Improving seasonal prediction of summer rainfall over southern China using the BCC_CSM1.1m model-circulation increment-based dynamic statistical technique

Abstract

A model-circulation increment-based dynamic statistical technique (MIDST) is proposed in this paper to improve the prediction of summer rainfall over southern China (SC) where quite low prediction skills have been found in the Beijing Climate Center Climate System Model version 1.1 with a moderate resolution (BCC_CSM1.1m). The results show that BCC_CSM1.1m can hardly represent the variability of the summer rainfall anomaly and its year-to-year increment over SC, and the skillful predictions are mostly confined over the middle reaches of the Yangtze River. Using the dynamic model output and statistical method, the MIDST is established to capture the coupled modes between the year-to-year increments of the summer rainfall anomaly and the associated simultaneous three-dimensional coupled air-sea circulation predictors. The cross-validation indicates that the prediction skills of the MIDST are evidently improved for both the summer rainfall increment prediction and summer rainfall anomaly prediction compared with the direct BCC_CSM1.1m prediction. The skillful prediction can persist for long forecast leads over most regions except southwestern China. As the major predictability source of seasonal prediction, the intense response to changes in the circulation related to the El Niño-Southern Oscillation (ENSO) is well captured, and thus, the performance improvement of the MIDST is primarily due to its more realistic representation of the incremental circulation related to the ENSO.