Ensemble analyses and predictions of Hurricane Katrina

Current operational data assimilation techniques have difficulty assimilating observations in and near tropical cyclones (TCs). Given this difficulty, many operational centers employ tropical cyclone bogusing schemes to represent the TC in conventional analyses. While bogusing improves short-term TC forecasts, the structure of the bogused TC is often not representative of the observed system. Furthermore, variational data assimilation techniques use an assumed background error-covariance matrix that is likely inappropriate for tropical cyclones. The ensemble Kalman filter (EnKF) is an attractive alternative for TC state estimation because this technique assimilates observations using flow-dependent covariances, including those of tropical cyclones. Moreover, the EnKF produces an ensemble of equally-likely analyses that are available for use in TC ensemble forecasting, without the need for perturbing deterministic analyses as is the current practice.

Here we study Hurricane Katrina using a 90 member EnKF based on the Weather Research and Forecasting (WRF) model. Observations are assimilated every six hours using conventional observations including RAINEX dropsondes (no satellite radiances or NHC best track data). Preliminary results show that the EnKF analyses accurately estimate Katrina's position (compared to NHC best track data), but overestimate the minimum central pressure. Assimilation of the RAINEX dropsonde data leads to significant positive corrections to the background field, especially when dropsondes were deployed symmetrically near the core. All 90 analysis ensemble members are integrated forward 60 hours from 00 UTC 27 August for an ensemble forecast of Katrina's landfall. Ensemble forecasts of Katrina's central pressure are too high; however, the NHC best track positions are within the span of the ensemble.