Regional Hybrid Ensemble-3DVAR Data Assimilation System: Sensitivity to Ensemble Initialization Scheme

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 6 January 2015
Juhui Ma, North Carolina State University, Raleigh, NC

A hybrid ensemble-three-dimensional variational (3DVAR) data assimilation method was used to produce analyses every 6-hr to initialize 24-hr Weather Research and Forecasting Model (WRF) model forecasts between 0000 UTC 1 and 0000 UTC 14 August 2013 over the contiguous United States domain at a 15-km grid spacing. The ensemble of 6-hr forecasts used to provide flow-dependent background error covariances for the hybrid were updated by interpolating global ensemble onto the experimental domain. The initial conditions of global ensemble were respectively generated by the ensemble Kalman filter (EnKF), the ensemble transform with rescaling (ETR), and the ETR applied within the EnKF (EnKF_ETR) methods. The 6-hr forecasts of the EnKF and ETR ensembles were already available from National Centers for Environmental Prediction (NCEP). Impacts of global ensemble initialization schemes on the regional hybrid data assimilation system were explored. Results demonstrated that the EnKF_ETR-3DVAR initialized forecasts produced lower errors than the other two methods. The forecasts generated by the EnKF-3DVAR method performed better at levels below 700hPa and worse at levels above than the ETR-3DVAR method. The possible causes of poor performances of EnKF- and ETR- hybrid methods were analyzed. Inflations applied within the EnKF and ETR made ensemble spreads decay or grow slowly from the initialization to 6-hr forecast.