Accounting for 3D radiative effects in MODIS aerosol retrievals near clouds using CALIPSO observations

Abstract

Retrieval of aerosol properties near clouds from passive remote sensing is challenging. Sunlight scattered by clouds into nearby clear regions can effectively enhance the clear area reflectance. These cloud 3D radiative effects may lead to large biases in aerosol retrievals if uncorrected, risking the incorrect interpretation of satellite observations for aerosol–cloud interaction in a cloudy atmosphere. In earlier studies, we developed a simple two-layer model (2LM) to estimate the cloud-induced clear-sky radiance enhancements in cloud fields. In this study, we take advantage of CALIPSO lidar observations, which should not be affected by the 3D radiative effect, to study passive aerosol retrievals in cloud fields in the Amazon region, specifically those produced by the operational Dark Target algorithm applied to Aqua-MODIS. From 2 years’ worth of co-located CALIPSO/MODIS aerosol retrievals, we find a larger increase in operationally retrieved MODIS AOD from clear to cloudy regions (∼0.075 or ∼40%) than for the CALIPSO AOD (∼0.021 or ∼20%). The much larger increase in MODIS AOD is mainly due to the 3D radiative effects. After using the 2LM model to account for cloud 3D radiative effects, the clear to cloudy increase in MODIS AOD was reduced to ∼0.043 (∼23%), which is much closer to CALIPSO observations. The 3D corrected average MODIS AOD for cloudy conditions is significantly larger than AOD for clear conditions, even for cloud fraction (CF) less than 0.1, suggesting aerosols in cloudy conditions are characteristically different from aerosols in clear conditions. Furthermore, the 3D correction of AOD (i.e., τ1D−τ3D) increases linearly with CF for a large range of CF. We have also examined the impact of the 3D effect on aerosol Ångström Exponent (AE) and fine model fraction (FMF) of AOD. We found that the uncorrected average AE and FMF depend strongly on CF, ∼25% increase in AE (decrease in particle size) and ∼60% increase in FMF as CF increases from 0.05 to 0.45. The 3D correction leads to smaller average AE (bigger particle size) and FMF that are almost independent of CF. Thus, the 3D corrected aerosol properties are expected to provide more accurate information for better understanding aerosol–cloud interactions.