A New Estimation Method of Ensemble Forecast Error in ETKF Assimilation with Nonlinear Observation Operator

Abstract

The estimation accuracy of ensemble forecast errors has a key influence on the assimilation results of all ensemble-based schemes. The ensemble transform Kalman filter (ETKF) assimilation scheme can assimilate nonlinear observations without using the adjoint of a dynamical model; however, the initially estimated ensemble forecast errors must be further adjusted. In this paper, the estimation of forecast error is improved using a self-generated analysis and a corresponding iterative procedure is then established for the ETKF with nonlinear observation operators. The improved assimilation scheme is validated using the Lorenz-96 model with a nonlinear and spatially correlated observation system as a test bed. The experiment results demonstrate that the improved ETKF assimilation scheme can effectively reduce the analysis error.