An Efficient Bedrock Landsliding and Runout Algorithm to Explore Hillslope-Channel Coupling in 2D Stochastic Landscape Evolution Models

Bedrock landsliding is thought to be one of the dominant hillslope erosion process in active mountain belts and is fundamentally stochastic. Owing to the heavy-tailed distribution of landslide volume, rare but very large landsliding events contribute significantly to long-term hillslope erosion and are major disturbances to drainage network: they can inhibit river incision for a long time by supplying large volumes of debris (e.g., following a large earthquake or major storm) or even dam rivers resulting in long-term storage of coarse material and catastrophic release. Bedrock landsliding is also the dominant mechanisms by which divides migrate. River incision on the other end is expected to play a key role in driving hillslope instability over long timescales. The dynamics of this interplay and its role in the long-term evolution of mountain topography remain largely unknown. Landscape Evolution Models resolving correctly the interaction between bedrock landsliding and river incision are needed to progress on this topic.

Surprisingly, only one algorithm of bedrock landsliding has been proposed for Landscape Evolution Models (LEMs). It reproduces the stochastic nature of landsliding, but has limited success in capturing the pdf of landslide area and has only a rudimentary runout algorithm. Here, we introduce a new reduced complexity bedrock landsliding algorithm (SLIDOS) that has been implemented in the stochastic 2D LEM €ROS. It is based on the probabilistic identification of unstable nodes according to local instability drivers (topographic slope, runoff, peak ground acceleration) and rock mass characteristics (cohesion, friction angle). From each of these nodes, a rupture plane is propagated upslope recursively. The landslide material above the rupture plane is then routed downslope on the topography by a particle method governed by a single tunable parameter.

We calibrate the 3 parameters of this algorithm and show that it can quantitatively reproduce (i) the pdf of source area, (ii) the volume-area scaling and (iii) the apparent decrease in runout friction coefficient with landslide volume observed in nature. We then illustrate the key differences between full-stochastic (floods+landslides) and non-stochastic simulations of landscape response to tectonic perturbations.