The similarity between optimal precursor and optimally growing initial error in prediction of Kuroshio large meander and its application to targeted observation

Abstract

The links between optimal precursor (OPR) and optimally growing initial error (OGIE) in the predictability studies of Kuroshio large meander (LM) are investigated using the Conditional Nonlinear Optimal Perturbation approach within a 1.5-layer shallow-water model. The OPR is a kind of initial anomaly that is the easiest to cause the occurrence of Kuroshio LM path. The OGIE refers to another kind of initial perturbation that has the largest effects on the prediction of the LM path. Numerical results show that the spatial structures of OPR and OGIEs are similar and their dominant amplitudes are localized upstream of the occurrence region of Kuroshio LM, i.e., southeast of Kyushu. Their spatial patterns are closely related to the potential vorticity distribution of the reference states. The main features of the nonlinear evolutions of both OPR and OGIEs are also similar, which implies that they have a similar development mechanism. The similarities between OPR and OGIEs indicate that targeted observations may be implemented to improve the prediction of the Kuroshio LM path. We define sensitive areas using the spatial patterns of OGIEs and investigate their usefulness. The results show that the initial errors in sensitive areas have greater impacts on the prediction of the Kuroshio LM path than those in other randomly selected areas. In addition, the prediction is more significantly improved when eliminating the part of initial error in the sensitive area than in other areas. KEYWORDS: predictability; similarity; precursor; initial error; targeted observation; Kuroshio