Piecewise Multiple Linear Models for Pavement Marking Retroreflectivity Prediction Under Effect of Winter Weather Events

Abstract

Most existing models for prediction of pavement marking retroreflectivity have been developed from data collected in locations with similar weather conditions; therefore, the effect of different weather conditions, such as winter weather events, has not been extensively studied. This study develops degradation models that can predict retroreflectivity of durable pavement marking materials under various winter weather conditions. Piecewise multiple linear models are proposed to explicitly account for the effect of winter weather events in this study. In comparisons of the proposed models with conventional multiple linear models (MLMs) developed from the same set of data, the proposed models outperformed the MLMs, and the overall root-mean-square error improved from 204.6 mcd/m2/lux for MLMs to 106.5 mcd/m2/lux for piecewise MLMs. The proposed models also show robust and consistent results in predicting retroreflectivity of different materials in different states. The results indicate that the proposed method can be adopted by various states and regions for comprehensive retroreflectivity prediction under various weather conditions.