Issues In Targeted Observation For The Prediction Study Of The Kuroshio Large Meander Path

The sensitive areas of targeted observation for the prediction study of the Kuroshio large meander(LM) path are determined using the conditional nonlinear optimal perturbation(CNOP) method and the first singular vector(FSV) method based on a 1.5-layer shallow-water(SW) ocean model. A series of sensitivity experiments and the ideal hindcasting experiments are designed to test the sensitivity and validity of sensitive areas within the numerical model. The observing system simulation experiments(OSSE) are used to evaluate the effects of targeted observations. Moreover, the effectiveness of sensitive areas obtained by the SW model are also tested by the multilayer and complicated Princeton Ocean Model(POM). The main conclusions and summarized of this paper are as follows:Firstly, the SW ocean model are used to simulate the Kuroshio path variations, the sensitive areas of targeted observation are identified by the CNOP and FSV method. The sensitivity experiments are designed to investigate the impacts of locations and patterns of initial errors on the developments of themselves. The results show as follows:(1) The kinetic energy developments of the CNOP-type and FSV-type initial errors are much greater than those random errors at the final forecasts, and the kinetic energy developments of initial errors with CNOP patterns are the greatest ones.(2) The kinetic energy developments of initial errors with CNOP patterns and FSV patterns superimposed into the CNOP sensitive areas are greater than that of themselves put in the FSV sensitive areas and randomly selected areas.(3) The kinetic energy developments of random errors introduced into the CNOP and FSV sensitive areas are greater than that superimposed into the randomly selected areas, and the initial errors in the CNOP sensitive areas have the greater impacts on the final forecasts. These indicate that the sensitive areas determined by CNOP are more sensitivity and validity than the FSV sensitive areas and other randomly selected areas. In addition, the results of the ideal hindcasting experiments indicate that the reductions(or eliminations) of CNOP-type initial errors in CNOP sensitive areas have more forecast benefits than the reductions(or eliminations) of FSV-type initial errors in FSV sensitive areas. These reveal that the eliminations of the possible existence of the CNOP patterns initial errors in the CNOP sensitive areas by using targeted observations can avoid the worst prediction results in principle.Secondly, the results of the OSSE show that the forecast skill of Kuroshio LM path via the targeted observations implemented in sensitive areas determined by the CNOP and FSV method have much larger improvements than in other randomly selected areas. Besides, the forecast benefits of Kuroshio LM by deploying additional observations over CNOP sensitive regions are largest.Finally, POM model is used to simulate the Kuroshio path variations south of Japan. Moreover, we demonstrate that the errors similarly to the patterns of initial errors obtained in SW model can also lead to a larger prediction error in POM model, the sensitive areas of targeted observation in the SW model are still effective in POM model. Besides, the CNOP sensitive areas are more sensitivity and validity than the FSV sensitive areas, which are in substantial agreement with the conclusions in SW model. This reveal that the targeted observation based on CNOP method is more effective and feasible approach to identify the sensitive areas for the prediction of the Kuroshio LM path south of Japan.