The ensemble-based data assimilation, commonly known as ensemble-Kalman filter (EnKF), has recently been demonstrated to be an effective and maturing assimilation technique with simulated and real observations for NWP across a range of scales. The present study uses a WRF-based EnKF to examine the impact of assimilatiing various data on the cloud-resolving prediction of Katrina (2005) with model grid resolution down to 1.5 km. These data include ground-based and airborne Doppler radar measurements, satellite dripwinds and dropsondes collected during the RAINEX field experiment in addition to standard upper-air and surface observations. In particular, we examine impact of using different data thinning schemes and varying radii of influence for different model domains that have different grid resolutions in the initiation and prediction of Hurricane Katrina at different stages.

Preliminary results show that assimilating both the ground-based and airborne radar observations is very beneficial for initializing the hurricane near its observed intensity with realistic asymmetry and for subsequent ensemble forecast. For example, forecast from a mean EnKF analysis that assimilates both the KMAX and KBYX radar Doppler observations at 00Z 26 August 2005 tracks the observed hurricane position very closely and brings the hurricane directly to New Orleans in 96 h. Moreover, the radar observations assimilated ensure the "hot-start" of a Category 2 hurricane in near observed intensity without commonly used bogussing or surgical relocation techniques. Ultimately, we plan to explore the structure, dynamics and predictability of hurricanes through assimilating all available observations and to determine minimum sufficient observations for monitoring and predicting tropical cyclones before diminishing returns.