MODELING THE DYNAMICS OF WEATHER-TRIGGERED LANDSLIDES: WHY A “HOT START” IS WRONG

The Oso, Washington, disaster of 22 March 2014 underscored the dependence of landslide hazards not only on slope stability but also on landslide speed and runout distance. Landslide dynamics models can provide useful tools for forecasting speed and runout, but many of these models have shortcomings that limit their physical relevance and explanatory power. A common problem stems from use of initial conditions that assume a landslide begins from a state with strongly unbalanced forces—a state that differs sharply from the statically balanced state that must exist prior to the onset of landsliding. A large initial force imbalance gives a simulated landslide a “hot start,” physically analogous to launching a landslide with a slingshot. Some investigators nevertheless employ a hot start because a physically flawed model may require it in order to reproduce an observed runout pattern. An example is provided by a recent use of DANW and DAN3D to assess the dynamics of the 2014 Oso landslide (Aaron et al., 2017, J. Geotech. Geoenviron. Eng. 143, 9: 05017005). This assessment assumed that landslide motion was resisted by a basal shear strength equivalent to that provided by a Coulomb friction angle of ϕ₁=4º for the leading, liquefied mass observed at Oso and by ϕ₂=12º for a stronger, trailing mass. These values are incompatible with the mean slope (θ=20º) of the initially static landslide mass (M~1.7×10¹⁰ kg). Indeed, use of ϕ=(ϕ₁+ϕ₂)/2=8º in 1-D static balance equations (MgH sin θ vs. MgH cos θ tan ϕ, where H = 72 m is the initial center-of-mass elevation above base level) shows that the hot-start condition used by Aaron et al. (2017) implied an instantaneous release of roughly 2.5×10¹² J of potential energy—enough to levitate the entire landslide mass about 15 m into the air.

Alternatives to the model of Aaron et al. (2017) exist. For example, the D-Claw model preserves a statically balanced initial state until rising pore-water pressure triggers slope failure. It then seamlessly simulates post-failure evolution of basal pore pressure and shear strength that may—or may not—cause long runout. Application of D-Claw to the 2014 Oso event also provides a realistic landslide energy budget (Iverson & George, 2016, Geotechnique 66, 3: 175–187). No instantaneous energy release is required to explain why the landslide’s long runout occurred.