A Look at Landsliding Statistics from a Physical Experiment
Olivia Beaulieu, Masters Candidate in Earth Sciences
Andrew Wickert, Department of Earth Sciences and St. Anthony Falls Laboratory, University of Minnesota
Landslides are a significant natural hazard, shape steep hillslopes, change rivers and their habitats, and impact soil-zone weathering and the global carbon cycle. Field inventories are used to produce snapshots of landslide distributions, but the temporal windows of these studies are either short (single-event) or poorly-defined (cumulative visible evidence of landslides), and in the latter case may not represent the initial distribution (due to erosion or revegetation). Furthermore, these data sets typically constrain landslide area but not landslide volume. Here we respond to the need to understand the long-term three-dimensional statistics of landsliding by building an experimental sandbox in which an incising and laterally-migrating river produces landslides by undercutting banks of moist sand. The forms of both the landslide area–frequency distribution and the landslide area–volume relationship are identical to those from analyses of field data. The modal landslide area in the experiment is controlled by cohesive strength due to capillary forces between the moist sand grains, and this area–cohesion relationship scales linearly to field conditions, where soil and root strength control cohesion. Together, these data demonstrate that it is possible to produce landslides at scale in a physical model with a realistic driver, and that this approach can elucidate first-order controls on the statistics and mechanics of landsliding.