Exposing biases in retrieved low cloud properties from CloudSat: A guide for evaluating observations and climate data

Global observations of low-level liquid clouds play an essential role in developing an understanding the state of the global climate. Cloud feedback processes, particularly in warm low-level clouds, are realized to be one of the most significant sources of uncertainty in predicting changes in climate using global climate model (GCM) data. To make useful comparisons between observations and simulated low-cloud properties, adequate knowledge about existing biases in the satellite record is needed. An inter-satellite comparison of low-cloud properties is examined using data from CloudSat, MODIS (Moderate Resolution Imaging Spectroradiometer), and the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO). We identify significant biases in the occurrence of retrieved LWP (Liquid Water Path) from CloudSat. Retrievals tend to be tied to problems involving cloud detection and algorithm retrieval failures related to precipitation and strict cloud screening procedures. In general, MODIS and CloudSat LWP data agree when carefully screened for lack of precipitation but significantly depart in precipitating clouds due to rainwater contamination of liquid water path in the CloudSat retrieval algorithm. The presence of drizzle and rain (occurring about 20% of the time) is associated with different mean LWP, mean particle sizes, and optical depths of all low clouds and therefore the radiative properties of the oceanic low clouds. Another more significant source of the LWP bias arises from the apparent lack of cloud detection. On average, CloudSat misses clouds with adequate liquid and ice water retrievals in approximately 45% of warm clouds when compared with observations from MODIS and CALIPSO. The bulk of the bias occurs in clouds below 1 km in the so-called “ground clutter zone.” By incorporating additional sensors such as MODIS, the following results suggest that this LWP bias can be greatly reduced.s MODIS, the following results suggest that this LWP bias can be greatly reduced.