Enhancing Rockfall Detection Using Permanent LiDAR Scanner (PLS) Data and Automated Workflows at St. Eynard Cliff (Grenoble, France)

The ANR C2R-IA (anrc2ria.fr) project aims to develop reliable decision-making tools for dynamic rockfall risk management, such as restricting access to hazardous zones during critical periods. To achieve this, we aim to develop a predictive model for observed rockfall events that relates them to weather conditions history using Artificial Intelligence tools. Training an artificial neural network requires a comprehensively labelled dataset of rockfall events. To build this dataset, we deployed various instruments, including a Permanent LiDAR Scanner (PLS), whose data is processed by an automated workflow to handle the large volume of hourly-acquired point clouds.

The workflow started with a pre-processing step that includes point cloud alignment (registration), quality control, cropping the area studied, and vegetation removal. During the processing phase, changes are identified using a multi-step approach:

• First, pairs of point clouds are aligned either globally or by spatial strips (Chanut et al. EGU 2025, poster session).
• Then, M3C2 distances (Lague et al, 2013) are calculated. For a pair of point clouds (N1, N2), the distance computation is made twice from N1 to N2 and from N2 to N1 to identify significant changes.
• Dense clusters of significant changes are extracted using DBSCAN clustering, and a spatial association between clusters from the two clouds is performed to track corresponding zones and ensure accurate changes in output.
• To refine block characterization, a local registration and comparison is further performed, followed by alphashape surface reconstruction for volume estimation.

The workflow was developed in Python, primarily using the CloudComPy library, and requires minimal operator intervention thanks to integrated quality metrics at each processing step.

This optimized workflow combined with a fixed point of acquisition (a reinforced concrete pillar) has significantly improved the detection threshold at the St. Eynard site (Grenoble, France), allowing for identifying rockfalls as shallow as 10 cm in depth and 0.01 m³ volume — an improvement from the previous 40 cm and 0.1 m³ (Verdier-Legoupil, 2023; Le Roy, 2020). Catalog completeness has also been improved, with the number of detected events increased thresholds from less than 50 events/month/km² to about 150 events/month/km². However, numerous false positives are generated, primarily due to persistent vegetation artifacts despite the vegetation removal step. To address this issue, future work will focus on integrating an automatic change validation method using criteria such as morphology, scalar field information, and additional point cloud comparisons to check the temporal persistence of changes.

Chanut, M.-A., Manceau, L., Levy, C., Dewez, T., Amitrano, D., 2025. Rockfall detection using lidar point clouds: identification of geometric distortions during acquisition and proposed processing to enable a low detection threshold. EGU 2025, Poster session.

Lague, D., Brodu, N., Leroux, J., 2013. Accurate 3D comparison of complex topography with terrestrial laser scanner: Application to the Rangitikei canyon (N-Z). ISPRS Journal of Photogrammetry and Remote Sensing 82, 10–26. URL https://doi.org/10.1016/j.isprsjprs.2013.04.009

Le Roy, G., 2020. Rockfalls multi-methods detection and characterization. Université Grenoble Alpes.

Verdier-Legoupil, M., 2023. Etude des chutes de blocs par la photogrammétrie, cas du St Eynard. Université Grenoble Alpes.