Study and tests of improved rain estimates from the TRMM precipitation radar

Rain rate R estimation from the 2A-25 profiling algorithm of the Tropical R ainfall Measuring Mission (TRMM) precipitation radar (PR) is analyzed in tw o ways. Standard results from the operating version-5 algorithm are compare d with those from the previous version 4. Also, various adjustments of the involved rain relationships in version 4 are explored, which leads to the p roposal of two alternatives to the standard rain rate (Rstd-V4). The first one, (R-N0), is based on N-0(*)-scaled relations exploiting the concept of normalized Gamma -shaped drop size distributions; the second one, (R-kR), r elies on using constant R-k instead of constant R-Z relation as in the stan dard, where Z is reflectivity and k is attenuation coefficient. Error analy sis points out a lower sensitivity of the alternative estimates to errors i n radar calibration, or initial relations, than the standard. Results from a set of PR data, over ocean and land, show that the version-4 alternatives , and version-5 standard (Rstd-V5), produce more rain than the version-4 st andard, which may correct for some reported underestimation. These approach es are tested via point-to-point comparisons of 3D PR-derived Z and R field s (versions 4 and 5) with "reference'' fields derived from airborne dual-be am radar on board a National Oceanic and Atmospheric Administration P3-42 a ircraft in Hurricanes Bonnie and Brett, for good cases of TRMM overpasses o ver the ocean. In the comparison domains, Bonnie is dominated by stratiform rain, and Brett includes convective and stratiform rain. In stratiform rai n, the mean difference in Z, accounting for different frequencies and scann ing geometries of both radars, lies within the uncertainty margin of residu al errors in the radar calibrations. Also, the PR mean rain-rate estimates, R-kR and Rstd-V5, agree fairly well with the P3 estimate, R-P3, whereas Rs td-V4 and R-N0 respectively underestimate and overestimate R-P3. In convect ive rain (Brett case), the PR estimates of Z and R largely exceed the P3 co unterparts. It is suggested that this may result from a corruption of the s urface-reference estimation of the total path attenuation caused by strong surface winds.