A comprehensive optimization methodology for strategic environmental sensor station locations

Adverse weather poses a significant threat to transportation safety. Road weather information systems (RWIS) aim to mitigate the impact of adverse weather by detecting spatiotemporal variations of weather and/or road pavement conditions in real time. Due to the lack of a detailed, unified guideline and diverse weather conditions across the United States, state and city transportation agencies follow different practices for choosing locations for environmental sensor stations (ESS) (the components that collect RWIS weather data). To fill this gap, this study proposes a comprehensive cell-based methodology that is data-driven, using crash records, weather data, and road network information. The contribution of the proposed methodology is that the model optimizes overall benefits derived from RWIS based on weather-sensitive crashes. Both normal and adverse weather crash data are used to derive cell-vulnerability rates in adverse weather. First, a sequential procedure is devised to identify the required number of stations for the region. Then, optimal weather station locations are identified using a genetic algorithm. The proposed approach is especially suited for optimizing region-wide ESS locations involving complex road networks or a large number of road segments. A case study was conducted using data from the Crash Records Information System (CRIS) between 2010 and 2013 in the Austin District, an area especially vulnerable to rain. It was found in the case study that ten ESSs would be a good choice to implement in the region. Their proposed global optimal locations layout would cover 94% of total crashes occurring in the region based on 20 miles of coverage for each station. The RWIS would have spatial coverage of 48% and 92% reliability should one ESS fail.