Abstract: By means of the Doppler radar measurements and automatic precipitation station data collected during nine precipitation processes in the Yishu River watershed in 2005 and 2006, the improved window probability matching method (WPMM), improved genetic algorithm (IGA) and the optimization method (OM) are employed to determine the relationship between radar echo intensity (Z) and precipitation intensity (R). The evaluation of the Z－R relation from WPMM, IGA and OM is analyzed and compared with that of the empirical Z－R relation. The optimum Z－R relation is used to estimate the regional precipitation. The regional precipitation is estimated and quantitatively compared with those observed by automatic precipitation observatories of the region by using average calibration, Kalman filter, optimum interpolation, variation method, optimum Kalman filter (combine Kalman filter with optimum interpolation) and variation Kalman filter (combine Kalman filter with variation method). The results suggest that if the value of ground raingauge is taken as truth, algorithms of the optimum Kalman filter and the variation Kalman filter perform best and excess the algorithm of the least root mean square errors, the Z－R relation method is superior than the largest root mean square errors algorithm, the precision of optimum interpolation and variation method is higher than that of average calibration and Kalman filter. For regional rainfall estimation, it is found that the spatial rainfall distributions are in good agreement with those interpolated by raingauge network measured precipitation to a large extent, however, the intense rainfall centers exhibit different patterns compared with those interpolated with raingauge network measurements, the algorithm of Z－R relation yields the largest mean relative error of about 70.51% among the selected algorithms. After adjustments are made to the radar estimated precipitation by using in site measurements with the automatic raingauge networks, radar rainfall estimations were improved dramatically on precision either in the spatial distribution or in the location of intense precipitation centers, the mean relative errors of the estimated regional rainfall by each above-mentioned algorithm adjusted respectively with in site raingauge measurements are evidently less than that by only Z－R relation algorithm being employed. The mean relative error for algorithm of the average calibration, Kalman filter, variation method, optimum interpolation, optimum Kalman filter and variation Kalman filter being about 16.55%, 16.27%, 13.44%, 13.86%, 13.16%, 13.51%, respectively. The optimum Kalman filter performs best, can truly reflect the precipitation status over the ground surface, and might be used to estimate the regional precipitation for the study region.