Using active spaceborne sensors to quantify bias in passive remote sensing of the cloud liquid water path

The cloud liquid water path is a first order descriptor of clouds, which links to Earth's water and energy cycles. Despite its importance, climate models show a large spread in their estimates of the cloud water path. One reason for this spread is the lack of accurate observations on the global scale. Over oceans, passive remote sensing techniques using emitted microwave or reflected solar radiation provide long term cloud liquid water path data sets. These data sets show large biases even when filtered for ideal remote sensing conditions. Reconciling these biases is of paramount importance to provide an accurate benchmark against which to compare global models. Here we use the active CloudSat radar and CALIPSO lidar instruments to understand and quantify biases between passive observations of cloud liquid water path derived from the MODerate resolution Imaging Spectroradiometer (MODIS) and the Advanced Microwave Scanning Radiometer for EOS (AMSR-E). This unique combination of instruments permits an accurate assessment of the bias in MODIS resulting from failed cloud detection or cloud property retrievals while also permitting an assessment of the influence of light precipitation on the AMSR-E estimates. The biases in MODIS due to retrieval failure are shown to be non-negligible but small due to the low liquid water paths associated with these cases. Precipitation biases in AMSR-E are also shown to be small due to clever but ad-hoc tuning of the AMSR-E retrieval algorithm and the relative infrequency of precipitation. These results confirm that it is not the difficult remote sensing cases from which the MODIS/AMSR-E biases result but rather subtle and persistent effects related to cloud morphology and environmental state. These comparisons highlight a path forward for multi-sensor cloud retrievals utilizing the full suite of passive and active sensors that could provide estimates of cloud liquid water path consistent with all the relevant observations.