Multiscale model for road weather information system location selection considering road weather characteristics and network topology

This study presents a novel approach to optimizing Road Weather Information System (RWIS) station locations by integrating and modeling complex topological and geometrical properties of diverse weather and traffic-related features. Existing RWIS network optimization methods primarily focus on weather-related characteristics without robust consideration of the multiscale features of regional weather and traffic-related factors. This study developed a novel algebraic topology-based Topological Machine Learning (TML) modeling approach that can capture the evolution of topological features— i.e., global properties —across multiple scales. The model represents topological properties using persistence homology graphs and a clustering algorithm utilizes these features that can be scaled for large RWIS networks. The model was demonstrated for the state of Iowa, and results show that the optimized RWIS network can provide about 18% more roadway coverage than the existing RWIS network with the same number of stations. The model was also applied to identify optimized locations for new RWIS stations to increase network level coverage. The expansion model findings showed that the coverage of the RWIS stations could be increased by 50% with 17 new RWIS stations in Iowa. Overall, the methodology presented in this study provide RWIS planners with a more robust and efficient RWIS location allocation methodthat optimize reliable road weather data collection system for cost-effective winter maintenance.