UTILIZATION OF REGRESSION MODELS FOR RAINFALL ESTIMATES USING RADAR-DERIVED RAINFALL DATA AND RAIN GAUGE DATA

Abstract

The procedure estimating hourly rainfalls by merging radar-derived rainfalls and gauge measurements is developed and tested. It uses simple linear regression, which is complemented by the normalization and correction of distribution. The data from radar Tulsa, Oklahoma, Weather Surveillance Radar-1988 Doppler version and rain gauge data from the radar domain are used. The quality of estimates is evaluated against independent rain gauges by the root-mean-square-error, bias and correlation coefficient in dependence on the density of a gauge network. The results indicate that even a sparse gauge network (about 50 gauges, i.e. 4000 km2 per one gauge) is sufficient to improve the radar-derived rainfalls. The improvement increases with the number of gauges.