[Thesis defence] 17/12/2025 – Hadrien Di Costanzo: «Drought and its uncertainties in the Mediterranean context. A Bayesian model for proactive management of water resources and their uses: application to the Gardon d'Anduze and Galeizon basins.» (ESPACE joint research unit)

Mr Hadrien DI COSTANZO will publicly defend his thesis entitled: «Drought and its uncertainties in the Mediterranean context. A Bayesian model for proactive management of water resources and their uses: application to the Gardon d'Anduze and Galeizon basins.» supervised by Cyrille GENRE-GRANDPIERRE and Ms Delphine BLANKE, on Wednesday 17 December 2025.

Laboratory

UMR 7300 ESPACE - Study of Structures, Adaptation Processes and Changes in Space

Composition of the jury

Mr Cyrille GENRE-GRANDPIERRE	Avignon University, ESPACE 7300 CNRS Joint Research Unit	Thesis supervisor
Mr Vincent MORON	Aix-Marseille University, CEREGE UM 34 CNRS	Rapporteur
Ms Anne JOHANNET	IMT-Mines Alès	Rapporteur
Mr Pierre-Alain AYRAL	UMR ESPACE 7300 CNRS (Joint Research Unit)	Examiner
Ms Delphine BLANKE	Avignon University, LMA	Thesis co-director
Mrs Caire DELUS	claire.delus@univ-lorraine.fr	Examiner
Mr Giovanni FUSCO	UMR ESPACE – University of the French Riviera	Examiner

Summary

This thesis proposes an innovative approach to modelling and managing drought in the strict sense, defined as a tension between water supply and demand, based on a Bayesian methodology applied to the Gardon d'Anduze and Galeizon Mediterranean basins. Given the complexity and uncertainty inherent in this phenomenon, a dynamic Bayesian network has been developed to integrate both environmental (precipitation, evapotranspiration, flow rates) and anthropogenic (drinking water, tourism) dimensions, while quantifying uncertainties related to data or scenario-based reasoning. This approach not only simulates interactions between variables, but also identifies critical levers for adaptive management, such as the seasonality of withdrawals or stochastic anticipation of the spread of water deficits in basins. By integrating machine learning techniques, expert advice and various public data (from Météo France, MODIS, INSEE, among others), the Bayesian model, called HydroSec, is a valuable operational tool for predicting periods of water resource stress and guiding the decisions of local managers. This is particularly crucial in the current context of global change, where aridity and water demand have already increased and are likely to continue to do so. The results highlight the effectiveness of Bayesian networks in transforming uncertainty into operational information, paving the way for more resilient and proactive drought management in the Mediterranean region.

Keywords :

Drought, Water resources, Bayesianism, Hydro-climatology, Physical geography, Bayesian network