Mapping global floods with 10 years of satellite radar data

Floods are among the deadliest and most costly forms of extreme weather, affecting millions of people and causing over $40 billion in damage annually. And they’re getting worse. As climate change increases the frequency and intensity of extreme precipitation, we need better data to understand where and when floods might strike—and to respond quickly when they do.

At Microsoft’s AI for Good Lab, our team set out to address a major challenge: the lack of consistent, long-term global flood maps. While satellites have long helped track floods, optical imagery is often obstructed by cloud cover, making it difficult to observe flood events—especially over time and at scale.

To overcome this, we developed a global flood detection model, published in Nature Communications, using ten years of Satellite Synthetic Aperture Radar (SAR) data. SAR can detect floods through clouds, at night, and in remote areas. This marks the first time a global, long-term flood dataset of this kind has been produced using SAR imagery.

https://www.nature.com/articles/s41467-025-60973-1

 This work is already making a difference.

 In Ethiopia, we used the model to detect nearly three times more flood-prone areas than existing datasets. It confirmed known flood zones and revealed new ones, offering a more complete view of flood risk across the country.

After devastating floods struck Kenya in May 2024, we used this tool to estimate that 75,000 hectares—about 2 percent of all cropland—were flooded or at risk. This aligned closely with government reports and helped support real-time disaster response efforts.

The tool is publicly available and designed to run on something as simple as a laptop. It can be used by researchers, planners, and governments anywhere to improve flood preparedness and resilience. We’re already sharing data and case studies with partners in Keny, Ethiopia, Spain, and the United Nations.

By filling a critical data gap with open, scalable AI tools, we hope this work helps communities better prepare for what’s ahead—and respond more quickly when disaster strikes.

#AIforGood #ClimateAdaptation #DisasterResponse #MicrosoftResearch #GeospatialAI