Page 211 - 《水资源与水工程学报》2024年第6期
P. 211

!35 "!6 #                         & ' ( ) & * + , -                               Vol.35No.6
               2024 $ 12 %              JournalofWaterResources&WaterEngineering                 Dec.,2024

            DOI:10.11705/j.issn.1672-643X.2024.06.21


                       ¬ƒ¬E|ûƒØÙÚÛY닂ƒ­Š


                                   8áABZ[———‚ƒ„…†‡

                                        : Q, STU, VWX, :YZ, 5 [

                                      ( ·¸noj& ³7l8³no&p, ·¸ TU 056038)
                 C D: '×1s9、 ’·«s9ØNeÄåS^xs9, Å´ŠÜéß(SVM)、 ÙAU(RF)、 ìé'‡s
                 VPmW(RBFNN) bª&-(ELM)4 šPµ,&-1}, W,“ 2001—2022 2·¸-ÕyD>·ÿ½
                 ˜DúÉîíÛü†§Ò»Ÿ8m§š›, ²½§ Sen xy‹v“·¸-Ò»§¨ž®ST。½¾¿À:RF、
                 RBFNNb SVM\§Â¾ìDúÉÕyîúØNÕyD>·ÿJ¿$aëûPQR,ELM\§Â¾(ë
                 µ, EF, '×ÕyD>·ÿ¯ø RF-RBF-SVMm§š›{W,·¸-†§Ò»Ÿ8š›; 際\§Rá
                 ì&ÌBCJŸ8áP¿$mŽ, ƨ¡‰ŒØ½˜DúÉ.ÁŸ8áP¿$mŽ; š›¤ËðMü‡
                85.15%, š\£l^xy8£2ž®xyXQRëÄ; BCÆÓ% SM10 Senxyü§¨ž®£‡
                0.174/a, eÄ(&ü<Ýjxy÷d¿BCÿ¤÷d 77.21%, Ò»ST€uɪ。
                 EFG: ;-s9; eÄ(&ü; ,&-; Ò»Ÿ8; ·¸-
                 HIJKL:TP181;P426.616   MNOPQ:A    MRSL:1672643X(2024)06020713

                 Integratedmodelingfordroughtmonitoringbasedonmultisourcedata
                      andmachinelearningalgorithm:acasestudyofHebeiProvince

                              WANGXiao,LIUHaixin,SUNZhenyu,WANGJialin,ZHUYan
                      (SchoolofMiningandGeomaticsEngineering,HebeiUniversityofEngineering,Handan056038,China)

                 Abstract:Basedonremotesensing,reanalysisandsoilmoisturesitedata,anintegrateddroughtmonito
                 ringmodelonmonthlyscalewasconstructedforvegetationofgrowingseasonsindifferentecologicalzones
                 ofHebeiProvincefrom2001to2022.Fourclassicmachinelearningmethods ,namely,supportvector
                 machine(SVM),random forest(RF),radialbasisfunctionneuralnetwork(RBFNN)andextreme
                 learningmachine (ELM)wereadoptedforthemodeling.ThencombinedwithSen’sslopeanalysis,the
                 spatialvariationofdroughtinHebeiProvincewasrevealed.Theresultsshowedthatthefittingeffectof
                 RF ,RBFNNandSVMpresentedstrongstabilityindifferentmonthsofthegrowingseason,aswellasin
                 differentecologicalzones ,whilethatofELMwasrelativelypoor.Therefore,theintegratedRFRBFNN
                 SVM modelwasselectedbasedondifferentecologicalzonesforthemodelingofcomprehensivedrought
                 monitoringinHebeiProvince.Theapplicationofthemodelindicatedthatitsfittingperformancewasbet
                 terinthewetenvironment,andinthemiddleofthevegetationgrowingseason.Theaverageaccuracyof
                 themodelwas85.15%,andthepredictedvaluewasingoodagreementwiththemeasuredvalue.The
                 averagespatialvariationof SM10was0.174/ainthestudyperiod,andsoilrelativehumidityin77.21%
                 oftheareashowedanupwardtrend ,andthedroughtsituationwasalleviated.
                 Keywords:multisourcedata;soilrelativehumidity;machinelearning;droughtmonitoring;Hebei
                 Province


                !"#$:20240611; %&#$:20240817
                '()*: !"#$%&'()* (42071246、42171212); · ¸ - # $ % & ' ( ) * (E2020402006、E2020402086、
                        D2022402030)
                +,-.: › •(1999—), V, ·¸Zâ?, @ABCD, BC1é‡ò7Ò»1。
                /0+,: Ir5(1980—), :, ñTSU?, YA, \], @AD^_, BC1é‡D>BClò7Ò»1。
   206   207   208   209   210   211   212   213   214   215   216