Deriving activity zones of the Äußeres Hochebenkar rock glacier through boulder tracking in multitemporal UAV-LiDAR point clouds

Monitoring rock glacier movement is crucial for understanding the response of mountainous permafrost to changing climatic conditions. This importance is acknowledged with the inclusion of Rock Glacier Velocity (RGV) as an Essential Climate Variable (ECV) product for permafrost. The RGV product focuses on tracking surface movement on an annual scale for rock glaciers whose movement is dominated by permafrost creep. While existing remote sensing methods provide valuable insights into RGV dynamics [1], they often fall short in capturing detailed 3D displacement patterns and distinguishing spatial variations in movement activity. This limitation becomes critical when studying destabilized rock glaciers [2], where overlapping processes, such as sliding on shear horizons, drive acceleration patterns and surface deformation. These dynamics result in complex movement behaviors that require more advanced monitoring techniques to fully understand. We introduce a novel neighborhood-based boulder-tracking approach that addresses these challenges by treating boulder faces as distinct objects on the surface of the rock glacier and tracking them over time within annual UAV-borne Laser Scanning (ULS) point clouds. Our method enables the derivation of 3D displacement vectors along actual movement paths, providing area-wide surface change information that allows for the differentiation of zones with similar movement activity.

We apply the method to the highly monitored Äußeres Hochebenkar rock glacier [3], utilizing high-resolution ULS datasets that cover the destabilized front section of the rock glacier in 2019, 2020, and 2021. A region growing segmentation is conducted to segment boulder faces, using the local normal vector as a growing criteria. The segmentation process achieves an F1 score of 0.72 in sample areas (Recall: 0.76, Precision: 0.69), and we identify approximately one segment per 3 m². For boulder tracking, we use a neighborhood-based matching approach, adapted from landslide monitoring. Our method identifies correspondences over time by focusing on the spatial relationship between neighboring segments and comparing them across epochs. Using this approach, we successfully track approximately one boulder per 47 m² between observation periods. K-means clustering is then applied to the 3D displacement vectors to identify distinct movement zones. This approach is used to assess variations in displacement magnitudes and directions across the destabilized section of the rock glacier. The analysis reveals differences in general movement patterns, particularly between the central flow line and the adjacent margin zones.

We demonstrate the strong potential of tracking boulder faces across multitemporal point clouds using spatial neighborhood information. This approach provides a robust solution for monitoring complex surface changes of rock glaciers and enables the differentiation of activity zones based on shared movement patterns. In the future, the demonstrated ability to track and quantify the movement of large numbers of individual blocks could contribute to assessments of flow coherence and its temporal changes in the context of RGV monitoring. This could help detect shifts in movement regimes, e.g. the transition of rock glaciers from a stable permafrost creep regime to destabilization.

REFERENCES

[1] Zahs et al. (2022): DOI: https://doi.org/10.1016/j.isprsjprs.2021.11.018  

[2] Marcer et al. (2021): DOI: https://doi.org/10.1038/s43247-021-00150-6

[3] Hartl et al. (2023): DOI: https://doi.org/10.5194/esurf-11-117-2023