83rd Annual

Tuesday, 11 February 2003
Preliminary Results From a Regression Based Technique for Multisensor Rainfall Estimation
Chandra R. Kondragunta, NOAA/NWS, Silver Spring, MD
In this study, an attempt was made to estimate rainfall using rainfall information from several sources including rain gauges, remotely sensed radars and satellites. The data used were collected over the West Gulf River Forecast Center (WGRFC) region. The rain gauge data used were the hourly reports received at the WGRFC for operational use. The radar rainfall estimates used were hourly products from the WSR-88D Precipitation Processing System. The satellite estimates used were operational rainfall products from the Geostationary Operational Environmental Satellite produced by the National Environmental Satellite Data and Information Service. In addition to these data, lightning data from the National Lightning Detection Network were also used. Hourly radar, satellite and lightning data collocated with rain gauges were collected during the warm season between June 2001 and April 2002, and were called combined cases. From these combined cases, rain gauges in the vicinity of lightning were called convective cases and the remaining ones were called non-convective cases. Multiple regression based models were built for each of these three types of cases. Radar, satellite, lightning and various multiplicative combinations of these data were used as model parameters. Preliminary results indicate that the convective model showed improved performance compared to the other two models. In terms of RMS error, the convective model showed an improvement of 12% and 35% when compared to radar and satellite estimates, respectively. In order to be able to estimate rainfall in situations when radar only or satellite only estimates are available, two more sets of models using parameters associated with radar only and satellite only will be developed. These results and some validation results will be presented.

Supplementary URL: