The global land fAOD dataset (2001–2020) using DLFE-Satellite

The fAOD is a vital proxy for the concentration of anthropogenic aerosols in the atmosphere. Global fAOD can be obtained from point-scale Aerosol Robotic Network (AERONET) measurements, but the spatial coverage is limited. To obtain spatially-continuous global fAODs, we applied the DLFE-Satellite method with Moderate Resolution Imaging Spectroradiometer (MODIS) data, ERA5 climate reanalyses, and AERONET data from 2001 to 2020 as inputs. For independent validation, we used fAOD retrievals obtained from six Surface Radiation budget network (SURFRAD) sites in the USA. Note that SURFRAD sites are in the vicinity of AERONET sites and were not used to train the model. Data from these sites could thus be used for the independent validation of fAOD estimates. The linear relation between DLFE-Satellite and AERONET retrievals had an R value of 0.93 and RMSE of 0.09 μg/m³; for validation with SURFRAD retrievals , R = 0.77 and RMSE = 0.14 μg/m³.fAOD是大气中人为气溶胶浓度的关键替代表征指标。全球尺度的fAOD可通过点尺度的气溶胶机器人观测网络（AERONET）观测获取，但其空间覆盖范围有限。为获取空间连续的全球fAOD数据，本研究以2001至2020年的中分辨率成像光谱仪（MODIS）数据、ERA5气候再分析资料及AERONET观测数据作为输入，应用DLFE-Satellite方法进行反演。为开展独立验证，本研究使用了美国境内6个地表辐射预算网络（SURFRAD）站点获取的fAOD反演结果。需注意，SURFRAD站点均位于AERONET站点附近，且未参与模型训练过程。因此，这些站点的数据可用于fAOD反演结果的独立验证。DLFE-Satellite方法反演结果与AERONET观测反演结果之间的线性相关系数R为0.93，均方根误差（Root Mean Square Error，RMSE）为0.09 μg/m³；与SURFRAD观测反演结果进行验证时，相关系数R为0.77，均方根误差为0.14 μg/m³。