General rightsThis document is made available in accordance with publisher policies. Please cite only the published version using the reference above. Full terms of use are available: http://www.bristol.ac.uk/pure/about/ebr-terms Correspondence to: andrea.ficchi@irstea.fr 2 AbstractWe investigate the improvement of the operation of a four-reservoir system in the Seine River basin, France, by use of deterministic and ensemble weather forecasts and real-time control. In the current management, each reservoir is operated independently from the others and following prescribed rule-curves, designed to reduce floods and sustain low-flows under the historical hydrological conditions. However, this management system is inefficient when inflows are significantly different from their seasonal average and may become even more inadequate to cope with the predicted increase in extreme events induced by climate change.In this work, we develop and test a centralized real-time control system to improve reservoirs operation by exploiting numerical weather forecasts that are becoming increasingly available.The proposed management system implements a well-established optimization technique, Model Predictive Control (MPC) and its recently modified version that can incorporate uncertainties, Tree-Based Model Predictive Control (TB-MPC), to account for deterministic and ensemble forecasts respectively. The management system is assessed by simulation over historical events and compared to the "no-forecasts" strategy based on rule-curves.Simulation results show that the proposed real-time control system largely outperforms the "no-forecasts" management strategy, and that explicitly considering forecasts uncertainty through ensembles can compensate for the loss in performance due to forecasts inaccuracy.

Abstract. Calibration of a conceptual distributed model is challenging due to a number of reasons, which include fundamental (model adequacy and identifiability) and algorithmic (e.g., local search vs. global search) issues. The aim of the presented study is to investigate the potential of the variational approach for calibrating a simple continuous hydrological model (GRD; Génie Rural distributed involved in several flash flood modeling applications. This model is defined on a rectangular 1 km2 resolution grid, with three parameters being associated with each cell. The Gardon d'Anduze watershed (543 km2) is chosen as the study benchmark. For this watershed, the discharge observations at five gauging stations, gridded rainfall and potential-evapotranspiration estimates are continuously available for the 2007–2018 period at an hourly time step. In the variational approach one looks for the optimal solution by minimizing the standard quadratic cost function, which penalizes the misfit between the observed and predicted values, under some additional a priori constraints. The cost function gradient is efficiently computed using the adjoint model. In numerical experiments, the benefits of using the distributed against the uniform calibration are measured in terms of the model predictive performance, in temporal, spatial and spatiotemporal validation, both globally and for particular flood events. Overall, distributed calibration shows encouraging results, providing better model predictions and relevant spatial distribution of some parameters. The numerical stability analysis has been performed to understand the impact of different factors on the calibration quality. This analysis indicates the possible directions for future developments, which may include considering a non-Gaussian likelihood and upgrading the model structure.

Adaptation strategies will be needed to cope with the hydrological consequences of projected climate change. In this perspective, the management of many artificial reservoirs will have to be adapted to continue to fulfil downstream objectives (e.g. flow regulation). This study evaluates the sustainability of the management rules of the artificial reservoirs on the Seine River basin, 5 France, under climate change scenarios. The Seine River basin at Paris (43,800 km 2) has major socioeconomic stakes for France, and the consequences of droughts and floods may be dramatic. In this context, four large multipurpose reservoirs were built on the basin during the XX th century for low-flow augmentation and flood alleviation. A hydrological modelling chain was designed to explicitly account for reservoir management 10 rules. It was calibrated in current conditions and then fed by the outputs of seven climate models in present and future conditions, forced by the A1B IPCC scenario, downscaled using a weather-type method and statistically bias-corrected. The results show that the hydrological model performs quite well in current conditions. The simulations made in present and future conditions indicate a decrease in water availability and 15 summer low flows, but no significant trends in high flows. Simulations also indicate that there is room for progress in the current multipurpose management of reservoirs and that it would be useful to define proper adaptation strategies.