Deciphering landslide behavior using large-scale flume experiments

Landslides can be triggered by a variety of hydrologic events and they can exhibit a wide range of movement dynamics. Effective prediction requires understanding these diverse behaviors. Precise evaluation in the field is difficult; as an alternative we performed a series of landslide initiation experiments in the large-scale, USGS debris-flow flume. We systematically investigated the effects of three different hydrologic triggering mechanisms, including groundwater exfiltration from bedrock, prolonged rainfall infiltration, and intense bursts of rain. We also examined the effects of initial soil porosity (loose or dense) relative to the soil’s critical-state porosity. Results show that all three hydrologic mechanisms can instigate landsliding, but water pathways, sensor response patterns, and times to failure differ. Initial soil porosity has a profound influence on landslide movement behavior. Experiments using loose soil show rapid soil contraction during failure, with elevated pore pressures liquefying the sediment and creating fast-moving debris flows. In contrast, dense soil dilated upon shearing, resulting in slow, gradual, and episodic motion. These results have fundamental implications for forecasting landslide behavior and developing effective warning systems.