Secure the resource today. Protect it for tomorrow.
AI-DANA: Design of an operational platform for a digital twin of the Valencian territory
Research on Physics-Informed Neural Networks (PINN) through the incorporation of generative AI and Uncertainty Quantification (UQ) for the prediction of floods in river basins of the Valencian Community
Or to put it another way… design of a digital platform for the early, explainable and probabilistic prediction of floods in river basins
Torrential rains and floods are becoming increasingly frequent and are affecting our cities, infrastructure, and, above all, people. AI-DANA was created to address this challenge: a project that uses artificial intelligence and digital innovation to anticipate risks.
What does this mean?
- Tools that help predict extreme events with greater accuracy.
- Systems that facilitate quick and effective decisions in critical situations.
- Solutions that strengthen the security and sustainability of our territory.
Project objectives
AI-DANA aims to develop an advanced digital platform to anticipate and assess the risk of flooding and overflows in river basins of the Valencian Community. The solution integrates meteorological, hydrological, and territorial information with physical models and AI to predict the evolution of extreme weather events and estimate their potential impacts on the region. Furthermore, it incorporates techniques for quantifying uncertainty and explainability to provide more reliable, understandable, and useful results for decision-making by managers, public administrations, and emergency services.
Progress and results achieved
During 2026, AI-DANA made progress in developing an intelligent platform to anticipate and assess flood risk. Advances were made in the collection, processing, and integration of hydrological, meteorological, and territorial information, as well as in the construction and adjustment of the hydraulic model for the study area, complemented by runoff and urban models. Furthermore, the connection between precipitation prediction models and hydrodynamic simulations was automated, yielding the first integrated results with historical data. In parallel, various AI techniques were evaluated to predict short-term rainfall patterns, and work began on uncertainty analysis, explainability, and the development of the georeferenced digital twin. These advances strengthen the project's technological foundation and allow for continued development into an operational tool that facilitates the early detection of risk situations and contributes to improved prevention and decision-making in the face of extreme weather events.