Different triggering mechanisms make grasslands more susceptible to clustered landslides than forests under extreme rainfall on the loess plateau.

Under extreme rainfall and vegetation restoration, clustered shallow landslides have occurred in some watersheds on the Loess Plateau. Accurately assessing landslide probability and understanding vegetation effects are critical for hazard prevention. Using an improved SINMAP model with two topographic wetness conditions representing distinct trigger mechanisms: w = 0 (loss of matric suction) and w = 1 (upper limit of pore water pressure), shear strength parameters were back-calculated from landslide data. Results show that observed landslides could be reproduced using the model with lower cohesion (≤2 kPa), indicating limited root reinforcement under extreme rainfall. At w = 0, increasing cohesion from 1 to 2 kPa reduced the proportion of landslide points in unstable area (Sujiagou: from 53.23% to 20.85%, Nanxiaohegou: from 81.78% to 48.26%, Renjiatai: from 85.66% to 61.76%), indicating that higher cohesion promotes a greater reliance on seepage forces for slope failure. The unstable area in grasslands responded more sensitively to effective rainfall than in forests: taking the unstable area under w = 1 as the baseline, over ΔPr = 0-90 mm d-1, for each 10 mm d-1 increase, the mean increases in unstable area in grasslands in Sujiagou, Nanxiaohegou, and Renjiatai (10.13%, 6.75%, and 10.28%) were significantly higher than those (6.16%, 3.74%, and 6.48%) in forests (P < 0.05). This is attributed to lower critical cohesion in grasslands, greater dependence on pore-water pressure buildup, and lower lateral hydraulic conductivity, making pore-water pressure more responsive to rainfall. This finding deepens understanding of clustered landslide mechanisms under vegetation effects.