INVERSE ANALYSIS AND DATA ASSIMILATION OF RIVER FLOW

Abstract

 This paper presents the inverse analysis of boundary conditions and parameters in river flow analysis, using the adjoint sensitivity method adopted as data assimilation method for weather forecasting and contributing to the improvement of forecasting accuracy. The adjoint equation and sensitivity coefficient were derived for the cross-section-averaged one-dimensional unsteady flow. The numerical simulation method was shown, and the applicability for the actual river was verified. The data assimilation using multi-point water gauges successfully estimated discharge at any point, and the accuracy was changed with the numbers of water gauges. In addition, in the data assimilation using water gauges at 8 points for the river channel network, the data assimilation results also showed high accuracy, reproducing observed discharge and water level. Furthermore, the forecasting simulation using the assimilation results as initial values showed highly accurate predicted water level up to 2 hours ahead. Next, its method of data assimilation was applied to the channel optimization. In the optimization of the channel shape considering two cases of river-bed excavation and channel widening, where the river water level is below the levee height, the inverse analysis was successfully applied to determine the optimized channel shape by one time simulation.