Flood Management in a Complex River Basin with a Real-Time Decision Support System Based on Hydrological Forecasts

During the last decades, the Upper Rhone River basin has been hit by several flood events causing significant damages in excess of 500 million Swiss Francs. From this situation, the 3rd Rhône river training project was planned in order to improve the flood protection in the Upper Rhone River basin in Vaud and Valais Cantons. In this framework, the MINERVE forecast system aims to contribute to a better flow control during flood events in this catchment area, taking advantage of the existing hydropower multi-reservoir network. This system also fits into the OWARNA national project of the Swiss Federal Office of Environment by establishing a national platform on natural hazards alarms. The Upper Rhone River basin has a catchment area with high mountains and large glaciers. The surface of the basin is 5521 km2 and its elevation varies between 400 and 4634 m a.s.l. Numerous hydropower schemes with large dams and reservoirs are located in the catchment area, influencing the hydrological regime. Their impact during floods can be significant as appropriate preventive operations can decrease the peak discharges in the Rhone River and its main tributaries, thus reducing the damages. The MINERVE forecast system exploits flow measurements, data from reservoirs and hydropower plants as well as probabilistic (COSMO-LEPS) and deterministic (COSMO-2 and COSMO-7) numerical weather predictions from MeteoSwiss. The MINERVE hydrological model of the catchment area follows a semi-distributed approach. The basin is split into 239 sub-catchments which are further sub-divided into 500 m elevation bands, for a total of 1050 bands. For each elevation band, precipitation, temperature and potential evapotranspiration are calculated. They are considered in order to describe the temperature-driven processes accurately, such as snow and glaciers melt. The hydrological model was implemented in the Routing System software. The object oriented programming environment allows a user-friendly modelling of the hydrological, hydraulic and operating processes. Numerical meteorological data (observed or predicted) are introduced as input in the model. Over the calibration and validation periods of the model, only observed data (precipitation, temperature and flows) was used. For operational flood forecast, the observed measurements are used to update the initial conditions of the hydrological model and the weather forecasts for the hydrological simulations. Routing System provides then hydrological predictions in the whole catchment area. Subsequently, a warning system was developed especially for the basin to provide a flood warning report. The warning system predicts the evolution of the hydrological situation at selected main check points in the catchment area. It displays three warning levels during a flood event depending on respective critical discharge thresholds. Furthermore, the multi-reservoir system is managed in an optimal way in order to limit or avoid damages during floods. A decision support tool called MINDS (MINERVE Interactive Decision Support System) has been developed for real-time decision making based on the hydrological forecasts. This tool defines preventive operation measures for the hydropower plants such as turbine and bottom outlet releases able to provide an optimal water storage during the flood peak. The overall goal of MINDS is then to retain the inflowing floods in reservoirs and to avoid spillway and turbine operations during the peak flow, taking into account all restrictions and current conditions of the network. Such a reservoir management system can therefore significantly decrease flood damages in the catchment area. The reservoir management optimisation during floods is achieved with deterministic and probabilistic forecasts. The definition of the objective function to optimise is realised with a multi-attribute decision making approach. Then, the optimisation is performed with an iterative Greedy algorithm or a SCE-UA (Shuffled Complex Evolution – University of Arizona) algorithm. The developed decision support system combines the high-quality optimisation system with its user-friendly interface. The purpose is to help decision makers by being directly involve in main steps of the decision making process as well as by understanding the measures undertaken and their consequences.