Global Modeling and Assimilation Office
NASA Goddard Space Flight Center
Last Updated: May 27 2011
Title: Evaluation of a hybrid ensemble-variational data assimilation scheme using an OSSE
Authors: Daryl Kleist (NOAA/NCEP/EMC, Univ. of Maryland-College Park, IMSG)
John Derber (NOAA/NCEP/EMC)
David Parrish (NOAA/NCEP/EMC)
Kayo Ide (Univ. of Maryland-College Park)
Jeff Whitaker (NOAA/ESRL/PSD)
The capability to incorporate flow-dependent, ensemble-based representations of a background error covariance matrix into variational data assimilation has recently been developed for use in the GSI by utilizing the extended control variable method. Preliminary experimentation with minimal tuning using a coupled, dual-resolution 3DVAR-EnKF hybrid system with the NCEP Global Forecast System model have shown that utilizing the hybrid paradigm can yield substantial forecast error reduction relative to a 3DVAR-based control system. In the dual resolution experiments, the ensemble is run at a coarser resolution relative to the driving, high-resolution, deterministic (hybrid) analysis and forecast components.
To further understand the impact of hybrid data assimilation systems on the quality of analyses and subsequent forecasts, experiments will be carried out using an observing system simulation experiment (OSSE). By utilizing an OSSE and having access to a truth, it is possible to better understand how and why the supplementary ensemble information is reducing analysis and forecast error. Additionally, it is possible to investigate how well the ensemble is representing the actual analysis and forecast error covariances.
Experiments will be carried out using the NCEP GFS model and GSI-based dual-resolution hybrid by using the Joint OSSE nature run and simulated observations. In addition to hybrid 3DVAR-EnKF experiments (as have already been carried out for real observation impact evaluation), additional runs will be made to evaluate the importance of ensemble size, localization, and coupling of the EnKF to the hybrid analysis. The hybrid 3DVAR-EnKF results will be compared to results from using a GSI-based ensemble 4DVAR (without tangent linear and adjoint). The importance of time localization and supplementing the ensemble with a full-rank static estimate within the 4D paradigm will be explored.
| |
Global Modeling and Assimilation Office NASA Goddard Space Flight Center |
Last Updated: May 27 2011 |