Northern conditions cause challenges to the trafficability of unpaved roads. In spring, melting ice and snow softens road structures causing thaw weakening. Unpaved roads play a significant role in everyday life, forestry and agriculture. For this reason, the Finnish Transport Infrastructure Agency uses a dedicated model to predict thaw weakening of unpaved roads. The current model forecasts the severity of thaw weakening of the upcoming spring, thus helping the road professionals and users to prepare and to avoid potential structural damage. The forecasting model was originally developed by the University of Oulu in 2004, assigned then by the Finnish Transport Agency. The model has been used for over 20 years now, and significant challenges regarding its validity have been identified. This study examines the specification and application of the current forecasting model. The study applies empirical analysis to compare forecasted and actual thaw weaking of unpaved roads. Comparative analysis is supplemented by interviews of practitioners. The results show that the current model significantly overestimates the severity of thaw weakening. Also, the validity of the model parameters must be re-evaluated, and the inclusion of new parameters should be considered. Moreover, the initial forecasting model does not apply to the current climatological conditions. As a conclusion, the study presents three strategies to develop the model: i) development of a completely new model, ii) development of an improved model relying partly on existing parameters and model architecture, iii) application of machine learning techniques to the improved model. The paper concludes that options ii) and iii) are the most feasible options to upgrade and improve thaw-weakening forecasting of unpaved roads.
The increasing trend of global warming necessitates precise quantification and understanding of its potential adverse impacts on the sustainability of deep foundations and infrastructure in cold regions. However, these interactions are inherently complex, and the current database suffers from limited methodological rigor and a weak theoretical basis, with only one available correlation for adfreeze shear strength and very few studies addressing the development of frost-induced stresses, which remain theoretically unjustified. Therefore, this paper presents the design and preparation of a unique large-scale laboratory setup for investigating frozen soil-pile interactions. By adopting the direct strain gauge method (SGM) instead of the conventional indirect reaction frame method (RFM), this setup overcomes the limitations of aggregate force measurements and enables localized monitoring of frost-induced stresses along the pile shaft. Combined with a controlled freezing technique and a large-scale test geometry, the framework offers a more representative and reliable approach for evaluating both adfreeze shear strength and frost-induced stresses in the laboratory. Results show that hollow steel piles embedded in ice-poor frozen soils can lose more than 80% of their existing load-carrying capacity under permafrost warming. The measured adfreeze shear strength agrees well with the correlation relating adfreeze strength to the long-term shear strength of frozen soils. The development of frost-induced stresses under surcharge loading was investigated, and the results show that surcharge pressures can limit frost-heave stresses unless they are in direct contact with the foundation element and participate in the frost-heave process. This finding challenges the prevailing assumption that frost-heave stresses are governed solely by the rupture of an adfreeze bond. The results show that peak frost-induced stresses can exceed the long-term adfreeze strength by approximately 27%. Further field and experimental investigations are warranted to evaluate the general applicability of the observed trends.
