Integrating soil, weather, and lightweight deflectometer data for dynamic bearing-capacity estimates on forest roads

Abstract

Trafficable forest roads are vital for providing timber for the forest industry. Dynamic trafficability models for forest roads are needed, because many forest roads are trafficable for only some part of the year. We studied 10 road segments in eastern Finland during non-frozen periods, collecting E-modulus measurements using a lightweight deflectometer from 2023 to 2025. Approximately 15.1% of variation in bearing capacity was between road segments, 8.4% within road segments, and 76.5% was dynamic, indicating that the theoretical maximal performance of static models would be limited to explaining 23.5% of the variation. Prediction models were built utilizing weather station data and openly available soil type information on subsoils. We found that moving averages of ambient temperature and rainfall can capture dynamic variation in bearing capacity. Re-graveling and surface freezing increase the bearing capacity precipitously. Systematic differences between road segments can be captured by calibrating the models. We found robust evidence that two calibration measurements carried out at separate times significantly improve predictions, while additional measurements provide little to no value.