Snowfall over Sea Ice: Learning from coincidences of GPM and CloudSat Satellites

Abstract:

Measuring global changes of snowfall using satellites is of paramount importance to understand global changes of cryosphere under a changing climate -- especially at high-latitudes where a dense network of ground-based gauges is lacking. Microwave remote sensing of precipitation in the solid phase is less understood than in the liquid phase, especially for light snowfall over frozen surfaces and in particular over sea ice. The reason is that the scattering signal of snowfall can be dynamically altered in the presence of snow-covered sea ice depending on its thickness, age, and snow-cover metamorphism. This study provides new insights into how the formation of sea ice and its overlying snow can affect the snowfall scattering signatures in microwave bands using the coincidences of GPM and CloudSat satellites. In particular, it is found that over sea ice, the polarization of snowfall signatures is reduced making the signatures more clustered and well distinguishable from background emission. To capture the changes of snowfall signatures over sea ice, a new Bayesian framework is introduced that equips the Goddard Profiling Algorithm (GPROF) with a k-nearest neighbor detection step to recover precipitation and its phase using a database populated with coincidences of the GPM Microwave Imager (GMI) and CloudSat Profiling Radar (CPR) observations. The results show that the new algorithm can passively detect the CPR snowfall with 92% probability while the root mean squared error (RMSE) of the retrieval remains below 0.16 mm.hr^-1 over ocean. The detection probability is increased by 5% and the RMSE is reduced to 0.11 mm.hr^-1 over sea ice.