542
Assimilation of Radar Radial Velocity Data with the WRF Ensemble-3DVAR Hybrid System for the Prediction of Hurricane IKE (2008)

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 26 January 2011
Assimilation of Radar Radial Velocity Data with the WRF Ensemble-3DVAR Hybrid System for the Prediction of Hurricane IKE (2008)
Washington State Convention Center
Yongzuo Li, University of Oklahoma, Norman, OK, Norman, OK; and X. Wang and M. Xue
Manuscript (1.4 MB)

Poster PDF (1.4 MB)

The hybrid ensemble-3DVAR data assimilation method developed for the Weather Research and Forecast (WRF) model (Wang et al. 2008ab) is applied to the assimilation of radial velocity data from two WSR-88D radars for the prediction of hurricane IKE (2008) before and during the landfall. Rather than using a static covariance as in traditional 3DVAR, with the hybrid method, ensemble covariance is incorporated into the WRF 3DVAR framework to provide a flow-dependent estimate of the background error covariance so that the radial velocity data can be more effectively used to correct the background hurricane forecast during the assimilation. This is the first time the hybrid system is applied for the assimilation of real radar data. In this study, WSR-88D radar radial velocity data are processed using a modified version of the radar data dealiasing algorithm of James and Houze (2001). The processed radial velocity data of KHGX (Houston, TX) and KLCH (Lake Charles, LA) are assimilated every 30 minutes for 3 hours immediately after IKE moving into the coverage of the two radars. Subsequent 21-h WRF ARW model forecast is performed to evaluate the effect of the data assimilation. Our results so far have shown that compared to the 3DVAR using the standard method to generate the static background error covariance, track and intensity forecasts initialized by the hybrid data assimilation method are much improved. Forecasts initialized by the 3DVAR where the static covariance is further tuned by reducing the scale of the static covariance are improved compared to using the default static covariance. The hybrid method in this case still shows smaller root mean square errors compared to the tuned 3DVAR in intensity forecast and when verified against radar radial velocity. Further experiments with different parameters in the hybrid system and comparison between the hybrid and tuned 3DVAR are being conducted and results will be reported at the conference.