Combining Land Use, Traffic and Demographic Data for Modelling Road Safety Performance in Urban Areas

Road accidents form a leading cause of death globally. Despite the recent progress that have been made, Greece continues to be among the worst performing countries in the EU, in respect to road safety. This research deals with the spatial analysis and modelling of road accidents, in the metropolitan area of Thessaloniki, Greece. Total accidents pertained to be the dependent variable whereas various land use, demographic and macroscopic traffic modelling data were considered as explanatory variables. As required, the model inputs were aggregated to the TAZ level. First, a properly specified OLS model was developed, followed by the application of the GWR method. Unlike OLS models that are considered to be global, GWR allows the relationships modelled to vary over space, in line with spatial non-stationarity of social processes. This latter approach, improves the goodness of fit statistics of the OLS model and is helpful for policy-making at a local scale. A number of interesting correlations have been found, between accidents and a variety of statistically significant factors, such as the number of leisure establishments, pedestrian volume and length of particular types of roads. The GWR model built, uncovered the spatially varying relationships, dictating specific areas where these explanatory variables are strong or low predictors of the dependent variable.