FROM SLOW TO FAST: ABRUPT DEBRIS-FLOW MOBILIZATION IN A HILLSLOPE EXPERIMENT

Rainfall-triggered shallow landslides that transform into rapid debris flows are a ubiquitous hazard in steep mountains worldwide. Not all shallow landslides, however, mobilize into deadly debris flows. Mobilization on natural hillslopes is poorly understood, yet early warning can depend on forecasting this hazardous behavior. The APERIF project (Ochiai et al., Landslides, 2004) performed an artificial rainfall (sprinkling) experiment to trigger a landslide on a planar 33º natural hillslope with decomposed granitic soil near Mt. Kaba-san, Japan. After ~6.8 hours of sprinkling at 78 mm/hr, a shallow landslide (~1 m thick and ~40 m³) mobilized into a debris flow and traveled down an adjacent channel. With 100-Hz sampling, we recorded subsurface pore pressures and ground surface motion preceding failure and through downslope debris-flow mobilization. Small surface motion commenced ~1 hour before rapid failure, in response to steadily rising positive pore pressures. Slow motion (< 0.001 m/s) continued until abrupt rapid failure transpired over a few seconds, when pore pressures rapidly dropped, then increased and brisk acceleration ensued. Over several seconds, pore pressures oscillated greatly as the mass fluidized. Velocities increased to 4-8 m/s in less than 0.5 s.

Abrupt change from slow to fast motion can present a mechanical conundrum. Loose soils can collapse quickly, yet many hillslope soils are dense, which typically dilate during shear and thus impede motion. We use a simple fully coupled pore-pressure/sliding model (Iverson, JGR, 2005), combined with measured and estimated parameters, to investigate the co-evolving motion and pore pressures observed during our experiment. Our simulations show that given either a contractive soil or a soil with no dilative effects, the slide would accelerate quickly. In contrast, dilation can greatly impede motion, as observed initially in our experiment. However, even dilative systems can experience swift acceleration if dilative effects are exhausted through sustained motion. Here, acceleration occurs during a rapid drop, then rise, in pore pressures, as was also observed. Our observations suggest that monitoring might detect precursory motion, but forecasting fluidization is dependent on soil properties and transient hydrologic conditions.