Spatial Analysis of Road Weather Safety Data Using a Bayesian Hierarchical Modeling Approach

The safety of highway travelers can be majorly impacted by adverse weather. Weather events and how they impact highways can be seen as incidents that display strong geographic patterns and are predictable and nonrecurring. Wisconsin counties with high relative crash risk under snow, rain, fog, and other inclement weather conditions are addressed by this study. The authors propose a Poisson model with a log link function including a spatial random effect within a Bayesian hierarchal modeling framework. Two spatial model types are considered in particular. A conditional autoregressive model specifying dependence via autoregression among neighboring counties is one. An exponential model assuming an exponential spatial dependence decline with an increase in the distance between two counties constitutes the other. There is also consideration of a spatially independent model as a baseline model. Fairly consistent crash patterns are shown with weather impact through Bayesian statistical inference. In the northern Wisconsin counties with more snowfall through the long winter, there were higher than expected snow-related crashes. The areas close to Lake Michigan form the rain-related crash clusters, since they had more rainfall than other parts of the state. In the mountainous valley terrain of the counties in the southwestern region, more foggy days were partially responsible for an overrepresentation of fog-related crashes. The modeling approach discussed by the authors can be recommended for ranking counties for road weather safety planning and programming.


  • English

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 01142774
  • Record Type: Publication
  • Files: TRIS, ATRI
  • Created Date: Oct 18 2009 2:47PM