Improving the prediction skill for China summer rainfall through correcting leading modes in Beijing Climate Center's Climate System Model

Abstract

In this study, the performance of Beijing Climate Center's Climate System Model (BCC_CSM) is analysed in terms of China summer rainfall prediction, especially its capability to reproduce the spatial structure and temporal variation of the leading modes of empirical orthogonal function decomposition. Results show that the BCC_CSM has predictability for the summer mean rainfall pattern, namely, reproducing more rainfall in South China and maritime area, and less rainfall in North China and inland area. The model can reproduce the spatial pattern of the first two leading modes of observed summer rainfall; however, it fails to show correct interannual variability of the first two principle components, which may lead to a low skill in summer rainfall anomaly prediction over most parts of China. The leading mode‐based correction (LMC) method is then applied for the post processing procedure to correct prediction errors. Cross validation confirms that the LMC method can integrate the historical spatial pattern information of observed summer rainfall with the original model prediction, making the corrected model output better reflect spatial structure of the leading modes and their interannual variation. Compared with the original model output, the positive temporal correlation coefficient area for the corrected China summer rainfall anomalies is thus increased, and the average spatial correlation coefficient during 1991–2015 is improved from −0.01 before the correction to 0.17. The LMC method shows its potential for improving the prediction skill of BCC_CSM, which can be used in operational prediction.