Page 130 - 2024年第55卷第2期
P. 130
imagedatawasusedtocalculatetheinundationfrequencyandvegetationfrequencyofeverypixelonthebarsur
face.Therelationshipbetweenthetwofrequencieswereanalyzed,andtheinundationfrequencyconditionsforthe
suitablevegetationgrowthareasweredetermined.Theresearchresultsindicatethatbothtypesofmid - channelbars
haveconcentratedlowandhighvegetationfrequencyzones ,whilethemediumvegetationfrequencyzonearemore
dispersed.Basedonthedistributioncharacteristicsofvegetationfrequency ,thebarsurfaceareacanbedivided
intohighcoveragezone,mediumcoveragezone,transitionalzone,andunsuitablevegetationzone,withthecriti
calvegetationfrequencythresholdsof80%,50%,and20%,respectively.Theinundationfrequenciesatdifferent
locationsonthebarsarethedominantfactoraffectingvegetationfrequency.Onaverageoveralongperiod,thein
undationfrequency - vegetationfrequencyrelationshipsonbothtypesofbarsfollowaunifiedlogisticcurve.Accord
ingtothiscurve ,criticalinundationconditionsforeachzonearedeterminedasanannualinundationdurationof18
days ,43days,and92days,respectively.Atlocationswheretheannualaverageinundationdurationexceed92
days,thevegetationfrequencyrapidlydecreasestobelow10%.Themethodsandfindingsinthispapercanprovide
referencesforpredictingthevegetationsuccessiontrendindam downstream reachesaftertheimpoundmentofthe
ThreeGorgesDam.
Keywords:UpperJingjiangRiver;mid - channelbar;vegetation;remotesensingimage;floodingfrequency;
criticalconditions
(责任编辑:耿庆斋)
(上接第 237页)
High - precisioncorrectionanduncertaintyanalysismethodfor
remotesensingprecipitationdownscaling
2,3
1,2,3
2,3
1,2,3
DONGJiaping ,YEYuntao ,GUJingjing ,HUANGJianxiong ,
2,3
GUANHaozhe ,CAOYin 2,3
(1.SchoolofCivilEngineering,TianjinUniversity,Tianjin 300072,China;
2.DepartmentofWaterResources,ChinaInstituteofWaterResourcesandHydropowerResearch,Beijing 100038,China;
3.KeyLaboratoryofRiverBasinDigitalTwinningofMinistryofWaterResources,Beijing 100038,China)
Abstract:Toeliminatetheinfluenceofhomogeneouspartsinprecipitationfieldsandenhancetheaccuracyofsta
tisticalprecipitationdownscalingresults,ahigh - precisioncorrectionmethodisproposedforremotesensingpre
cipitationdownscalingbasedontheBayesianHigh - AccuracySurfaceModeling(Bayes - HASM)algorithm.This
methodintroducesahigher - precisionsurfacemodelingapproachandcombinesitwithBayesianoptimizationalgo
rithmstoachieveautomaticoptimizationofmodelparametersandhigh - precisiondownscalingcorrection.Itad
dressestheerrorsandmulti - scaleissuespresentinexistingdownscalingresidualcorrectionmethods.Theresults
indicatethatBayesianoptimizationsignificantlyreducestheuncertaintyofhigh - precisionsurfacemodeling ;after
Bayes - HASM residualcorrection ,thescatterdistributionofthedownscaledresultsisclosertothe1∶1line.The
accuracyindicatorsattheannual ,seasonal,monthly,andten - dayscaleshavebeensignificantlyimproved,
withCCandIAindicatorsreachingaround0.9,RMSEhassignificantlydecreased,andRBsignificantlyreduced.
Theaboveresultsindicatethatthismethodcansignificantlyreducethemodel ’suncertaintyandeffectivelyeliminate
theimpactofhomogeneouspartsoftheprecipitationfield ,therebyimprovingtheaccuracyofthedownscaledprecip
itationresults.
Keywords:digitaltwinwatershed;highaccuracysurfacemodeling;Bayesianoptimization;statisticaldown
scaling;remotesensingprecipitationdata
(责任编辑:韩 昆)
— 2 5 —
2