An optimization strategy for identifying parameter sensitivity in atmospheric and oceanic models

Abstract

A new optimization strategy is proposed to identify the sensitivities of simulations of atmospheric and oceanic models to uncertain parameters. The strategy is based on a nonlinear optimization method that is able to estimate the maximum values of specific parameter sensitivity measures; meanwhile, it takes into account interactions among uncertain parameters. It is tested using the Lorenz'63 model and an intermediate complexity 2.5-layer shallow-water model of the North Pacific Ocean. For the Lorenz'63 model, it is shown that the parameter sensitivities of the model results depend on the initial conditions. For the 2.5-layer shallow-water model used to simulate the Kuroshio large meander (KLM) south of Japan, the optimization strategy reveals that the prediction of the KLM path is insensitive to the uncertainties in the bottom friction coefficient, the interfacial friction coefficient, and the lateral friction coefficient. Rather, the KLM prediction is relatively sensitive to the uncertainties of the reduced gravity representing ocean stratification and the wind stress coefficient.