Predicting Retroreflectivity Degradation of Pavement Marking Using a Bayesian Ordered Logit Model

Abstract

Adequate pavement marking retroreflectivity R_L is crucial in enhancing road safety, and cost-effective marking restriping planning requires R_L deterioration models and service life estimation. This paper presents the Bayesian ordered logit model that incorporates Markov Chain Monte Carlo simulation to predict the degradation trend in R_L using data collected from representative highways in Wyoming. The results indicate that the marking age and characteristics, geometric locations, pavement surface, traffic volume, and snowplow operations are the significant factors affecting pavement marking degradation. The results indicate that the lane line and center line are more prone to degradation compared with the edge line. In addition, the degradation rate is higher for white markings than for yellow markings and in the curve section compared with the straight section. Cumulative traffic, truck traffic passage, and snowplow operation also significantly affect R_L degradation. The outcomes derived from the distinct models reveal substantial differences among the three functional classification models and justify investigating separate models. By integrating the random intercept into the modeling process, the analysis found a notable presence (from 22% to 31%) of unobserved heterogeneity within the same route, a crucial aspect that ensures the attainment of accurate and unbiased results. This paper also proposed a prediction equation to determine pavement marking service lives and reveals that the service life of different materials varies based on road classification, marking material, and line position. The proposed model is readily implementable and can assist pavement marking management by effectively scheduling restriping and maintenance periods for different materials.