NOWCAST AND FORECAST MODELING OF WATER-LEVEL PROFILES IN RIVERS BASED ON IMPROVED DIEX-FLOOD AND DEEP LEARNING

抄録

 We improved the fundamental equation and the data assimilation algorithm of the DIEX-Flood, which can estimate the longitudinal profile of the water level in the current and future situation in order to reduce the computational cost and maintain the robustness of the model. As a result of applying the improved DIEX-Flood to some flood events of the Kinu River, we can simulate the longitudinal profile of the water level in the current situation with high accuracy, good stability and low computational cost. Furthermore, we proposed a new flood forecasting method combining the DIEX-Flood and deep learning. This prediction method can calculate the longitudinal profile of the water level in the future, so it makes the evaluation of the spatial and temporal distribution of flood risk possible. As a result of the flood forecast simulation using this system, we can predict water level profiles 6 hours ahead with high accuracy.