Group seminar on 9. July, 14:15
Variable-dependent and selective multivariate localization for ensemble-variational data assimilation in the tropics
Joshua Lee
In this study, two limitations of standard implementations of ensemble-variational data assimilation (EnVar) are addressed. These are the inability to (i) prescribe different spatial localization length-scales for different variables; and (ii) control multivariate localization relationships. These limitations are addressed by introducing and evaluating variable-dependent and selective multivariate localization within the EnVar framework applied to a simplified non-hydrostatic model, the ABC model. A tropical setting of the ABC model is used where impacts of these limitations may be more pronounced than with other settings. Multi-cycle observation system simulation experiments are conducted with a 100-member ensemble. The results reveal that selective multivariate localization reduces the cycle-averaged root-mean-square error (RMSE) in the experiments when multivariate covariances associated with hydrostatic balance are retained but when multivariate zonal wind/mass errors are knocked-out. When variable-dependent horizontal and vertical localization are incrementally introduced, the cycle-averaged RMSE is further reduced. Overall, the best performing experiment using both variable-dependent and selective multivariate localization led to a 3-4% reduction in cycle-averaged RMSE compared to the standard EnVar experiment.