Abstract:With finely zoning of the Yangtze River Basin and based on ECMWF， NCEP， CMA-GFS， GERMAN，CMA-MESO and CMA-SH9 precipitation forecast products, the rolling tests of finely zoning, time division and precipitation classification are carried out to obtain the corresponding model performance ranking. The “neighborhood optimization method” is used for heavy rain and above, and the “point-to-point optimization method” is used for moderate rain and below to establish the optimal integration scheme. The ensemble forecast and statistical analysis results are used to reduce the empty forecast rate of heavy rainfall and light rain. Finally, the multi-model dynamic integrated precipitation products (MDI) are obtained. MDI-14, MDI-21, MDI-28 and MDI-35 are produced during the 14 d, 21 d, 28 d and 35 d sliding training periods respectively. The application analysis shows that MDI have obvious advantages in precipitation forecasts, among which MDI-28 has the best effect in predicting heavy rain and above, and especially, its 24 h TS-score is 0.051-0.141 higher than that of the others; MDI-35 has the best effect in predicting moderate rain and below, with the 24 h TS-score being 0.006-0.117 higher than that of the others, and empty and missing forecast rate are also controlled well. Comparing the spatial distribution of TS scores of 24 h precipitation forecasts between MDI and ECMWF, we find that the former is higher than that of the latter in most areas of the Yangtze River Basin. Especially, for heavy rain and above, it is 0.05-0.26 higher in the east regions of the upper reaches of Yangtze River as well as in the middle and northern regions of the middle and lower reaches of Yangtze River.