Polarimetric Radar Verification of GPM Satellite-Based Retrievals of the Raindrop Size Distribution

The Global Precipitation Measurement (GPM) mission is an international mission led by NASA and the Japan Aerospace Exploration Agency (JAXA). Of central importance to GPM and its multi-satellite constellation is the GPM “Core” Observatory, a reference platform carrying a multi-frequency passive microwave imaging radiometer (GMI) and the Dual-Frequency Precipitation Radar (DPR). Relative to the DPR, NASA GPM mission science requirements define specific measurement error standards for retrieved precipitation parameters such as rain rate, raindrop size distribution (DSD), and falling snow detection at spatial resolutions ranging from instantaneous instrument fields of view [IFOV], to grid scales of 50 km x 50 km. Of particular importance to GPM rainfall retrieval algorithms at all scales are DPR-based estimates of DSD parameters such as the mass-weighted mean diameter (D_m) and the normalized intercept (N_w) of a three-parameter gamma raindrop size distribution (shape parameter fixed). Hence, in this study we discuss GPM Ground Validation (GV) polarimetric radar estimates of the DSD and their subsequent use for assessing relative agreement between GV- and GPM DPR-based DSD estimates.

GV comparisons to DPR DSD retrievals rely heavily on a) the use of 65 U.S. National Weather Service WSR-88D S-Band polarimetric radars; and b) networked data collections from numerous NASA GPM 2D Video Disdrometers (2DVD) deployed in myriad field locations and precipitation regimes. 2DVD data are processed to develop empirically-based DSD retrieval equations based on observed DSDs and modeled radar reflectivity (Z) and differential reflectivity (ZDR). In turn, the 2DVD-derived DSD equations are used with polarimetric radar-observed Z and ZDR to provide spatially distributed DSD estimates under individual GPM Core overpasses. In this context the U.S. operational WSR-88D radars provide a constant heartbeat of volume scanning across a wide range of latitudes, longitudes, precipitation rates, and DSD characteristics under a large sample of GPM satellite overpasses. The broad sample of 2DVD field datasets ensures inclusion of representative DSD regimes that can be used to create ensemble “global” relationships between Z, ZDR, D_m and N_w, or as a means to model specific regional or meteorological DSD regimes. Supplemental research radar datasets (e.g., NASA NPOL, CSU-CHILL etc.) are also used on a target of opportunity basis to examine GPM-diagnosed DSDs in specific field campaigns or sampled precipitation regimes. All polarimetric radar data are automatically quality controlled, DSDs computed, and the data subsequently geometrically volume-matched to DPR sample volumes for rain observed in the layer below the melting layer and above ground clutter for each ray of a given IFOV, using GPM Validation Network (VN) ground radar and satellite data processing software.

GPM Level-1 (L1) science requirements for DSD retrieval are specified with respect to Dm. Specifically, the L1 requirement is that DPR-retrievals estimate Dm to within +/- 0.5 mm of GV. Indeed, our results indicate that GPM is meeting its L1 DSD estimation requirement. Version-4 GPM DPR D_m products typically compare to ground-based products with a systematic bias of order +0.1 to +0.2 mm and random error of ~0.2-0.3 mm, depending on algorithm type and version (DPR version-5 being 0.1 mm more positively biased than version-4). Importantly, we find that agreement between GV and DPR D_m (and N_w) estimates is also a strong function of precipitation type (convective or stratiform); a slightly larger positive bias being noted in convective precipitation overall and larger secondary modes of disagreement occurring in convective precipitation at large DPR-Dm (2.5 mm or greater). Though a metric for N_w is not defined in GPM science requirements, similar comparisons between DPR and GV-diagnosed N_w are noisier than those of D_m. Here the DPR algorithms tend to cluster around LOG(N_w) values of ~3.0 - 3.5 in stratiform rain, but exhibit more dynamic range and covariance with GV in convective precipitation. Collectively, we suggest that the comparison between the ground-based polarimetric radar and space-based radar diagnosed Dm is reasonable, meeting GPM Level 1 science mission requirements. Comparisons between the GV and space-based N_w estimates are not as robust, but encouraging nonetheless. It is expected that improved methods of estimating the DSD for both ground based radar and satellite-based retrieval algorithm approaches will result in even stronger convergence between ground and space-based estimates. Given the fundamental nature of the DSD in radar-based rainfall estimation this should also result in superior ground and space-based remote sensing measurements of rainfall across the globe.