Artificial intelligence at the service of flood risk management.
Ex Machina is among the partners of PoC 3.7 “Simulation models for river and watercourse flooding”, one of the pilot projects of the Italian interregional program REG4IA dedicated to innovative flood risk management. A PoC, or proof of concept, is a small-scale experimentation that serves to demonstrate the feasibility of a solution before adopting it structurally. We are participating by bringing the experience gained with our AIRAS – Artificial Intelligence Response on Alert System, the decision support platform developed for protecting territories from extreme meteo-hydrogeological events.
The project is promoted by the Emilia-Romagna Region through the Directorate General for Resources, Europe, Innovation and Institutions, the Regional Agency for Territorial Security and Civil Protection (ARSTPC), and the Regional Agency for Prevention, Environment and Energy (ARPAE). Working alongside us are the University of Padua, the University of Parma, the Politecnico di Milano, and the CIMA Foundation. The activities formally kick off in spring 2026, with closure expected by the end of the same year.
The idea from which the PoC starts is to replicate and enhance an already operational application: the real-time simulation of flooding scenarios from levee breaches developed for the Lamone River after the May 2023 flood, the result of agreements with the universities of Padua and Parma. That system couples a hydrological model (RHYME, developed by UNIPD) and a two-dimensional hydrodynamic model (PARFLOOD, by UNIPR) within the FEWS interface, leveraging the computing power of the regional supercomputer “MarghERita”. With MarghERita, it is possible to simulate in real time the scenario of three simultaneous levee breaches.
The objective of the new project is to extend this approach to another interregional watercourse between Tuscany and Emilia-Romagna. For technical reasons, the choice fell on the Senio River, the last right-bank tributary of the Reno: a stretch of about 101 kilometers, continuously embanked downstream of Castelbolognese and therefore particularly exposed to the risk of breaches. The ultimate goal is to improve forecasting through preventive warning and evacuation actions, leveraging the opportunities offered by artificial intelligence.
Our contribution focuses on the most innovative components of the workflow. We are involved in testing AI for the simulation of levee breaches and in the exploratory assessment to identify embankment vulnerabilities, in collaboration with university hydraulic and geotechnical groups. We also participate, together with the Politecnico di Milano and the CIMA Foundation, in the development of the Impact-Based Forecast approach, both for the real-time assessment of flood impact and for “what if” scenarios useful for planning and prevention.
The ability to create an AI-based support system for those operating on the territory, capable of integrating the outputs of the models developed by the various universities involved in the project, and the development of a dedicated conversational agent stem precisely from the experience gained with AIRAS.
AIRAS is in fact a conversational agent trained on existing civil protection plans and procedures, capable of assisting decision-makers during alert phases. The tool is based on generative models instructed, through fine-tuning techniques, on regional regulatory documentation, municipal civil protection plans, and standardized operating procedures.
The system also correlates various data collected on the territory, such as rainfall data, with historical flood models, providing operational indications for possible preventive road closures or the dispatch of resources when alarm thresholds are reached, or other territorial safety activities.
Among the expected outcomes of the REG4IA PoC is the preparation of guidelines to replicate the procedure in other territories. The system developed on the Senio, in fact, is not intended as an isolated solution, but as a transferable methodology to other basins and other regions. For us, it is an opportunity to bring the experience gained with AIRAS into a real and complex case, putting artificial intelligence at the service of those who, in the field, must decide quickly and well when the water rises.
