Loess landslides represent a prevalent and severe type of geological disaster in the Loess Plateau and its surrounding regions. Their frequency and intensity are notably exacerbated under rainfall conditions. This study focuses on investigating the destabilization mechanism of loess landslides induced by rainfall preferential infiltration on the northern slope of the Xining Haihu Bridge. A combination of on-site monitoring, soil property testing, and numerical simulations was employed. The findings reveal that during rainfall events, water rapidly infiltrates into the deep soil layer through pre-existing preferential pathways, such as cracks. This alters the internal water distribution within the soil, leading to localized slope saturation. The subsequent increase in pore water pressure and substantial reduction in soil shear strength emerge as critical factors in triggering loess landslides. Additionally, numerical simulation models were utilized to analyze slope stability under varying rainfall scenarios. The analysis identifies key factors influencing the stability of loess landslides, namely rainfall intensity, duration, and the position and depth of cracks. Furthermore, this study innovatively integrates quantitative analysis of rainfall-induced preferential infiltration with dynamic simulations of landslide stability. This approach offers a more robust theoretical foundation for predicting, assessing, and mitigating loess landslides. By quantifying the relationship between landslide stability and rainfall infiltration patterns, the study provides vital technical support for early warning systems, disaster prevention strategies, and the optimization of engineering measures aimed at addressing loess landslide hazards.

来源平台：FRONTIERS IN EARTH SCIENCE