Development and Application of a Winter Weather Traffic Imputation Model: A Comparative Study Against Machine Learning Techniques During the Winter Season

This study examines how winter weather conditions influence traffic patterns for both passenger vehicles and trucks, using data collected from weigh-in-motion (WIM) stations and nearby weather monitoring sites along Alberta’s Highways 2 and 2A. To explore how snowfall and temperature affect traffic volumes, we developed Ordinary Least Squares Regression (OLSR) models. The findings indicate that passenger car volumes drop more sharply than truck volumes under increased snowfall, with the decline being particularly notable on Highway 2, a rural stretch. In contrast, Highway 2A showed an uptick in truck traffic, likely due to detours from adjacent routes with less winter maintenance. For estimating missing traffic data during severe weather, we employed both OLSR and a machine learning technique, k-Nearest Neighbor (k-NN). In comparing the two approaches, OLSR demonstrated superior accuracy and consistency, making it more effective for filling in missing traffic data throughout the winter season. The performance of the OLSR model underscores its reliability in addressing data gaps during adverse winter conditions. Additionally, this study contributes to sustainable transportation by improving data accuracy, which aids in better resource allocation and enhances road safety during adverse weather. The findings support more efficient traffic management and maintenance strategies, including optimizing winter road maintenance and improving sustainable infrastructure planning, thereby aligning with the goals of sustainable infrastructure development.