%0 Journal Article %T 基于MODIS/AQUA 数据的地表短波净辐射反演 %T Estimation of net surface shortwave radiation from MODIS/AQUA %A 应王敏 %A 吴 骅※ %A Ying Wangmin %A Wu Hua※ %J 中国农业信息 %J China agricultural informatics %@ 1672-0423 %V 31 %N 1 %D 2019 %P 24-34 %K MODIS/AQUA;地表短波净辐射;MODTRAN %K MODIS/AQUA;net surface shortwave radiation;MODTRAN %X 【目的】地表短波净辐射是许多全球、区域气候、水文、陆面过程模型中的重要参数, 但运用遥感技术对地表短波净辐射精确实施监测一直是难点,希望构建具有一定泛化能力的 高精度地表短波净辐射反演模型。【方法】利用大气传输模型MODTRAN 模拟了大量复杂情 况下的地表短波净辐射和通道辐射亮度。拟合地表短波净辐射与大气顶层观测值的非线性关 系,分为两个步骤:(1) 窄通道表观反射率向宽通道反照率转换;(2) 利用大气顶层向上辐 射通量与地表短波净辐射的非线性关系反演地表短波净辐射。构建完地表短波净辐射反演模 型后,利用地表辐射观测网(SURFRAD)7 个地面实测站点长时间序列数据,结合MODIS/ AQUA 遥感数据,对地表短波净辐射反演模型进行验证。【结果】反演模型构建过程中的步 骤一结果表明,其偏差Bias 为0,均方根误差RMSE 为0.011,决定系数R2 为0.997;步骤 二的偏差Bias 为0 W/m2,均方根误差RMSE 为7.29 W/m2,决定系数R2 为0.999。利用地 面站点数据验证反演模型的结果表明,总体偏差Bias 为-4.8 W/m2,均方根误差RMSE 为 77.1 W/m2,决定系数R2 为0.9106,并且超70% 样本的反演误差低于50 W/m2。【结论】以 MODIS/AQUA 为驱动数据源,地表短波净辐射反演模型的构建精度和验证精度良好,具有一 定的泛化能力和普适性。 %X [Purpose]Net Surface Shortwave Radiation(NSSR)is a key component of many global,regional climate,hydrological,and surface process models. However,it is difficult to accurately estimate NSSR from remote sensing data,so a high accurate NSSR retrieval model with a certain generalization ability is needed.[Method]The atmospheric transmission model MODTRAN was used to simulate NSSR and band radiation in various conditions. Consequently,the nonlinear relationship between NSSR and observations at the top of atmosphere(TOA)was fitted by two steps:(1)Converting narrowband reflectivity to TOA broadband albedo;(2)Retrieving NSSR from TOA upward shortwave flux. After building NSSR retrieval model,the measurements of seven SURFRAD stations were used to validate the proposed model with MODIS/AQUA data.[Result]The result of first step in the NSSR retrieval model shows that the bias,root mean square error(RMSE)and R2 for all simulated data in this model are 0,0.011 and 0.997,respectively. The bias,RMSE and R2 for second step of the proposed model are 0 W/m2,7.290 W/m2 and 0.999,respectively. The proposed retrieval method of NSSR was also validated by applying it to MODIS/AQUA data and ground-based measurements,which has an overall bias value of -4.8 W/m2,an RMSE value of 77.1 W/m2 and an R2 value of 0.9106. The error histogram of validation shows that approximately 70% of the absolute difference are below 50 W/m2.[Conclusion]With MODIS/AQUA data as the driving source,the accuracy of development and validation of NSSR retrieval model are both pretty good,which shows that the proposed model is an efficient model with certain generalization ability and universality. %R 10.12105/j.issn.1672-0423.20190103 %U http://www.cjarrp.com/zgnyxx/ch/reader/view_abstract.aspx %1 JIS Version 3.0.0