AUROrA

AI-Driven Urban Flood Resilience: Integrating Earth Observation and Architectural Innovation

Principal Investigators 

Prof. Stefan Bauer,
Chair of Algorithmic Machine Learning & Explainable AI, TUM School of Computation, Information and Technology

Prof. Markus Disse,
Chair of Hydrology and River Basin Management, TUM School of Engineering and Design

Prof. Thomas H. Kolbe,
Chair of Geoinformatics, TUM School of Engineering and Design

Prof. Xiaoxiang Zhu,
Chair of Data Science in Earth Observation, TUM School of Engineering and Design

Abstract Urban areas are increasingly vulnerable to flooding due to climate change and rapid urbanization, which decrease natural water retention and alter hydrological cycles. Despite strategies like green infrastructure and improved drainage systems, a key challenge remains in how to systematically assess and enhance urban flood resilience with limited resources. Existing approaches often work in isolation, highlighting the need for integration. This project aims to develop a framework for urban flood analysis by integrating multi-sensor Earth observation (EO) data, hydraulic-hydrological models, urban digital twins (including critical infrastructure such as sewer and metro networks), and AI methodologies. The system will 1) enhance flood detection by linking observed parameters with digital twins, 2) quantify resilience through an AI-based surrogate hydrodynamic model, 3) generate what-if digital twin scenarios with AI-driven, quality assured methods, and 4) automate scenario creation via representative learning from EO data. The integrated approach will improve and accelerate flood resilience planning with minimal investment. Calibration and validation will use data from significant flood events in Germany (2021), Brazil (2024), and Spain (2024), ensuring robustness across contexts. Transferability will be supported by open standards such as CityGML, 3DTiles, and SensorThings API.