The study presents the ability of seasonal forecast models to represent the observed teleconnection associated with the major climate variability modes including the El Niño-Southern Oscillation (ENSO), Indian Ocean Dipole (IOD), North Atlantic Oscillation (NAO), and Western Pacific (WP) pattern. The prediction skill of the two ocean modes (ENSO and IOD) is much better, in terms of temporal variation, than that of the atmospheric modes (NAO and WP) in seven individual model reforecasts, although the indices for all four modes are mostly well represented in their dominant season.

Firstly, many of the features seen in observations are generally well reproduced in terms of the remote teleconnection associated with the ENSO. However, in some regions, the models simulate air temperature/precipitation responses that are too strong. The errors in temperature/precipitation response for North America, in particular, increase significantly during the ENSO peak season. This leads to temperature/precipitation forecast errors during the development of the ENSO. In most models, a strong linear relationship is maintained between the temperature/precipitation anomalies associated with both the El Niño and the La Niña phases of the ENSO; however, this relationship is weak in observations. One possible reason for the strong linearity in the models is that the remote atmospheric teleconnection pattern between the phases is too linear. Another is that the models cannot realistically simulate the impact of the Arctic variability on the mid-latitude climate, which may be because of the cold bias of the Arctic Ocean in models.

Secondly, the models tend to predict the response of temperature and precipitation to the IOD well over regions adjacent to Indian Ocean. Countries west (east) of the Indian Ocean become warm and wet (cold and dry) during the positive phase of the IOD, while its impact on mid-latitude regions is less distinct. Furthermore, the teleconnection pattern over the Pacific Ocean is similar to that of the ENSO. The variability of sea surface temperature (SST) over the Indian Ocean is predominated by two main modes: the Indian Ocean Basin (IOB) and the IOD modes. Models in which the ENSO variability is stronger than the observation show a larger variance in the IOB and IOD modes. The spatial air-sea coupled pattern associated with the developing phase of the IOD is well represented in models despite a slight difference among them. By excluding the ENSO effect, the influence of the IOD over East Asia becomes stronger, but the models fail to show this increased impact.

Third, the impact of the NAO is observed broadly over the region from eastern North America across the north Atlantic to far eastern parts of Asia, and a significant temperature/precipitation impact area is well represented in many of the models and ensembles. In some cases, however, there is a clear inter-model diversity for the impact of the NAO. For instance, in April the NAO temperature response over Europe and the Asian continent varies from model to model, and it seems to be associated with the mean bias of the SST in the North Atlantic. Interestingly, the models tend to show a one-month delayed regional response in the NAO during spring (e.g., the model response in March is more similar to observed responses in February than in March) and the NAO peak season is also delayed to March in the model simulations. This delayed response seems to be associated with the exaggerated relationship between the ENSO and NAO during spring.

Fourth, the temperature response to the WP mode is very pronounced during the boreal spring and winter seasons in East Asia, the Bering Sea, and North America. Interestingly, the model's skill score is relatively high during the transitions between seasons in September and October. However, there are large errors in the remote temperature response in October. The precipitation response in the tropical Pacific and over the North American continent is also simulated adequately by climate models. However, most of the models have errors where the positive temperature and precipitation response in tropical regions in December is maintained until the spring of the following year.

Lastly, this study presents a map of climate mode impacts that collectively displays all significant modes (from the ENSO, IOD, NAO, and WP) that affect temperature/precipitation at individual grids for 12 months. The statistical summary of this map is also given for six verification regions. The map will provide supplementary information for monthly consensus predictions in the APCC and a tangible guide to the researches of tailoring multi-model ensemble forecasts.

1. 서론

2. 자료 및 분석 방법

2.1 계절 예측 모형과 과거 재현 자료

2.2 관측 자료

2.3 기후모드 정의

2.4 모형 검증 지역 및 지수

3. 기후모드 지수변동성 및 지역반응

3.1 기후모드 지수 예측성

3.2 기후모드의 지역반응 모의 특성

3.2.1 다중모형/앙상블 간의 일관성 및 관측 대비 오차

3.2.2 기후모드 지역반응에 대한 다중모형간 모의 성능 비교

4. 기후모드 영향에 대한 전지구 통합 가이던스맵

5. 결론

REFERENCES