[p2] Model-related issues


Comparison of Kalman Filter and Hybrid assimilation methods in the spectral barotropic general circulation model

T. Kurihana (University of Tsukuba)


The effect of hybrid data assimilation (hybrid-DA) method becomes one of the cutting edge research trends in the recent DA field. Especially given that ensemble forecasting within hybrid?DA system, Ensemble Kalman Filter (EnKF) provides the background error covariance for three dimensional data assimilation (3DVar) and four dimensional data assimilation (4DVar). Meanwhile, the predictability of ensemble forecasting sometimes could be deteriorated. This research, however, runs these hybrid-DA methods by coupling Kalman Filter (KF) in the spectral barotropic general circulation model in University of Tsukuba, called S-model, under the perfect model configuration. The S-model enables to implement KF because of the relatively lower dimension of model parameter, which avoids the extremely expensive computational cost of KF in a typical general circulation model (GCM). As a result, according to the difference between truth and analysis of barotropic height, the performance of hybrid-3DVar and hybrid-4DVar was higher than that of non-hybrid-DA methods. Root mean square error of these hybrid methods also showed smaller. Following these consequences, hybrid with not EnKF but KF could also work well in lower dimensional GCM.