中国区域MODIS陆上气溶胶光学厚度产品检验

Validation of Aerosol Optical Thickness Product over China with MODIS Data Operated at NSMC

以我国MODIS共享网站积累的MODIS L1B数据和美国威斯康辛大学提供的IMAPP软件包气溶胶产品软件为基础,经过产品运行本地化改进处理,在国家卫星气象中心建立了气溶胶产品业务化生成和发布机制。为支持气溶胶遥感产品算法改进以及潜在用户对产品的合理应用,给出对国家卫星气象中心运行的MODIS气溶胶遥感产品质量检验分析结果。利用2005年1月—2007年5月AERONET地基气溶胶监测网的L2.0级气溶胶光学厚度产品作为真值,用它匹配MODIS陆上气溶胶光学厚度产品开展检验。检验结果表明:以卫星过境前后30 min地基观测时间平均值匹配地基站点位置10 km半径范围内的卫星反演结果空间平均值开展检验,总体样本的气溶胶光学厚度均方根误差约为0.25;满足产品误差要求(±0.05±0.20τ)的样本占总样本数的44%;气溶胶光学厚度反演结果精度具有季节和地域差异,干季(秋、冬、春)的气溶胶光学厚度误差较小,而雨季气溶胶光学厚度误差较大,云是雨季气溶胶光学厚度反演结果误差较大的主要影响因素。

The aerosol optical thickness producing software provided by Cooperative Institute for Meteorological Satellite Studies （CIMSS） of University of Wisconsin has been modified and operationally run at NSMC. And the MODIS/AOT product is shared at National MODIS Data Center. The MODIS aerosol optical thickness product has been validated with AERONET Level 2.0 aerosol optical thickness product so that the product can be improved and popularized. The MODIS/AOT product from January 2005 to May 2007 has been matched with L2.0 AOT product from AERONET stations in east Asia during the same period. The spacial average value of the MODIS/AOT within the 10 km distance from the site of AERONET sta- tion have been compared with the temporal average value of the AERONET/AOT within 30 minutes period the satellite passing the station. The validation result show that RMSE of all validation AOT samples over land is near 0.25, and about forty-four percent of the test samples meet the expected uncertainty of --4-0.05± 0. 20r. The precision of MODIS/AOT is different according to seasons and areas. Usually, RMSE of MODIS/AOT is smaller at drought season than that at rain season. Cloud is the main factor im- pacted on the large RMSE at rain season. At south evergreen vegetable area, MODIS/AOT RMSE is bet- ter than that of north seasonal changed area. MODIS/AOT at 658 nm has obviously systematical over-re- trieval since the surface reflectivity of red band is over-estimated. Absolute error of MODIS/AOT at 466 nm retrieved with blue band data has about half of the positive value and RMSE of MODIS/AOT at 466 nm is higher than that of 658 nm. The cause is that MODIS/AOT obtained with blue band is more sensitive for aerosol model used in retrieval process than it done with red band, though the estimated sur- face reflectivity of blue band has smaller error than that of red band. In a word, error from aerosol model is main cause of high random error of MODIS/AOT at 466nm. Correlation among the surface reflectivity at 0.47, 0.66 μm and 2.1 μm bands are discussed using se- lected 79 items of vegetation spectrum reflectivity observed in China by I.RCVES/CMA, in order to ana- lyze effects of surface reflectivity for retrieval MODIS/AOT and get some ideas that could improve the precision of MODIS/AOT. The vegetation spectrum reflectivity is obtained from different vegetations and from different growing period of the same vegetation. The analysis show that there is high linear correla- tion （86. 430%） between reflectivity at 0.47 μm bands and reflectivity at 0.66 μm bands for the vegetations. And linear correlation between vegetation reflectivity at 2. 1 μm and vegetation reflectivity at O. 66 μm is 68.58%. Linear correlation between vegetation refiectivity at 2.1 μm and vegetation refleetivi- ty at 0.47μm is 59.79%. It can be concluded that the scheme which decides vegetation surface reflectivity of two visible bands in algorithm of Collection 5 MODIS aerosol products agrees well with the statistics, and is instrumental for algorithm of AOT retrieval with other similar satellite sensors.