[p1] Non-Gaussianity and nonlinearity


4D-EnVAR with iterative calculation of non-linear model

S. Yokota (Meteorological Research Institute), M. Kunii (Meteorological Research Institute), K. Aonashi (Meteorological Research Institute), and S. Origuchi (Fukuoka Aviation Weather Station)


4D-EnVAR is a useful ensemble-based variational data assimilation method, which does not use adjoint matrices of tangent linear observation operator (H) and forecast model (M). However, 4D-EnVAR analyses are generally worse than Hybrid 4D-VAR probably because 4D-EnVAR does not iteratively calculate non-linear H and M. Yokota et al. (2016, SOLA) developed 4D-EnVAR with observation space localization that iteratively calculates non-linear H, and showed that this method can make better analyses than 4D-LETKF. In this study, we improved this 4D-EnVAR to iteratively calculate M as well as H (hereafter, 4D-EnVAR-M), and showed the advantage of 4D-EnVAR-M with single-observation assimilation experiments and observation system simulation experiments (OSSEs) using a low-resolution AGCM. In the single-observation assimilation experiments, 4D-EnVAR-M analyses were closer to the observations than 4D-EnVAR analyses without iterative calculation of M. In the OSSEs, observations of zonal and meridional winds, temperature, relative humidity, and surface pressure were created by adding random errors to "true" values (results of free-run simulation) and assimilated. Biases and root mean square errors from "true" values were smaller in the forecasts from 4D-EnVAR-M analyses than those from 4D-EnVAR analyses without iterative calculation of M. Therefore, iteratively calculating M is likely to be effective to make better analyses.