An effective method for improving the accuracy of Argo objective analysis

Abstract

Based on the optimal interpolation objective analysis of the Argo data, improvements are made to the empirical formula of a background error covariance matrix widely used in data assimilation and objective analysis systems. Specifically, an estimation of correlation scales that can improve effectively the accuracy of Argo objective analysis has been developed. This method can automatically adapt to the gradient change of a variable and is referred to as “gradient-dependent correlation scale method”. Its effect on the Argo objective analysis is verified theoretically with Gaussian pulse and spectrum analysis. The results of one-dimensional simulation experiment show that the gradient-dependent correlation scales can improve the adaptability of the objective analysis system, making it possible for the analysis scheme to fully absorb the shortwave information of observation in areas with larger oceanographic gradients. The new scheme is applied to the Argo data objective analysis system in the Pacific Ocean. The results are obviously improved.