Study the impact of ensemble covariance in hurricane analysis and predictions using a hybrid ETKF-3DVAR data assimilation method

Current operational data assimilation scheme relies on using vortex bogussing and relocation to initialize hurricane forecast. Ensemble based error covariance estimate can extract information from the actual observations flow-dependently and thus can effectively initialize hurricane forecast.

In this study, we explore the impact of using ensemble covariance in the analysis and forecast of hurricanes using a hybrid 3DVAR-ETKF data assimilation scheme that we developed for the WRF model. In our first experiment, we applied the method for hurricane Rita 2005. Data assimilation cycle was conducted by running WRF 3.1. Conventional observations from the real time GTS archive were assimilated every 6 hours starting from when Rita was a tropical depression till it made landfall. In the hybrid method, we can conveniently adjust the weight to the static error covariance and the ensemble covariance. The benchmark run was defined as placing full weight on the static covariance. We call experiments where we incorporated ensemble covariance generated by the ETKF to the background error covariance as a hybrid run. Data from best-track archive was used as verifications. Our results showed that the hybrid run produced more accurate track analysis than the benchmark run. Detailed analysis on how and why using ensemble covariance improved the analysis will be conducted and presented in the conference.