Using spatial interpolation to determine impacts of annual snowfall on traffic crashes for limited access freeway segments.

Snowfall affects traffic safety by impacting vehicle performance, driver behavior, and the transportation infrastructure. Depending on intensity snowfall can reduce visibility, pavement friction performance, vehicle stability and maneuverability. Based on this premise, the objective of this study was to use spatial interpolation to analyze the effects of annual snowfall on crash occurrence at non-interchange freeway segment locations in Michigan. Using the geostatistical method of Ordinary Kriging, site specific historical snowfall values were estimated based on data obtained from a series of weather stations during the primary winter months (December, January, and February) for the years of 2004 through 2014 along Michigan's entire limited access freeway network. These weather data were spatially matched with historical crash data and roadway inventory data for each freeway segment. A Negative Binomial regression model was developed to quantify the effects of snowfall on crashes. Explanatory variables included annual average daily traffic, segment length, horizontal curvature, and snowfall. The results indicated that annual snowfall has a statistically significant positive effect on winter crashes for all types of crashes analyzed. With respect to vehicle type, crashes involving a truck or bus experienced the strongest relationship with annual snowfall. Considering crash severity, property damage crashes possessed a stronger relationship with snowfall as opposed to injury crashes.