Forest Road Trafficability Prediction

This study explores methods for predicting forest road trafficability in Eastern Finland by combining falling weight deflectometer (FWD) bearing capacity measurements with remotely sensed data including terrain, hydrological, climatic, and structural road attributes. Using eight statistical models across two methodological frameworks: mixed-effects models and intra-road prediction approaches. The research evaluated various predictor sets such as depth-to-water metrics, terrain wetness indices, canopy structure, and usable road width. Results showed that road-level effects account for over 90% of variability in bearing capacity, with usable road width being the most significant predictor and depth-to-water at 0.5 ha consistently useful across models. However, while mixed-effects models performed well on training data, predictive performance for new roads was limited, with marginal R² values ranging from 0-27%. The findings highlight both the importance of road-level field data and the current limitations of remote sensing data for predicting forest road trafficability.