Abstract
This thesis aims to propose an approach and tools for developing flash flood forecast bulletins that are tailored to the needs of crisis managers and aid decision-making in crisis situations. To achieve this goal, a characterization of the link between crisis management and hydrological forecasting is conducted through a scientific review. This review highlights the lack of understanding of hydrological processes and functioning crisis managers in small communities, which prevents them from fully benefiting from this information. Conversely, hydrologists show little interest in specific variables that could be directly applied to crisis management.To bridge this gap, a workshop methodology was developed to deepen the understanding of hydrometeorological forecast bulletins by crisis managers and their needs. Neural networks we employed to forecast hydrological variables and to classify crisis management variables. The evaluation of the forecast quality was focused on their relevance in crisis situations.The results indicate that the link between crisis managers and forecasters is insufficient, with both parties having different objectives and priorities. Regarding hydrological forecasts, better quality is observed in level forecasts compared to discharge forecasts, both from a hydrological and crisis management perspective. The workshop helped define the desired and useful forecast bulletins for managers at various levels. The evaluation of forecast quality in relation to crisis management demonstrated that these forecasts are inseparable from the objectives of the managers. Finally, the feasibility of predicting crisis management variables through classification using artificial neural networks was demonstrated.The results of this work are promising as they show that it is possible to apply simple, yet highly useful, models that are directly exploitable in crisis management. This is made possible by the properties of neural networks. This work also proposes an approach to defining the needs of managers. In the future, these models and approaches can be improved and applied to other regions. Research efforts to explore the links between hydrology and crisis management should continue due to the complexity of these links.