author:徐天河,李 耸,王帅民,江 楠

１．山东大学空间科学研究院,山东威海 ;

２．河北工程大学矿业与测绘学院，河北邯郸；

来源出版物：测绘学报 第51卷第8期 文献号：DOI:10.11947/j.AGCS.200.20210480．出版年：Aug 2022

abstract：SingleＧlevelmeteorologicalproducts (ERA５singleＧleveldataand measured meteorological parameters)andmultiＧlevelmeteorologicalproducts (ERA５pressureleveldataandCOSMICdata)are usedtoestimatetheZTD of 236 CMONOCstations,namelyERA５S_ZTD,MET_ZTD,ERA５P_ZTDandRO_ZTD,basedonthe modelmethodandtheintegration methodrespectively．FourZTDestimationsare evaluatedwiththereferenceofGNSS_ZTD;theresultsshowthattheaveragemonthlyRMSEare42.8,53.6,16.1and62.3mm,respectively．TheaccuracyofERA５P_ZTDestimatedbytheintegralmethodisthe highest,thatofERA５S_ZTDandMET_ZTDcalculatedbythemodelmethodarethenext．EstimatingRO_ZTDwiththeintegralmethodhasthelowestaccuracy．InordertofurtherimprovetheaccuracyofZTD estimations,improvedmodeloftroposphericdelayisproposedwiththeRBFneuralnetworkinthispaper．Thecalculationresultsshowthat:theaveragemonthlyRMSEbetweentheZTDfromfourimprovedmodels andGNSS_ZTDare23.5,32.1,14.2and40.8mm,whicharereduced43.4％,36.3％,10.0％and34.4％than rawZTDestimations．Theoverallmodifiedeffectoftheimprovedmodelisobvious,andtheimprovement rateisrelatedtothedensityofstationdistribution．

keywords：GNSS;RBFneuralnetwork;ERA５;COSMIC;troposphericdelay

citation：徐天河,李耸,王帅民,等．顾及气象数据的中国区域对流层延迟 RBF 神经网络优化模型[J]．测绘学报,２０２２,５１(８):１６９０Ｇ

１７０７．DOI:１０．１１９４７/j．AGCS．２０２２．２０２１０４８０．

XU Tianhe,LISong,WANGShuaimin,etal．ImprovedtroposphericdelaymodelforChinausingRBFneuralnetworkand meteorologicaldata[J]．Acta Geodaeticaet Cartographica Sinica,2022,51(8):1690G1707．DOI:10.11947/j.AGCS.200.20210480.