干旱区生态环境知识资源中心

This study evaluates Moderate Resolution Imaging Spectroradiometer (MODIS) Aerosol Optical Depth (ACID) retrievals with ground measurements collected by the China Aerosol Remote Sensing NETwork (CARSNET). In current stage, the MODIS Collection 5 (C5) AODs are retrieved by two distinct algorithms: the Dark Target (DT) and the Deep Blue (DB). The CARSNET AODs are derived with measurements of Cimel Electronique CE-318, the same instrument deployed by the AEROsol Robotic Network (AEROENT). The collocation is performed by matching each MODIS ADD pixel (10 x 10 km(2)) to CARSNET AOD mean within 7.5 min of satellite overpass. Four-year comparisons (2005-2008) of MODIS/CARSNET at ten sites show the performance of MODIS AOD retrieval is highly dependent on the underlying land surface. The MODIS DT AODs are on average lower than the CARSNET AODs by 6-30% over forest and grassland areas, but can be higher by up to 54% over urban area and 95% over desert-like area. More than 50% of the MODIS DT AODs fall within the expected error envelope over forest and grassland areas. The MODIS DT tends to overestimate for small ADD at urban area. At high vegetated area it underestimates for small ADD and overestimates for large ADD. Generally, its quality reduces with the decreasing ADD value. The MODIS DB is capable of retrieving ADD over desert but with a significant underestimation at CARSNET sites. The best retrieval of the MODIS DB is over grassland area with about 70% retrievals within the expected error. The uncertainties of MODIS ADD retrieval from spatial temporal collocation and instrument calibration are discussed briefly. (C) 2011 Elsevier Ltd. All rights reserved.