摘要
遥感降水产品具有受地形影响较小,且覆盖面积广泛的优势,对研究大尺度水量平衡具有重要价值.但遥感降水反演受传感器精度、反演算法等影响,数据精度具有区域性、系统性、季节性的偏差,研究遥感降水数据的区域特征及融合校正方法对遥感降水数据的应用具有重要意义.本文结合地形因子和气象站实测降水数据对黄河源区TRMM3843遥感降水数据进行融合校正,并对校正后的遥感降水数据的精度进行评价.研究表明,①纬度、高程、坡度、坡向等21个因子与TRMM3843月降水显著相关;②采用逐步回归、BP神经网络、GWR地理加权回归3种校正方法,校正了TRMM3843月降水数据,并采用6个指标对校正后的月降水数据进行评价;③3种方法校正后的降水数据比原始TRMM3843数据精度有所提高.3种校正方法在暖季(4~10月)校正效果优于冷季(11月到次年3月).GWR地理加权回归校正效果最好,校正后的TRMM3843月降水在暖季各月R^2接近0.8,且对偏差的调整效果更为显著,校正后的降水显著的降低了相对偏差,相对偏差接近0,各月的绝对误差也降低了1~5mm.
Remote sensing precipitation is of great value to study large scale water balance,which is less affected by terrain and has extensive coverage.However,remote sensing precipitation is affected by sensor precision and inversion algorithm,and data accuracy has regional,systematic and seasonal deviations.It is very important to study the correction method of remote sensing precipitation due to the accuracy of precipitation data.In this study,TRMM 3B43precipitation data was used as the data source,precipitation correction experiment was conducted in the Source region of the Yellow River located northeast of the Qinghai-Tibet Plateau.The Analysis shows:①The correlation analysis shows that 21 topographic factors including latitude and longitude,elevation,slope and aspect in the source region of the Yellow River are related significant to the TRMM 3B43 precipitation.②This study combined the terrain factor,using the stepwise regression method,BP neural network method,GWR geographic weighting regression method to correct the TRMM 3B43 monthly data,and result of the corrected precipitation data were evaluated by 6index. ③The precision of the precipitation data corrected by the three methods have been improved compared with the original TRMM 3B43data,the three correction methods were better in the warm season from April to October,and the result was poor in the cold season from November to March.GWR geography weighted regression is the best,the R2of precipitation is close to 0.8in the warm season,which is higher than that of the original precipitation data 0.7.The deviation is significant reduced,the relative deviation is nearly 0, and the absolute error of each month is also reduced by 1~5mm.
作者
李琼
魏加华
安娟
张博
任燕
LI Qiong;WEI Jiahua;AN Juan;ZHANG Bo;REN Yan(State Key Laboratory of Plateau Ecology and Agriculture,Qinghai University,Xining 810016,China;State Key Laboratory of Hydro Science and Engineering,Tsinghua University,Beijing,100084,China;School of Water Resources and Electric Power,Qinghai University,Xining 810016,China;Institute of Economic-Technology,Qinghai Electric Power Company,Xining 810016,China)
出处
《应用基础与工程科学学报》
EI
CSCD
北大核心
2018年第6期1147-1163,共17页
Journal of Basic Science and Engineering
基金
国家十三五重点研发计划项目(2017YFC0403600)
水沙科学水利水电工程国家重点实验室开放课题(sklhse-2017-A-04)
省部共建三江源生态与高原农牧业国家重点实验室自主课题(2017-ZZ-02)
国家电网公司科技项目(52283014000T)
关键词
降水校正
TRMM
黄河源区
逐步回归法
BP神经网络
GWR
precipitation correction
TRMM 3B43
the source region of the Yellow River
stepwise regression method
BP neural network
GWR
作者简介
李琼(1986-),女,博士,讲师.E—mail:liqiong1118@126.com;通信作者:魏加华(1971-),男,博士,研究员.E-mail:weijiahua@tsinghua.edu.cn.
