High-frequency UAV LiDAR survey for monitoring active earthflows at the slope scale by using Digital Image Correlation, Homologous Point Tracking and DEM of Differences (Baldiola landslide, Northern Italy)

The Baldiola landslide (Panaro river valley, Northern Italy) is an active earthflow that has experienced continuous toe advancement and source area widening/retrogression for over 40 years, as evidenced by archive aerial and satellite imagery. This long-term evolution now poses a potential hazard to a residential area near the crown and for slope-river interaction, emphasizing the need for innovative monitoring strategies to assess displacement dynamics and support risk management. For such an objective, during 2024, we have implemented high-frequency (i.e. from weekly to bi-weekly) UAV-based LiDAR & Photogrammetric surveys, in order to obtain a detailed characterization of landslide kinematics.

More specifically, the UAV-derived datasets collected throughout 2024, i.e. Digital Elevation Models (DEMs) and high-resolution Orthomosaics, were processed in order to obtain spatially distributed slope displacement values across the entire landslide by using Digital Image Correlation (DIC) & Homologous Point Tracking (HPT) for horizontal displacement and DEM of Difference (DoD) for vertical variations. Results have been validated in specific key points by using time series from continuous Robotic Total Station (RTS) monitoring.

Results of archive aerial and satellite imagery analysis showed more than 120 meters retrogression of the main scarp since 1978, with 30 to 50 meters occurring between 2006 and 2024). Results of DIC and HPT evidence differential movement patterns across the landslide body, with higher displacement rates from 5 to 10 m/month recorded along the main channel, particularly in the middle-lower channel and toe area, and extensive retrogression recorded in part of the source area (14 meters between April and November 2024). The comparison between RTS and HPT-derived displacements showed a strong correlation (R² > 0.99 in most cases), confirming the reliability of UAV-based tracking methods. Additionally, DIC analysis successfully captured displacement trends comparable to RTS and HPT, demonstrating the potential of automated image processing for large-scale motion detection. The DoD analysis was essential for tracking and monitoring local reactivations, particularly in the source area, where a depletion of several meters was observed. Furthermore, and altogether, the results unravel mass transfer processes at the slope scale mand the spatial and temporal pattern of progressive acceleration of the landslides from the source area, down into the channel and finally to the toe zone, as well as the peculiar pattern of progressive deceleration of the phenomenon.

This integrated approach allowed a detailed assessment of the landslide’s kinematics, providing valuable insights into its spatial variability and temporal evolution and, ultimately, the processes governing earthflows. The high-frequency UAV dataset proved particularly useful in detecting small-size accelerations and minor reactivations that were not always evident in RTS data alone. Future research will focus on examining the relationship between rainfall events and acceleration phases, aiming to improve the understanding of triggering mechanisms and short-term response dynamics.