A New Machine Learning−Based Cloud Phase Discrimination Algorithm Designed for Passive Infrared Satellite Sensors

Clouds play critical roles in the Earth’s energy budget due to their large coverage and strong radiative effect. Among various important cloud properties, cloud thermodynamic phase is a critical midway to link cloud microphysical properties with cloud optical and radiative properties. In this study, we developed a novel cloud thermodynamic phase algorithm based on a Random Forest (RF) classifier. The training (75%) and validation (25%) datasets are generated using a 3-year coincidental MODIS (onboard both Aqua and Terra) and CATS (onboard the ISS) observations. The “true” cloud thermodynamic phases are provided by a lidar on CATS and inputs are solely from MODIS thermal infrared (IR) observations and surface temperature from reanalysis. An independent 1-year cloud phase dataset from CALIPSO/CloudSat is also used for validation purpose. Our preliminary results show that the RF-based phase algorithm performs much better than current MODIS MOD06 IR Phase 1km product. In the near future, we intend to apply a similar RF-based phase algorithm to daytime clouds with a more complete spectral information (e.g., using IR and shortwave observations).