Discover Spatiotemporal Association and Causation between Potholes and Flat Tire Frequency

Abstract

This study attempts to untangle the association and causation between pavement potholes and vehicle flat tire frequency. To this end, the study collected pothole data from crowdsourced user-generated reports and extracted flat tire incidents from official operation notes. To account for spatiotemporal variations in potholes and flat tire incidents, a geographically and temporally weighted negative binomial regression (GTWNBR) model was created to explore their spatiotemporal association, alongside a negative binomial fixed-effects regression model (NBFER) for causal inference of flat tire incidents. The GTWNBR model revealed that potholes are positively correlated with flat tire frequency, with a significant spatiotemporal variation. The potholes in the central city and interchanges of highways were found to have a higher association with flat tire frequency than in other areas. Besides, the flat tire frequency was most strongly correlated with potholes that outbreak during the winter months. Further, the NBFER model substantiated that increments in potholes can significantly cause more flat tire incidents. The NBFER and GTWNBR models differ in their strengths: the NBFER model is better suited for making causal inferences, whereas the GTWNBR model can provide a more comprehensive understanding of covariates and achieve higher predictive performance. The study findings recognize tangible connections of how pavement conditions trigger flat tire incidents, and they can pave the way toward prioritizing infrastructural investment decisions related to roadway maintenance.