Abstract: Background From July 30 to August 1, 2023, the Mentougou District of Beijing experienced an unprecedented extreme rainfall event, later classified as the “23.7 Catastrophic Basin Flood.” This event, driven by the remnants of Typhoon Doksuri and the West Pacific Subtropical High, brought cumulative rainfall exceeding 1,000 mm in some locations. Unlike typical flash floods in the region, the 23.7 event triggered widespread shallow landslides and soil-driven debris flows, highlighting a rare but increasingly significant hazard pattern in the semi-humid, semi-arid climate zone of North China. The event’s uniqueness lies in its compound triggering mechanism: a combination of prolonged antecedent rainfall and short-duration, high-intensity precipitation episodes. Methods This study integrates high-resolution BJ-2 satellite imagery and UAV photogrammetry to identify 7,011 shallow landslides across Mentougou. A total of nine environmental variables—including slope, aspect, NDVI, lithology, land use, and distances to roads, rivers, and faults—were analyzed using kernel density estimation and a Random Forest model. High-accuracy DEM data from UAV coverage (151.82 km²) was used to extract refined terrain metrics for 1,995 landslide samples. The Random Forest algorithm was employed to evaluate the relative importance of each variable in shaping landslide spatial patterns, while kernel density mapping provided insights into distributional clustering.Results 1) The rainfall event exhibited dual-peak characteristics with extreme intensities. Cumulative rainfall exceeded 1,000 mm over 66 hours, with hourly maxima surpassing 130 mm. These values exceeded the 100-year return period thresholds and met both key triggering conditions for shallow landslides: prolonged saturation and peak-hour intensities far above regional thresholds. 2) Landslides showed strong spatial clustering along river valleys, especially in areas near the Yongding River and its tributaries. High-density zones (up to 88 events per km²) were concentrated in towns like Zhaitang and Yanhecheng. Landslide density decreased with distance from river channels, indicating the significant role of valley morphology and river erosion in landslide initiation. 3) Random Forest analysis revealed that slope (17.5%), NDVI (17.0%), and aspect (14.9%) were the dominant factors controlling landslide distribution. Southeast-facing slopes and gradients between 40° and 50° were particularly prone to failure. Anthropogenic variables such as distance to roads (11.2%) and night light intensity (10.8%) also played substantial roles, suggesting a strong influence from human activity. 4) Slope, vegetation, and aspect interacted in controlling landslide behavior. On slopes less than 40°, high NDVI values (>0.7) were associated with larger landslide areas, likely due to higher soil water retention. On steeper slopes (40°-70°), NDVI had a stabilizing effect, with higher values linked to smaller landslide areas, reflecting enhanced root cohesion and surface protection. 5) Comparison of slope data from 12.5 m DEM and UAV-derived DEM revealed deviations of up to 20°, with coarser data significantly underestimating slope steepness. This highlights the critical need for high-resolution topographic data in landslide susceptibility mapping, especially in complex mountain terrains.Conclusions The 23.7 rainfall event demonstrated a compound hazard mechanism that triggered a rare, widespread outbreak of shallow landslides in northern China's mountainous terrain. The landslides exhibited clear spatial clustering, closely tied to topographic, hydrological, vegetative, and anthropogenic variables. Slope, vegetation cover, and aspect emerged as dominant controls, while road construction and urban development also exacerbated local susceptibility. The findings stress the critical importance of high-resolution elevation data in such analyses and provide a valuable foundation for improving early warning systems and hazard mitigation planning in similar environments under a changing climate.