Date Presented:

10-13 Nov.

Abstract:

Landslides represent a persistent hazard in Mediterranean en-vironments, necessitating reliable monitoring techniques for risk assessment. Recent advancements in 3D close-range re-mote sensing, particularly through the integration of Un-manned Aerial Systems (UAS) and Terrestrial Laser Scanning (TLS), have enhanced our ability to capture surface changes. Object-based analysis offers a robust framework for seg-menting and classifying landslide features from high-resolution spatial data, improving the detection and under-standing of landslide processes and rockfall dynamics be-yond traditional pixel-based methods. This work describes the use of repeated UAS and TLS surveys at several active landslide and rockfall sites, referring to the challenges that arose and the solutions that were provided. Point clouds were processed to generate high-resolution digital elevation models (DEMs), and object-based techniques were applied for sys-tematic segmentation and classification of landslide objects. Data fusion was performed following rigorous co-registration procedures, drawing on ground control points to align da-tasets from different sensors. Spatial and geomorphological attributes were extracted from classified objects for temporal change detection and feature analysis. UAS-derived models demonstrated efficient coverage and high surface detail, while TLS data provided increased point density and vertical accuracy in targeted sectors. Object-based analysis enabled the identification and delineation of distinct landslide fea-tures, such as scarps, displaced blocks, and accumulation zones, with improved accuracy compared to manual delinea-tion. The combined UAS-TLS approach reduced data gaps and enhanced the representativeness of classified objects. The segmentation process, however, was sensitive to vegeta-tion cover and point density variations, requiring parameter optimization for each dataset. Integrating advanced data analysis with fused UAS-TLS datasets improved the identifi-cation and monitoring of morphological changes, facilitating a more process-based understanding of landslide and rockfall dynamics. The object-based framework allowed for con-sistent feature extraction across temporal sequences, support-ing detailed change analysis and the recognition of geomor-phological evolution patterns. Challenges included managing data heterogeneity and ensuring segmentation reliability across varying surface conditions. This study demonstrates that combining object-based analysis with multi-sensor 3D data acquisition enhances landslide and rockfall feature de-tection and temporal analysis, providing a scientific basis for more effective landslide hazard assessment and management in Mediterranean terrains.