Statistical Prediction of East Asian Summer Monsoon Rainfall Based on SST and Sea Ice Concentration

Abstract

The canonical ensemble correlation (CEC) prediction method is applied to build a statistical prediction model of East Asian summer monsoon (EASM) rainfall. The predictors are regional sea surface temperature (SST) and sea ice concentration (SIC) fields in selected optimal periods of time. Results show this CEC model is much more skillful than the commonly used canonical correlation analysis (CCA) model which uses only winter tropical Pacific SST as predictor. The skillful regions of the new model cover almost half of the area of East Asian land, and from 1980 to 2005, there are 20 years bearing significant spatial pattern correlation. The careful selection of predictors is the reason of the high skill achieved. By dividing the global ocean into five ocean basins and selecting five regions with large variation of SIC, the CEC prediction based on these regional SSTs or SICs can recognize more regional and weaker forcing from the sea. Moreover, for each SST and SIC region, the optimal period of time leading to the most skillful CCA forecast of EASM rainfall is a special month or combination of two months from winter and spring respectively instead of seasons. The selection of optimal periods of time is the key for the good performance of the CEC prediction by improving the forecast skill of its ensemble members.