[A-CG30] 熱帯におけるマルチスケール大気海洋相互作用

[ACG30-07] Understanding tropical interbasin interaction using linear inverse modelling

*木戸 晶一郎¹、Ingo Richter¹、東塚 知己^2,1、Ping Chang³ (1.海洋開発研究機構 付加価値創生部門　アプリケーションラボ、2.東京大学大学院　理学系研究科　地球惑星科学専攻、3.Texas A&M University)

キーワード：Interbasin interaction、Linear inverse modeling

Many observational and modelling studies have recently underlined the importance of tropical interbasin coupling in understanding climate variability and predictability. The coupling among tropical basins can be separated into three components; the interaction between Pacific and Indian Oceans (the PO-IO interaction), that between Pacific and Atlantic Oceans (the PO-AO interaction), and that between the Atlantic and Indian Oceans (the AO-IO interaction). Though many previous studies have discussed the significance of individual components, the relative importance of these coupling components has not been carefully evaluated and fully understood. To address this issue, we have constructed a linear inverse model (LIM) based on observed sea surface temperature (SST) anomalies in the tropical Pacific, Atlantic, and Indian Oceans, and performed a series of prediction experiments using this LIM. We found that our LIM has a good skilll in forecasting tropical SST variability, including those associated with the El Niño and Southern Oscillation (ENSO). To assess the impact of interbasin interaction, we have removed individual coupling component by modifying off-diagonal elements of the linear operator. Using this “decoupled” operator, we have conducted several prediction experiments. We find that the decoupling leads to a substantial decrease in prediction skill of ENSO and related SST variability, especially at longer lead times. Partial decoupling experiments that nullify specific coupling components suggest that the PO-IO interaction has the largest impact on the prediction skill of ENSO-related variability, whereas the PO-AO interaction also has a nonnegligible contribution. On the other hand, the impacts of the AO-IO interaction seem to be smaller than those of the other two coupling components. Results from the LIM simulations with white noise forcing, as well as an analysis of optimum initial conditions will be also discussed. These are aimed to examine the underlying statistical relations and physical processes.