2.清华大学水沙科学与水利水电工程国家重点实验室，北京100084
3.青海大学省部共建三江源生态与高原农牧业国家重点实验室，青海西宁810016
2.StateKeyLaboratoryofHydro-scienceandEngineering，TsinghuaUniversity，Beijing100084，China
3.StateKeyLaboratoryofPlateauEcologyandAgriculture，QinghaiUniversity，Xining810016，China
收稿日期:2021-01-14修回日期:2021-05-13网络出版日期:2021-07-22
Received:2021-01-14Revised:2021-05-13Online:2021-07-22
作者简介Aboutauthors
卫星降水产品在一定程度上为地表观测稀疏地区的降水提供了参照值，但在具体应用时仍需进行适用性和精度评价。为降低不确定性，通行做法是利用地面观测数据对卫星产品进行融合校正，评价和修正卫星降水产品。以TRMM为例，对研究中国大陆范围的融合TRMM降水的公开成果进行了检视，发现不同研究成果结论之间存在不可忽视的方向性相背情况，极大地影响着生产实践应用中对融合产品的可靠性判断。研究认为，这种融合结果相背的现象与融合校正方法关系不大，而与地面降水参照站点选取的范围有密切关系。研究表明，仅靠TRMM卫星降水自身及与地面融合方法的创新，尚不能降低卫星降水产品的不确定性。目前仍需加强对卫星降水融合中地面观测数据的完整性要求，采用多种独立检验方法验证融合结果的一致性和可靠性。
关键词：遥感降水融合;TRMM;高估;低估;一致性检验
Althoughsatelliteprecipitationhasprovidedavaluablereferencefortheprecipitationestimationinthesparseregion，itsuncertaintiesinthepracticalapplicationarestillchallenging.Researcherstriedtofusegroundobservationandsatelliteprecipitationtoevaluateandmodifysatelliteprecipitationproductsforuncertaintyreduction.Unfortunately，thispaperexaminedmanypublicationsoffusionTRMMprecipitationproductsinmainlandChinaandfoundanon-negligibledirectionalcontradictionbetweentheapproaches.Thefindingswillsignificantlyaffectthereliabilityofsatelliteprecipitationinpracticalapplications.Itwasfoundthatthecontradictionhasnotrelatedtothesatelliteprecipitationproductsandfusionmethodsbutsignificantlyrelatedtotheselectionofprecipitationobservationstations.ItwasalsofoundthattheuncertaintyishardtobereducedonlybyinnovatingthefusingmethodsofTRMMprecipitationwithgroundobservationdata.Therefore，itisnecessarytostrengthentheintegrityofgroundobservationandadoptindependentpracticestoverifyconsistencyandreliability.
Keywords：Satelliteprecipitationfusing;TRMM;Overestimate;Underestimate;Consistencyvalidation
本文引用格式
表1常用卫星降水产品一览
Table1Listofcommonsatelliteprecipitationproducts
偏差校正法（BiasCorrection）是指用观测数据计算估计值的加性或乘性等偏差因子对原始估计值重新标定以减少偏差，使修正后的估计值逼近测量值的方法。典型的偏差校正法包括加性偏差校正法（AdditiveBiasCorrection）、乘性偏差校正法（MultiplicativeBiasCorrection）和分位数映射法（QuantileMapping）。
相对于偏差校正，插值（Interpolation）则是将点数据向线数据和面数据的展布，通常先用离散数据拟合连续函数，再用这个连续函数对缺测的离散点进行插补。在遥感数据和地面观测数据融合中，一般保留地面观测数据精度，引入遥感数据作为参考进行局部校正。
回归分析法（RegressionAnalysis）本属于插值法，但多元回归分析法（MultipleRegressionAnalysis）因其是结合物理机制中的多个影响因子来估计或预测目标值而被单列。
表2降水数据融合校正方法特点及分类表
Table2Featuresandcategoriesofprecipitationdatamergingmethods
偏差校正法
（BiasCorrection）
插值展布法
（Interpolation）
多元回归法
（MultipleRegressionAnalysis）
机器学习法
（MachineLearning）
卫星遥感降水与地面观测降水相比，通常有3种情形：高估、低估和吻合。高估即认为所选用的卫星降水产品在研究区高估了实际降水；低估即认为所选用的卫星降水产品在研究区低估了实际降水；若卫星降水产品与真实降水值误差在±10%以内可视为与实际降水吻合，称之为吻合。
表3TRMM遥感降水产品对中国大陆区域降水估计的结论统计
Table3StatisticsofTRMMproductsforprecipitationestimationinmainlandChina
由上可知，对于中国大陆高海拔地区，多数研究结论认为TRMM卫星降水的表达是高估的，占55.5%；认为吻合的占25.4%，而认为低估的，占19.1%。对于中国大陆低海拔地区，多数研究认为TRMM卫星降水的表达是吻合的，占48.8%；认为高估的占39%，而认为低估的仅占12.2%。可见，不同研究成果的结论存在较大的分歧。
图1TRMM遥感降水产品研究在中国大陆各分区高低估及吻合情况分析
Fig.1Analysisofoverestimate，underestimate，andcoincidenceofTRMMproductsindifferentregionsofmainlandChina
图2TRMM遥感降水产品融合校正及精度评价研究内容分析
（a）TRMM产品评价结果与高程范围关系；（b）TRMM产品评价结果与融合方法关系
Fig.2AnalysisofmergingandaccuracyevaluationresultsofTRMMproducts
（a）RelationshipbetweenTRMMproductsevaluationresultsandelevationrange；（b）RelationshipbetweenTRMMproductsevaluationresultsandmergingmethods
不难发现，“背论”的出现与高程范围没有明显的统计关系，无论是海拔较高、海拔中等还是海拔较低的地区均会出现“背论”现象；同样，无论是利用偏差校正、插值展布、多元回归和机器学习中的哪类方法的研究成果，也均存在“背论”现象；此外，单纯精度评价的研究成果也显示有高估、低估和吻合3种结论并存的现象。可见，这种“背论”现象与融合校正方法也无直接联系。
表4降水观测数据源差别统计表
Table4Statisticsofdifferentprecipitationobservationdatasources
Table5StatisticsofdifferentprecipitationobservationdatasourcesinQinghai-Tibetregion
以上可见，TRMM降水卫星融合成果的结论中“背论”是普遍存在的。这带给我们一个启示：在以方法推动的遥感降水融合研究中，方法创新性和自洽性尚不能消除结果的相背性，也不能消除由此带来的成果所表达的知识混乱性，更不能对实际应用中的数据带来可靠的辅助功能。这种“背论”现象既表明了卫星遥感降水传感器及其反演算法本身还需进一步改进，也表明了卫星遥感降水产品与地表观测数据的融合方法及可靠性仍属初步阶段，还表明了遥感降水产品的使用还应处于谨慎乐观状态。在自然科学研究层面，应寻找更为可靠的方法；在生产应用层面，应更加重视多渠道相互印证。
综上，本文认为除常用的RB和RMSD等评价指标之外，可以添加一致性系数方法、基于其他降水产品进行检验的方法进行多重检验，提高研究结果的可靠性。
WILHEITTT，CHANGEATC，RAOMSV，etal.Asatellitetechniqueforquantitativelymappingrainfallratesoveroceans
FERRARORR.Specialsensormicrowaveimagerderivedglobalrainfallestimatesforclimatologicalapplications
LIUYuanbo，FUQiaoni，SONGPing，etal.Satelliteretrievalofprecipitation：anoverview
刘元波，傅巧妮，宋平，等.卫星遥感反演降水研究综述
HUFFMANGJ，ADLERRF，ARKINPA，etal.TheGlobalPrecipitationClimatologyProject（GPCP）combinedprecipitationdataset
XIEP，ARKINPA.Globalprecipitation：a17-yearmonthlyanalysesbasedongaugeobservations，satelliteestimates，andnumericalmodeloutputs
KUMMEROWC，BARNESW.TheTropicalRainfallMeasuringMission（TRMM）sensorpackage
HUFFMANGJ，BOLVINDT，NELKINEJ，etal.TheTRMMMultisatellitePrecipitationAnalysis（TMPA）：quasiglobal，multiyear，combined-sensorprecipitationestimatesatfinescales
SOROOSHIANS，HSUK，GAOX，etal.EvaluationofPERSIANNsystemsatellite—Basedestimatesoftropicalrainfall
JOYCERJ，JANOWIAKJE，ARKINPA，etal.CMORPH：amethodthatproducesglobalprecipitationestimatesfrompassivemicrowaveandinfrareddataathighspatialandtemporalresolution
HUFFMANGJ，BOLVINDT，NELKINEJ.IntegratedMulti-satellitERetrievalsforGPM（IMERG）TechnicalDocumentation
HONGY，HSUK，MORADKHANIH，etal.UncertaintyquantificationofsatelliteprecipitationestimationandMonteCarloassessmentoftheerrorpropagationintohydrologicresponse
YONGB，HONGY，RENL，etal.AssessmentofevolvingTRMM-basedmultisatellitereal-timeprecipitationestimationmethodsandtheirimpactsonhydrologicpredictioninahighlatitudebasin
JIANGS，RENL，HONGY，etal.Improvementofmulti-satellitereal-timeprecipitationproductsforensemblestreamflowsimulationinamiddlelatitudebasininSouthChina
XUH，XUC，SAELTHUNNR，etal.Evaluationofreanalysisandsatellite-basedprecipitationdatasetsindrivinghydrologicalmodelsinahumidregionofSouthernChina
DEMAEYER，LUOM，LIUT，etal.ProportionalcoefficientmethodappliedtoTRMMrainfalldata：casestudyofhydrologicalsimulationsoftheHotanRiverBasin（China）
LEANDERR，BUISHANDTA.Resamplingofregionalclimatemodeloutputforthesimulationofextremeriverflows
TEUTSCHBEINC，SEIBERTJ.IsbiascorrectionofRegionalClimateModel（RCM）simulationspossiblefornon-stationaryconditions？
ZHANGX，TANGQ.Combiningsatelliteprecipitationandlong-termgroundobservationsforhydrologicalmonitoringinChina
MATHERONG.Principlesofgeostatistics
KRAJEWSKIWF.Cokrigingradar-rainfallandraingauge
GRIMESD，PARDO-IGUZQUIZAE，BONIFACIOR.Optimalarealrainfallestimationusingraingaugesandsatellitedata
TODINIE.ABayesiantechniqueforconditioningradarprecipitationestimatestorain-gaugemeasurements
ZHAOY，XIEQ，LUY，etal.HydrologicevaluationofTRMMmultisatelliteprecipitationanalysisforNanliuRiverBasininHumidSouthwesternChina
QIWenwen，ZHANGBaiping，PANGYu，etal.SpatialandseasonaldistributioncharacteristicsofprecipitationovertheQinghai-TibetPlateaubasedonTRMMdata
齐文文，张百平，庞宇，等.基于TRMM数据的青藏高原降水的空间和季节分布特征
SUNMeiping，ZHANGHaiyu，GONGNinggang，etal.StudyonmaximumprecipitationheightzoneinQilianMountainsareabasedonTRRMprecipitationdata
孙美平，张海瑜，巩宁刚，等.基于TRMM降水订正数据的祁连山地区最大降水高度带研究
SHIT，YANGX，GEORGEChristakos，etal.SpatiotemporalinterpolationofrainfallbycombiningBMEtheoryandsatelliterainfallestimates
LIUJ，YAND，LIUS，etal.EvaluationofTRMMprecipitationanditsapplicationtodistributedhydrologicalmodelinNaquRiverBasinoftheTibetanPlateau
GUOS，WANGJ，CHENH，etal.TrackingtheerrorsourcesofspatiotemporaldifferencesinTRMMaccuracyusingerrordecompositionmethod
BRILLINGERDR.Spatio-temporalmodellingofspatiallyaggregatebirthdata
NIES，LUOY，WUT，etal.Amergingschemeforconstructingdailyprecipitationanalysesbasedonobjectivebias-correctionanderrorestimationtechniques
XIEP，XIONGA.Aconceptualmodelforconstructinghigh-resolutiongauge-satellitemergedprecipitationanalyses
SHENY，ZHAOP，PANY，etal.Ahighspatiotemporalgauge-satellitemergedprecipitationanalysisoverChina
YONGB，RENL，HONGY，etal.Firstevaluationoftheclimatologicalcalibrationalgorithminthereal-timeTMPAprecipitationestimatesovertwobasinsathighandlowlatitudes
JIASF，ZHUWB，LVAF，etal.AstatisticalspatialdownscalingalgorithmofTRMMprecipitationbasedonNDVIandDEMintheQaidamBasinofChina
BRUNSDONC，FOTHERINGHAMAS，CHARLTONME.Geographicallyweightedregression：amethodforexploringspatialnonstationarity
BRUNSDONC，AITKINM，FOTHERINGHAMS，etal.Acomparisonofrandomcoefficientmodellingandgeographicallyweightedregressionforspatiallynon-stationaryregressionproblems
WHEELERDC.Diagnostictoolsandaremedialmethodforcollinearityingeographicallyweightedregression
ZHOUL，CHENY，LIANGN，etal.DailyrainfallmodeltomergeTRMMandgroundbasedobservationsforrainfallestimations
CHAOLijun，ZHANGKe，LIZhijia，etal.Geographicallyweightedregressionbasedmethodsformergingsatelliteandgaugeprecipitation
GAOYC，LiuMF.Evaluationofhigh-resolutionsatelliteprecipitationproductsusingraingaugeobservationsovertheTibetanPlateau
CHENF，LIUY，LIUQ，etal.SpatialdownscalingofTRMM3B43precipitationconsideringspatialheterogeneity
LVA，ZHOUL.ArainfallmodelbasedonageographicallyweightedregressionalgorithmforrainfallestimationsovertheAridQaidamBasininChina
LIUJ，ZHANGW，NIEN，etal.SpatialdownscalingofTRMMprecipitationdatausinganoptimalsubsetregressionmodelwithNDVIandterrainfactorsintheYarlungZangboRiverBasin，China
ZHANGY，LIY，JIX，etal.Fine-resolutionprecipitationmappinginamountainouswatershed：geostatisticaldownscalingofTRMMproductsbasedonenvironmentalvariables
ZHAON，YUET，CHENC，etal.AnimprovedstatisticaldownscalingschemeoftropicalrainfallmeasuringmissionprecipitationintheHeiheRiverBasin，China
CHENF，GAOY，WANGY，etal.Adownscaling-mergingmethodforhigh-resolutiondailyprecipitationestimation
WANGY，KongY，ChenH，etal.ImprovingdailyprecipitationestimatesfortheQinghai‐TibetanPlateaubasedonenvironmentalsimilarity
BAEZ-VILLANUEVAOM.RF-MEP：anovelRandomForestmethodformerginggriddedprecipitationproductsandground-basedmeasurements
WUHongcai，YANGAQinli，LIUJiaming，etal.AspatiotemporaldeepfusionmodelformergingsatelliteandgaugeprecipitationinChina
XUG，WangZJ，XiaT.MappingarealprecipitationwithfusiondatabyANNmachinelearninginsparsegaugedregion
LIZ，YANGD，HONGY.Multi-scaleevaluationofhigh-resolutionmulti-sensorblendedglobalprecipitationproductsovertheYangtzeRiver
SHIYuli，SONGLei.CalibrationofTRMM
石玉立，宋蕾.1998—2012年青藏高原TRMM3B43降水数据的校准
YANGY，LUOY.UsingthebackpropagationneuralnetworkapproachtobiascorrectTMPAdatainthearidregionofNorthwestChina
XIAT，WANGZ，ZHENGH.TopographyanddataminingbasedmethodsforimprovingsatelliteprecipitationinmountainousareasofChina
ZHUQ，XUANW，LIUL，etal.EvaluationandhydrologicalapplicationofprecipitationestimatesderivedfromPERSIANN‐CDR，TRMM3B42V7，andNCEP‐CFSRoverhumidregionsinChina
SHENY，PANY，YUJ，etal.QualityassessmentofhourlymergedprecipitationproductoverChina
TANGG，ZENGZ，LONGD，etal.StatisticalandHydrologicalcomparisonsbetweenTRMMandGPMlevel-3productsoveraMidlatitudeBasin：isDay-1IMERGagoodsuccessorforTMPA3B42V7？
TONGK，SUF，YANGD，etal.EvaluationofsatelliteprecipitationretrievalsandtheirpotentialutilitiesinhydrologicmodelingovertheTibetanPlateau
MAY，TANGG，LONGD，etal.SimilarityanderrorintercomparisonoftheGPManditspredecessor-TRMMmultisatelliteprecipitationanalysisusingthebestavailablehourlygaugenetworkovertheTibetanPlateau
WEIG，LVH，CROWW，etal.Evaluationofsatellite—BasedprecipitationproductsfromIMERGV04AandV03D，CMORPHandTMPAwithgaugedrainfallinthreeclimatologiczonesinChina
WANGYuhan，CHENJinsong，YANGDawen.Bayesianassimilationofmultiscaleprecipitationdataandsparsegroundgaugeobservationsinmountainousareas
XUR，TIANF，YANGL，etal.GroundvalidationofGPMIMERGandTRMM3B42V7rainfallproductsoversouthernTibetanPlateaubasedonahigh-densityraingaugenetwork
ZENGQ，WANGY，CHENL，etal.Inter-comparisonandevaluationofremotesensingprecipitationproductsoverChinafrom2005to2013
JINQiu，ZHANGZengxin，HUANGYuhan，etal.AccuracyevaluationofprecipitationanalysisinYangtzeRiverBasinbasedonsatelliteTRMM
金秋，张增信，黄钰瀚，等.基于TRMM卫星产品的长江流域降水精度评估
LIUJie，XIAJun，ZOULei，etal.Applicabilityanalysisofmulti-satelliteremotesensingprecipitationdatainTarimRiverBasin
刘洁，夏军，邹磊，等.多卫星遥感降水数据在塔里木河流域的适用性分析
JIANGL，BAUER-GOTTWEINP.HowdoGPMIMERGprecipitationestimatesperformashydrologicalmodelforcing？evaluationfor300catchmentsacrossMainlandChina
LIHui，YANGTao，HEQisheng，etal.ResearchonaccuracyvalidationandcalibrationmethodsofTRMMinTianshanMountainsofXinjiang
李慧，杨涛，何祺胜，等.新疆天山山区TRMM卫星降水数据的精度检验和校正方法研究
QINFuying，JIAGensuo，YANGJie，etal.AccuracyevaluationandseasonaldistributionofprecipitationovertheMongolianPlateaubasedonTRMMdata
秦福莹，贾根锁，杨劼，等.基于TRMM卫星数据的蒙古高原降水精度评估与季节分布特征
YONGB，RENL，HONGY，etal.Hydrologicevaluationofmultisatelliteprecipitationanalysisstandardprecipitationproductsinbasinsbeyonditsinclinedlatitudeband：acasestudyinLaohaheBasin，China
BHARTIV，SINGHC.EvaluationoferrorinTRMM3B42V7precipitationestimatesovertheHimalayanregion
YUANHuiling，SUNRuochen.ApplicationofsatellitestationfusiondataevaluationandhydrologicalsimulationintheupperreachesofHuaiheRiverbasin
袁慧玲，孙若辰.淮河上游流域卫星站点融合资料评估及水文模拟应用
CAIY，JINC，WANGA，etal.ComprehensiveprecipitationevaluationofTRMM3B42withdenseraingaugenetworksinamid-latitudebasin，northeast，China
RENM，XUZ，PANGB，etal.Assessmentofsatellite-derivedprecipitationproductsfortheBeijingregion
BIANWei，CHENYaning，LIZhi，etal.AssessingtheapplicabilityofdifferentsourcesofprecipitationdatainAksuRiverBasinonsouthslopeofTianshanMountains
卞薇，陈亚宁，李稚，等.不同源降水数据在天山南坡阿克苏河流域适用性分析
LUX，WEIM，TANGG，etal.EvaluationandcorrectionoftheTRMM3B43V7andGPM3IMERGMsatelliteprecipitationproductsbyuseofground—BaseddataoverXinjiang，China
CAOY，ZHANGW，WANGW，etal.EvaluationofTRMM3B43dataovertheYangtzeRiverDeltaofChina
MAZ，SHIZ，ZHOUY，etal.AspatialdataminingalgorithmfordownscalingTMPA3B43V7dataovertheQinghai-TibetPlateauwiththeeffectsofsystematicanomaliesremoved
HUANGGuiping，CAOYanping.AccuracyanalysisofTRMM
黄桂平，曹艳萍.TRMM卫星3B43降水数据在黄河流域的精度分析
ZHUG，QIND，LIUY，etal.AccuracyofTRMMprecipitationdatainthesouthwestmonsoonregionofChina
JIX，CHENY.CharacterizingspatialpatternsofprecipitationbasedoncorrectedTRMM3B43dataovertheMidTianshanMountainsofChina
ZHANGC，CHENX，SHAOH，etal.Evaluationandintercomparisonofhigh-resolutionsatelliteprecipitationestimates—GPM，TRMM，andCMORPHintheTianshanMountainarea
PENGB，SHIJ，NI-MEISTERW，etal.EvaluationofTRMMMultisatellitePrecipitationAnalysis（TMPA）productsandtheirpotentialhydrologicalapplicationatanaridandsemiaridbasininChina
CHENGLizhen，YANGMeixue，WANGXuejia，etal.AnalysisoftheaccuracyofTRMM3B42rainfalldataintheupperandmiddlereachesofTaoheRiver
程立真，杨梅学，王学佳，等.TRMM3B42降水产品在洮河中上游的精度评估分析
LIQiong，YANGMeixue，WANGuoning，etal.AnalysisoftheaccuracyofTRMM3B43precipitationdatainthesourceregionoftheYellowRiver
李琼，杨梅学，万国宁，等.TRMM3B43降水数据在黄河源区的适用性评价
MENGJ，LIL，HAOZ，etal.SuitabilityofTRMMsatelliterainfallindrivingadistributedhydrologicalmodelinthesourceregionofYellowRiver
WANGS，LIUJ，WANGJ，etal.EvaluationofGPMIMERGV05BandTRMM3B42V7precipitationproductsoverhighmountainoustributariesinLhasawithdenseraingauges
WANGL，CHENR，HANC，etal.Animprovedspatial-temporaldownscalingmethodforTRMMprecipitationdatasetsinalpineregions：acasestudyinNorthwesternChina'sQilianMountains
WANGZ，ZHONGR，LAIC.EvaluationandhydrologicvalidationofTMPAsatelliteprecipitationproductdownstreamofthePearlRiverBasin，China
LIX，ZHANGQ，XUC，etal.SuitabilityoftheTRMMsatelliterainfallsindrivingadistributedhydrologicalmodelforwaterbalancecomputationsinXinjiangcatchment，PoyangLakeBasin
YANGN，ZHANGK，HONGY，etal.EvaluationoftheTRMMmultisatelliteprecipitationanalysisanditsapplicabilityinsupportingreservoiroperationandwaterresourcesmanagementinHanjiangBasin，China
SUNR，YUANH，LIUX，etal.Evaluationofthelatestsatellite-gaugeprecipitationproductsandtheirhydrologicapplicationsovertheHuaiheRiverBasin
CHENC，CHENQ，DUANZ，etal.MultiscalecomparativeevaluationoftheGPMIMERGv5andTRMM3B42v7precipitationproductsfrom2015to2017overaclimatetransitionareaofChina
CHENY，HUANGJ，SHENGS，etal.Anewdownscaling-integrationframeworkforhigh-resolutionmonthlyprecipitationestimates：combiningraingaugeobservations，satellite-derivedprecipitationdataandgeographicalancillarydata
GUANX，LIUG，CHENT，etal.EvaluationofprecipitationproductsbyusingmultiplehydrologicalmodelsovertheupperYellowRiverBasin，China
MAQ，JIC，XIONGL，etal.SpatiotemporalpatternsofsatelliteprecipitationextremesintheXijiangRiverBasin：fromstatisticalcharacterizationtostochasticbehaviourmodelling
PENGF，ZHAOS，CHENC，etal.EvaluationandcomparisonoftheprecipitationdetectionabilityofmultiplesatelliteproductsinatypicalagricultureareaofChina
SUNS，SHIW，ZHOUS，etal.Capacityofsatellite—Basedandreanalysisprecipitationproductsindetectinglong-termtrendsacrossMainlandChina
WANGN，LIUW，SUNF，etal.Evaluatingsatellite-basedandreanalysisprecipitationdatasetswithgauge-observeddataandhydrologicalmodelingintheXiheRiverBasin，China
YANGX，LUY，TANM，etal.Nine-yearsystematicevaluationoftheGPMandTRMMprecipitationproductsintheShuaishuiRiverBasininEast-CentralChina
YANGY，WUJ，BAIL，etal.ReliabilityofgriddedprecipitationproductsintheYellowRiverBasin，China
YUL，MaL，LiH，etal.Assessmentofhigh-resolutionsatelliterainfallproductsoveragraduallyelevatingmountainousterrainbasedonahigh-densityraingaugenetwork
ZHOUZ，MAL，LIH，etal.MULTIdimensionalevaluationoftheTRMM3B43V7satellite-basedprecipitationproductinmainlandChinafrom1998-2016
QIS，LVA.ApplicabilityanalysisofmultipleprecipitationproductsintheQaidamBasin，NorthwesternChina
WANGW，LINH，CHENN，etal.Evaluationofmulti-sourceprecipitationproductsovertheYangtzeRiverBasin
GUORuifang，LIUYuanbo.Strategyandmethodforsatelliteprecipitationdataevaluation：anoverview
郭瑞芳，刘元波.遥感降水数据精度检验策略及检验方法综述
GUORuifang，LIUYuanbo.Multi-satelliteretrievalofhighresolutionprecipitation：anoverview
郭瑞芳，刘元波.多传感器联合反演高分辨率降水方法综述
插值展布法...
插值展布法...IsbiascorrectionofRegionalClimateModel（RCM）simulationspossiblefornon-stationaryconditions？22013...加性偏差校正是用估计值与观测值之间的差值作为因子进行校正，乘性偏差校正则是通过一个比例因子使降水估计值具有与观测值相同的平均值.可以单独使用［12~15］，也可联合使用［16］，各有所长.分位数映射法利用历史数据建立观测数据和遥感产品数据的概率分布函数，通过概率映射（ProbabilityMapping）或直方图均衡化（HistogramEqualization）使两者分布匹配［17，18］.该方法能够匹配估计值和观测值之间在月尺度的完整分布，同时设定了降水阈值，可避免因存在过多的极值而造成分布严重扭曲....
插值展布法...Combiningsatelliteprecipitationandlong-termgroundobservationsforhydrologicalmonitoringinChina32015...加性偏差校正是用估计值与观测值之间的差值作为因子进行校正，乘性偏差校正则是通过一个比例因子使降水估计值具有与观测值相同的平均值.可以单独使用［12~15］，也可联合使用［16］，各有所长.分位数映射法利用历史数据建立观测数据和遥感产品数据的概率分布函数，通过概率映射（ProbabilityMapping）或直方图均衡化（HistogramEqualization）使两者分布匹配［17，18］.该方法能够匹配估计值和观测值之间在月尺度的完整分布，同时设定了降水阈值，可避免因存在过多的极值而造成分布严重扭曲....
多元回归法...HydrologicevaluationofTRMMmultisatelliteprecipitationanalysisforNanliuRiverBasininHumidSouthwesternChina22017...经典的插值方法是克里金法（Kriging）［19］，克里金法又常被称作为地统插值法（GeostatisticalInterpolation）.克里金法后来发展出多种变形［20~22］和为减小测量稀疏区域融合过程中随机误差的辅助方法［23，24］，均有诸多应用［25~29］.为使插值更加光滑、纠偏更加有效、计算更加快速，核平滑法（KernelSmoothing）［30］、最优插值法［31］和最优概率插值法（ProbabilityDensityFunction-OptimalInterpolation）［32，33］等陆续被引入到卫星遥感产品插值融合中，提高了插值效果....
多元回归法...
多元回归法...Amergingschemeforconstructingdailyprecipitationanalysesbasedonobjectivebias-correctionanderrorestimationtechniques22015...经典的插值方法是克里金法（Kriging）［19］，克里金法又常被称作为地统插值法（GeostatisticalInterpolation）.克里金法后来发展出多种变形［20~22］和为减小测量稀疏区域融合过程中随机误差的辅助方法［23，24］，均有诸多应用［25~29］.为使插值更加光滑、纠偏更加有效、计算更加快速，核平滑法（KernelSmoothing）［30］、最优插值法［31］和最优概率插值法（ProbabilityDensityFunction-OptimalInterpolation）［32，33］等陆续被引入到卫星遥感产品插值融合中，提高了插值效果....
多元回归法...Firstevaluationoftheclimatologicalcalibrationalgorithminthereal-timeTMPAprecipitationestimatesovertwobasinsathighandlowlatitudes32013...诸多研究发现，卫星降水产品的质量与纬度、海拔、坡向、坡度和下垫面组成等因素有关［34，35］.多元回归分析法结合物理机制，挖掘大气（如温度、空气湿度、云层厚度等）、地理（如地形、经纬度、海拔、坡度、坡向等）、地表（如植被、土壤含水量、冰雪厚度等）等和地面观测数据之间的关系，以提高融合的精度.多元回归法可给出具体关系式，变量间的关系很容易被理解和解释，有诸多方法分支及应用［36~49］....
机器学习法...
机器学习法...Acomparisonofrandomcoefficientmodellingandgeographicallyweightedregressionforspatiallynon-stationaryregressionproblems01999Diagnostictoolsandaremedialmethodforcollinearityingeographicallyweightedregression02007DailyrainfallmodeltomergeTRMMandgroundbasedobservationsforrainfallestimations22016...本文分析了部分发表在WebofScience核心集合数据库和中文核心期刊上的研究中国大陆的TRMM卫星降水代表性成果［12~15，18，27~29，34，35，39~47，50~105］，发现不同研究成果的结论颇有差异，出现了不可忽视的结论相背，称其为“背论”.此处“背论”不同于通常所说的“悖论”.前者是指对同一问题用相同或不同方法研究得到的结论相反的情形，而后者多指某种科学假设下，观测到的结果与预期结果相反的情形....
机器学习法...Evaluationofhigh-resolutionsatelliteprecipitationproductsusingraingaugeobservationsovertheTibetanPlateau12013...（MultipleRegressionAnalysis）
机器学习法...SpatialdownscalingofTRMM3B43precipitationconsideringspatialheterogeneity12014...（MultipleRegressionAnalysis）
机器学习法...ArainfallmodelbasedonageographicallyweightedregressionalgorithmforrainfallestimationsovertheAridQaidamBasininChina02016SpatialdownscalingofTRMMprecipitationdatausinganoptimalsubsetregressionmodelwithNDVIandterrainfactorsintheYarlungZangboRiverBasin，China12018...（MachineLearning）
在众多卫星降水产品中，TRMM降水产品应用最广泛，包括TRMM3B42和TRMM3B43全球格点化数据集［57］.其中，中国地区有194个测站数据加入了GPCC（GlobalPrecipitationClimatologyCentre）数据集以辅助TRMM卫星降水产品校正［58］....