Advanced Data Assimilation For Cloud-Resolving Hurricane Initialization And Prediction

Data assimilation is a technique aimed at decreasing errors in the initial conditions, which are one of the primary sources of uncertainties in hurricane prediction. The current study examines the performance of three of the most advanced data assimilation techniques by assimilating inner-core, high-resolution Doppler radar observations for cloud-resolving hurricane initialization and forecasting for Hurricane Katrina (2005), one of costliest and deadliest natural disaster in the US history. The ensemble Kalman filter (EnKF), three-dimension variational method (3DVar) and four-dimensional variational method (4DVar), all of which are based on the next-generation Weather Research and Forecast model (WRF), are the algorithms of focus. We demonstrate that the EnKF has great promise in delivering more accurate forecast along with a realistic estimate of forecast uncertainties of this event. We also show a clear advantage of 4DVar over 3DVar, the latter of which is used in the current-generation U.S. operational hurricane prediction models.