The predictability of seasonal prediction for surface air temperature (SAT) in present is not competent over the mid-latitude, especially over East Asia including the Korean Peninsula. One of the reasons for low predictability of seasonal prediction is closely associated with the unreliable predictability of the operational dynamic models with 3 ~ 4 weeks lead time over the East Asia region. Moreover it is unfeasible that the performance of the dynamic model will be improved within the short-term. Therefore, more information such as the statistical relationship between the predictant (i.e. SAT in the Korean Peninsula) and other climate factors based on the historical dataset is able to be critical for improving the predictability of the operational seasonal prediction.

In this study, the statistical prediction model for the SAT over the Korean Peninsula in individual months of the winter season is developed based on multiple regression analysis. For the stable predictability of the model, we tried to discover the climate predictors that are linked physically and dynamically to the SAT in the Korean Peninsula. The possible mechanisms of the relationship between the winter SAT in the Korean Peninsula and its predictors are also suggested in this report.

For predicting the SAT in December, three predictors such as the sea ice concentration (SIC) over the Kara-Laptev sea, the tendency of ENSO-related sea surface temperature (SST) over the eastern Pacific and Eurasian snow cover information are used. The relationship between the predictors and the SAT in December is stable during the training period. The model shows the high correlation

skill between the prediction and observational SAT (correlation coefficient : 0.75). The SAT predicted in cross validated hindcast is also highly correlated with the observational SAT (correlation coefficient : 0.71).

The models for the SAT in January and February are also constructed using predictors linked dynamically to the SAT in the Korean Peninsula such as the SIC over the upstream region, the SST variability in the tropical Pacific, the SST variability over the North Atlantic, and circulation in upstream of East Asia. Even though the correlation skills of the models for the SAT in January and February are somewhat less than the model for the SAT in December, the models also show useful skill in the correlation between predicted SAT and observation (correlation coefficients : both 0.66 in January and February).

We expect that the prediction information produced using our statistical prediction model will be useful and helpful in the operational seasonal prediction for the SAT in the Korean Peninsula.

1. 서론

2. 연구 자료 및 방법

2.1 연구 자료

2.2 연구 방법

3. 우리나라 겨울철 기온 변동과 연관된 기후감시인자

3.1 전 지구 대기순환장 및 기후요소(해수면온도, 해빙, 눈덮임)와의 동시상관성

3.2 저주파 변동(Low-frequency variability)과의 동시상관성

3.3 주요 기후인자 위상에 따른 우리나라 기온 경향

3.3.1 ENSO(El-Nino Southern Oscillation)

3.3.2 AO(Arctic Oscillation)

3.3.3 카라-바렌츠 해 해빙

3.3.4 WP(Western Pacific pattern)

4. 우리나라 겨울철 기온 예측인자 및 통계예측 모델 구축

4.1 12월 기온 예측

4.1.1 12월 기온 예측인자

4.1.1.1 카라-랍테프 해(Laptev Sea) 해빙농도

4.1.1.2 가을철 동태평양 해수면온도 경향성(Tendency) : ENSO

4.1.1.3 10월 유라시아 눈덮임속도

4.1.2 12월 기온 통계예측 모델 및 검증

4.2 1월 기온 예측

4.2.1 1월 기온 예측인자

4.2.1.1 카라 해(Kara Sea) 해빙농도

4.2.1.2 가을철 태평양 해수면온도 경향(Tendency)의 동서경도(Gradient)

4.2.1.3 타미르반도 상층 지위고도

4.2.2 1월 기온 통계예측 모델 및 검증

4.3 2월 기온 예측

4.3.1 2월 기온 예측인자

4.3.1.1 10월 바렌츠(Barents) 해 해빙농도

4.3.1.2 가을철 호주 북쪽 해수면온도 경향

4.3.1.3 10월 걸프해류 해수면온도

4.3.2 2월 기온 통계예측 모델 및 검증

5. 결론

6. 토론-선형 통계예측 모델의 한계

REFERENCE