An Analysis of Traffic Incidents on an Australian Urban Road Network

• Main Authors: Hojati, A.T, Fereirra, L., Charles, P., Kabit, M.R
• Format: Proceeding
• Language: English
• Published: Australasian Transport Research Forum 2011
• Subjects: TA Engineering (General). Civil engineering (General); TF Railroad engineering and operation
• Online Access: http://ir.unimas.my/id/eprint/2732/1/An%20Analysis%20of%20Traffic%20Incidents%20on%20an%20Australian%20Urban%20Road%20Network.pdf; http://ir.unimas.my/id/eprint/2732/; http://www.scopus.com.remotexs.unimas.my/record/display.uri?eid=2-s2.0-84916897520&origin=resultslist&sort=plf-f&src=s&sid=2fc8b887e9b2d09c17847ece811c70bd&sot=aut&sdt=a&sl=18&s=AU-ID%2835769299900%29&relpos=4&citeCnt=2&searchTerm

Assessing and prioritising cost-effective strategies to mitigate the impact of traffic incidents on non-recurrent congestion on major roads are currently a major challenge for road network operations. There is a lack of relevant local research in this area. Several incident duration models developed from international research are not considered appropriate for Australian conditions due to different driver behaviour and traffic environment contexts. A comprehensive data mining research project was undertaken to analyse traffic incident data, obtained from the Queensland Department of Transport and Main Roads’ STREAMS Incident Management System (SIMS) for a one year period ending in November 2010. Various factors that contributed to frequency, type, characteristics, duration and location of traffic incidents were examined and the findings are discussed in this paper. Results indicate that breakdown, multiple vehicle crash and debris were the major sources of incidents. Although incident frequency dropped sharply on weekends, the average incident duration was similar or longer than those of weekdays. Also, rainfall increased the incident duration in all categories. Furthermore, a variety of probability distribution functions were employed in order to test the best model for each category of incident duration frequency distribution. Log-normal distribution was inferred to be appropriate for crash and stationary vehicle incidents and gamma distribution for hazard incidents. Future research directions have been identified, particularly the estimation of the impact (cost) of traffic incidents, to assist in prioritising investment.