RIVER FLOW INVERSE ANALYSIS AND DATA ASSIMILATION

This paper presents the inverse analysis of boundary conditions and parameters of river flow. The adjoint variable method is adopted for data assimilation for weather forecasting and is found to improve forecasting accuracy. The adjoint equation and sensitivity were derived for one-dimensional unsteady flow, a numerical simulation method was illustrated, and the applicability of the method to an actual river was verified. Data assimilation using multi-point water gauges successfully estimated discharge at any point and the accuracy changed with the number of water gauges. In the case of a river channel network, the data assimilation results also showed high accuracy. Furthermore, the forecasting simulation using the assimilation results as initial values showed highly accurate predicted water levels up to two hours in advance. The data assimilation method was then applied for channel shape optimization. In the optimization of the channel shape, considering two cases of riverbed excavation and channel widening where the river water level was below the levee height, the inverse analysis was successfully applied to determine the optimized channel shape via one-time simulation.