中国地区MODIS气溶胶光学厚度产品综合验证及分析

讨论了2000年02月~2019年08月不同分辨率、算法中等分辨率成像光谱仪（MODIS）气溶胶光学厚度（AOD）产品在不同季节、区域、下垫面条件下中国区域的精度和适用性，在像元尺度上对比了重采样降低分辨率后的多角度大气校正（MAIAC）AOD数据与其他产品的精度.研究表明：在同等验证条件下，相较于暗目标算法（DT）、深蓝算法（DB）和暗目标与深蓝结合算法（DTB），1km MAIAC AOD产品在中国地区与AERONET站点AOD观测数据整体一致性最高，R²达到0.891，均方根误差（RMSE）仅为0.126，超过75%的验证样本落在期望误差线（EE）范围内；同时，该产品受季节、区域和下垫面变化影响也最小，其中秋季R²达到0.917，RMSE为0.111，样本落在EE内的比例达到80.11%.3km DT算法AOD产品在植被覆盖率较大的森林和农田区域优于1km MAIAC AOD产品，在植被覆盖率较小的草地和城市区域则差于1km MAIAC AOD产品，且该产品在不同季节均存在AOD高估问题，其中，夏季高估程度最高（平均相对误差（RMB）=1.622，AOD值高估62.2%）.DB在长三角和珠三角地区存在AOD被低估的现象.DTB算法兼顾了DT算法和DB算法的优缺点，DTB算法AOD产品的相关性一般高于DB算法AOD产品，样本被高估程度一般低于DT算法AOD产品.通过重采样方法降低1km MAIAC AOD产品分辨率后，相同尺度下的MAIAC AOD数据精度优于DT算法、DB算法和DTB算法AOD产品，因此，MAIAC算法更适用于小尺度城市群集中区域的大气环境监测.

Validation and analyzation of MODIS aerosol optical depth products over China

A new multi-angle atmospheric correction (MAIAC) algorithm was applied for the Moderate-resolution Imaging Spectroradiometer (MODIS) to provide global aerosol optical thickness (AOD) product with 1km resolution. The dark target (DT), the dark blue (DB) and the combined DT and DB (DTB) algorithms have also been optimized and updated. In this paper, the accuracy and applicability of 5MODIS AOD products with different resolutions over China were verified and analyzed in terms of season, region and landcover type. Moreover, the resampled MAIAC AOD data was compared with other AOD products on the pixel scales. Our results showed that compared with DT, DB and DTB AOD products, 1km MAIAC AOD product had a best agreement with the ground-based AERONET measurements with overall R² of 0.891, Root Mean Squared Error (RMSE) of 0.126 and=Expected Error (EE) larger than 75%, and its accuracy was least impacted by season, region and landcover type. Especially, the R² of 1km MAIAC AOD reached to 0.917 and the RMSE was only 0.111, and there was 80.11% of 1 km MAIAC AOD falling in the expected error envelops. The reliability of 3km DT AOD was higher than that of 1km MAIAC AOD over the regions larger fraction of vegetation cover (e.g. forest and cropland), while it was lower than that of 1km MAIAC AOD for grassland and urban regions with smaller fraction of vegetation cover. 3km DT AOD was always overestimated, especially, the most overestimation of 3km DT AOD by 62.2% (RMB=1.622) was in the summer. Although the DB algorithm was more applicable for the regions with bright surface (e.g. urban regions), the DB AOD product showed a significant underestimation over Yangtze River Delta (YRD) and Pearl River Delta (PRD) regions. The DTB AOD product combined the advantages of DT and DB AOD products. The correlation between the DTB AOD and the AERONET was always higher than that between the DB AOD and the AERONET AOD, and the overestimation of DTB AOD retrievals was lower than that of DT ones. With the same resolutions, the accuracy of the MAIAC AOD was better than DT, DB and DTB AOD products. Therefore, the MAIAC algorithm was more suitable to the air monitoring of urban agglomerations.