Land use and traffic collisions: A link-attribute analysis using Empirical Bayes method

• The hot spot and hot zone methodologies give different spatial patterns.
• The best fit estimation method in hot spot and hot zone methodologies are different.
• The relationship between land use and road safety is best understood at the local scale.
• The relationship can also be context-specific.

Road traffic collisions represent one of the major public health problems among the leading causes of deaths globally. This paper examines several approaches in detecting hazardous road locations, and discusses the spatial distribution of these locations as well as their relationships with different land uses in Hong Kong. Two most commonly used methodologies in detecting hazardous road locations are used: the hot spot and hot zone methodologies. Both methodologies are performed using raw collision count, excess collision count and Empirical Bayes (EB) estimations. The EB estimation uses land use characteristics near the road network in defining the reference groups. Finally all the approaches are compared by a test to assess their stability. The results show that for different hazardous road location detection methodologies, the best fit estimation methods on sites are different. The results confirm some land use impacts in previous studies, and suggest some further patterns on road safety. The findings are useful in understanding the complex interrelationships between land use and road safety, and in facilitating planners and policy makers to build safer cities.