Our purpose is to investigate a possible adaptation strategy for multi-objective reservoirs management on the Seine river basin, upstream of Paris, by using a centralized Tree-Based Model Predictive Control (TB-MPC). This method uses all the information available in real-time, including ensemble weather forecasting, to obtain an adaptive control. In TB-MPC, a tree is generated from an ensemble of weather forecasts. The tree structure summarizes the information contained in the ensemble, specifying the time, along the optimization horizon, when forecast trajectories diverge and thus uncertainty is expected to be resolved. This information is then used in the model predictive control framework, to optimize an objective function over a finite receding horizon, using a model to predict the evolution of the system in response to the forecasted inputs. The TB-MPC controller is implemented in combination with the integrated model of the Seine river basin, including a semi-distributed hydrologic model of the watershed, a simplified hydraulic model of the river network, and the four reservoir models. The controller optimizes a global cost function that takes into account the costs associated to high and low flows, based on critical thresholds at some key downstream stations, and a penalty cost based on the final reservoir storages, to guarantee a sustainable management in the long term. The reservoir's management is tested under the evolution of hydro-climatic conditions due to climate change, using statistically downscaled and bias corrected climate simulations produced by seven Global Circulation Models (GCMs). For the present conditions, time series of ensemble weather forecasts are available and will be used to statistically generate time series of ensemble forecasts for the future. The performances of the TB-MPC controller will be simulated and compared to the actual decentralized management performances over present and future scenarios.