The Development of Ensemble Statistical Prediction Model for Changma (EASM) Rainfall

Thursday, 18 December 2014
JinYong Kim, Kyong-Hwan Seo and Jun-Hyeok Son, Pusan National University, Busan, South Korea
Several statistical forecast models for predicting Changma rainfall are developed. The forward-stepwise regression method selects the potential predictors that satisfy a few criteria. The potential predictors selected are springtime sea surface temperature (SST) anomalies over the North Atlantic (NA1), the North Pacific (NPC) and the tropical Pacific Ocean (CNINO), the Northern Indian Ocean (NIO), the Bering Sea (BS), the Western North Pacific (WNP), and the spring Eurasian snow cover anomalies (EUSC). Using these predictors, four regression models are developed and compared with observations. The correlation skill score for the weighted ensemble mean forecast amounts to as high as ~0.89 for last 20 years. The Gerrity skill score calculated from a 3 by 3 contingency table is also as high as ~0.86. The physical process associated with each predictor has been discussed. The five predictors (NA1, NPC, CNINO, WNP and NIO) affect Changma precipitation by tropical and extra-tropical forcing. These factors tend to develop or strengthen the anticyclonic anomalies to the south or southeast of the Korean peninsula. In contrast, the two predictors (EUSC and BS) tend to strengthen the Okhotsk high. Both anticyclonic circulation anomalies act to strengthen the Changma front.