Laboratory Experiments on Steady State Seepage-Induced Landslides Using Slope Models and Sensors

A thorough understanding of the failure initiation process is crucial in the development of physically-based early warning system for landslides and slope failures. Laboratory-scale slope models were constructed and subjected to instability through simulated groundwater infiltration. This is done by progressively increasing the water level in the upslope tank and allowing water to infiltrate laterally towards the toe of the slope. Physical changes in the slope models were recorded by tilt sensors and video cameras. When the model slope was destabilized, the chronology of events occurred in the following sequence: (1) bulging at the toe, (2) seepage at the toe, (3) initial failure of soil mass, (4) piping, (5) retrogressive failure, (6) formation of tension cracks and (7) major failure of soil mass. Tension cracks, piping and eventual failure are manifestations of differential settlements due to variations in void ratio. Finite element analysis indicates that instability and subsequent failures in the model slope were induced primarily by high hydraulic gradients in the toe area. Seepage, initial deformation and subsequent failures were manifested in the toe area prior to failure, providing a maximum of 36 min lead time. Similar lead times are expected in slopes of the same material as shown in many case studies of dam failure. The potential of having a longer lead time is high for natural slopes made of materials with higher shear strength thus evacuation is possible. The tilt sensors were able to detect the initial changes before visual changes manifested, indicating the importance of instrumental monitoring