TJ24.4 Development and Evaluation of a Global Hydrological Ensemble Forecasting System

Tuesday, 8 January 2013: 2:15 PM
Room 10B (Austin Convention Center)
Yu Zhang, Universtiy of Oklahoma, Norman, OK; and Y. Hong, J. J. Gourley, X. Wang, and X. Xue

A Global Hydrological Ensemble Forecasting System (GHEFS) driven by both Real Time (RT) Tropical Rainfall Measurement Mission (TRMM) Multi-satellite Prediction Analysis (TMPA) precipitation ensembles and Global Ensemble Forecast System (GEFS) Quantitative Precipitation Forecast (QPF) ensembles for near real time streamflow simulation and seven-day lead time forecast respectively is under development and evaluation using a physical based Coupled Routing and Excess STorage (CREST) distributed hydrological model. This system provides deterministic and probabilistic (e.g. 95% confidence boundaries of streamflow ensembles) modeling streamflow at both near real time streamflow simulation and seven-day lead time forecast. A sophisticated error model that first quantify the uncertainty in TRMM RT in both temporal dynamics of bias and spatial variability of precipitation estimation error and then generate real time rainfall ensembles by taking TRMM V7 research product as benchmark for the nowcasting mode. For the forecasting mode, the system is forced by GEFS precipitation products which utilize the Ensemble Transform method to account for the uncertainty of forecasts from initial condition errors. Furthermore, a sequential data assimilation approach - Ensemble Kalman Filter (EnKF) is applied to count in the observation error and the uncertainties in the model initial condition. EnKF updates all the internal states for the CREST model whenever the observation (e.g. streamflow) is available. The GHEFS is validated in several basins inside the U.S. and other continents in terms of flood detection capability (e.g. NSCE, Peak Timing), indicating improved prognostic capability by increasing lead time and providing probabilistic outputs, thus offering more time for responding agencies and yielding unique uncertainty information about the magnitude of the forecast impacts.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner