Comparison of WRF-DART Ensemble Adjustment Kalman Filter (EAKF) and Rank Histogram Filter (RHF) in Analyses and Simulations of a Convective Case over the Great Plains Region

The mesoscale community Weather Research and Forecasting (WRF) model is used in conjunction with the NCAR Data Assimilation Research Testbed (DART) to perform cycled data assimilation for a convective case on 23-24 May, 2011 over the Southern Great Plains Region. The case featured low predictability in terms of location, intensity, and timing of the convection. The Ensemble Adjustment Kalman Filter (EAKF) is compared to the Rank Histogram Filter (RHF) to determine how the different filters could help the model reproduce associated mesoscale characteristics of the convective system during its initiation and evolution.

In general, the EAKF is considered to be one of the state-of-the-art data assimilation methods. However, this particular convective system made the data assimilation a nonlinear and a non-Gaussian likelihood problem. Specifically, the prior states of the ensemble members diverged from each other without data assimilation. In such a case the EAKF may not perform optimally. The RHF, on the other hand, is better able to handle non-Gaussian priors and outlier observations. Therefore, the study case represents an ideal opportunity to test if the RHF can perform better than the EAKF. Comparisons are also made with variations in localization, ensemble perturbations, and the inflation.

Early results show that both filters have considerable value in improving the simulations of the convective system. Overall, the EAKF proved to be superior by most metrics with larger and more realistic spread, but lower root-mean-square (RMS) errors and bias of atmospheric state variables. More detailed studies and additional results, especially the relative strengths and weaknesses of each filter, will be presented at the conference.