A combination of social media and satellite data improves flood monitoring in China

Global climate change has increased flood frequency worldwide, yet many urban and flash floods remain poorly captured by satellite remote sensing. Here we integrate Global Satellite Mapping of Precipitation data with 92.98 million Sina Weibo posts to trace spatiotemporal flood footprints across China from 2012 to 2024. Using a cascading threshold method, we identify 6,018 rainstorm events across 370 cities. Topic modeling of posts within rainstorm affected areas detects 1,094 flood events, far exceeding the 114 and 45 events recorded in Emergency Events Database and Dartmouth Flood Observatory. Comparison with satellite imagery on Google Earth Engine shows that approximately 50% of these events were unobservable by remote sensing. While satellites capture riverine flooding in open areas, social media fills critical gaps in densely populated urban settings. This social sensing framework reveals previously undocumented flood events and their propagation patterns, offering a complementary approach that enhances traditional flood monitoring capabilities. In China, about half of the flood events were unobservable by remote sensing, capturing mostly riverine flooding in open areas, while social media fills gaps in densely populated urban settings, according to a social sensing framework that combines satellite precipitation mapping and Sina Weibo posts