Evaluation of an Ensemble Data Assimilation System with Airborne and Ground-based Doppler Radar Winds for Atlantic Hurricanes

Hurricanes are among the costliest and deadliest natural disasters, yet our current generation of hurricane prediction systems often fails to produce accurate and timely forecasts. A lack of sufficient observations over the tropical oceans and the dynamically inconsistent depiction of tropical cyclones (TC) by operational data assimilation systems may be two of the largest factors contributing to inadequate vortex initialization in numerical forecasting efforts. Since 2007, the Penn State University (PSU) hurricane team has been developing a WRF based ensemble data assimilation and prediction system that utilizes Doppler radar winds for cloud-resolving hurricane initialization and forecasting.The system has great potential for improving hurricane track and intensity forecasts, as demonstrated by the landmark case study in which ground-base Doppler radar winds were assimilated for hurricane Katrina (2005) and Humberto (2007), and NOAA P-3 airborne Doppler radar winds for Katrina. In 2008, under the supports of HRD/NOAA, Texas Advanced Computing Center (TACC), National Science Foundation and NOAA Hurricane Forecast Improvement Program (HFIP), the team planted a superobbing program on P-3 airplanes and applied the system in real-time for Atlantic hurricane analysis and prediction. For real-time forecasts of hurricane Ike (2008), the system outperforming deterministic operational forecast models initialized from corresponding three-dimensional variational analyses as well as the NHC official forecasts issued at similar times. In this study, experiments are conducted for both hindcast and real-time tropical cyclone events to evaluate our analysis and prediction system. The hindcast experiments include several Atlantic storms in 2005 (Emily, Katrina, Rita, Philippe and Wilma), 2007 (Humberto, Ingrid, Karen and Felix) and a storm in the 2008 hurricane season (Dolly). The real-time runs include most of the storms in the 2008, 2009 and 2010 Atlantic hurricane seasons when airborne radar data available. Multi-case experiments using our system are evaluated against the NHC best-track analysis as well as NHC operational forecasts, GFDL operational forecasts, HWRF forecasts and WRF forecasts without EnKF assimilation of radar winds. The accuracy of hurricane intensity and track forecasts and hurricane vortex initialization will be analyzed on a case-by-case basis for storms of varying intensity and level of development.