Performance of the New Version of the GPM/DPR Rain Retrieval Algorithm

More than three years have passed since the launch of the GPM core satellite which carries the Dual-frequency Precipitation Radar (DPR) and the GPM Microwave Imager (GMI). The DPR consists of Ku-band radar (KuPR) and Ka-band radar (KaPR). This paper reviews the performance of the latest version (V5) of the DPR rain retrieval algorithm.

There are many changes and improvements implemented in the latest version of the DPR algorithm from the at-launch version (V3). Major changes are:

1. New calibration parameters in level 1 products

2. Robust method for estimating the drop-size distribution (DSD) parameters

3. Introduction of default initial offsets to DSD parameters

4. Use of differential path-integrated attenuation (PIA) as a constraint

5. Introduction of a few new variables and flags

The DPR has been calibrated from data obtained during several satellite overpasses of the active radar calibrator located on the ground at Tsukuba, Japan. In 2016, the data were re-examined and improved calibrations were derived. The radar parameters such as receiver gains, beam widths and pulse widths in level 1 products were adjusted in V5 based on these calibrations. As a result, values of the measured radar reflectivity factor (Z_m) increase by about 1.3 dB and 1.2 dB for KuPR and KaPR, respectively.

In the dual-frequency algorithm, a parameter called epsilon is used to adjust the DSD parameters. This epsilon parameter adjusts the default R-D_m relationship. All other DSD parameters were simultaneously adjusted in conformity with the adjusted R-D_m relations by assuming a DSD model function that follows a Gamma distribution and a fall speed model. The epsilon parameter is adjusted from its nominal value of one in such a way that the attenuation-corrected effective radar reflectivity factor (Z_e) in KuPR reproduces the corresponding measured Zm profile of KaPR. In this process, we apply an additional constraint that the PIA difference between KaPR and KuPR calculated from the retrieved Z_e profiles should agree with the corresponding PIA measured by the surface reference technique (SRT).

The epsilon values that are estimated are accumulated each month over 5 deg by 5 deg boxes and their averages are classified by surface type (land or ocean) and rain type (stratiform or convective). This constitutes the epsilon database.

This epsilon database is used to define the default initial DSD parameter (or R-D_m relationship) that is used in the single-frequency (Ku-only and Ka-only) rain retrievals. Because this database gives significantly smaller values of epsilon over land in comparison with the original default value of unity used in previous versions, the rain estimates over land in V5 have in general decreased substantially from the corresponding V4 estimates.

Some of the new flags added in V5 are flagIceHeavyPrecip, flagSurfaceSnowfall, and flagAnvil. flagIceHeavyPrecip is used to identify cases in which heavy solid precipitation exists above the -10 degrees isotherm height. flagSurfaceSnowfall indicates that the precipitation is snow (not rain) at the surface. flagAnvil flags the angle bins where the precipitation profile shows an anvil echo. These flags should be useful additions to users of the DPR data.. The classification module was improved to identify instances of precipitation associated with lake effect convection.

The paper will explain these changes and show some results of improvement realized in the V5 algorithm.