Floods are among the most devastating natural disasters, and scientists need fast, reliable ways to map where flooding occurs using satellite data. Current methods have major drawbacks: they either require extensive training data that takes time to collect, or they struggle to tell the difference between flood water and normal water bodies like rivers and lakes. We developed a new approach using Bayesian change point analysis to automatically detect floods from radar satellite images. Our method works by spotting sudden changes in how the ground reflects radar signals back to the satellite—when an area floods, this reflection pattern changes dramatically. The key advantage is that our method needs no training data and works immediately in any location worldwide. We tested it on three major flood events in China, Somalia, and Ukraine, where it significantly outperformed existing methods. It works especially well in open areas like farmland and plains, though it struggles more in complex urban environments. This breakthrough means emergency responders could potentially map flood extent within hours of a disaster anywhere in the world, without waiting to train complex computer models or gather local data. This rapid response capability could save lives and improve disaster management.