Classification of Road Icing Mechanisms and Analysis of Precursory Characteristics Using Fixed Road Weather Information System (RWIS) Observations

고정식 도로기상관측자료 기반 도로결빙 발생 메커니즘의 유형화 및 전조 특성 분석

Road icing is a major winter hazard that significantly increases the risk of traffic accidents. Previous studies have primarily focused on predicting road surface temperature or icing occurrence, whereas the temporal evolution and formation mechanisms of road icing have received less attention. To address this gap, this study used long-term observations from a fixed Road Weather Information System (RWIS). A total of 1,880 icing events observed at 66 of the 67 RWIS stations were analyzed. Superposed-epoch analysis was applied to characterize the 24-h evolution of surface temperature, air temperature, and relative humidity before icing onset. Based on antecedent meteorological and road surface conditions, icing events were classified into four mechanism types using physically based rules. The separability of the proposed mechanisms was evaluated using linear discriminant analysis (LDA) and a Light Gradient Boosting Machine (LightGBM) classifier. SHapley Additive exPlanations (SHAP) were applied to identify the dominant meteorological factors and their transition characteristics. The results showed that, despite distinct formation mechanisms, all icing types exhibited common precursory features characterized by progressive cooling and increasing relative humidity before icing onset. Surface temperature and relative humidity were identified as the most influential variables governing icing occurrence. In particular, surface temperatures near 2℃, combined with relative humidity around 80%, consistently marked a transition zone where the icing risk increased markedly. These findings demonstrate that road icing is a process-driven phenomenon governed by combined cooling and moisture supply, and provide a scientific basis for developing mechanism-informed prediction models and process-based early warning systems.