18 Aug Stefan Szylkarski
Stefan Szylkarski, Australia
Presentation Title: A combined hydrological-hydrodynamic data assimilation approach for improving river flow forecasting.
Data assimilation is an important part of river forecast systems to improve forecasts and increase forecast lead time. State-of-the-art forecast systems include assimilation of water level and discharge measurements for updating the hydrodynamic state of the river model. In this case, the impact of the data assimilation is limited to a time horizon where the improved initial river state is washed out. Improvement in forecast skills can be obtained by updating also the hydrological state variables to benefit from the hydrological memory of the catchment. A combined hydrological-hydrodynamic data assimilation approach has been developed in the MIKE HYDRO River modelling system (successor of the MIKE 11 river modelling system). The data assimilation is based on the ensemble Kalman filter. It supports data assimilation for both the hydrodynamic river model and the catchment rainfall-runoff models using available catchment runoff and river water level measurements. Besides updating the hydrological-hydrodynamic state, the Kalman filter provides an estimate of the uncertainty of the state, which is used for provision of confidence intervals of the flow forecasts. The new data assimilation approach is demonstrated on a case study from the Murrumbidgee River in New South Wales, Australia. Discharge observations are available for several catchments for assimilation in the rainfall-runoff models. It is demonstrated that the combined hydrological-hydrodynamic data assimilation approach provides better forecast skills compared to assimilation of the hydrodynamic river model only.