Model test study on the hydrological mechanisms and early warning thresholds for loess fill slope failure induced by rainfall

Abstract Under rainfall conditions, the characteristics of hydraulic changes related to the failure evolution of fill slopes are critical to the early warning of slope failure. In this study, a series of model tests are carried out on loess fill slopes with different slope types. The matric suction, volumetric moisture content, pore water pressure and deformation of the slope are monitored in real time during the rainfall process. Meanwhile the development of cracks and the failure modes of the slope are analyzed, and then the early warning strategy of fill slope failure induced by rainfall is studied. The results show that loess landslides induced by rainfall exhibit a hysteresis. However, the volumetric water content and the matrix suction respond earlier than the pore water pressure. The slope failure is produced when the volumetric water content and the matrix suction are respectively at their maximum and minimum values. The evolution mechanism of cracks on the slope starts when tensile cracks are first generated near the leading edge surface and then extend backwards. In the next stage, shear fracture is produced near the flanking of the slope and finally, the trailing edge produces a transfixion tensile crack from the top to toe of the fill slope, which not only provides the advantage of an infiltration channel to control the hydrological process, but also evolves into the trailing edge boundary of each failure of fill slope. With the increase of rainfall duration, the filling slope presents different failure characteristics, revealing the hydraulic behavior of a partially saturated slope. In addition, combined with the infinite slope stability analysis in Mohr–Coulomb soil, the matrix suction and volumetric water content thresholds of the fill slopes with different slope angle are determined. The early warning threshold model for rainfall induced slope instability is proposed and a new framework of early warning is provided.