Evaluating the Performance of High-Resolution Hurricane Prediction Modeling System

An experimental version of the Hurricane Weather Research and Forecasting System (designated HWRFx) is currently being used at NOAA's Hurricane Research Division (HRD) to evaluate possible improvements to the prediction of both track and intensity of tropical cyclones. Observations obtained from NOAA's aircraft missions are combined with the HWRFx output for a specific storm, using a technique called data assimilation (DA). In what is known as cycling, the DA system sequentially updates the dynamic and thermodynamic fields of the model in the vicinity of the storm. In each cycle the update is computed by an Ensemble Kalman Filter algorithm in which ensemble-based covariances between observations and model fields are utilized. The combined data in a form of gridded fields from the final update cycle, called analysis, are then used to initialize HWRFx to generate a forecast. In this study, the performance of different experiments that were carried out with the DA system for a case of Hurricane Bill of 2009 is analyzed. Graphical displays of related physical variables are generated as diagnostics and compared with real observations when available. In addition, the ensemble covariance data are analyzed for the experiment that shows the best performance, to understand the behavior of the model when updating the physical variables within the cycles. A sample of the results, including a comparison of the final- cycle analysis of each experiment, the time-cycles of the best experiment, and the normalized covariance (correlations) will be shown in this presentation.