Mitigation of flash floods in arid regions using adjoint sensitivity analysis

This paper presents an analysis of the sensitivities of flood wave propagation to variations in certain control variables and boundary conditions by means of the adjoint method. This uses a variational technique to find the relationships between changes in predicted flood water levels and changes in control variables such as the inflow hydrograph, bed roughness, and bed elevation. The sensitivities can be used for optimal control of hydraulic structures, for data assimilation, for decision makers' procedures, for the analysis of the effects of uncertainties in control variables on the predictions of floods water levels, and for investigating both the sensitivities of model flood forecasts to model parameters, boundary and initial conditions. Example of the last application of the sensitivity analysis is presented and discussed These methods are developed and implemented through a numerical hydraulic model of channel flow based on the Shallow Water Equations (SWEs) and the corresponding adjoint model. The equations are integrated using finite difference methods and a new modified method of characteristics is used to define the open boundaries. Results of validation tests on both the forward hydraulic model and on the adjoint model are presented