GIS aided multi-level fatal traffic accident analysis in Hong Kong

Abstract

A Geographical Information Systems (GIS) aided accident analysis algorithm, which integrates various statistical and spatial techniques, has been developed in this study. The algorithm is aimed to (1) reconize the effects of potential causal factors on the occurrences of fatal accidents; (2) evaluate the fatal accident risk of study areas and (3) locate areas which have the highest fatal accident risk. A GIS aided fatal accident analysis graphical user interface prototype program has been developed based on the algorithm. The program is aimed to facilitate and speed up the fatal accident analysis. In order to demonstrate the effectiveness of the algorithm, it is applied to perform a multi-level fatal accident analysis. The multi-level fatal accident analysis is conducted at territorial, zonal and street levels. The findings show that the occurrence of fatal accidents follows a Poisson distribution. In territorial level accident analysis, the results show that population, employment, vehicle kilometer travelled, number of car ownerships and median monthly domestic household income are found to have significant correlation with the occurrence of fatal accidents. In zonal level accident analysis, population, employment, land area and various land use factors are found to have significant correlation with the occurrence of fatal accidents. In street level accident analysis, the results show that vehicle flow and speed limit are found to have significant correlation with the occurrence of fatal accidents. This study has demonstrated that the GIS aided multi-level analysis algorithm can provide analysts a broader view to understand the occurrence of fatal accidents. The results can serve as a reference for town planners considering road safety in the town planning process. In zonal and street level fatal accident analysis, fatal accident black-sites are ranked according to their fatal accident risk, which combines historical fatal accident records with the fatal accident potential hazards. The results show that the proposed algorithm has improved the information of the ranking list comparing to that based on historical accident records. Employing this algorithm, the un-reported zones which have high potential fatal accident hazards can also be included in the risk ranking list. This information is certainly useful to the authority in giving priority to the road safety improvement plan.