Hydrological ensemble forecasts: developments to improve their quality and estimate their utility

The last decade has seen the emergence of streamflow probabilistic forecasting as the most suitable approach to anticipate risks and provide warnings for public safety and property protection. However, beyond the gains in security, the added-value of probabilistic information also translates into economic benefits or an optimal management of water resources for economic activities that depend on it. In streamflow forecasting, the uncertainty associated with rainfall predictions from numerical weather prediction models plays an important role. To go beyond the limits of classical predictability, meteorological services developed ensemble prediction systems, which are generated on the basis of perturbations of the initial conditions of the models and stochastic variations in their parameterization. Equally probable scenarios of the evolution of the atmosphere are proposed for forecasting horizons up to 10-15 days. The integration of weather ensemble predictions in the hydrological forecasting chain is an interesting approach to produce probabilistic streamflow forecasts and quantify the total predictive uncertainty in hydrology. Last and final summary in the thesis. (author)