Towards a model chain for an automated thalweg identification

To quantitatively assess properties of gravitational mass flows (GMFs), such as snow avalanches, we investigate the potential of an automated thalweg identification. In this context, a thalweg is defined as the main flow direction of a flow path in three-dimensional terrain. Along the main flow direction, two-dimensional representations, such as the relation of altitude to distance along the thalweg, provide a simplified representation of the three-dimensional GMF properties. Additionally, single quantitative characteristics describing the flow path can be identified from the thalweg. These thalweg analyses are used, for example, to compare different avalanche events or simulation outcomes and therefore to statistically analyze the underlying avalanche terrain. So far, automated thalweg identification has mainly been limited to individual flow paths on local scale.

The goal of the presented approach is to identify the two-dimensional thalweg representations of the GMFs from three-dimensional terrain on a regional scale automatically. For this, we extend an open-source model chain that delineates potential release areas (PRAs) based on three-dimensional terrain and computes their potential mobility and runout. In a further analysis, we can derive scalar characteristics along each thalweg, such as runout quantities. For example, these values can be used to analyze and validate the simulation of snow avalanches by comparing them to existing size classification approaches. 

The challenge is to identify the thalweg for every flow path automatically on a regional scale. For this purpose, PRAs are delineated from terrain-derived characteristics such as slope, surface roughness, wind exposure, and forest cover. Subsequently, PRAs are constrained by aspect criteria and subdivided into hydrologically meaningful subcatchments. For each PRA, we compute the runout using com4FlowPy, which is a computational module within the open-source avalanche simulation framework AvaFrame. It has been used to simulate several GMF processes (e.g., snow or rock avalanches) on a regional scale. With a well known, conceptual approach, it simulates the runout and intensity of GMFs. Within the runout simulation, the thalweg's location is computed as main flow direction for each release area by averaging the respective flow quantities. This model chain enables automatic thalweg delineation, requiring a digital elevation model and a suitable parametrization as model input. 

These thalwegs allow us to derive quantitative characteristics, such as runout length, corresponding angle and maximum flow velocity, for every flow path on a regional scale. These characteristics can be used for statistical analyses of the avalanche terrain.  When simulating snow avalanches of potential size, the applied parameterization can be validated by comparing these simulated thalweg characteristics with established avalanche size classification approaches. In addition, the analysis of simulated thalwegs enables comparisons of terrain characteristics between different study regions in terms of, for example, runout length, destructiveness, and the potential avalanche size. Identical input parameters are applied across the regions, ensuring that differences arise from terrain properties.