This work presents how machine learning models predict discharge and outperforms the 6 h lead-time empirical model used operationally in Toulouse. The 40 k points database includes discharge and rainfall data, for 36 flood events. The approach also provides a reliable solution for extended 8 h lead-time. The scarcity and the heterogeneity of the data, especially in presence of outliers, heavily weigh on the learning strategy. Rainfall data processing increases the predictive performances.

Predicting water resource evolution is a key challenge for the coming century. Anthropogenic impacts on water resources, and particularly the effects of dams and reservoirs on river flows, are still poorly known and generally neglected in global hydrological studies. A parameterized reservoir model is reproduced to compute monthly releases in Spanish anthropized river basins. For global application, an exhaustive sensitivity analysis of the model parameters is performed on flows and volumes.

The flow of freshwater in rivers is commonly studied with computer programs known as hydrological models. An important component of those programs lies in the description of the river environment, such as the channel resistance to the flow, that is critical to accurately predict the river flow but is still not well known. Satellite data can be combined with models to enrich our knowledge of these features. Here, we show that the coming SWOT mission can help better know this channel resistance.

Here, the first assessment of future changes in extreme precipitation in small Mediterranean watersheds is done through three watersheds frequently subjected to flash floods. Collaboration between Spanish and French laboratories enabled us to conclude that the intensity of high precipitation will increase at the end of the century. A high degree of confidence results from the multi-model approach used here with eight regional climate models (RCMs) developed in the Med and Euro-CORDEX project.