Pedestrian Collision Prediction Models for Urban Intersections
- 1 January 2002
- journal article
- Published by SAGE Publications in Transportation Research Record: Journal of the Transportation Research Board
- Vol. 1818 (1), 102-107
- https://doi.org/10.3141/1818-16
Abstract
In more recent safety analysis methods, collision prediction models can be used for identifying intersections with promise of safety improvement and for evaluating the effects of treatment. Considerable effort has been directed at developing collision prediction models, but little has been directed at pedestrian collisions. Collisions involving motor vehicles and pedestrians pose a significant safety problem, principally in urban areas, where the levels of vehicle-pedestrian conflicts are high. Data from Toronto, Canada, are used in the development of pedestrian collision prediction models for three- and four-legged urban intersections, with and without signal control. These models, which relate safety to pedestrian and vehicle traffic volumes, can be used to identify locations that might be targeted for treatment and to help evaluate treatment effects. Models are developed by using pedestrian and vehicular volumes and vehicle volumes only. It is seen that the use of pedestrian volume information results in a much richer model, emphasizing the importance of collecting this information in routine traffic counting programs. An important issue for collision prediction models is transferability to other jurisdictions. This is especially important in the case of pedestrian collision models, because many jurisdictions may not have data sets containing sufficient collisions and pedestrian volume counts with which to calibrate reliable models. Data from the city of Hamilton, Ontario, were used to test the transferability of the Toronto four-legged signalized intersection model. The test was successful: the recalibrated Toronto models predicted collision numbers that were very close to those predicted by the model calibrated directly for the Hamilton data.Keywords
This publication has 5 references indexed in Scilit:
- Accident Prediction Models With and Without Trend: Application of the Generalized Estimating Equations ProcedureTransportation Research Record: Journal of the Transportation Research Board, 2000
- Empirical Bayes Procedure for Ranking Sites for Safety Investigation by Potential for Safety ImprovementTransportation Research Record: Journal of the Transportation Research Board, 1999
- Disaggregate Safety Performance Models for Signalized Intersections on Ontario Provincial RoadsTransportation Research Record: Journal of the Transportation Research Board, 1998
- Pitfalls of Using R2 to Evaluate Goodness of Fit of Accident Prediction ModelsTransportation Research Record: Journal of the Transportation Research Board, 1996
- Models for predicting accidents at junctions where pedestrians and cyclists are involved. How well do they fit?Accident Analysis & Prevention, 1993