Inflation adjustment on error covariance matrix of ensemble Kalman filter

In ensemble Kalman filter assimilation, the estimated forecast error covariance matrix and prior observational error covariance matrix could be far from the truth. This is likely to significantly affect the assimilation results. To compensate, this paper introduce two inflation factors to adjust forecast and observational error covariance respectively and estimate them simultaneously in one assimilation circle. The proposed schemes are tested using Lorenz-96 model, with a class of nonlinear observational operators. It illustrates that the improved assimilation schemes perform better than the original scheme.