This post continues on The Trouble With Landslides by investigating in more detail why predicting how landslides will behave is challenging.
Small landslides are fairly easy to predict: rockfalls essentially follow trajectories that can be predicted with relatively straightforward physics, and small flowing landslides can be evaluated by balancing the impact of gravity and friction on an idealized mass sliding down an idealized slope. The details can get painful when accounting for the complications of working with reality instead of idealized fictions, and working with centers of mass is of limited utility when you can still get killed by spread beyond that point, but good first approximations of landslide motion can be calculated using techniques taught in first-year university physics courses.
Larger landslides, on the other hand, are downright weird. How big is “large” is hotly debated, as is even the validity of declaring a firm lower-limit on volume, but any landslide of a volume roughly equal to a pyramid, the Hoover Dam, or 100 football fields filled with players buried nose-deep is probably going to behave strangely. (For the technically-inclined, that’s roughly 1 million cubic meters.)
Strange behaviour in landslides is defined as running out either farther or shorter than the distance predicted by using the physics of friction and gravity alone. Catastrophically large landslides are renowned for their excessive runout — traveling significantly farther than makes sense if friction and gravity are the only influences on its behaviour. Trying to understand why large landslides are different than small ones is one of the major theoretical riddles in the field. The major theories proposed (and sometimes rejected) fall into four categories:
- Reducing internal friction, decreasing energy lost within the landslide;
- Reducing basal friction, decreasing energy lost between the landslide and what it is traveling over;
- Requiring favourable geomorphology, decreasing energy lost to encountering hills; or
- Consequences of volume, with larger-volume landslides spreading over larger areas.
The reality is probably complex, with no one answer but instead some combination of processes each adding to increased run-out distance of catastrophic landslides beyond what is expected from simple prediction techniques. The result is the same: it is complicated to predict the motion of larger landslides.
Without a full physical understanding of what is causing excessive run-out, current methods of predicting how a large landslide will behave depend heavily on statistics and models. The premise is that by looking at landslides that have already happened, models can be tuned to predict the behaviour of similar future events. The simplest models are pure statistical correlations between volume and run-out distance, while more complex software models apply equations of fluid motion to the prospective event.