Modelling machine-induced soil deformation in forest soils using stump proximity and machine learning

Soil deformation is a key challenge in sustainable timber harvesting, particularly in environments with low bearing capacity. In mechanised forestry, this issue is especially pronounced in peatlands, where rutting arises from soil displacement and root shearing within the soft, organic substrate. While tree roots are known to reinforce soil, the specific role of stump-root systems in mitigating rut formation remains underexplored. This study examines the influence of stump presence on rut depth using Unmanned Aerial Vehicle (UAV) based digital terrain models (DTMs), manual field measurements, spatial modelling, and machine learning techniques. UAV-derived rut depth estimates were first compared with manual data, revealing slightly lower values in deeper ruts, particularly in curved trails, with mean discrepancies of 3 cm. Statistical analysis confirmed that cumulative stump influence significantly reduced rut depth, with a small to medium effect in straight trails (${\varepsilon }^2$ = 0.04–0.20) and a moderate to large effect in curved trails (${\varepsilon }^2$ = 0.02–0.32). Machine learning models achieved high predictive accuracy ($R^2$ = 0.69–0.85), identifying stump-related variables and soil shear modulus as key predictors of rut formation. These findings emphasise the importance of incorporating stump-root reinforcement into forest planning to optimise machine path selection and minimise soil disturbance. Future research should refine species-specific reinforcement models and explore advanced root mapping techniques, such as ground-penetrating radar (GPR), to strengthen decision-support tools for sustainable forestry.

Science4Impact statement (S4IS)

This study presents a spatially informed methodology to evaluate the influence of tree stump-root systems on rut formation in peatland soils. By integrating UAV mapping and machine learning, this study enables the predictive identification of low-impact areas, reducing site disturbance and supporting climate-smart forestry. These findings offer a practical starting point and a potential tool for optimising skid trail layout, improving operational efficiency, and minimising soil disturbance and site damage. The approach supports evidence-based decision-making in peatland conservation, helping align forest operations with broader environmental and climate goals.