Multi-Source Remote Sensing Identification and Spatial Distribution Analysis of Floods and Geohazards Triggered by the Extreme Rainfall in Beijing, July 2025

Objectives: From 24 July 2025, Beijing experienced an episode of extreme rainfall that triggered natural disasters, including floods and landslides. These events affected more than 300000 people, damaged over 24000 houses, and attracted widespread national and international attention. Methods: Multi-source remote sensing images, including multi-temporal C-band Sentinel-1 and Gaofen-3 images, were employed for the rapid and intelligent extraction of inundated water bodies using the Multiscale Feature Fusion and Attention Mechanism (MF2AM) model, followed by manual validation. In addition, multi-temporal optical satellite images were used to compile a rainfall-induced geohazards inventory, while terrain and geomorphic factors were incorporated to analyze their spatial distribution and developmental characteristics. Results: The extreme rainfall increased the surface water area by approximately 39.56 km² in the Huairou and Miyun districts. At least 1300 rainfall-induced geohazards were identified, covering a total area of about 5.44 km², mainly classified as small to medium, and primarily concentrated in central Huairou and northwest Miyun. Spatial analysis revealed that 80% of the geohazards occurred at elevations between 250 and 700 m, about 90% developed on slopes of 10°–35°, and most were concentrated on southeast-facing aspects (60°–220°). Overlay analysis with OpenStreetMap road data revealed that natural disasters caused damage to at least 105.98 km of roads, of which floods and geohazards contributed approximately 97.18 km and 8.8 km, respectively. Conclusions: Multi-source remote sensing proved effective in capturing both flood inundation and rainfall-induced geohazards and assessing their impacts on infrastructure. The findings provide critical data support for emergency response, post-disaster reconstruction, and secondary disaster risk assessment.