Bias in MODIS cloud drop effective radius for oceanic water clouds as deduced from optical thickness variability across scattering angles

The Moderate Resolution Imaging Spectroradiometer (MODIS) represents our most advanced system for long-term, global observations of cloud optical depth (τ) and cloud drop effective radius (Re). Here we showthat the both zonal means of (1) MODIS-retrieved τ and Re and (2) τ retrieved fromtheMulti-angle Imaging Spectroradiometer (MISR) using MODIS-retrieved Re as input exhibit a large, local negative bias in the rainbow scattering direction relative to adjacent scattering angles for oceanic water clouds. Using radiative transfer simulations, we demonstrate that these local biases are explained by overestimates in MODIS-retrieved Re. While 3-D factors on the angular variability of τ are evident, we conclude that a Re bias is itself an important contributor to this variability. With 3-D factors present, we show that upper and lower bound estimates on the biases in the three MODIS Re products, namely Re_1.6, Re_2.1, and Re_3.7 retrieved from 1.6, 2.1, and 3.7 µm channels, respectively, are possible from a combination of MISR and MODIS data. The midpoints of the upper and lower bounds in zonal mean biases are estimated to be 3-11 µm for Re_2.1 and Re_1.6 and 2-7 µm for Re_3.7, depending on latitude, with smaller biases at latitudes wheremore stratiformclouds contribute to the total low cloud cover. Removing the bias from the zonal means reveals that Re_1.6, Re_2.1, and Re_3.7 are more similar with each other compared to their original values. The change in zonal mean of Re as a function of latitude is also very different between original and bias-corrected products. These differences change the interpretation of the vertical and horizontal distribution of Re from the original MODIS product.