Based on the station data, reanalysis data and dynamic climate model hindcast data, a dynamic-statistical downscaling prediction method of annual precipitation anomaly in the Yangtze River Basin and its application skill are discussed by using the empirical orthogonal decomposition (EOF) iteration and interannual increment method.Results show that based on the annual scale circulation field of reanalysis data, a statistical downscaling prediction scheme for annual scale precipitation anomaly increment over the Yangtze River Basin is established.The average anomaly correlation coefficient (ACC) of 26-year hindcast test can reach 0.6, which proves that the scheme has high predictability.A dynamic-statistical downscaling prediction scheme of annual precipitation anomaly increment is furth erestablished by using the annual scale circulation predicted by the model.The average ACC is 0.42, showing a high hindcast skill.The skill is much better than that of the directly output precipitation of the model.By analyzing the factors affecting the skill of annual precipitation prediction, it shows that when the annual average SST anomaly in equatorial central and eastern Pacific is negative, the prediction skill is higher, and the average ACC is more than 0.5.Under the cold water background of La Niña development year or La Niña duration year, more eigenvectors are selected by EOF iteration, which are incorporated into the multi-scale atmospheric circulation information as the prediction signal, and the prediction skill of annual precipitation anomalyis improved.