The FLOWS project: Improving flood crisis management through Earth observation solutions

Kurzfassung

Europe is increasingly experiencing devastating floods, highlighting the severe impacts of climate change and exposing vulnerabilities in land use and population safety. Earth Observation (EO) technologies offer significant potential for flood crisis management, but the catastrophic floods in Belgium and Germany in July 2021 revealed gaps in the effective use of EO data, largely due to a lack of methods adapted to the needs of crisis managers and insufficient awareness of available tools. The FLOWS project addresses these challenges by determining how and when EO data and derived products can optimally support flood crisis management across three phases: crisis response, aftermath, and reconstruction. The project builds on the experiences of crisis managers during the 2021 floods and a comprehensive analysis of EO data acquired during the event. Using a problem-tree approach, the project identifies geospatial challenges faced by first responders and stakeholders, including crisis centers, emergency services, authorities, municipalities, and water managers. These stakeholders are actively involved in the process, providing input to guide development and validation through Agile prototyping. Key innovations include methodologies to enhance situational awareness, such as leveraging multi-sensor EO data through optimized algorithms for UAVs, commercial SAR, and Sentinel-1/2 data. Real-time computer vision pipelines enable adaptive UAV flight adjustments and efficient onboard processing of RGB imagery, ensuring rapid detection of flooded areas and victims, and providing first responders with essential data. Social media and mobile phone data are integrated into GIS-based solutions to map population dynamics at fine spatial and temporal scales. Additionally, deep learning methods are applied to assess flood-induced damage across diverse environments, including urban and rural areas. Using very high-resolution (VHR) and Sentinel data, the project develops automated tools to classify and map impacts on buildings, transportation networks, vegetation, and riverbanks, supporting both immediate response and long-term recovery planning. Finally, heterogeneous data sources are integrated into disaster hot spot maps using probabilistic fusion techniques. These maps provide an up-to-date situational overview of the most affected areas, enabling crisis managers to prioritize response efforts and allocate resources efficiently. By integrating EO methodologies and engaging key stakeholders, FLOWS aims to advance flood preparedness and crisis management, ultimately supporting flood-affected populations and fostering resilient communities.