In this study, we assessed the prediction skills of climate variables in the tropical Pacific and Indian Oceans using retrospective forecasts from the new APCC in-house seasonal forecast model, the Seamless Coupled Prediction System (SCoPS). In the retrospective forecasts, a 7-month integration was made for each of the 12 calendar months over a period of 33 years, from 1982 to 2014. The prediction skills fromSCoPS are compared with those from the existing APCC in-house model, CCSM3. Firstly, SCoPS has better overall skill for prediction of global 2m air temperature and precipitation. Particularly, SCoPS has significant skill in predicting precipitation at 6-month lead time over the tropical Pacific Ocean, whereas the CCSM3 has no skill at all. However, precipitation and air temperatures over the land regions beyond the Tropics fail to be predicted in both models, even with an early lead forecast time.

Next, we evaluated the prediction skills for El Niño-Southern Oscillation (ENSO). With SCoPS, there is improvement in predicting all of the ENSO related indices from 1 to 6-month lead times. In terms of the prediction of the extreme 1997/98 El Niño event, when SCoPS starts from July and September initial conditions, it can successfully simulate the amplitude of the Niño 3.4 index, while CCSM3 predicts a much weaker 1997/98 El Niño event. However, when SCoPS is initialized in March and May, namely spring season, it cannot simulate the intensity of the observed Niño 3.4 index. It is speculated that the prediction error in ENSO amplitude is partly related to the cold bias in the tropical Pacific Ocean, which peaks in the summer season after the forecast model starts from the initial conditions in Spring.

This study also estimates the skill in predicting two types of ENSO, Eastern Pacific (EP) and Central Pacific (CP) type. It was found that the EP type is better predicted in SCoPS and the CP type is better predicted in CCSM3. In both models, the two

types of ENSO are not separated well, as they were in the actual observations. When estimating skills in the Pacific Island Countries, in addition to the prediction skill of the dominant mode in the tropical Pacific, we separated the Pacific Island countries into four climatic regions; the Western Pacific (WP), North Western Pacific (NWP), South Western Pacific (SWP), and Equatorial Pacific (EQP). Overall, SCoPS has better performance when predicting 2m air temperature and precipitation over their regions from 1 to 6-month lead times. It was found that the EQP region has better skills than other climatic regions for both precipitation and 2m temperatures. In terms of seasonal dependence of the prediction skills, SCoPS has better capability when predicting air temperature and precipitation in the Spring or Winter seasons.

We also evaluated the skills for the Indian Ocean Dipole (IOD) and Indian Ocean basin-wide (IOBW) mode related SST anomalies, and found that SCoPS significantly improves prediction capability for the IOD and IOBW modes when compared with CCSM3. We also verified the prediction skill for ENSO and IOD impact on local precipitation change, and identified that there is increased teleconnection error with forecast lead times.

Lastly, the SCoPS prediction skills were compared with those in the nine APEC Climate Center (APCC) Multi Model Ensemble (MME) system seasonal forecast models in terms of ENSO, IOD, IOBW mode, and precipitation over the globe, South Asia, East Asia, and the five climatic regions of the Pacific island countries. Overall, we found that SCoPS has comparable prediction skills to the APCC MME models.

1. 서론

2. 연구 자료 및 방법

2.1 관측 자료

2.2 과거 예측 자료 생산

2.2.1 모형 개요

2.2.2 과거 예측 런

2.3 예측성 검증

2.4 다른 계절 예측 모델과 비교 검증

3. APCC in-house 모형들의 예측 성능 비교

3.1 전 지구 기온과 총 강수량

3.2 태평양에서 예측 성능

3.2.1 ENSO 예측

3.2.2 ENSO 다양성 예측

3.2.3 태평양 도서국 지역의 예측

3.3 인도양에서 예측 성능

3.4 주요 기후 모드와 지역 강수변화 관계 예측성능

3.4.1 ENSO와 지역 강수 변화

3.4.2 IOD와 지역 강수변화

4. APCC MME 시스템 모형들과 예측성 비교

5. 요약 및 결론

REFERENCE