Precision salt application using advanced machine learning algorithms to achieve improved road safety and reduced environmental impacts

Abstract

The application of de-icing salts to improve winter road safety, although necessary in cold climates, may adversely affect groundwater resources and degrade aquatic life in urban streams, if over-prescribed, and cause an increase in crash rates, if under-prescribed. The main objective of this research is to develop algorithms for precision salt application rate (SAR) using advanced machine learning methods to achieve the desired road safety with less adverse environmental effects. This study highlights the importance of accurate real-time monitoring of pavement surface temperature and meteorological variables (i.e., storm duration, hourly precipitation rate, and air temperature) as key factors in prescribing salt application rates during winter storm events. A new SAR model was trained/tested using a decade of historic salt application rates from a range of winter storm events on three different road classes. The application of this model can help road authorities to achieve greater road safety and reduce adverse environmental impacts, especially in the identified and mapped salt vulnerable areas.