Evaluation of Aerosol Optical Depth Products From Multiangular and Polarized Satellite Measurements Over Mountainous Areas

This article aims to determine the factors that introduce more errors to mountainous aerosol optical depth (AOD) in the hope that such results can guide aerosol inversion over mountainous areas. AODs retrieved by the Generalized Retrieval of Aerosol and Surface Properties (GRASP) algorithm from POLDER-3 are compared with those obtained from Aerosol Robotic Network (AERONET) over global mountainous areas between 2005 and 2013. The comparison between the AODs from GRASP/POLDER-3 (AODGP) and AERONET (AODA) is used to test the performance of the GRASP algorithm. It was found that the accuracy of AODGP is significantly different over variable underlying surfaces with different terrain types, and mountainous AODGP is seriously overestimated. The AODGP values are much less consistent with AODA over mountainous than urban areas for almost all aerosol types, which suggests that the terrain size has a significant impact on aerosol remote sensing. Mountainous AODGP in low latitudes performs better than that in midlatitude as the former is less disturbed by shelters and the shadows of mountain bodies. The performance of AODGP as a function of altitude and slope is similar, where the correlation decreases and the overestimation percentage increases quickly with an increase of slope or altitude. The observation geometry and aerosol profile change as a function of slope and altitude, which impart inaccuracies in the retrieved AOD. Overall, the terrain size is dominant in the multiangular and polarization remote sensing of aerosols and the POLarization and Directionality of the Earth's Reflectances (POLDER) measurements are sensitive to mountain terrain size. Mountainous slope and altitude disturb the AODGP most seriously, followed by latitude and seasonal variation.