With ongoing global climate warming and humidification, extreme rainfall events have become increasingly frequent in southern China, and the resulting clustered shallow landslides have emerged as one of the dominant types of geological hazards in the region. Based on a combination of YOLOv11-based automatic detection and manual correction, this study investigates two clustered shallow landslide events triggered by extreme rainfall in May 2022 and June 2024 in Wuping County, Fujian Province. A total of 2,407 shallow landslides were identified and interpreted, and their area characteristics, planform geometries, and spatial distribution patterns were systematically analyzed. In addition, the relative contributions of twelve influencing factors, including elevation, were quantitatively evaluated using the interpretability of a LightGBM model. The results indicate that rainfall, slope gradient, elevation, and the topographic wetness index (TWI) are the primary controlling factors governing the spatial distribution of shallow landslides, whereas the normalized difference vegetation index (NDVI) and land-use type play a moderating role in local landslide occurrence. Field investigations further reveal six common characteristics of the two clustered shallow landslide events: (1) the landslide bodies are generally thin, with 97% exhibiting thicknesses of less than 2 m; (2) some landslides lack a continuous and unified slip surface, and their movement directions vary with slope aspect, indicating the combined effects of rainfall-induced surface hydraulic erosion and gravitational erosion; (3) landslides predominantly initiate from ridge crests or gully heads, and unlike most landslides in Fujian Province that occur near residential areas, the large relative height difference at the slope toe in these events resulted in high destructive potential and severe casualties and property losses; (4) landslide occurrence closely coincides with peak rainfall intensity, with numerous landslides developing within a short period, thereby severely constraining early warning and evacuation efforts; (5) the original slope surfaces in landslide-prone areas are densely vegetated and highly concealed, dominated by tree species such as Schima superba, Pinus massoniana, bamboo, and Cunninghamia lanceolata, making potential hazards difficult to detect through routine inspections; and (6) exposed slip surfaces after failure exhibit almost no vegetation roots, indicating shallow root penetration that does not cross the slip surface, and thus the reinforcing and anchoring effects of vegetation roots contribute little to slope stability in the study area. This study enhances the understanding of the formation mechanisms of clustered shallow landslides in southeastern coastal China under climate change and provides a scientific basis for landslide risk early warning and disaster prevention under extreme rainfall conditions.
 

Clustered shallow landslides,Extreme rainfall,Spatial distribution,Multiscale Data