Effects of Ice Particle Representation on Passive Microwave Precipitation Retrieval in a Bayesian Scheme

A physically based Bayesian passive microwave precipitation retrieval requires an accurate forward radiative transfer model along with realistic database representation of hydrometeors, atmospheric properties, and surface emission. NASA's Global Precipitation Measurement (GPM) Mission provides an unprecedented opportunity for the development of such databases, matching a well-calibrated radiometer with dual-frequency radar. Early versions of passive microwave products from GPM utilized a physically constructed database in a Bayesian retrieval scheme, assumed ice particles to be spheres, and used Mie radiative transfer. A large body of recent work demonstrates that this is insufficient for retrieval at the GPM radiometer frequencies. In this paper, the retrieval is updated to use nonspherical particles. Simulated brightness temperature (Tb) agreement with observations is shown to be significantly improved across the high frequencies, decreasing biases significantly and increasing correlations to observed Tb. This is compared with a second identical retrieval performed with the assumption of spherical ice particles, and retrieval results are compared globally, seasonally, and instantaneously for a case study at the rain rate level. While not at the high level of improvement shown in Tb space, the precipitation retrieval is improved as compared to one using observed Tb in correlation, bias, and root-mean-square error. Reported improvements, while modest in magnitude, advance the retrieval to more physical consistency which allows for deeper insight into ice particle properties associated with precipitation.