Monday, 23 January 2012: 4:00 PM
New GSMaP Over-Land Precipitation Retrieval Algorithm
Room 256 (New Orleans Convention Center )
The GSMaP (Global Satellite Mapping of Precipitation Project) over-land algorithm finds surface precipitation rates that give forward-calculated brightness temperature (TB) depressions at 37 GHz (TB37) and 85 GHz (TB85) best fit with the Microwave Imager (MWI) observation. The forward calculation part includes radiative transfer model (RTM) and statistical models of precipitation-related variables (profile, convective/stratiform ratio, particle size distribution (PSD), etc) derived mainly from the TRMM observation. The current algorithm underestimates the TRMM Precipitation Radar (PR) surface rain (Rainsurf), in particular, for shallow precipitation events. This algorithm also shows differences in retrieval biases between stratiform and convective precipitation. The main causes of these systematic biases are the overestimation of TB37 depression in the forward calculation part and lack of consideration of variations in vertical precipitation profiles for different precipitation events with various precipitation top heights in the retrieval part. The objective of the present study is to develop the novel forward calculation part, and to introduce indices of the frozen precipitation depth and stratiform rain ratio from MWI TBs into the retrieval part. For the first purpose, we performed forward calculation experiments with variations in the precipitation-related variable models. The results showed that TB37 depression was sensitive to PSD of rain drops and depth of frozen precipitation while it had little sensitivity against other frozen precipitation properties. It was also found that TB85 was much sensitive to the frozen precipitation than TB37, and that the TB85 depression became larger than the TB37 depression for deeper frozen precipitation. Then, we applied the experimental forward calculation to the over-land precipitation retrievals from TRMM Microwave Imager (TMI). Their validation with PR Rainsurf for 1998 showed that the overall underestimation was alleviated by changing rain-drop PSD. For the second purpose, we introduced the ratio of TB85 depressions to TB37 depressions (R8537) as the index of the frozen precipitation depth. (We expressed R8537 in terms of ratio of precipitation retrieved from TB85 depression (rain85) to those from TB37 (rain37) using the conventional GSMaP algorithm.) As the index of stratiform rain ratio, we introduced the horizontal precipitation inhomogeneity derived from Rain85 (Sigma85). Then we classified the TMI retrievals with R8537 and Sigma85, and compared them with PR Rainsurf for 1998. The results show: 1) Both Rain37 and Rain85 greatly underestimated Rainsurf for cases with R8537 <1 and Sigma85 >1. 2) Precipitation with Sigma85 <1 tended to have larger scattering for cases with R8537>1, compared to those with Sigma85 >1. 3) The above relations had little seasonal variations and held good for all precipitation types, except for high land and winter-time extratropical low types that showed great underestimation of Rainsurf for all R8537 and Sigma85 ranges. Then, we derived linear fitting coefficients between Rain37, Rain85 and Rainsurf for each R8537 and Sigma85 class and precipitation type for 1998. The new retrieval part used these fitting coefficients for the calibration. We validated the performance of the new over-land algorithm using TRMM data sets for 2004. The results show that the calibration using R8537 and Sigma85 alleviated negative bias of the precipitation retrievals for shallow precipitation and high land precipitation.