthevolcanoes.Contoursaredrawnandcoloredatthe5th,16th,33th,67th,84th,and 95thpercentileboundaries.Forexample,giventhatavolcaniceventoccurswithinthemappedarea,thereisa50%chanceitwilloccurwithin theareadefinedbythe1.7102km-2contour,basedonthismodelofthespatialdensity. Connor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 Page9of19

PAGE 10

thelavaflowcodestochasticallychoosesatotalerupted lavavolumefromatruncatednormaldistributionwitha meanof7.2,astandarddeviationof0.5,andtruncatedat 6and 9(Table3)).Thisrangefavorseruptionswith smaller-volumeflows,butalsoallowsrare,comparatively larger-volumeflows. Theinputparameterstothelavaflowcodethatare usedtoestimatetheprobabilityofinundationofthesite areshowninTable4.TheboundaryoftheANPPsiteis takenasarectangulararea,2.6km2.Forthepurposesof thesimulation,itisassumedthatifalavaflowcrosses thisperimeter,thesiteisinundatedbylava.Thelava flowsimulationisbasedontheeruptionofonelavaflow, orcoolingunit,fromeachvent.Basedonthedistribution offlowthicknessvaluesfrom15observedlavaflows, withinandsurroundingtheShamiramPlateau,thecode stochasticallychoosesavalu eformodallavaflowthicknessfromatruncatednormaldistributionhavingamean of7.0m,astandarddeviationof3.0m,andtruncatedat 4mand 15m(Figure5).Lavaresidualistheamount oflavaretainedineachactivecell,andisdirectlyrelated tothemodalthicknessofthelavaflow. Inreality,morethanonelavaflowmayeruptduring thecourseofformationanddevelopmentofasingle monogeneticvolcano.However,thefirstlavaflowto formduringthiseruptionwilltendtohavethelongest lengthandgreatestpotentialtoinundatetheANPPsite. Experimentswereconductedtosimulatetheformation ofmultiple(upto10)lavaflowsfromasinglevent,or groupofcloselyspacedvents.Itwasdeterminedthat thelaterlavaflowstendtobroadentheflowfield,but notlengthenit.Thisresultisinagreementwith observationsoflavaflowfielddevelopmentonMt.Etna (KilburnandLopes,1988).FortheANPPsite,theconditionalprobabilityofsiteinundationwassensitiveto lavaflowlength,butinsensitivetobroadeningofthe lavaflowfield.Therefore,onlyonelavaflowwassimulatedpereruptivevent.Nevertheless,forsomesitesthe potentialforbroadeningtheareaofinundationbysuccessiveflowsmaybeanimportantfactor.SimulationresultsAtotalof10000simulationswereexecutedinorderto estimatetheprobabilityoflavaflowinundationresulting fromtheformationofnewmonogeneticventsonthe ShamiramPlateau.Outof10000events,2485ofthe simulatedflowscrossedtheperimeterofthesite,or 24.9%percentofthetotalnumberofsimulations. Thedistributionofsimulatedventlocationsforthelava flowsimulationisshowninFigure6.Lavaflowserupting fromthecentralpartoftheShamiramPlateau,upto6km northoftheANPPsite,haveamuchgreaterpotentialof inundatingthesiteareathanlavaflowsoriginatingfrom south,east,orwestofthesite.Thecentralpartofthe Table3LavaflowsimulationinputparametersParameter RangeNotes ANPPsiteboundary Boundariesusedinanalysis East(km) 428.2 West(km) 426.0 North(km) 4449.0 South(km) 4447.0 Lavathickness(m)4-15Truncatednormaldistribution; Mean=7.0m StandardDev.=3.0m Lavaflowvolume(m3)106-109Truncatednormaldistribution; (log)Mean=7.2 (log)StandardDev.=0.5 Iterationvolume 105Lavavolumeaddedatsource ventineachiteration Numberofsimulations10000InputparametersusedintheMonteCarlosimulationoflavaflowinundation oftheANPPsitebyflowsoriginatingonorneartheShamiramPlateau.Flow thicknessandvolumearebasedonobservedthicknessesandvolumesoflava flowslocatedonandsurroundingtheShamiramPlateau.Aprobability distributionisassignedtoeachofthesetwoparametersbasedonthebinned distributionofmeasuredobservations(Figure5). Table4Configurationfileforlavaflowsimulationof ventsontheShamiramPlateauParameter=ValueExplanation Inputs DEM_SOUTH=4440N,S,E,W DEM_NORTH=4470boundaries DEM_EAST=440 oftheDEM DEM_WEST=410 DEM_SPACING=0.1DEMgridspacing(km) DEM_FILE=file (ASCIIformat) rowsofelevationvalues (masl) RESIDUAL_AV=8.0 Lavathickness(m):Average RESIDUAL_SD2=1.0StandardDeviation (highervalue = higherlavaviscosity) ERUPTED_LAVA=1e5Volumeoflavadistributed periterationorpulse(m3) TOTAL_LAVA_AV=1e7Lavavolume(m3):Average TOTAL_LAVA_SD2=0.5StandardDeviation FLOWS=1 Numberoflavaflowstosimulateper run RUNS=10000 Numberoflavaflowruns (forstatistical analysis) AOI_WEST=426.0 Areaofinterest AOI_EAST=428.2 AOI_SOUTH=4447.8 AOI_NORTH=4449.0 SPATIAL_DENSITY_FILE=fileXYZformat,gridofspatialdensity valuesforthepotentialofvolcanism SPATIAL_DENSITY_SPACING=.1spacingofspatialdensitygrid(km)Configurationfileforsimulatedlavaflows.TheformatofthisASCIIfileis parameter=value.Theshownvaluesreflecttherangeofvaluesusedforthe lavaflowhazardassessmentontheShamiramPlateau.Connor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 Page10of19

PAGE 11

ShamiramPlateauisthemostlikelylocationoffuture eruptions,basedonthespatialdensityanalysis.Substantial topographicbarrierstothesouth,east,andwestblocklava flowsfrominundatingthesitefromthesedirections,and theprobabilityofventformationintheselocationsis muchlower. Inordertotestmodelvalidityagainstavailablegeologic datafromtheregion,acomparisonwasmadeofmeasuredthickness,area,andl og(volume)versuslavaflow lengthforeachobservedlavaflow(Figure7).Thesame comparisonwasmadeforeachsimulatedlavaflow.Lava flowlengthforeachflow,simulatedandobserved,was calculatedasfollows.First,thelavaflowmid-pointwas estimatedalongE-Wlines egmentsdrawnacrossthe flowatregularintervals.Thedistancebetweenthese mid-pointswassummedalongtheN-Sextentofthelava flow.Second,thesameprocedurewasusedbutmidpointswerecalculatedalongN-Slinesegmentsandthe distancebetweenmid-pointswassummedalongtheE-W direction.Thelongerofthetwodistanceswastakento bethelengthofthelavaflow.Thismethodprovidedan objectivecomparisonbetweenobservedandsimulated flowlengths.AsshowninFigure7,thesimulatedlava flowvolumes,thicknesses,andarealextentsallfallwithin therangesofvaluesmeasuredinthefield. Larger-volumelavaflowsweresimulatedforflank eruptionsofAragatsvolcano.Forthesesimulationsatrachyandesitetotrachybasaltcompositionwasassumed. Thisflowregimemimicsaeffusionrate-limitedlava flows,withlavathicknesses(orlavaresiduals)ranging fromapproximately6 9m.Thisflowgeometryisconsistent,forexample,withtheTirinkatar-1,Ashtarak,and Paroslavaflows.Thetotalvolumesofthesesimulated flowsrangefromapproximately5108m3(0.5km3)to 8.7108m3(0.87km3).Anadditionalspatialdensity estimatewasmadetodefinetheprobabilityoffuture ventformationontheflanksofAragatsvolcano.This modelisbasedonthelocationsof27ventslocatedon theflanksofAragatsvolcano(Table5).Thisspatialdensityestimatewasusedtoinitializesimulatedlavaflows originatingfromflankventstoassessthehazardoflargevolume,effusionrate-limitedflanklavaflows.Sincethe detailsoftheseflanklavaf lowshavebeenverypoorly documented(only5havebeenclassifiedbythickness, volume,andlength)anaccuratestatisticalanalysisof theseparameterswasnotconsidered.Rather,valuesfor volumeandthicknesswererandomlyselectedfromthose trachyandesitetotrachybasaltflankflowsthatweremeasuredinthefield.Theconfigu rationparametersforthis flanklavaflowsimulationregimeisdetailedinTable6. Approximately1000flowsweresimulatedbasedona patternofvolcanismdefinedbythespatialdensitymodel showninFigure8.Theseflowsrequiredmorerun-time thanthesmaller-volumeShamiramflowsbecauseofthe Figure5 Histogramsshowinglavaflowthickness,volume,and log(volume) .Histogramsshowingtherangesofobservedand simulatedlavaflowthickness,volume,andlog(volume).Blackbins characterize15observedlavaflows.Flowthicknessfollowsanormal distributionandvolumefollowsalog-normaldistribution.These fieldobservationsaresummarizedinTable1.Redbinscharacterize 10000flowthicknesses,randomlyselectedfromatruncatednormal distributionwithameanof7andastandarddeviationof3, truncatedabove4mandbelow15m.Similarly,flowvolumes,were generatedbyrandomselectionoftheirlogarithmsfromatruncated normaldistributionwithameanof7.2andastandarddeviationof 0.5,truncatedabove6andbelow9(Table3).Theseplotsshowthat thedistributionschosenfortheMonteCarlosimulationreasonably matchtherangeofobservedvalues. Connor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 Page11of19

PAGE 12

greaternumberofgridcellsinundated.Noneofthe simulatedflowseruptedontheflanksresultedininundationoftheANPPsite.TheShamiramPlateaucreatesan effectivetopographicbarriertotheselavaflowsdiverting drainageoflavawestoreastoftheplateau.Therefore, althoughimpressiveinlengthandvolume,theANPPsite isnotlikelytobeinundatedbylonglavaflowsemitted fromtheflanksofAragatsvolcano.Sincetheselonglava flowsdonotrepresentacrediblehazardtotheANPP site,alargerMonteCarlosimulation(greaterthan1000 runs)andseparatestatisticalanalysisofeffusionrate-limitedlavaflowshighontheflanksofMt.Aragats,wasnot Figure6 Plotsoflengthversusarea,thickness,andlog(volume)forobservationsandsimulations .Plotsoflavaflowlengthversusarea, thickness,andlog(volume)includefieldobservations(graydots)andcomputersimulations(redpoints).Eachplotshowsresultsof10000lava flowsimulations,generatedusingtheprobabilitydistributionsshowninFigure5andspecifiedinTable3.Fieldobservationsof20lavaflowsare giveninTable1.Thelargestobservedlavaflowsplottotherightofthegrayline,marking > 20kmlength,justbeyondtherangeofthe simulatedvalues.These5flowswerenotconsideredwhendeterminingtheparameterrangesforthelavaflowsimulationsbecauselavaflowsof thislengthareeffusion-ratelimited,associatedwithveryinfrequentflankactivity,andnotfoundontheShamiramPlateau.Theseresultsshow thatthevolumes,thicknesses,andarealextentsofnearlyallobservedflowsfallwithintherangesofthesimulatedvalues. Connor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 Page12of19

PAGE 13

Figure7 ResultsofMonteCarlosimulationoflavaflowinundationofthesite .ResultsofMonteCarlosimulationoflavaflowinundation ofthesite(whitebox).VentlocationsforlavaflowsthatinundatedtheANPPsiteareshownasreddots.Bluedotsindicatetheventlocations fromwhichlavasdidnotinundatetheANPPsite.MostlavaflowsthatinundatethesiteoriginateonthecentralpartoftheShamiramPlateau, northoftheANPPsite. Connor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 Page13of19

PAGE 14

performed.Twoexamplesofe ffusionrate-limitedflows arediagrammedinFigure9.ConclusionsWedemonstrateamethodolo gyforsite-specificlava flowhazardassessment.Thistwo-stageprocessusesa two-dimensionalellipticalG aussiankernelfunctionto estimatespatialdensity.TheSAMSEmethod,amodifiedasymptoticmeansquarederrorapproach,usesthe distributionofknowneruptiveventstooptimallydetermineasmoothingbandwidthfortheGaussiankernel function.Potentialventlocations(N=10000)arestochasticallysampledfromthe resultingspatialdensity probabilitymap.Foreachrandomlysampledventlocation,alavaflowinundationmodelisexecuted.Lava flowinputparameters(volumeandmodalthickness)are determinedfromdistributionsfittofieldobservationsof thelowviscositytrachybasalttotrachydacitelavaflows ofthearea.Theareasandflowextents(aquantitative measureoflavaflowlength)ofthesesimulatedlava flowscomparereasonablywiththoseofmappedlava flows.Thisapproachyieldsaconditionalprobabilityof lavaflowinundation,giventheopeningofanewvent, andprovidesamapofventlocationsleadingtosite inundation. LavaflowhazardsexistattheANPPsitebecause potentialeruptionsontheShamiramPlateaumayproducelavaflowsthatinundatethesite.ThisMonteCarlo analysishasshownthat,giventhenumberofrelatively small-volumelavaflowsoccurringontheShamiramPlateau,approximately25%ofalleruptions,resultingfrom theformationofanewvent,mightalsoproducelava flowsthatinundatetheANPPsite.Althoughverylong andvoluminouslavaflowsoccurintheAragatsvolcanic system,thisanalysisdemonstratesthatthesetypesof flowsdonotpresentacrediblehazardforthesite,as thetopographyoftheShamiramPlateauwoulddivert suchpotentialflowsawayfromthesitearea. Table527MappedventsontheflanksofAragats VolcanoEasting Northing 430920 4485826 422295 4488512 414366 4498480 439898 4478024 440441 4476970 425896 4491003 421407 4471589 418534 4469462 408119 4495051 408990 4481638 414068 4471495 427253 4483296 424558 4482259 423136 4480327 411159 4469329 423682 4494414 405800 4477396 406683 4476948 418530 4494870 424111 4495248 408363 4492635 415964 4497175 422344 4491454 428042 4474090 428225 4474806 424775 4492714 399806 4491891Thelocationof27volcaniceventsusedinthespatialdensityanalysisof futurevolcanismontheflanksofAragatsvolcano,unitsareUTMmeters. Theseventlocationsdeterminethecloser-to-optimalbandwidthusingthe SAMSEbandwidthestimationmethod. Table6Configurationfileforsimulationoflavaflows fromflankventsParameter=ValueExplanation Inputs DEM_SOUTH=4441N,S,E,W DEM_NORTH=4482boundaries DEM_WEST=408 oftheDEM DEM_EAST=448 DEM_SPACING=0.1DEMgridspacing(km) DEM_FILE=file (ASCIIformat) rowsofelevationvalues (masl) MIN_RESIDUAL=1Maptoobservedflowthicknesses(m): MAX_RESIDUAL=41=6,2=7,3=8,4=9, (lowervalue=lowerlavaviscosity) ERUPTED_LAVA=1e6Volumeoflavadistributed periterationorpulse(m3) MIN_TOTAL_LAVA=1Maptoobservedflowvolumes(km3): MAX_TOTAL_LAVA=31=51082=5.3108,3=8.7108FLOWS=1 Numberoflavaflowstosimulateper run RUNS=1000 Numberoflavaflowruns AOI_WEST=426.0 Areaofinterest AOI_EAST=428.2 AOI_SOUTH=4447.8 AOI_NORTH=4449.0 SPATIAL_DENSITY_FILE=fileXYZformat,gridofspatialdensity valuesforthepotentialofvolcanism SPATIAL_DENSITY_SPACING=.1 spacingofspatialdensity grid(km)ConfigurationfileforsimulatedlavaflowsfromtheflanksofAragatsvolcano. TheformatofthisASCIIfileisparameter=value.Theshownvaluesreflect therangeofvaluesusedforthelavaflowsimulationforhazardassessment fromaflankeruptiononAragats.Connor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 Page14of19

PAGE 15

Anintegratedhazardassessmentalsodependsonthe estimationoftherecurrencerateofeffusivevolcanism. Assumingarecurrencerateofeffusiveeruptionsonthe ShamiramPlateauof4.110-7yr-1and3.510-6yr-1, basedoncurrentlyavailableradiometricagedeterminations(Chernyshevetal,2002),theannualprobabilityof siteinundationbyrenewedeffusivevolcanismonthe ShamiramPlateauisapproximately1.010-7to8.8107. Figure8 Spatialdensitymodelfor27eventsontheflanksofAragatsvolcano.Thespatialdensitymodelofthepotentialforvolcanismis shownforanarealocatedabovetheANPPsite(blackbox),basedon27mappedvolcaniccenters(whitecircles)locatedontheflanksof Aragatsvolcano.TheSAMSEestimatorisusedtogenerateanoptimalsmoothingbandwidthbasedonthisclusteringofvolcanicvents.Contours aredrawnandcoloredatthe5th,16th,33th,67th,84th,and95thpercentileboundaries.Thisspatialdensitymodelwasstochasticallysampledfor ventlocationsforlavaflowsimulationontheflanksofAragatsvolcano.TheblacktrianglemarksthelocationofthesummitofAragats. Connor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 Page15of19

PAGE 16

Figure9 TwosimulatedlargevolumelavaflowsonthesouthflankofAragatsvolcano .Simulatedlarge-volumeflowsoriginatinghigher uptheflanksofAragatsvolcanodivertaroundthetopographicbarrierpresentedbytheShamiramPlateau.Theselavaflowsaresimulatedwith aofvolume0.5km3andathicknessof3m,similartotheTirinkatar-1andAshtaraklavaflows(Table1)).TheANPPsiteisindicatedbytheblack box. Connor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 Page16of19

PAGE 17

MethodsSpatialdensityanalysisThelavaflowhazardassessmentbeginswithaspatial densityanalysisinvolvingthelocationsof18volcanic eventslocatedontheShamiramplateau.Thisanalysis willhelpdeterminethemostlikelylocationsoffuture volcaniceventswhichwillthenbecomethesourcelocationsforpossiblelavaflows.Theseeventsarelistedin Table2.Usingthese18eventsanoptimalbandwidthis determinedusingtheSAMSEmethodinthe ks packagewithinthestatisticalprogram, R .Therequired R commandsarethefollowing: library(ks) vents18<-read.table( events_zoom. wgs84.z38.utm ) show(vents18) bw_samse_18vents<-Hpi(x=vents18, nstage=2,pilot= samse ,pre= sphere ) show(bw_samse_18vents) where ks isthenameofthe R packageneededto performtheanalysis, vent18 isalocal R variableholdingventlocations, vents 18_ wgs84.z38.utm istheinput textfilecontainingtheventlocations(eastingand northingseparatedbyaspace), bw samse 18vents isa local R variableholdingtheoutputfromthe Hpi routine,thebandwidthmatrixinmeters: [,1][,2] [1,]844328.34-13235.75 [2,]-13235.752113393.17 SpatialdensityanalysisisaccomplishedusingaPERL script(seeAdditionalfile1).Parametersforthescript areinserteddirectlyatthetopofthescriptasshownin thefollowingcodesection: #################################### ######################### #INPUTSECTION:Thesevariablescanbe adjustedbytheuser ###################################### ###################### ##Thisisthecompletesetofevents: #events_zoom.wgs84.z38.utm :N=18 <425053/430618><4442102/4454652> my$west=420000; my$east=436000; my$south=4439000; my$north=4463000; my$Grid_spacing=100; #ThebandwidthmatrixviaSAMSE2-stage pre-transformation sphering #units=squaremeters,for18events nearANPP #[,1][,2] #[1,]844328.34-13235.75 #[2,]-13235.752113393.17 #units=squarekilometers my$H=pdl[ [.84432834,-.01323575], [-.01323575,2.11339317] ]; #Theinputfileofeventlocations my$in= events_zoom.wgs84.z38.utm ; # Th e outputfileforthespatialintensitygrid my$out1= spatial_density_samse_events_zoom.wgs84.z38.utm.2 ; where, $north,$south,$east,$west arethemapboundariesinUTMmeters, $Grid spacing isthemapgridspacing,unitsinkm, $H isthekernelbandwidth,units convertedtokm2, $in isthenameoftheinputfileofvolcaniceventlocations(ASCII format:eastingnorthing), and $out1 isthenameoftheoutputfileofthespatial densitygrid(ASCIIformat:eastingnorthingdensity). $HisamatrixanditsstructureinthescriptiscontrolledbythePERLpackage pdl .The4valuesforthe matrixarederivedfromtheoutputofthe Hpi routine (asnotedabove).Torunthescriptfromthecommand linetype: perlgausXY.pl where gausXY.pl isthenameofthescript.Allparametersareinserteddirectlyatthetopofthescriptas indicatedabove. AsecondPERLscriptdrivesthelavaflowsimulation (seeAdditionalfile2andAdditionalfile3).Theinputs forthisscriptarecontainedinaconfigurationfile.To runthecodefromthecommandlinetype: perllava_flow.pllavaflow.conf0 where lava_flow.pl isthenameofthescript, lavaflow.conf isthenameoftheconfigurationfile,and 0 is thestartingrunnumber.Eachrunofthescriptsimulatesonecompletelavaflowsimulation.Thetotal numberofsimulatedlavaflowsissetintheconfigurationfile.Theconfigurationfileparametersarelistedin Tables4and6. ThePERLlavaflowsimulationscriptproduces3outputfiles: lava_flow_stats.# Thisfileisacompilationofall simulatedlavaflows(where # referstotheinitialrun number).Thistextfilecontains6columns: Easting(units=km) Northing(units=km) Hit(1=hit;0=miss) TL(units=cubicmeters) Residual(units=meters) Total(units=cubicmeters)Connor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 Page17of19

PAGE 18

where Easting and Northing refertothelocationof theeruptingvent, Hit iseither1or0,where1means thatthelavaflowpenetratedtheareaofinterest(i.e.the boundaryofthesite)and0indicatesthatitdidnot, TL isthetotalvolumeoferuptedlava, Residual referstoa flow smodalthickness,and Total isalsothetotal volumeerupted,butcalculatedinadifferentway.The totalnumberoflavaflowsimulationsarerecorded. flow.#.utm Thisfilerecordsthegridlocationand thicknessoflavaineachinundatedcell(where # refers toanindividualrunnumber).Thistextfilecontains3 columns: XYthickness ,where XYreferstotheinundatedgridcell,and thickness referstothethickness(m) oflavainthatcell.Thisfileisusedtocalculatethe lengthandareaofeachsimulatedlavaflow. vents.utm Thistextfilerecordstheventlocationof eachlavaflowsimulation.Thefilecontainstwocolumns: EastingNorthing .AdditionalmaterialAdditionalfile1:PERLscriptthatestimatesspatialdensity .This codedependsoninputsgeneratedbytheSAMSEbandwidthestimation routinefromthe ks librarypackageaspartofthe R programming package.ThisPERLscriptisanASCII(text)filethatcanbeviewedwith anytexteditor.Itisrunfromthecommandline:perl. Additionalfile2:PERLscriptthatsimulatesvolume-limitedlava flowsfromventsonandaroundtheShamiramPlateau.Thislava flowscriptdependsontheoutputspatialdensitygridfilegeneratedby theabovementionedspatialdensityscript(additionalfile1).Itisan ASCIIfilethatcanbeviewedwithanytexteditor.Itisrunfromthe commandline:perl. Additionalfile3:Perlscriptthatsimulateseffusionrate-limitedlava flowsfromventslocatedontheflanksofAragats .Thislavaflow scriptdependsontheoutputspatialdensitygridfilegeneratedbythe abovementionedspatialdensityscript(additionalfile1).ItisanASCIIfile thatcanbeviewedwithanytexteditor.Itisrunfromthecommand line:perl. Acknowledgements Theauthorsgratefullyacknowledgethelogisticalandtechnicalsupportof StaffattheInstituteofGeologicalSciencesofArmenianNationalAcademy ofSciences.DiscussionswithArkadiKarakhanianregardingArmenian geologyandfieldmappinggreatlyenhancedtheauthors overall understandingofthegeologicalsettingofArmenia.Reviewsofearlyresults ofthisstudybyBrittHill,WillyAspinall,andAntonioGodoy,allrepresenting theInternationalAtomicEnergyAgency,ledtoimprovementsinthe methodspresentedhere.Thisresearchwaspartiallysupportedbyagrant fromtheUSNationalScienceFoundation(DRL0940839).ReviewsbyBritt HillandAntonioCostaimprovedthemanuscript. Authordetails1UniversityofSouthFlorida,4202E.FowlerAve,Tampa,FL33620,USA2InstituteofGeologicalSciencesofArmenianNationalAcademyofSciences, Yerevan,Armenia3SchoolofEarthandEnvironment,TheUniversityofLeeds, Leeds.LS29JT,UK Authors contributions LJCwrotespatialdensityandlavaflowinundationcomputercodesand carriedoutlavaflowsimulations.CBCconceivedofthestudyand participatedincodedevelopmentandanalysis.LJCandCBCdraftedthe manuscript.KMandISmappedlavaflowsontheShamiramPlateau, developedthedatasetonlavaflowparameters,andprovidedrelated geologicalandgeochemicaldata.Allauthorsreadandapprovedthefinal manuscript. Competinginterests Theauthorsdeclarethattheyhavenocompetinginterests. Received:20June2011Accepted:25January2012 Published:25January2012 References1.AbrahamsonN(2006)Seismichazardassessment:Problemswithcurrent practiceandfuturedevelopments.FirstEuropeanConferenceon EarthquakeEngineeringandSeismology,Geneva,Switzerland 2.BarcaD,CrisciGM,GregorioSD,NicolettaF(1994)Cellularautomatafor simulatinglavaflows:AmethodandexamplesoftheEtneaneruptions. TransportTheoryandStatisticalPhysics23:195 232 3.BebbingtonMS,CroninSJ(2010)Spatio-temporalhazardestimationinthe Aucklandvolcanicfield,NewZealand,withanewevent-ordermodel. BulletinofVolcanology73(1):55 72 4.CappelloA,VicariA,DelNegroC(2011)Assessmentandmodelingoflava flowhazardonMt.Etnavolcano.BollettinodiGeofisicaTeoricaed Applicata52(2):10 20 5.ChernyshevIV,LebedevVA,ArakelyantsMM,JrbashyanR,GhukasyanY (2002)GeochronologyoftheAragatsvolcaniccenter,Armenia:Evidence fromK-Ardating.DokladyEarthSciences384(4):393 398.(inRussian) 6.ConditCD,ConnorCB(1996)Recurrencerateofbasalticvolcanismin volcanicfields:AnexamplefromtheSpringervilleVolcanicField,AZ,USA. GeologicalSocietyofAmericaBulletin108:1225 1241 7.ConnorCB,ConnorLJ(2009)Estimatingspatialdensitywithkernel methods.In:ConnorC,ChapmanN,ConnorL(ed)VolcanicandTectonic HazardAssessmentforNuclearFacilities.CambridgeUniversityPresspp 331 343 8.ConnorCB,HillBE(1995)ThreenonhomogeneousPoissonmodelsforthe probabilityofbasalticvolcanism:ApplicationtotheYuccaMountainregion. JournalofGeophysicalResearch100:12107 10125 9.ConnorCB,StamatakosJA,FerrillDA,HillBE,OfoegbuGI,ConwayFM, SagarB,TrappJ(2000)Geologicfactorscontrollingpatternsofsmallvolumebasalticvolcanism:Applicationtoavolcanichazardsassessmentat YuccaMountain,Nevada.JournalofGeophysicalResearch105(1):417 432 10.ConnorCB,SparksRSJ,DezM,VolentikACM,PearsonSCP(2009)The natureofvolcanism.In:ConnorCB,ChapmanNA,ConnorLJ(ed)Volcanic andTectonicHazardAssessmentforNuclearFacilities.CambridgeUniversity Presspp74 115 11.CostaA,MacedonioG(2005)Computationalmodelingoflavaflows:A review.GeologicalSocietyofAmericaSpecialPapers396:209 218 12.DelNegroC,FortunaL,VicariA(2005)Modellinglavaflowsbycellular nonlinearnetworks(CNN):preliminaryresults.NonlinearProcessesin Geophysics12:505 513 13.DiggleP(1985)Akernelmethodforsmoothingpointprocessdata. AppliedStatistics34:138 147 14.DragoniM(1989)Adynamicalmodeloflavaflowscoolingbyradiation. BulletinofVolcanology51:88 95 15.DragoniM,TallaricoA(1994)Theeffectofcrystallizationontherheology anddynamicsoflavaflows.JournalofVolcanologyandGeothermal Research59:241 252 16.DuongT(2007)Kerneldensityestimationandkerneldiscriminantanalysis formultivariatedatainR.JournalofStatisticalSoftware21(7):1 16 17.DuongT,HazeltonML(2003)Plug-inbandwidthselectorsforbivariate kerneldensityestimation.JournalofNonparametricStatistics15:17 30 18.GriffithsRW(2000)Thedynamicsoflavaflows.AnnualReviewofFluid Mechanics32:477 518 19.HarrisAJL,RowlandSK(2001)FLOWGO:akinematicthermo-rheological modelforlavaflowinginachannel.BulletinofVolcanology63:20 44 20.HarrisAJL,RowlandSK(2009)Controlsonlavaflowlength.In:Thordarson T,SelfS,LarsenG,RowlandSK,HskuldssonA(ed)StudiesinVolcanology, TheLegacyofGeorgeWalker,SpecialPublicationsforIAVCEINo.2,The GeologicalSociety,London,33 51 21.HillBE,AspinallWP,ConnorCB,GodoyAR,KomorowskiJC,NakadaS(2009) RecommendationsforassessingvolcanichazardsatsitesofnuclearConnor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 Page18of19

PAGE 19

installations.In:ConnorCB,ChapmanNA,ConnorLJ(ed)Volcanicand TectonicHazardAssessmentforNuclearFacilities.CambridgeUniversity Presspp566 592 22.HornikK(2009)TheRFAQ.http://CRAN.R-project.org/doc/FAQ/R-FAQ.html. ISBN3-900051-08-9 23.IAEA(2011)Volcanichazardsinsiteevaluationfornuclearpowerplants. DraftSafetyGuideDS405.InternationalAtomicEnergyAgency,Vienna, Austria 24.JaquetO,ConnorCB,ConnorLJ(2008)Probabilisticmodelingforlong-term assessmentofvolcanichazards.NuclearTechnology163(1):180 189 25.JarvisA,ReuterHI,NelsonA,GuevaraE(2008)Hole-filledseamlessSRTM data.InternationalCentreforTropicalAgriculture(CIAT),v4.availablefrom http://srtm.csi.cgiar.org 26.KilburnCJR,LopesRMC(1988)ThegrowthofaalavafieldsonMt.Etna, Sicily.JournalofGeophysicalResearch93:14,759 14,722 27.KilburnCJR,LuongoG(1993)ActiveLavas:MonitoringandModelling.UCL Press,London,UK 28.KiyosugiK,ConnorCB,ZhaoD,ConnorLJ,TanakaK(2010)Relationships betweentemporal-spatialdistributionofmonogeneticvolcanoes,crustal structure,andmantlevelocityanomalies:AnexamplefromtheAbu monogeneticvolcanogroup,southwestJapan.BulletinofVolcanology71. doi:10.1007/s00,445-009-0316-4 29.LeBasM,LeMaitreRW,WoolleyAR(1986)Theconstructionofthetotal Alkali-Silicachemicalclassificationofthevolcanicrocks.Mineralogyand Petrology46(1):1 22 30.LuhrJ,SimkinT(1993)Paricutin:theVolcanoBorninaMexicanCornfield. GeocsiencePress,Inc,Phoenix,AZ 31.MalinMC(1980)LengthsofHawaiianlavaflows.Geology8:306 308 32.MartinAJ,UmedaK,ConnorCB,WellerJN,ZhaoD,TakahashiM(2004) Modelinglong-termvolcanichazardsthroughBayesianinference:example fromtheTohokuvolcanicarc,Japan.JournalofGeophysicalResearch 109(B10208).doi:10.1029/2004JB003,201 33.MiyamotoN,SasakiS(1997)Simulatinglavaflowsbyanimprovedcellular automatamethod.ComputersandGeosciences23:283 292 34.MolinaS,LindholmCD,BungumH(2001)Probabilisticseismichazard analysis:Zoningfreeversuszoningmethodology.BollettinodiGeofisica TeoricaetApplicata42(1-2):19 39 35.PinkertonH,WilsonL(1994)Factorscontrollingthelengthsofchannel-fed lavaflows.BulletinofVolcanology56:108 120 36.RowlandSK,GabrielH,HarrisAJL(2005)LengthsandhazardsofchannelfedlavaflowsonMaunaLoa(Hawaii),determinedfromthermaland downslopemodelingwithFLOWGO.BulletinofVolcanology67:634 647 37.SilvermanB(1986)DensityEstimationforStatisticsandDataAnalysis.No. 26inMonographsonStatisticsandAppliedProbability,ChapmanandHall 38.SilvermanBW(1978)Choosingthewindowwidthwhenestimatinga density.Biometrika65:1 11 39.StasiukMV,JaupartC(1997)Lavaflowshapesanddimensionsasreflections ofmagmasystemconditions.JournalofVolcanologyandGeothermal Research78(1-2):31 50 40.StockC,SmithEGC(2002)Comparisonofseismicitymodelsgeneratedby differentkernelestimations.BulletinoftheSeismologicalSocietyofAmerica 92:913 922 41.ThordarsonT,SelfS(1993)TheLaki(Skaftarfires)andGrimsvotneruptions in1783-85.BulletinofVolcanology55:233 263 42.VolentikACM,ConnorCB,ConnorLJ,BonadonnaC(2009)Aspectsof volcanichazardassessmentfortheBataannuclearpowerplant,Luzon Peninsula,Philippines.In:ConnorCB,ChapmanNA,ConnorLJ(ed)Volcanic andTectonicHazardAssessmentforNuclearFacilities.CambridgeUniversity Presspp229 256 43.WadgeG,YoungPAV,McKendrickIJ(1994)Mappinglavaflowhazards usingcomputersimulation.JournalofGeophysicalResearch99:489 504 44.WalkerGPL(1973)Lengthsoflavaflows.PhilosophicalTransactionsofthe RoyalSociety,London274:107 118 45.WandMP,JonesMC(1995)KernelSmoothing.No.60inMonographson StatisticsandAppliedProbability,ChapmanandHall 46.WardSN(1994)Amultidisciplinaryapproachtoseismichazardinsouthern California.BulletinoftheSeismologicalSocietyofAmerica84:1293 1309 47.WooG(1996)Kernelestimationmethodsforseismichazardareasource modelling.BulletinoftheSeismologicalSocietyofAmerica88:353 362doi:10.1186/2191-5040-1-3 Citethisarticleas: Connor etal .: Probabilisticapproachtomodeling lavaflowinundation:alavaflowhazardassessmentforanuclear facilityinArmenia. JournalofAppliedVolcanology 2012 1 :3. Submit your manuscript to a journal and bene t from:7 Convenient online submission 7 Rigorous peer review 7 Immediate publication on acceptance 7 Open access: articles freely available online 7 High visibility within the eld 7 Retaining the copyright to your article Submit your next manuscript at 7 springeropen.com Connor etal JournalofAppliedVolcanology 2012, 1 :3 http://www.appliedvolc.com/1/1/3 Page19of19