DEVELOPMEN TAN DAPPLICATIO NO FSPATIALL YPARAMETERIZE D DEPT HDURATIO NFREQUENC YMODE LFO RESTIMATIO NO FDESIG N RAINFAL LFO ROROMI AREGIONA LSTATE ,ETHIOPI A

Abstract

DepthDurationFrequency(DDF)relationshipsar ecurrentlyconstructedbase donatsite frequenc yanalysisofrainfalldat aseparatelyfordifferentdurations.Theserelationshipsar e notaccurat eandreliablesincetheydependonassumptionssuc ha sdistributionselection foreac hduration;theyrequirealarg enumberofparameters,experienceintensive equationsandregionalizationisalsoverypoorandcoarse.Inthisstudy,scalingproperties ofextremerainfalldepthserie sar eexaminedtoestablishscalingbehaviorofstatistical momentsandquantileestimatesoverdifferentdurations.Theannualextremeserie sof precipitationmaximaforstormdurationrangingfromO.Sh rto24hrobserveda tnetworkof raingauge ssitedi nOromiaregionalstat ewereanalyzedusinganapproachbasedon moments.Theanalysisinvestigatedth estatisticalpropertieso frainfallextremesand detectedthatth estatisticsofth erainfallextremesfollowsapowerlawrelationwithits iduration.Moreover,thevariationsofth edistributionparameterswithdurationsofannual • jmaximumrainfalldepthserie swereexploredandfoundthatth elogEVl ,EV landlogistic !distributionparametersexhibitapowerlawrelationshipwithdurations.Followingth e analysis,scal einvarianceofextremerainfalldepthserie sisinvestigatedanddissipative 'i(multiplescaling)natureofextremerainfalldepthserie sisconsidered,thu sintroducinga !generaldistributionfree frameworktodevelopDepthDurationfrequency(DDF)model.A ^DepthDurationFrequency(DDF)modelwithgriddedse tofparametersisdevelopedfor estimationofpointrainfallfrequenciesforarang eofdurationforan ylocationi nOromia nationalandregionalstate .ADDFmodelwasfittedtoserie sofannualmaximaandits parametersweredeterminedbyaleas tsquare smethodandthes eparameterswere interpolatedandmappedona1kmgrid.Themodelallowsforaparsimoniousandefficient parameterizationofDDFrelationships,anditsperformanceisshowntoimprovethe reliabilityandrobustnes so fdesignstormpredictionsa scomparedwiththos eachievableby interpolatingthequantilepredictionsofextremerainfalldataforspecificdurations. Moreover,designrainfallestimatesfoundfromthescalingDDFmodelar ecomparableto estimate sobtainedfromtraditionaltechniques ;however,th escale dapproachwasmore efficientandgivesmorereliableestimatecomparedwithth eobservedrainfalldeptha tall stations.