originalseasonalwinds

TrmMctionsinGIS,1997,d.
2,no.
I,p.
25Anintegratedregionalizationofearthquake,flood,anddroughthazardsinChinaJINFENGWANG,STEPHENWISE,ANDROBERTHAlNlNGJinfengWangStateKeyLaboratoryofResourcesandEnvironmentalInformationSystems,InstituteofGeography,ChineseAcademyofSciences,Beijing100101,China.
e-mail:wangjf@www.
Ireis.
ac.
cnStephenWiseDepartmentofGeography,UniversityofShefield,S102TN,UnitedKingdom.
RobertHainingDepartmentofGeography,UniversityofShefield,S102TN,UnitedKingdom.
Earthquake,flood,anddroughtdatafromdifferentsourcesarecombinedinasingledatasetusingthesamedatastructure,projection,andscale.
Theintensityandfrequencyofeachhazardisclassifiedintosevere,heavy,modest,andlight,producingadassificationwith64combinedstatesforthethreekindsofhazard.
Theseclassesarethenrankedaccordingtoseverity.
Thethreehazardcoveragesarcoverlaidandthepolygonsthatareproducedarecodedbytheclassificationsystem.
Amapisproducedthatshowsthedistributionofthese64classesinregionsandtheirareasmeasuredfromthespatialtopologicaldatafileintheGIs.
Spatialanalysisrevealsthespatialassociationamongthethreehazardsandbetweenthethreehazardsandhumanfactors.
Thereisabriefdiscussionoftheimplicationsoftheregionalizedmapforhazardmonitoring.
1IntroductionHazardscanbedefinedas'extremeeventswhichmayaffectdifferentplacessinglyorincombinationatdifferenttimes'(Blaikieetal1994:21)andnaturalhazardsas'thoseelementsofthephysicalenvironmentharmfultoManandcausedbyforcesextraneoustohim'(BurtonandKates1964quotedinSmith1992:8).
Strictlyspeaking,anextremeeventcanonlybecalledahazardifithasthepotentialtocausedeathordamagetohumans-asSmith(1992:9)notes:'asevereearthquakeinaremoteunpopulatedregionisanextremenaturalevent,ofinteresttoseismologistsandnomore'-butitiscommonpracticetousethetermhazardtorefertoanyextremeeventwhichhasthepotentialtocausedamageandthisistheusagewhichwillbeadoptedhere.
Hazardsoccurwithvaryingdegreesoffrequencyandseverityandalthoughitisoftendifficulttopredicttheoccurrenceofanindividualhazardevent,itispossibletodeterminethestatisticallikelihoodofpatternsofoccurrenceintimeandspace.
Whennaturalhazardsoccurinornearpopulatedareas,thisinformationcanbeusedtocalculatetheriskforthepopulationconcernedofsufferingdamageasaconsequenceofahazardousevent(Cutter1993).
Riskismeasuredinrelationtothenumberofpeopleaffectedbyandthefrequencyofoccurrenceofthehazardousevent.
When'significantnumbersofvulnerablepeople(peopleatrisk)experienceahazardandsufferseveredamageand/ordisruptionoftheirlivelihoodsysteminsuchawaythatrecoveryisunlikelywithoutexternalaid(Blaikieetal1994:21)thenadisasterissaidtohaveoccurred.
Theoccurrenceofadisasteristhusaconsequenceoftheinteractionbetweenphysicalprocessesandhumansystems.
Whetherornotadisasteroccursdependsontheseverityofthenaturalevent(thehazard)inrelationtothehumansystemthatexistswithintheareawheretheeventoccurs.
Chinasuffersfrommanydisasterscausedbynaturalhazards,especiallyearthquakes,floods,anddroughts,losingonaverage10billionRh4B(US$1billion)ofpropertyannually(or1-2percentoftotalannualGNP).
Thedeathtollreached240000intheTangshanearthquakein1976;23million1361-1682/97/2-01-0250PcarsonProfessiondLimited199725-IWhg,SWise,andRHoininghectaresareaffectedbydroughteachyear,12millionhectaresofwhicharepopulated;11millionhectaresarecoveredbyfloodeachyear,7millionhectaresofthisareabeingpopulated(People'sRepublicofChina1994).
TheseeventsareaconsequenceofChina'slocationatthejunctionofthehugeEurasianlandmassandthePacificOcean:thestrongthermalgradientbetweenlandandoceaninducesseasonalwindsthatareresponsibleforthefrequentfloodsanddroughts;thecollisionbetweentheEurasianandPacifictectonicplatesareresponsibleforthefrequentearthquakes.
Thesethreemajorhazardscausemanyotherkindsofassociatedproblemssuchaslandslides.
ThemagnitudeandfrequencyofthesehazardswhichimpactonlargenumbersoftheChinesepopulationaresuchthatdisasterreliefisanimportantandurgenttaskinChina.
Wherehazardsposeathreattohumanlifeorproperty,itisimportanttomonitorandwherepossiblepredictthem,thusminimizingtheriskofdisaster(Alexander1993).
Threemajorapproachestotheproblemmaybeidentifiedintheliterature,withInformationTechnology(IT)andGISplayinganincreasingroleinallthree(GatrellandVincent1991;Rhind1991):(1)predictingtheoccurrenceofhazards;(2)generalriskassessment;and(3)developingimprovedtoolstomanageadisastershoulditoccur.
Thefirstapproachisthepredictionoftheoccurrenceofparticularevents.
Insomecases,suchasfloods,apredictionmaybemadewithsomeaccuracybaseduponmonitoringoffactorssuchasrainfallandsoilmoisturelevels.
Indeed,weatherpredictiongenerallyisoneareawheremoderncomputers,coupledwithimproveddatacollectionfromsatellites,haveledtoenormousimprovementsintherangeandaccuracyofpredictions(Tyler1989).
Someevents,suchasearthquakes,canonlybepredictedbythecontinuousmonitoringofknownprecursorssuchasearthmovements,andthishasbeengreatlyaidedbythedevelopmentofremotedatacollectionviatelemetry(Alexander1991).
Theemphasishereisonreal-timemonitoringandnumericalanalysis,socurrentlyGISdoesnotfeatureverymuchinthiswork.
Thesecondapproachisthemoregeneralassessmentofrisk,asopposedtothepredictionofindividualevents.
Thisisimportantforplanningdefencestrategiesforexistingsettlementsandforlocatingnewdevelopmentsawayfromhazardousareas.
Itinvolvestwoelements-themappingofthelikelyintensityofthehazard,followedbythecombinationofthiswithinformationonthelocationofpopulationoreconomicactivities,andincreasinglyGISisplayingaroleinboththeseareas.
Hazardintensitymaybeestimateddeductively,baseduponaknowledgeoftheprocessesatwork,orinductively,basedonananalysisoftheoccurrenceofthehazardintheregion.
Inbothcases,itisthespatialvariationinintensitywhichisimportantandGISisthereforeofgreatpotentialbenefit.
Thedeductiveapproachisonlypossiblewheretheunderstandingoftheprocessesatworkissufficient,asinthecaseoflandslidehazard(vanWestenandTerlien1996),flooding(Mejianavarroeta11994),andgroundwatercontamination(Merchant1994),andinmanycasesinvolvesacombinationofcomputermodellingandGIs.
TheinductiveapproachbenefitsfromtheabilityofGIStobringtogetherinformationonawidevarietyofpossibleexplanatoryfactorswhichmaythenbeusedtoderiveempiricalmodelsorprovidesurrogatesforhazardlevelswheremeasurementsaresparse.
Thisapproachhasalsobeenusedfortheanalysisoflandslides,especiallyatregionalscaleswherethemodellingofindividuallandslidelocationswouldnotbefeasible(McKeanetal1991;Dikauetal1996).
AkeyelementofthisapproachistheuseofGIStointerpolatefrompointmeasurementstoprovidedataestimatesforwholeareas,asinthecaseofairqualityandsoilcontaminationmeasurements(SenguptaandVenkatachalam1994;Szucs1995).
Finally,GIShastheabilitytooverlaytheestimatedvaluesonthehazardintensitywithinformationonpopulationdistribution,thusidentifyingtheareasmostatrisk(Masliaetal1994;SenguptaandVenkatachalam1994;Hiscocketal1995;Lowryetal1995).
Thethirdapproachisthedevelopmentofimprovedtoolsforthemanagementofthesituationwhenadisasterdoesactuallyoccur.
Thisnormallyinvolvestheorganizationofwhateverresourcesarerequiredtodealwiththedisaster(emergencyservices,specializedequipment,etc)andtheorganizedevacuationoftheaffectedarea.
Bothofthesehaveastrongspatialelement,andgreatinterestisbeingshowninspatialdecisionsupportsystems,basedaroundGIS,toassistinthesetasks(Alexander1991;GatrellandVincent1991;Wadgeeta11993).
R~imdimtionofhazardsearthquakesInsomecasestheremaybegoodreasonstoexpectrelationships-forexample,earthquakeswilloftenleadtolocalizedfloodingbecauseofthedamagetoflooddefensesystems.
Inothercases,andespeciallyattheverybroadscale,theassociationmaybepurelyanaccidentofgeography-forexamplethecoastalregionofChinaliesclosetothetectonicallyactivezone,butisalsotheareamostaffectedbymonsoonstorms.
Tohelpdesignasamplingstrategyforhazardmonitoring.
Despitetheadvancesinautomateddatacollection,especiallywiththeuseofremotesensing(Alexander1991),thecollectionofdataonnaturalprocessesstillreliesonpointsampling.
Thismeansthatthelocationofthestationsisanimportantissue,sinceitwouldnotonlybeexpensivetoattempttomonitornaturalhazardsinthiswayoverawholecountrysuchasChina,butinefficientaswell,leadingtooversamplinginareasoflowrisk.
Itmakessensetoconcentratemonitoringeffortsinareaswhichsufferhighlevelsofriskfrommorethanonehazard,bothbecauseofthepotentialeconomiesandbecauseofthegreaterpotentialfordisasterinsuchareas.
Thesubjectofthispaperfallswithinthesecondoftheseapproaches-theassessmentofhazardlevelsatanationalscaleforChinaforthreenationalhazards-earthquake,flood,anddrought.
ThistypeofanalysishasbeenperformedforanumberofcountriesincludingtheUnitedKingdom(Perry1981),Japan(Nakano1974),theUSSR(GerasimovandZvonkova1974),andItaly(Alexander1987),butinmostcaseseachhazardistreatedseperately,withnoattempttoproduceanassessmentofthevariationinriskfromallhazardsforthecountry.
However,asAlexander(1993)pointsout,foranyonelivinginanarea,thedegreeofrisktheyaresubjecttoarisesfromthecombinationofallpotentialhazards.
Theeffectisnotsimplyadditive(themorehazardsthemorerisk)butmultiplicativesincesomehazardsaremoredamagingincombinationthansingly,forexamplecoldtemperaturesarefarmoredamagingwhenalliedwithstrongwinds(GerasimovandZvonkova1974)andsomehazardsactuallycauseothers,asinthecaseofearthquakesandtsunami(Alexander1991).
ThefocusonasinglehazardwasalmostuniversaluntiltheworkofHewittandBurton(1971)inLondon,Ontario,whoattemptedtoassessthelevelsofriskfortheinhabitantsfromallpotentialhazards.
Thisapproachhassincebeenappliedelsewhere(Cooke1984),butatanationalleveltheonlyattempttoapplythesameideashasbeentheworkofGerasimovandZvonkova(1974),althoughPerry(1981)makespassingreferencetotheimportanceoftryingtodothis.
GerasimovandZvonkova(1974)brieflyreviewedthepatternofawiderangeofhazardsintheUSSR,andthenproducedamapinwhichthewholeareawasgroupedintofourclasses,rangingfromareassufferingcatastrophiceventscausinglossoflifeanddamage,throughtoareassufferingonlylocalhazardssuchasfrostandfog.
Unfortunately,nodetailsareprovidedofhowtheseclassesaredefined,orofhowtheavailabledatawasanalysedinordertoproducethemap,whichshowstheUSSRdividedinto29regionsonthebasisofthisclassification.
Therearetworeasonswhyitisofinteresttoattempttoproduceanintegratedregionalizationofhazards:1Toassessthedegreetowhichtheoccurrenceofhazardsisspatiallyrelated:forexample,doareaswhichsufferseverefloodingalsosufferfrom2Chinaisfortunateinhavingaverylongrecordofdatacollectionontheoccurrenceofmanynaturalhazardswhichcanbeanalysedtoproduceestimatesoftheprobabilityofhazardsofagivenmagnitudeoccurringatanygivenlocation.
Thisanalysishasalreadybeencarriedoutinthecaseofearthquakes,floods,anddroughtandpublishedintheformofamapforeachhazardforthewholeofChina.
Themainaimoftheworkreportedherehasbeentocombinethesethreesetsofdataintoanintegratedregionalization,whichshowsthelevelofallthreehazardswhichmightbeexperiencedacrossthecountry.
ThisinformationisthencombinedwithinformationonpopulationandeconomicactivitytoproduceabroadassessmentofriskacrossthewholeofChina.
Wefirstdescribetheoriginaldatawhichwasusedfortheprojectandthenthetechniquewhichwasusedtoproducetheregionalization.
Thereliabilityoftheregionalizationisdiscussedbeforeitsuseinvarioustypesofanalysisisdescribed.
27JWong,SWse,andRHaining2ThedatabaseusedinconstructingtheregionalizationInconsideringnaturalhazards,itisimportanttodistinguishtwoaspectsoftheoccurrenceofhazardevents:theirmagnitudeandtheirfrequency.
Thetwoarecloselyrelated,sinceeventsoflargemagnitudetendtooccurlessfrequentlythanthoseofsmallmagnitude.
Fromdataforasinglerecordingstationitispossibletoconstructamagnitude/frequencycurvefromwhichthefollowingtwoquestionscanbeanswered:1Howfrequentlywillaneventofsizexoccur(Normallyexpressedastherecurrenceinterval,ornumberofyearsbetweenoccurrencesofeventsofthismagnitude.
)recurrenceintervalofyyears2HowlargewillaneventbewhichhasaInundertakingananalysisofriskatasinglepoint,thefirstquestionisoftenofinterest-forexampleifitisknownthatexistingflooddefenceswillwithstandfloodsuptoacertainmagnitude,itisimportanttobeabletoestimatethelikelihoodofthembeingbreached.
Inlookingatthespatialvariationofhazardmagnitude,itismorecommontoselectarecurrenceinterval(e.
g.
50years)andlookatthevariationinthemagnitudeofeventsofthisfrequencyacrossthearea.
Asdescribedabove,ourinitialdatasourceswerethreemapseachshowingthepatternofoccurrenceofonehazardacrossthewholeofChina-twoshowthevariationinmagnitudeofeventsatagivenrecurrenceintervalwhilethethirdcombinesmeasuresofmagnitudeandfrequency.
Themapsarenotreproducedhere,butwedescribetheirderivationbaseduponpersonalcommunicationsfromtheirauthors(whosenamesaregivenintheacknowledgementssection).
2.
1FloodsTheoriginaldataonfloodingweretakenfrom460gaugingstations,eachwitha500-yearrecord.
Suchalengthofrecordmeansthatestimatingthemagnitudeofthe50-yearfloodisaprocessofinterpolationratherthanextrapolation(aswouldbethecaseinmanyothercountries).
Itiswellknownthatflooddischargeishighlycorrelatedwith28catchmentarea(GregoryandWaling1972),andsothe50-yearflooddischargewasdividedbythecatchmentarea.
Thisanalysisproducedaseriesofpointestimatesoffloodmagnitude,whichweretheninterpolatedmanuallytoproduceanisolinemapofthemagnitudeofthe50-yearfloodacrossChina.
Themanualinterpolationallowslocal,expertknowledgeoffactors,suchastopographyandbroadscaleclimaticpatternswhichaffectfloodmagnitude,tobeusedindrawingthecontours.
Thecontourintervals(100,500,1000,2000,and4000m3s-'h2)werechosenbasedonaknowledgeofthedamagecausedbyfloodsofdifferentmagnitudes,andhencerepresentimportantbreaksbetweenminorandmajorfloodevents.
2.
2EarthquakesTheoriginaldatabaseconsistedofrecordsfrom2000recordingstations,eachwitha20-yearrecordplushistoricalrecordsspanningover4000years.
Fromthis,theintensityofanearthquake(measuredontheRichterscale)witha10-yearrecurrenceintervalwasestimatedforeachareaof1degreeoflatitudeby1degreeoflongitude.
2.
3DroughtTheoriginaldataweretakenfromtheChinaYearbook,whichrecordsdroughtincidencesataprovinciallevel.
Inalltherewere40yearsofdataforeachofthe31provincesofChina.
Provinceswereclassifiedintermsofbothfrequencyandseverityofdroughtstodistinguishbetweenareaswithfrequentdroughts(anaverageofatleastoneayearovera40-yearrecord)andlessfrequentdroughts,andbetweenareaswithextremedroughts(croplossofmorethan80percentoccurringmorethan20timesin40years)andlessextremedroughts.
Thismeantthateachprovincefellintooneoffourpossibleclassesbasedonthefourposssiblecombinationsofdroughtfrequencyandseverity.
Provincialboundariesarenotverysuitableforreportingdroughtoccurrence,sincetheactualareasaffectedwillbedeterminedbytopographicandclimaticfactorswhichwillspantheseboundaries.
Therefore,localknowledgeoftopographicandclimaticfactorswasusedtodelimitthosepartsofChinawhichfellintoeachofthefourclasses,producingamapinwhichtheareasarefelttobeamoreaccuratereflectionofdroughtoccurrencethanamapbasedonprovincialboundaries.
RagionalimfionafhazardsmeasuredinquitedifferentunitsintoasinglemeasureforanareaisaproblemdiscussedbrieflybybothPerry(1981)andHewittandBurton(1971)whobothconcludethatthebestapproachisonebasedonestimatingtheintensityofeachhazardwhichislikelytocauseasimilaramountofdamageordisruption.
InthecaseofthedataforChina,theauthorsoftheoriginalmapshadchosentheisolineandclassintervalsbasedonanunderstandingoftheeffectsofhazardsatdifferentlevels,anditwasthuspossibletogroupthehazardintensityvaluesoneachmapintofourclasses-Severe,High,Medium,andLight.
TheactualclassintervalsusedforeachmapareshowninTable1-inthecaseoffloodandearthquaketheintervalsarebasedonmagnitudeatasetrecurrenceinterval,inthecaseofdroughtonacombinationofmagnitudeandfrequency.
Eachoftheoriginalmapscanbethoughtofasafield(avariablewhichvariescontinuouslyacrossspace)althoughtherearedifferencesbetweenthem.
Floodmagnitudeisavariablewhichisonlymeaningfulalongthecourseoftherivers-atallpointsinbetween,themagnitudeofthe50-yearfloodisameaninglessquantity.
Thereforethefloodmapisinfactatrendsurfacemapandtheisolinescannotbetakenasliteralestimatesoftheintensityofthe50-yearfloodatallpoints(atopicwhichwereturntolateron).
Earthquakeanddroughtintensitycanbeconceptualizedascontinuousfieldswithavalidvalueatallpointsacrossthesurface,butagaingiventhescaleoftheanalysisthemapsareperhapsbestthoughtofastrendsurfacemaps.
ThethreemapsbetweenthemmadeuseofthreeofthesixmethodsusedtorepresentfieldsinGIS-isolinesforthefloodmap,gridcellsforearthquakes,andpolygonsfordrought(Kemp1993).
Beforecombiningthethreemaps,weusedtheLATTICECONTOURcommandinARC/INFOtoproduceaninterpolatedversionoftheearthquakedatathatwasmorecomparablewiththeothertwomaps.
Clearlythisintroducesanotherelementofestimationintotheprocedure,andadifferentcontouringalgorithmwouldprobablyproduceaslightlydifferentresult.
However,ithastheeffectofremovingtheobviousartefactsofgridcellsfromthefinalmap.
ThetwoisolinemapswerethenreclassifiedaccordingtothevaluesinTable1,toproducepolygonmapsshowingtheareasfallingintoeachofthefourclasses,andthesewereoverlaidwiththe2.
4ThemapsEachmapisthereforearepresentationofthespatialvariationintheintensityofthenaturalhazardinquestion.
Inthecaseoffloodandearthquakethishasbeenaccomplishedbymappingtheintensityofeventsatadefinedrecurrenceinterval,whilethedroughtmapshowsvariationsinbothmagnitudeandfrequency.
Notethatthemapsshowthevariationinthenaturalevents-theydonotnecessarilyshowthefullextentoftheareawhichmightbeaffectedbyanindividualcovereventsuchasasingleflood.
Thisisparticularlytrueinthecaseoffloods,whereafloodeventwillcoveramuchgreaterareainthelowlandsthaninamountainousregion.
However,thisisnotaseriousproblemgiventheverybroadscaleoftheanalysisanditisreasonabletoassumethatthesemapsindicatethosepartsofChinawheretheintensityofeachnaturalhazardispotentiallyhighest.
3ConstructionoftheregionalizationTheaimoftheprojectwastocombinethesethreesetsofdatatoproduceonemapof'hazardregions'(i.
e.
regionssufferingsimilarlevelsofallthreehazards)whichcouldbeusedwhenconsideringthedesignofasamplingschemeforhazardmonitoring.
Therewerethreestagestothis,allofthemsuitableforimplementationinGIs:1conversionofthemapstoacommonmapprojection2combinationoftheinformationfromthethreemapsandconstructionofaregionalization3combinationofthemapofhazardregionswithdataattheprovinciallevelonpopulationandlevelsofeconomicactivity.
Thefirststageoftheanalysiswasrelativelystraightforward.
AllthreemapsweredigitizedandinputintotheARC/INFOGIS(EnvironmentalSystemsResearchInstitute,Redlands,California).
Twoofthemaps(floodanddrought)werebasedontheAlbersconicprojection(lowerlatitudetangentof30"N,upperlatitudetangent40"N,andacentrallongitudelineof105"E)whiletheearthquakedatawerereportedonalatitude-longitudegridbasis.
ThiswasthereforeconvertedtoAlbersconicusingARC/INFO.
Thecombinationofinformationonhazards29JWang,SWise,ondRHainingEarthqwk.
(l)PwkFlood(m3/rxkm)Y-)[70,1201[40000,899991W-V)[60,691[Zoo00,399991M(odamto)(50,591[5000,199991L[W)[o0.
491[00,49991NahrFreq=Frequency,ExiDry=ExtremelydryTable1.
Fourlevelsofintensityofearthquake,flood,anddrought.
hughFreq>=ltime/ywr8,ExtDry>=1time/2yearFreq=1time/2yeorFreq>=ltime/ywr8,ExtDryc1time/2yearFreqosite-d.
ag=06i41(b)EarthquakeanddroughtDroughtDroushfAm5HMLAreaS+HM+LS001600004300073CI344EarthquakeH0048700085003C202457S+H0077504215M00156000610060'20222L00000000250006501977M+L0024204767Arearatiomain-diog=02825opposite-oiag=01746Arearotiomon-diog=05543opposite-diog=04457(c)FloodanddroughtD-whWhtAmSHMLAreaS+HM+LS00027000000003500!
8@FloodH00184000040048101832S+H0021602537M00312000710061'0150GL0061700229004iO03486M+L0I22906018Arearatioman-diog=04131opposite-diog=01358Arearotiomoil-diag=06234oppsite-atag=03766Table6.
Thespatialassociationbenvccnearthquakc,flood,anddrought.
Regionolizdionofbrdsby2tables;Table6(c)showsthatthedistributionoffloodhasanassociationwithdrought.
Theassociationisdominatedbythelargecommonareawithmediumandlightintensityforbothfloodanddroughthazards.
sensitivetheinferencetheoryistotherequirementthattherowandcolumntotalsareequal.
AccordingtoNorcliffe(1373,somerelaxationispossibleandthisisanareawhichdeserveshrtherresearch.
Theproportionofthemapwherethelevelsoftwohazardsare'identical'isobtainedfromthefollowingmeasure:(sumofthediagonalvaluesinthe4by4table)Tocheckfor'similarity'ofassociationcollapsethe4by4tableto2by2(mergeS&Hcounts,andmergeM&Lcounts)andthenrepeat:(sumofthediagonalvaluesinthe2by2table)(AgainitappearsthatCourt'smethodcannotbeusedsincetherowandcolumntotalsdonotsumto0.
5asthelevelscannotbemanipulatedtoapproximatethemedian.
)TheresultsarelistedinTable6.
Forthe4by4and2by2tables,theexpectedvaluesoftheproportionsassuminganequaldistributionare4/16=0.
25and2/4=0.
5respectively.
Thelargertheproportionofthemaindiagonalthestrongerthetendencyforspatial(positive)associationbetweenthetwohazards,andthelargertheproportionoftheoppositediagonalthestrongerthetendencyforspatialseparation(negativeassociation)betweenthetwohazards.
Therelativemagnitudesofvalueswithinindividualcolumns(rows)arealsoindicativeofpatternsofassociationatgivenlevelsofhazard.
Table6(a)showsthatthereisnoapparentspatialassociationbetweenearthquakeandfloodfromthe4by4table,butthereisevidenceofsomenegativeassociationfrominspectingthe2by2table.
Severeearthquakesandfloodstendnottooccurinthesameareas;Table6(b)showsthatthedistributionofearthquakestendstoassociatepositivelywithdroughtfromboththe4by4and26.
3SpatialcorrelationbetweenhumanfactorsandhazardsThepopulation(POP),grossnationalproduct(GNP),provincialinvesrment(IN)andtotalprovincialarea(AREA)areavailableforeachoftheprovincesofChina(seeAppendix2:TableA2).
Ifweassumethatthethreesocio-economicfactors(POP,GNP,andINV)areuniformlydistributedwithinaprovincethentheamountofthefourfactorsthataresubjecttodifferentlevelsofhazardindifferentprovincesandinthewholecountrycanbeestimated.
TheprovincialtotalsaregiveninTableA1(inAppendix2)whilstTable7showsthenationalfigures.
Foramoreaccurateassessmentinthecaseofthethreesocio-economicfactorsitwouldbenecessarytousemoredetailedinformationontheintra-provincialdistributionofpopulationandeconomicactivity.
Fromtheprovincialleveldata(Appendix2)thePearsoncorrelationbetweenpopulation,investment,GNP,andthehighestlevelofhazardwithrespecttoalltypesofhazard(level1inTable2)wasevaluated.
Thehazardlevelsarecoded,forpurposesofcorrelationassevere(4)downtolight(1).
ThePearsoncorrelationisonlyindicativeoftheassociationbecausethehazardvariableisonlymeasuredattheordinalscale.
TheresultsaregiveninTable8.
Thisshowsthatthethreehumanfactorsarepositivelycorrelated.
ThereisanegativecorrelationhoweverbetweeneachofthethreehumanfactorsandtheseverityofhazardswhichmeansthatthehazardlevelsappeartobemuchHadArwGNPINVPOPArea%GNP%INVXPOPYlevel(1Wbn2J(mRMB)(mWBJ(10OOO)Severe21828883708832539328316123261724High438341158346558259060939546494952Modemk17835772320827058924295919232321Light10463097016343083645711333Sum9457GO32070001i985001'77360100100100100Table7.
Humanfactorsoverriskareas(1-level)inChina.
37IWong,SWise,ondRHaininghigherinremoteareaswithlesshumanactivitythaninareaswithdensehumanactivity.
ThissuggeststhatthedistributionofhumanactivitywithinChinahas,withsomeimportantexceptionssuchasBeijingprovince,adaptedtothedistributionofhazardsattheprovincialscale.
However,giventhesimplisticassumptionsmadeaboutthedistributionofpopulationandeconomicactivitytheseresultsmustbeviewedwithconsiderablecaution,andmoreworkneedstobedoneifamorerobustestimateistobemade.
Itispossible,forexample,thateveninprovincesthatexperiencehighlevelsofhazard,theintra-provincialdistributionofhumanactivityislocatedawayfromthehazardousareas,sothattheassumptionsmadeherewouldunderestimatetheextenttowhichhumanactivityavoidsareasofseverehazard.
6.
4ImplicationsoftheregionalizationforhazardmonitoringWebrieflyconsidertheimplicationsoftheregional-izationforhazardmonitoringthroughthesettingupofasamplingnetwork.
Samplingisunavoidablewhenmonitoringanyaspectofthenaturalenviron-mentsinceitwillneverbepossibletomonitoranynaturalprocesscontinuouslyintimeandspace.
Itisthereforeimportantthatthesamplingstrategyadoptedisdesignedonthebasisofsomeknowledgeorunderstandingofthedistributionofthenaturalprocessesatworksothatthemaximuminformationcanbederivedfromagivennumberofsamplingsites.
Therearealsodistinctadvantagestosampling.
Itreducescosts,itshouldfacilitatetherapidcollectionofthemostvaluabledata(whichcanbeofparticularimportancewhendealingwithnaturaldisasters)andbyconcentratingresourcesithelpstoensurethatthedataareofhighquality.
Ifthevolumeofdatatobeprocessedisreducedthroughsampling,thenpersonnelofhigherqualitycanbeemployedandgivenintensivetrainingandthisshouldleadtomoreaccuratedataprocessing.
Theregionalizationprovideslargescaleinformationonhowmonitoringsitesmightbedistributed(orhowresourcesformonitoringmightbedistributedbetween,forexample,provinces).
Theissueofexactsitingofmonitoringsiteswilldependonthetypeofhazardandanunderstandingoftheprocessmechanismsgivingrisetohazardevents(e.
g.
thelocationoffaultlinesforearthquakes).
Itisarguablethatifresourcesarelimitedmonitoringshouldberestrictedtoareaswherehazardlevelsaresevere(oratleasthigh).
TheevidenceofTables4and5aresuggestiveofwhere(ataprovinciallevel)monitoringresourcesmightbeconcentratedifthecriteriaarebasedonwhetherhazardlevelsaresevere.
Forexample,anareaofnorth/north-eastChina(includingtheprovincesofBeijing,Tianjin,Hebei,andShaanxi)fallswithinthiscategory,whereastheareafurthernorthandincludingtheprovincesofInnerMongolia,Liaoning,Jilin,andHeilongjiangarerelativelyfreeofhazards.
TheprovincesofHenan,Hubei,Hunan,andGuangdonginanareaofSouthernChinaalsogenerallyhavelowlevelsofhazard.
Otherfactorswillalsoneedtobeconsideredinrelationtoresourcedistributionforhazardmonitoring.
Theoverallpopulationdistributionwillbeimportantsincetheremaybelittlepurposeinintensivelymonitoringareasoflowpopulation.
Inaddition,iftheregionalizationisspatiallyhighlyfragmentedthenmoremonitoringsitesmaybeneededinanareathaniftheregionalizationgenerateslargeareas(becausemoreregionsmeanmorestrata).
Themorespatiallyautocorrelatedthelevelofahazard,themoreinformationaboutanareaanyonemonitoringsiteprovides.
TheextenttowhichmonitoringeffortcanbereducedasafunctionofthespatialautocorrelationinthePopulationONPIWOrtnnnt0.
55560.
9508Population0.
7487ONPNohrValuesorePeorson'sproduct-momentcorrelationco-efficientTable8.
Spatialcorrelationbetweenhumanfactorsandhazards.
38Hazardlevel0.
27220.
29320.
2803Regiondizatitmcdkdsrepresentstratathatjustifydifferentlevelsofmonitoringbecauseofthepotentialseriousnessofthehazardeventsineachregion.
Thedetailsofanyintra-provincialallocationofresourcesandthesubsequentgeographicallocationofmonitoringsitesare,however,topicsrequiringfurtherresearch.
variabletobemonitoredisdiscussedinRipley(1981)andHaining(1990).
7ConclusionsThispaperhaspresentedasingle,unifiedclassificationofearthquakes,floods,anddroughts.
Thecomprehensiveregionalization,builtusingGIs,showsthelarge-scalespatialdistributionandareasofthehazardsandtheirintensitylevels.
Majorhazardareasand'islandsofsafety'areillustratedintheregionalizationmap.
ThereliabilityoftheintegratedregionalizationdependsonthereliabilityoftheoriginalregionalizationmapsofindividualhazardsandtheeffectsofGISmanipulations.
WhilstGIShandlingdoesnotappeartoproduceseriouserrorstotheareastatistics,theuseofsharpboundariesratherthanzonesoftransitionoverstatestheextenttowhichclearlydefinedhazardregions,distinctfromconditionsinneighbouringareas,existinreality.
Evenatthisbroadscaleofanalysisthisisanareaofresearchwhichjustifiesfurtherconsideration.
Thespatialanalysisontheregionalizationmaprevealedthatwhilstsevereearthquakesandfloodsdonotoccurinthesameareas,severeearthquakesanddroughtsfrequentlyhappeninthesameregions(althoughnotnecessarilysimultaneously).
Thereisstrongspatialassociationbetweenfloodsanddroughts.
Furthermore,hazardsaremoreintenseinremoteareaswithlesshumanactivitythaninareaswithdensehumanactivity,indicatingthatattheprovincialscaleofanalysis,humanactivitieshaveadaptedtothegeographicaldistributionofhazards.
Thisseemsentirelyreasonable,andsuggeststhatthespatialdistributionofhazards,atleastatthisscale,hasbeenfairlyconstantovertime.
Asnotedabove,moredetailontheintra-provincialdistributionofhumanactivitywouldgiveaclearerpictureoftheextenttowhichadaptationhasinfacttakenplace.
Theregionalizationsuppliesaframeworkfortheconstructionofamonitoringnetworkofthenaturaldisastersonanationwidescale.
Othertypesofhazardcouldbeintegratedintotheregionalizationusingtheproceduresdescribedinthepaper.
TheultimatepurposeofthisresearchistoimprovehazardmonitoringinChina,andasacontributiontothisobjectivetheworkreportedhereprovidesafirstindicationofhowresourcesformonitoringmightbeallocated.
ThehazardregionsAcknowledgementsThisworkwassponsoredbygrantsfromtheEUinternationalscientificco-operationprogrammeandtheStatePlanningCommitteeofChinaandwasundertakenwhilstJinfengWangheldaMarieCurieFellowshipintheDepartmentofGeographyattheUniversityofSheffield.
Thefollowingindividualsprovidedtheoriginaldatasetsonwhichtheworkreportedherewasbased:XinlianChen,SeismicAnalysisandPredictionCenter,StateSeismicBureauofChina,Beijing100080,China(earthquakedata);QichenTang,InstituteofGeography,ChineseAcademyofSciences,Beijing100101,China(flooddata);andKeranLi,InstituteofGeography,ChineseAcademyofSciences,Beijing100101,China(droughtdata).
JinfengWangwouldliketothankShupengChenforhishelpfuladvice.
ARC/INFOwastheGISusedfortheworkreportedhere.
Twoanonymousrefereesprovidedextremelyhelpfuladviceonthepresentationofthisworkandtheauthorswishtorecordtheirappreciation.
ReferencesAlexanderD1987Landofdisasters.
GeographicalMagazine,59:22631AlexanderD1991Informationtechnologyinreal-timeformonitoringandmanagingnaturaldisasters.
ProgressinPhysicalGeography,15:238-60AlexanderD1993NaturalDisasters.
London,UniversityCollegeLondonPressAltmanD1994Fuzzysettheoreticapproachesforhandlingimprecisioninspatialanalysis.
InternationalJournalofGeographicalInformationSystem,8:271-89BlaikieP,CannonT,DavisI,andWisnerB1994AtRisk.
London,RoutledgeBurroughPA(1992)AreGISdatastructurestoosimplemindedComputersandGeosciences,18:39540039IWang,SWise,ondRHainingBurroughPAandFrankAU(eds)1996GeographicObjectswithIndeterminateBoundaries.
London,Taylor&FrancisBurtonI,KatesRW,andWhiteGF1978TheEnvironmentasHazard.
NewYork,OxfordUniversityPressChrismanNR1989Modellingerrorinoverlaidcategoricalmaps.
InGoodchildMandGopalS(eds)AccuracyofSpatialDatabases.
London,Taylor&Francis:21-34CookeRU1984GeomovphologicalHazardsinLosAngeh.
London,AllenandUnwinCourtA1970Mapcomparisons.
EconomicGeography,46:435-8CutterSL1993LivingwithRisk.
London,EdwardArnoldDikauR,CavallinA,andJagerS1996DatabasesandGISforlandslideresearchinEurope.
Geomorphotogy,15:227-39GatrellA1987Onputtingsomestatisticalanalysisintogeographicalinformationsystems:Withspecialreferencetoproblemsofmapcomparisonandmapoverlay.
ResearchReportNo.
5NorthernRegionalResearchLaboratory.
Lancaster,UniversityofLancasterManagingnaturalandtechnologicalhazards.
InMasserIandBlakemoreM(eds)HandlingGeographicalInformation.
Harlow,Longman:148-80GerasimovIPandZvonkovaTB1974NaturalhazardsintheterritoriesoftheUSSR:Study,control,andwarning.
InWhiteGF(ed)NaturalHazards.
NewYork,OxfordUniversityPress:243-51GoodchildMandGopalS(eds)1989AccuracyofSpatialDatabases.
London,Taylor&FrancisGregoryKJandWalingDE1973DrainageBasinFormandProcess.
London,EdwardArnoldHainingRP1990SpatialDataAnalysisintheSocialandEnvironmentSciences.
Cambridge,CambridgeUniversityPressHainingRandArbiaG1993ErrorpropagationthroughMapoperations.
Echnometrics,35:HewittKandBurtonI1971TheHazardnowofaPlace-ARegionalEcoloaofDamagingEvents.
Toronto,UniversityofToronto,DepartmentofGeographyHiscockKM,LovettAA,BrainerdJS,andParfittJP1995GroundwatervulnerabilityGatrellAandVincentP199193-305assessment-twocasestudiesusingGISmethodology.
QuarterlyJournalofEngineeringGeology,28:179-94KempKK1993EnvironmentalmodellingandGIs:Dealingwithspatialcontinuity.
InKovarKandNachtnebelH(eds)ProceedingsHydroGIS93.
Wallingford,InternationalAssociationofHydrologicalSciences,Publication211:107-15LeungY1987Ontheimprecisionofboundaries.
GeographicalAnalysis,19:125-51LewisP1977MapsandStatistics.
London,MethuenLowryJH,MillerHJ,andHepnerGF1995AGIS-basedsensitivityanalysisofcommunityvulnerabilitytohazardouscontaminantsontheMexico/USborder.
PhotogrammetricEngineeringandRemoteSensing,61:MaffiniG,ArnoM,andBirterlichW1989ObservationsandcommentsonthegenerationandtreatmentoferrorindigitalGISdata.
InGoodchildMandGopalS(eds)AccuracyofSpatialDatabases.
London,Taylor&Francis:55-67MasliaML,AralMM,WilliamsRC,SustenAS,andHeitgerdJL1994Exposureassessmentofpopulationsusingmodelling,demographicanalysis,andGIs.
WaterResourcesBulletin,30:1025-41McKeanI,BueckelS,andGaydosL1991Remotesensingandlandslidehazardassessment.
PhotogrammetricEngineeringandRemoteSensing,57:1185-93MejianavarroM,WohlEE,andOaksSD1994Geologicalhazards,vulnerabilityandriskassessmentusingGIS-modelforGlenwoodSprings,Colorado.
Geomorphology,10:MerchantJW1994GIS-basedgroundwaterpollutionhazardassessment-acriticalreviewoftheDRASTICmodel.
PhotogrammetricEngineeringandRemoteSensing,60:1117-28NakanoT1974Naturalhazards:ReportfromJapan.
InWhiteGF(ed)NaturalHazard.
NewYork,OxfordUniversityPress:23143NorcliffeG1977InferentialstatisticsfirGeographers.
London,HutchinsonPeople'sRepublicofChina,StateStatisticalBureau19941794StatisticalYearbookofChina.
1347-59331-5440RegionalimtionofhamrdsPerryBeijing,ChinaStatisticalPublishingHouseAH1981EnvironmentalHazardintheBritishIsles.
London,George,Allen,andUnwinRejeskiD1993GISandrisk:Athreecultureproblem.
InGoodchildMF,ParksB0,andSteyaertLT(eds)EnvironmentalModellingwithGIs.
NewYork,OxfordUniversityPress:318-31RhindDW1991Environmentalmonitoringandprediction.
InMasserIandBlakemoreM(eds)HandlingGeographicalInformation.
Harlow,LongmanRipleyB1981SpatialStatistics.
NewYork,JohnWiley&Sons,Inc.
SenguptaSandVenkatachalamI'1994HealthhazardassessmentinanindustrialtownwiththehelpofGRAM-GIs.
EnvironmentalMonitoringandAssessment,32:155-60SmithK1992EnvironmentalHazard-AssessingRiskandReducingDisaster.
London,RoutledgeSzucsM1995SoiltraceelementsavailabilitymapsofHungary.
Computers,Environments,andUrbanSystems,19:117-22TylerC1989Climatology-technology'sanswertotheBritishweather.
GeographicalMagazine,61:44-7vanWestenCJandTerlienMTJ1996AnapproachtowardsdeterministiclandslidehazardanalysisinGIS-acasestudyfromManizales,Colombia.
EarthSu$aceProcessesandLandfom,21:853-68Errormodellingforthemapoverlayoperation.
InGoodchildMandGopalS(eds)AccuracyofSpatialDatabases.
VereginH1989WadgeZadehLondon,Taylor&Francis:3-18G,WislockiA,PearsonEJ,andWhittowJB1993Mappingnaturalhazardswithspatialmodellingsystems.
InMatherP(ed)GeographicalInfirmationHandling-ResearchandApplications.
Chichester,JohnWiley&Sons,Ltd:239-50LA1965Fuzzysets.
InformationandControl,8:338-53Appendix1Provinceacronyms11BJ-Beijing12TJ-Tianjin13HB-Hebei14SX-Shanxi15NM-InnerMongolia21LN-Liaoning22JL-Jilin23HLJ-Heilongjiang31SH-Shanghai32JS-Jiansu33ZJ-Zhejiang34AH-Anhui35FJ-Fujian36JX-Jiangxi37SD-Shandong41HEN-Henan42HUB-Hubei43HUN-Hunan44GD-Guangdong45GX-Guangxi46W-Hainan51SC-Sichuan52GZ-Guizhou53YN-Yunnan54XZ-Tibet61SAX-Shaanxi62GS-Gansu63QH-Qinghai64NX-Ningia65XJ-Xinjiang71TW-Taiwan41IWq,Swise,ondRHoiningAppendix2PIPOP-11747.
95123399913310.
3714343681526.
782122233132333435363767408414243444546512221525379294546162636488.
2865580.
8281.
7366.
7284.
76138.
1115.
69.
013.
79196.
482.
68467.
28269.
95113.
6353.
38308412.
9876.
7622.
173624.
46896.
46258.
9482.
1119.
2434.
4469674172.
3349.
7773.
6826.
811691.
35362.
62131.
9333.
217.
122.
59913.
25195.
871.
245.
842.
48338.
5273.
831.
3422.
324.
872.
07301.
3265.
6927.
9392.
65175.
6568.
19346.
5154.
9860.
163377.
961511.
1586.
6236.
052.
231078.
78283.
71104.
8379.
9223.
657.
26863.
911011.
04336.
13716.
63194.
33210.
869.
584447.
751205.
7332.
91872.
61482.
823.
921.
163470135.
253.
561235.
2365.
73354.
63833.
78166.
6917.
567.
87216.
15259.
81341.
56402.
4455.
83233934.
391390.
9213.
2976.
46914.
14155.
07612.
09210.
5610.
5663.
291274137.
7442.
9978.
01138.
53185.
3716.
5684.
399.
62459.
1262.
5422.
86464.
07114.
79181.
4383.
445.
1722.
1634.
6616.
4519.
2741.
1741.
816.
53212.
335.
26367.
5119.
9527.
764.
91668.
3208.
99160.
828.
4490.
4520.
77303.
557.
35336.
8164.
443877.
7894.
5430.
37965.
7170.
3495.
4511.
44404.
468.
94165.
329.
56.
710.
4643.
281.
3768.
998.
346.
2115.
6183.
4626.
8722.
4850.
492.
8727.
3210.
334856191988243589969582785246265205867172374291662431953096107629097651475521332119742601048I135633292171607122238339745123433964503646888411588191903979221669214693236119108734105812931242234701785785817066399919217778.
66446.
6137.
21242.
36185.
21201.
61150.
92358.
02309.
12815.
86137.
329.
35263.
3347.
7466.
1411.
56119.
3933.
834.
247.
65101.
41TableAl:Areaofhumanfactorsfallingineachhazardregion(level2)brokendownbyprovince.
42(:5152799.
I78'38ji;,3316384649968I8072LA60970412678I0509191635713239355122302599103983084283061235935694611140232933875I4'58424833857383134070118251415632852457504186289992366887451690351881116330II1013134915166424119147102191806332093147847021394494482132436253627837258763375198I86502XY5L37066108682556094521813258WC7148656247,3353237I819138205678423442219394I369775517243511'6662I1753481313471573839483991478212(166279531518628317512581833295I44WX11IAAAl5055122232182764079716A84483028462612592893655285'j4,:080491683335I5pop-illpp-1im-Ill95012356791120244I1037347891463381867552516964221645175462413886732817355211275pl11121314152122233132333435363741424344454651525354616263616543IMng,SWise,andRHainingAppendix2I'12I314152122233'3233343536374142434445465152535461626364651112864928536633415673012646223248740421808255567236401077134915!
2696727544266169858979803150102839667028642270289491583565312986311119266072225443878870:225111041959340940838856622323734436152345358467106L959816054824192264522352077022483316421117782291371196892464388325I6372721675611c226218215c74644254AREA(1aoOarq.
w168113160'561200'4018046062i00100130120160150I67180210175230347560170390120020045072060160011355LI57E810311788761L4182113616124237742737611777222510649623911175796124134012061601257I1072622591070213719817473121162112111776I1010669757147614141784111511116339315432916011012001698722271161468614872544767924069637601961111156I53511046381I'0612153142131271394142414L;8260215364312'1356342452272214123623623102:713218933138621393435364411I3739I47741424326144692198II54546493345165519952468187539485447716120758181624466307216446117365NokrA-SSdenotesthesizeofareawithSSseveritylevelsinaprovince,ctcAblankmeans10recordTableA2.
Spatialdistributionofhumanfactorsandhazardseverity(Level2)byprovince(dataonprovinciallevelsofhumanfactorsfromPeople'sRepublicofChina1994StatisticalYearbookofChina).
44