2009 crash data from the State of Michigan was combined with weather data from four Clarus weather stations in the Upper Peninsula of Michigan. Crashes were monitored within a 50 mile radius and associated with weather conditions at the Clarus station. From this data, a series of regression models were then created based on critical tipping points of weather data, as well as continuous weather observations. This provides an algorithm consisting of seven risk equations which are used under differing weather conditions. The crash risk algorithm was then combined with a time based algorithm in order to recommend a route. The relative weighting of crash risk and time is established by the user. Using an open source geospatial routing tool and open source road network software, a recommended route is defined. (Author/publisher)
Abstract