In this study, a physically-based statistical model was developed to predict the summer (June-August) precipitation in the mid-latitude East Asian region. A Multivariate Empirical Orthogonal Function (MEOF) model was used to build the predictand, the East Asian summer monsoon. The four leading modes were used to define the East Asia summer monsoon and as predictands of the prediction model. The first two modes mainly represent a teleconnection forcing of Western North Pacific Subtropical High (WNPSH) on the mid-latitude East Asia while the rest represent the mid-latitude processes. The reconstructed data using four modes explains 54% of the total variance.

Predictors for each mode were selected by a stepwise forward regression method. Predictors were selected from sea level pressure, surface air temperature, sea surface temperature, snow cover, and sea ice fields in winter and spring seasons. The leading correlation maps between these leading variables and MEOF principle components (PCs) were analyzed to find suitable predictors. Predictions were conducted through a multiple regression method using these selected predictors for verification of prediction. Hindcast data was generated through the 3-leave-out cross-validation method. Hindcast data and observation precipitation are compared and the correlation coefficients of these four modes are 0.79, 0.68, 0.82, and 0.64, respectively. The predicted PCs and spatial patterns from observation were reconstructed to generate the prediction spatial field of precipitation and the Temporal Correlation Coefficients (TCC) in this domain between observation and prediction fields were averaged (0.41).

The prediction spatial fields can give information in a smaller regional scale in the given domain. The forecast skill from TCC over South Korea indicates moderate value (0.40). However, the ground-truth Korean station precipitation observations is inhomogeneous with the satellite observation data, which is used to build the statistical prediction model. Thus, the Korean station precipitation data was optimized using predicted PCs to reproduce the Korean precipitation in this prediction model. The correlation coefficient between Korean station observation and predicted precipitation is 0.38, which gives a way to utilize the East Asian summer monsoon prediction model for local prediction.

1. 서론

2. 연구 자료 및 방법

2.1 연구 자료

2.2 방법

2.2.1 동아시아 여름몬순의 정의

2.2.2 전방향 걸음별 회귀 방법

2.2.3 후측모의 검증

3. 중위도 동아시아 여름몬순과 역학적 메커니즘

3.1 첫 번째 모드: 북서태평양 고기압

3.2 두 번째 모드: WNPSH와 남북방향 파동 전파

3.3 세 번째 모드: 오호츠크해 저기압, 중태평양 엘니뇨와 NAO

3.4 네 번째 모드: 북태평양 고기압의 강화

4. 중위도 동아시아 여름몬순의 예측

4.1 예측변수의 선정

4.2 예측식과 검증

5. 한국 강수 예측 모형

6. 요약 및 결론