9:30 AM - 9:45 AM
[MGI24-03] Introduction of global error covariance to nested ensemble variational assimilation
Keywords:Ensemble variational assimilation, Nesting system
This study attempts to optimize a nested regional analysis by making the best use of global analysis information.
A regional atmospheric model requires lateral boundary conditions obtained typically from a global model. The regional analysis sometimes suffers from deterioration of the large-scale structure compared to that of the global analysis due to limitations in the domain size and observations. The large-scale error may cause the displacement error for disturbances such as typhoons or synoptic-scale fronts and degrade the performance of convective-scale data assimilation (DA). Although several scale-dependent blending methods of global and regional analyses have been proposed to alleviate those large-scale errors, these blending methods may hinder the optimality of individual DA.
Guidard and Fischer (2008) and Dahlgren and Gustafsson (2012) introduced the augmented information vector by the global analysis or forecast into the regional variational assimilation and reported promising results. However, their formulations require several assumptions for the error correlations to simplify the implementation and ignore the cross-covariance between the global and the regional forecast errors.
In this study, we relax these assumptions and extend their augmented variational formulation to an ensemble framework to account for the flow-dependency of the forecast error. We utilize a global ensemble to estimate the global error covariance in the augmented cost function in addition to the regional forecast error covariance. This alteration also allows us a straightforward extension to the cross-covariance formulation.
We first compare the proposed method against the conventional independent DA and the previous studies in the single assimilation experiments for one-dimensional periodic sine waves referring to the experimental settings in Baxter et al. (2011). In these experiments, the conventional DA worsens the errors in the low wavenumbers. The augmented method of the previous studies yields statistically significant error reduction in these large-scale structures especially for the limited observation volume. Our proposed method also ameliorates the large-scale errors as expected, though the error reduction is statistically less significant than that of the previous studies probably due to a lack of flow-dependency and sampling error. We also perform the cycled assimilation experiments in the one-way nesting system using one-dimensional chaotic models proposed by Lorenz (2005). The consideration of the flow-dependency has a significant impact on the analysis in these experiments. The performance of the augmented methods of this study and of the previous studies are found to depend largely on the quality of the global DA. We will also discuss the effect of the cross-covariance between the global and the regional forecast errors.
A regional atmospheric model requires lateral boundary conditions obtained typically from a global model. The regional analysis sometimes suffers from deterioration of the large-scale structure compared to that of the global analysis due to limitations in the domain size and observations. The large-scale error may cause the displacement error for disturbances such as typhoons or synoptic-scale fronts and degrade the performance of convective-scale data assimilation (DA). Although several scale-dependent blending methods of global and regional analyses have been proposed to alleviate those large-scale errors, these blending methods may hinder the optimality of individual DA.
Guidard and Fischer (2008) and Dahlgren and Gustafsson (2012) introduced the augmented information vector by the global analysis or forecast into the regional variational assimilation and reported promising results. However, their formulations require several assumptions for the error correlations to simplify the implementation and ignore the cross-covariance between the global and the regional forecast errors.
In this study, we relax these assumptions and extend their augmented variational formulation to an ensemble framework to account for the flow-dependency of the forecast error. We utilize a global ensemble to estimate the global error covariance in the augmented cost function in addition to the regional forecast error covariance. This alteration also allows us a straightforward extension to the cross-covariance formulation.
We first compare the proposed method against the conventional independent DA and the previous studies in the single assimilation experiments for one-dimensional periodic sine waves referring to the experimental settings in Baxter et al. (2011). In these experiments, the conventional DA worsens the errors in the low wavenumbers. The augmented method of the previous studies yields statistically significant error reduction in these large-scale structures especially for the limited observation volume. Our proposed method also ameliorates the large-scale errors as expected, though the error reduction is statistically less significant than that of the previous studies probably due to a lack of flow-dependency and sampling error. We also perform the cycled assimilation experiments in the one-way nesting system using one-dimensional chaotic models proposed by Lorenz (2005). The consideration of the flow-dependency has a significant impact on the analysis in these experiments. The performance of the augmented methods of this study and of the previous studies are found to depend largely on the quality of the global DA. We will also discuss the effect of the cross-covariance between the global and the regional forecast errors.