Advancing Passive Microwave Retrievals of Precipitation Using CloudSat and GPM Coincidences: Integration of Machine Learning with a Bayesian Algorithm

Abstract Integration of machine learning with a classic Bayesian algorithm is investigated for passive microwave precipitation retrievals using coincidences from the Global Precipitation Measurement core satellite and the CloudSat cloud profiling radar (CPR). Among several machine learning models, the extreme gradient-boosting decision tree (XGBDT), equipped with a weighted cross-entropy loss function, exhibits the highest accuracy in the detection of precipitation occurrence and phase with a true positive rate greater than 94 (98)% and a false positive rate smaller than 1 (1)% for rainfall (snowfall) over land and oceans with no frozen surfaces. Bayesian retrievals in the embedding space of a fully connected multilayer perception (MLP), equipped with a focal loss function, provide the most accurate estimates of the rates with a mean absolute error of less than 1.80 (0.15) mm h −1 for rainfall (snowfall). Mutual information analysis unravels that beyond the near-surface air temperature, the 37 and 183 ± 7(3) GHz are the most informative channels for phase detection over the ocean (land). The physical consistency of the retrievals and new explanations of the precipitation passive microwave signatures are provided through partial dependence analysis and annual comparison with the reanalysis data. Significance Statement The cloud profiling radar (CPR) onboard the CloudSat satellite provided information about light precipitation and snowfall on a quasiglobal scale with unprecedented accuracy for more than 15 years. This study examines how machine learning (ML) and its integration with classic Bayesian statistical approaches can learn from coincidences of CPR and the radiometer onboard the Global Precipitation Measurement (GPM) core satellite to expand our algorithmic tools for a better understanding of precipitation global variability. New approaches are used to interpret the results of the ML model in connection with the underlying physics controlling the precipitation phase change processes and their microwave signatures. Comparisons with observationally constrained simulations from Earth system models unravel where satellite observations can provide new insights or need further improvements. The outcome paves the way to expand our understanding of the impacts of climate change on snowfall patterns and the hydrologic water cycle.