We developed a simplified laboratory landslide setup using a flat inclined plane, a housing box for grains from which they can be released, and a flat section where they eventually come to rest. Using optical techniques developed in our laboratory, we were also able to measure in real time quantities such as the landslide thickness and speed, which are especially relevant for theories of granular rheology. We also began to develop a method for measuring the pressure underneath the landslide, a key feature of the theory for long landslide runout we were originally motivated to test. We accompanied our experiments with large-scale numerical simulations of granular landslides using the open-source DEM code LIGGGHTS.
We started by doing studies of landslide runout for a single fall height and grain size as well but changing systematically the landslide total volume. For a single grain size and fall height we observed the classical result that the runout distance increases with landslide volume. When we either changed the fall height or changed the grain size we found the same classical behavior, but when plotted in the traditional way (fall height divided by runout distance versus volume) we ended up with many distinct curves. This lack of agreement between the data caused by different fall height and diameter meant that the traditional plot, which is an attempt to account for the influence of the fall height and the volume on the runout distance, does not work. We thus performed a standard scaling analysis often used in the fields of applied mathematics, fluid mechanics, and physics, and revealed the true dependence of the runout distance on fall height, landslide size (which became the number of grains), and the average landslide grain size. Plotting in our new way, the normalized landslide runout distance for a wide range of parameter values collapse onto a common, universal curve.
In fact, we observed that for small landslide volumes the scaling analysis breaks down, which is not uncommon. A given scaling does not always work for all ranges of a parameter. We were able to determine when this breaks down as a function of landslide size, fall height, grain diameter, and the viscosity of the surrounding air, by assuming that the breakdown is caused by a transition from a granular gas to granular liquid behavior. This observation of a granular phase transition is novel in its own right, but the scaling result that we found is also of practical value in designing laboratory experiments to study natural landslides. The scaling of the landslide runout as well as its breakdown are in a publication under review.
Finally, using the new scaling that we found as a foundation, we began to add new parameters into our experiments that are also of relevance to natural landslides. Our original experiments were performed on a flat, smooth substrate, but natural landslides have a variety of complicated terrain. To account for this systematically we performed a large number of experiments for which we varied the roughness of the substrate. We did this by attaching standard sandpaper, which we characterized using a microscope, as well as our own laboratory-made rough substrates. Again using the same standard scaling analysis revealed a new dependence on the ratio of grain size to roughness. This result is not only of interest to landslide runout prediction, but it also continuously bridges the gap between two standard laboratory approaches: the smooth and rough substrates. This work is being written up into a manuscript for publication.
