Satellite dual-frequency millimeter-wave radar retrieval algorithm

In this study, we formulate and investigate the performance of an algorithm to retrieve precipitation from satellite radar reflectivity observations at 36 (Ka-band) and 96 GHz (W-band). The algorithm is based on 1) a Hitschfeld–Bordan attenuation correction procedure that yields generic Ka-only precipitation profile estimates and 2) an optimization procedure that adjusts the Ka-band estimates to be physically consistent with coincident W-band reflectivity observations and surface reference technique–based path-integrated attenuation estimates at both Ka and W bands. The algorithm is investigated using airborne multiple frequency radar observations collected during the Integrated Precipitation and Hydrology Experiment (IPHEX) field experiment. The IPHEX airborne radar dataset is particularly insightful because three NASA airborne radar systems, i.e. the HIWRAP Ku-Ka band, the CRS W-band and the EXRAD X-band radars, have been deployed during the experiment, which facilitate the derivation of accurate precipitation estimates that may be use to quantify the uncertainties in dual-frequency millimeter wave radar retrievals. Specifically, triple frequency (X, Ka and W-band) ice phase retrievals are expected to be more accurate than dual-frequency Ka-W-band retrievals because the X-Ka dual frequency reflectivity ratio (DFR) is more unequivocally related to particle sizes than the Ka-W DFR. In addition, the attenuation at X-band is very small for stratiform precipitation, which facilitates the derivation of more accurate estimates than from Ka and W-band-only observations below the freezing level. Therefore, triple frequency airborne retrievals, derived from X-band generic rain estimates optimized to be consistent with both Ka- and W-band radar observations are expected to be less uncertain than dual-frequency Ka-W radar retrievals and may be used as reference in evaluating the performance of dual-frequency Ka-W radar retrievals. Based on the differences between triple and dual frequency retrievals, it is concluded that the Ka-W retrieval algorithm performs satisfactorily if appropriate regularizations terms describing the “a priori” distribution of retrieved variables are incorporated.