Abstract:Based on statistical verification and forecast bias analysis on the rainstorm processes of landed typhoons in 2018, we improve the observation based ensemble subsetting technique that has been applied on operational forecast, to obtain objective products of typhoon precipitation with higher resolution and forecast accuracy, using multi model forecasts from several centers, the real time typhoon locations and track forecasts from National Meteorological Centre. The multi model forecasts include precipitation and typhoon track ensemble forecasts and operational high resolution forecasts from ECMWF and NCEP, the forecasts of main regional models (GRAPES Meso, GRAPES 3 km, RMAPS, SMS WARMS) in China. The results show that the spatial resolution of the improved objective product is significantly promoted compared with the previous version. The forecast accuracy is also significantly improved compared with the operational deterministic forecast from ECMWF and NCEP, and the improvement rate is 15%-20% roughly. The threat score of the improved objective product is also higher than the forecast of regional models, and slightly higher than the QPF of forecaster. In addition, studies have shown that the accuracy of the improved objective product has a high correlation with the selection of reference track. The threat score of objective product based on the average track of 3 to 5 best members is higher than the one based on the average track of all members. The observation based ensemble subsetting and QPF fusion technique based on probability matching, is helpful for improving 〖JP2〗the typhoon rainstorm forecast of the global and regional models.