A comparative analysis of Kuroshio Extension indices from a modeling perspective

Dynamical indices previously introduced to characterize the Kuroshio Extension (KE) decadal variability (DV) fromaltimeter data are comparatively analyzed to assess their ability to reveal the same phenomenon from model outputs. Based on this analysis, a new combined index identifying each stage of the KE DV in numerical simulations is then proposed. The analysis begins by recognizing that the numerical simulation of the nonlinear frontal KE DV is very sensitive to changes in model implementation, whereas the linear broad-scale baroclinic Rossby wave field known to trigger the KEDV is a robust model feature. This selectivemodel sensitivity poses a subtle problem concerning the diagnosis of the KE DV from model outputs, and requires the use of indices that unequivocally identify the KE frontal structure. The capability of six indices to achieve this task is thus investigated. Two model outputs representing paradigms of a fairly realistic simulation and of an unrealistic simulation in which only the broad-scale variability is present are used. A synthesized index is well captured by both simulations: it is therefore recognized to be unsuitable formodel studies.An integrated SSH index does not resolve explicitly the frontal variability.Among the remaining indices, the KE path length (modified through the application of the wavelet transform) and the meanKElatitudinal position are shown to provide, in combination, an unambiguous identification of each stage of the KE DV (recognized to exhibit chaotic hysteresis) and are therefore suggested to be used with model outputs.