The search for path-dependency mechanisms using physically-based soil-landscape modelling of landslides

The spatial pattern of landslide susceptibility is a key input for decision making by many natural hazard agencies. Therefore, the estimation of landslide susceptibility maps has received much attention in the last decades. Increasingly, such maps are produced by statistical methods that relate the locations of observed landslides to geofactors such as slope steepness or vegetation density. Almost without exception, these susceptibility assessments are entirely spatial. At the same time, recent studies of large multitemporal landslide datasets have shown empirically that landslide susceptibility changes over time as well as space, as a result of the impact of recent nearby landslides. In at least two study sites, places near previous landslides are temporarily more susceptible to landsliding, sometimes substantially so. Several candidate mechanisms underlie this form of complexity (called path-dependency) in the landslide system, and targeted field measurements in landslide-prone study sites should be recorded to fully understand which mechanism is most important.

Awaiting such measurements, physically-based mechanistic modelling of landslide impacts in the soil-landscape system can help explore the possible mechanisms. Here, we report on our development of landslide simulation capabilities in soil-landscape evolution model LORICA. In this model, landslides affect not only surface elevation, but also local soil and vegetation properties. Since other processes in the model also affect these properties, the impact of landslides is not permanent. Applied to a hypothetical soil-landscape, this model allows us to explore whether a) local topographic effects such as oversteepening, b) temporarily changed soil hydraulic parameters, or c) disruption of vegetation and roots, are the most likely mechanisms behind landslide path-dependency.