Simplicity vs. complexity – a long-standing challenge in the simulation of landslide runout

First approaches to analyze the propagation of hazardous, extremely rapid, landslides go back at least to the 1930s, when data-driven approaches such as angles of reach were used to estimate runout distances. Various much more sophisticated tools for the simulation of flow-type landsides including debris flows or rock avalanches, or cascading effects involving more than one type of phenomenon, have been developed since then, particularly throughout the last two decades. They build on increasingly complex physical-mathematical models, starting from comparatively simple and straightforward, depth-averaged Voellmy-type mixture models, moving to still depth-averaged two- and three-phase models able to simulate the interaction of landslides with water bodies, and currently proceeding to highly complex and highly flexible full 3D models. Phenomena such as erosion, entrainment, deposition, phase separation, or non-hydrostatic effects are increasingly well understood and incorporated into operational mass flow simulation tools.

However, this trend of increasing model complexity, supported by increasing physical process understanding and enhanced computational capacities, is not necessarily in line with the demand by natural hazard practitioners, who need capable but easy-to-handle simulation tools for their work. Besides the computational costs, it is mainly the multitude of often unknown, rather conceptual model parameters representing a barrier for practitioners towards using the more or even the most advanced approaches. Further, more complex models do not necessarily provide more accurate results than simpler ones, this always depends on the scope, purpose, and phenomenon. If only estimates of runout distances or impact areas are needed, very simple data-driven models may do the work. Even when it comes to flow thicknesses, velocities, or impact pressures, ordinary debris flows may still be better represented by comparatively simple mixture models than by parameter-hungry two- or three-phase models. Yet, more complex models are needed for more complex processes such as landslide-lake interactions or other types of process chains with dynamically changing material composition. Therefore, intelligent approaches have to be found to find an appropriate balance between simplicity and complexity.

Mainly based on seven years of experience with the depth-averaged multi-phase mass flow simulation framework r.avaflow, we will discuss the main challenges and ideas to find a useful balance between simplicity and complexity in the simulation of landslide runout.