Mapping Skid Trails and Evaluating Soil Disturbance From UAV-Based LiDAR Surveys in Mediterranean Forests

Soil disturbance resulting from forest harvesting activities can have significant and lasting environmental consequences, particularly in sensitive ecosystems such as Mediterranean forests. Skid trails, the routes used by machinery to extract timber, are among the most critical areas of impact, and their detection is critical for assessing post-harvest impacts and informing future planning. Traditional ground-based methods for detecting these trails are often labor-intensive and inefficient. This study evaluates the potential of Unmanned Aerial Vehicle (UAV)-based Laser Scanning (ULS) surveys for the accurate detection of skid trails across five Mediterranean forest sites subjected to different silvicultural treatments. Four analytical techniques were tested: Hillshading (Hill), Local Relief Model (LRM), Relative Density Model (RDM), and the machine learning-based SkidRoad_Finder (SRF). Accuracy was assessed using a ground-truth dataset obtained through Global Navigation Satellite System (GNSS) field mapping. Of the tested techniques, RDM performed best overall, achieving approximately 73% accuracy, 66% sensitivity, and a Cohen's kappa value of 0.50. LRM performed best in even-aged beech forests, due to its ability to capture microtopographic changes. In contrast, SRF consistently underperformed, likely due to its reliance on training data not representative of the Mediterranean context. Our findings highlight that former skid trails and related soil disturbance can be effectively detected using ULS data. However, the accuracy and reliability of detection vary depending on site-specific factors such as forest type, vegetation structure, and terrain complexity. Overall, this study underscores the utility of ULS for operational forest monitoring and supports the tailored selection of detection techniques based on local stand characteristics and available computing resources.