Evaluation of passive satellite remote sensing of cloud liquid water

We compared one year of AMSR-E Wentz and MODIS cloud liquid water path estimates in warm marine clouds. In broken scenes AMSR-E increasingly overestimated MODIS, and retrievals became uncorrelated as cloud fraction decreased, while in overcast scenes the techniques showed generally better agreement but with a MODIS overestimation. We found microwave and optical retrievals being most consistent in marine Sc clouds with correlations up to 0.95 and typical rms differences of 15 g/m². The overall MODIS high bias in overcast domains could be removed, in a global mean sense, by adiabatic correction; however, large regional differences remained. Most notably, MODIS showed strong overestimations at high latitudes, which we traced to 3D effects in plane-parallel optical retrievals over heterogeneous clouds at low Sun. In the tropics/subtropics, microwave-optical differences also depended on cloud type, with MODIS overestimating in stratiform and underestimating in cumuliform clouds, resulting in large-scale coherent bias patterns where marine Sc transitioned into trade wind Cu. We noted similar geographic variations in Wentz cloud temperature errors and MODIS 1.6-3.7 ¿m droplet effective radius differences, suggesting that microwave retrieval errors due to cloud absorption uncertainties, and optical retrieval errors due to cloud vertical stratification might have contributed to the observed liquid water path bias patterns. Finally, cloud-rain partitioning was found to introduce a systematic low bias in Wentz retrievals above 180 g/m² as the microwave algorithm erroneously assigned an increasing portion of the liquid water content of thicker non-precipitating clouds to rain.