Observing Array Designed for Improving the Short‐Term Prediction of Kuroshio Extension State Transition Processes

Given the essential implications of Kuroshio Extension (KE) bimodality on oceanic dynamical environment and climate, the present study investigates the targeted observation schemes, based on the conditional nonlinear optimal perturbation (CNOP) method and a reduced‐gravity shallow‐water model, to improve the forecast skills of transition processes of KE bimodal states. To obtain a suitable observing array, the observation schemes, with different numbers of observation sites and observation distances between two sites, are designed. Furthermore, to demonstrate the superiority of the observing networks in predicting KE transition processes, two existing observation schemes and six random observation schemes are compared with the CNOP‐determined observing array. Based on this, a relatively optimal observing array with three sites and observation distance of 90 km is established, which is mainly located between 31°N and 33°N in the south of Japan. This targeted observing network is universal for two KE transition processes. The removal of initial errors on this array results in the mean prediction improvements of about 9.2% and 22.5% for KE transition processes from the low‐ to the high‐energy state and from the high‐ to the low‐energy state, respectively. Plain Language Summary Kuroshio Extension (KE) path has a remarkable bimodality between a low‐energy and a high‐energy state, which leaves the important imprints on oceanic dynamical environment, climate, and marine ecosystems. Hence, the dynamics and predictability of KE bimodality have been widely investigated. However, due to the complexity of its physical mechanisms, the prediction of KE transition is limited. To improve its forecast skills, the targeted observing array with three sites and observation distance of 90 km is determined based on the conditional nonlinear optimal perturbation method and a reduced‐gravity shallow‐water model. Due to the closely relationship between Kuroshio meander south of Japan and KE bimodality, the targeted observing network obtained universally for two KE transition processes is mainly located in the south of Japan. The removal of initial errors on this observing network can significantly improve the prediction skills of about 9.2% and 22.5% for KE transition processes from the low‐ to the high‐energy state and from the high‐ to the low‐energy state, respectively. Key Points The observing networks are designed to improve the forecast skills of transition processes of Kur