Towards an automated acquisition and parametrization of debris-flow prone torrent channel properties based on photogrammetric-derived uncrewed aerial vehicle data

Debris flows are a hazard in mountainous regions. Cost-effective, long-term studies of debris-flow torrents, however, are rare, leading to uncertainties in hazard assessment and hazard prevention. Here, we address the question of whether cost-effective remote sensing techniques can be applied for assessment of mountain torrents and possibly further gather accurate, long-term information on the evolution of the catchment. Torrents prone to debris flows are often devoid of vegetation in the near channel area and hence can be well captured with photogrammetrically derived methods using uncrewed aerial vehicle (UAV) surveys. The possibility of automatically extracting specific torrent parameters from high-resolution terrain models, such as cross-section area or gradient, is investigated. The presented methodology yields continuous and automatically derived geometrical parameters such as torrent bed width, inclination and cross-section area, which is a major advantage compared with point-based, often dangerous field surveys. Their cross-validation with field measurements shows strong agreement. Those parameters are accurate along sharply incised sections with strong limitations along sections with steep adjacent slopes and/or dense vegetation. The information along the torrent allows fast identification of key sections and weak spots which can be precisely evaluated in the field. The study highlights that proper classification of real ground points poses the key challenge. We show that photogrammetric routines to derive a high-resolution digital terrain model (DTM) are limited in the case of dense vegetation coverage. In such cases, LiDAR surveys have clear advantages even though they are also limited by very dense vegetation. We find that UAV data can be used for an objective method of estimating debris-flow torrent geometric properties. And the introduced approaches therefore build a stepping stone towards a more comprehensive, reproducible and objective assessment of torrent processes and predispositions. However, ground-referencing fieldwork remains essential, and further research on remote sensing supported hazard assessment of debris-flow-prone torrents is indispensable.