A hybrid of four-dimensional variational and ensemble-based data assimilation methods for precipitation prediction in hydropower planning and risk management

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Thursday, 8 January 2015: 9:15 AM
224B (Phoenix Convention Center - West and North Buildings)
Meng Zhang, IBM, Yorktown Heights, NY; and B. Xie, H. wang, W. Zhuang, C. Chen, L. Liu, and W. Yin

The ensemble Kalman filter is coupled with four-dimensional variational data assimilation to implement an optimal state and uncertainty analyses for numerical weather prediction, which utilizes the rain gauge observation data of a major river basin over Southwest China in real-time to improve the forecast skills of precipitation prediction. The advantages of hybrid data assimilation are derived both from the flow-dependent and static background error covariance of the two components, respectively. In this case study of a typical heavy rainfall event in June 2014, the accuracy of the precipitation forecasting, is modeled by an ensemble group for its uncertainty estimation and a deterministic variational analysis for its optimal state estimation. The results show that the 24h rainfall forecast is significantly improved by the evaluation of the TS score, especially the center and strength of rainfall systems are altered to be close to real rainfall location after refreshed cycles of the rain gauge data assimilation and Weather Research and Forecast (WRF) model forecast. The accurate precipitation forecasts are also applied to support the daily operations of hydropower planning and risk management.