Anna Shlyaeva;
Jeff Whitaker;
Chris Snyder
NOAA/ERSL and Colorado University/CIRES
The 8th EnKF Data Assimilation Workshop
Model space localization in serial ensemble filters
Talk:
EnSRFmodloc_Shlyaeva.pdf
Ensemble-based data assimilation systems typically use covariance localization to dampen spurious correlations associated with sampling error while increasing the rank of the covariance estimate. Variational methods use model-space localization, where background error covariances are localized based on distances between model variables, while ensemble filters typically use observation-space localization based on distances between model variables and observations. It has been shown that for non-local observations, such as satellite radiances, model-space localization can be superior.
In this presentation a new method for performing model-space localization in serial ensemble filters using the linearized observation operators (or Jacobians) will be demonstrated. Results of radiance-only assimilation in a global forecast system show the benefit of using model-space localization relative to observation-space localization. Results with the new method using linearized forward operators are nearly identical to the previously proposed 'modulated ensemble' approach. The EnKF with vertical model-space localization gives results similar to those of the EnVar system (without outer loops or extra balance constraints), while increasing the computational cost by a factor between 2 and 8, depending on how sparse the Jacobian matrices are.
Ensemble-based data assimilation systems typically use covariance localization to dampen spurious correlations associated with sampling error while increasing the rank of the covariance estimate. Variational methods use model-space localization, where background error covariances are localized based on distances between model variables, while ensemble filters typically use observation-space localization based on distances between model variables and observations. It has been shown that for non-local observations, such as satellite radiances, model-space localization can be superior.
In this presentation a new method for performing model-space localization in serial ensemble filters using the linearized observation operators (or Jacobians) will be demonstrated. Results of radiance-only assimilation in a global forecast system show the benefit of using model-space localization relative to observation-space localization. Results with the new method using linearized forward operators are nearly identical to the previously proposed 'modulated ensemble' approach. The EnKF with vertical model-space localization gives results similar to those of the EnVar system (without outer loops or extra balance constraints), while increasing the computational cost by a factor between 2 and 8, depending on how sparse the Jacobian matrices are.
Thanks to McGill's Atmospheric and Oceanic Sciences department for hosting this web-site.