A new flood prediction method with data assimilation for water-level data

To deal with flood risks which have been increasing as a result of climate change, it is important to evaluate the nowcast and forecast of the streamwise distribution of water level in rivers with high accuracy. This study presents a new flood prediction method with data assimilation technique for water-level data which is referred to as Dynamic Interpolation and Extrapolation method for Flood prediction (DIEX-Flood). In the present method, the observed water-level data at several points are interpolated and extrapolated in the streamwise direction with satisfying 1-D momentum equation and continuity equation. Furthermore, water-level profiles in future time are predicted by unsteady flow calculation with using current time's hydraulic / channel conditions. In order to validate the performance of the present method, it was applied to water-level data at several points under high-flow conditions in Edo River in Japan. The results indicate that the present method can smoothly interpolate water-level profiles from the observed data, and calculated maximum profile is in good agreement with observed high-water-mark profile. Furthermore, high accurate prediction is possible in the entire river when accurate boundary conditions are given. In addition, even if boundary conditions include errors, prediction accuracy of the downstream can be maintained due to the lead time of the flood propagation.