The Role of Three-Dimensional Boundary Stresses in Limiting the Occurrence and Size of Experimental Landslides

The occurrence of seepage-induced shallow landslides on hillslopes and steep channel beds is important for landscape evolution and natural hazards. Infinite-slope stability models have been applied for seven decades, but sediment beds generally require higher water saturation levels than predicted for failure, and controlled experiments are needed to test models. We initiated 90 landslides in a 5 m long laboratory flume with a range in sediment sizes (D = 0.7, 2, 5, and 15mm) and hillslope angles ( = 20 degrees to 43 degrees), resulting in subsurface flow that spanned the Darcian and turbulent regimes, and failures that occurred with subsaturated and supersaturated sediment beds. Near complete saturation was required for failure in most experiments, with water levels far greater than predicted by infinite-slope stability models. Although 3-D force balance models predict that larger landslides are less stable, observed downslope landslide lengths were typically only several decimeters, not the entire flume length. Boundary stresses associated with short landslides can explain the increased water levels required for failure, and we suggest that short failures are tied to heterogeneities in granular properties. Boundary stresses also limited landslide thicknesses, and landslides progressively thinned on lower gradient hillslopes until they were one grain diameter thick, corresponding to a change from near-saturated to supersaturated sediment beds. Thus, landslides are expected to be thick on steep hillslopes with large frictional stresses acting on the boundaries, whereas landslides should be thin on low-gradient hillslopes or in channel beds with a critical saturation level that is determined by sediment size. Plain Language Summary Shallow landslides on hillslopes and steep riverbeds commonly occur when rainwater saturates soil enough to outweigh its frictional stability. Landslides play a key role in shaping mountains and pose natural hazards to those living in steep terrain. To predict the occurrence of landslides we often rely on models that have not been verified experimentally. To test these models, we initiated 90 landslides on a laboratory hillslope under a wide range of soil conditions and inclination angles. We found that landslides on very steep hillslopes were nearly as stable as on low gradient hills but were much larger in size. At small hillslope angles, landslides could be as thin as a single sediment grain, and the amount of water required to initiate these slides was determined by sediment size. Landslide models that consider the frictional force acting only at the base of the landslides underpredict stability in all experiments. Frictional forces acting at the boundaries of landslides are often ignored, but including these forces can explain the increased stability of our experimental hillslopes. However, these models are unable to predict the size of the experimental landslides, which were, on average, 10-fold smaller than predicted.