Enhancing GPM Satellite Precipitation Retrievals with Dynamic Surface Information

Accurate precipitation measurement and monitoring on a global scale are of great importance for societal and hydrological applications, as well as for larger understanding of global energy and water cycles and their variability. Due to the sparsity of available measurements at the surface, satellite platforms continue to be the best current avenue for precipitation assessment on a global scale, yet satellite rainfall products often show substantial disagreement with regional validation data, particularly at the lowest and highest rain and snow rates. This is particularly an issue over land surfaces, where commonly used passive microwave algorithms are sensitive primarily to ice scattering signals, which may not necessarily correlate with precipitation at the surface in a direct quantitative or consistent way. For this reason algorithms have historically been empirical in nature and validation projects inevitably find that various schemes perform well in areas similar to where they are calibrated and tuned, and poorly in others. This type of approach is necessary due to the high emissivity of the surface in the microwave channels, along with its highly dynamic variability, particularly when ice, snow, or liquid water is present on the surface. NASA’s Global Precipitation Measurement Mission (GPM) offers an important and unique opportunity to improve upon empirical passive microwave retrieval techniques by enhancing a constellation of passive radiometers with a core satellite that includes a collocated active precipitation radar. In this work the GPM constellation retrievals are enhanced with dynamic land surface information from a land surface model and satellite-derived vegetation index with the goal of improving algorithms by adding information beyond the atmospheric ice scattering signal. The Goddard Profiling Algorithm (GPROF) is utilized for the GPM constellation radiometers, employing a Bayesian scheme. The retrieval is evaluated as a function of the surface parameters. Algorithm changes suggested by this evaluation are tested and compared, including indexing retrievals by dynamically changing vegetation and soil moisture values and using these parameters to forward model emissivity for use as an index.