This master project's objectives are: - Understanding of the main approaches commonly used for discharge forecasting. - Understanding of key concepts of machine learning and evolutionary optimization. - Learning how to interpret discharge forecasts, both deterministic and probabilistic, and evaluate their performance. - Work on the acquisition of near-real time satellite rainfall estimates and use them to perform discharge forecasts. - Address the optimization of reservoir operations based on probabilistic forecasts. - Study the limits of applicability of the methodology (e.g. nature of the series being forecasted, data requirements). - Study how the methodology can be applied to selected catchments in Southern Africa. - Compare the results and use of the proposed forecasting methodology with those obtained by Bayesian inference (Differential Evolution Adaptive Metropolis, DREAM).