Planar and Network Spatial Analyses of Road Traffic Accidents - A Review of Methods and Tools

Abstract

Developed and developing countries are observing increasing rate of road traffic accident (RTA) and they are investing in efforts trying to reduce the frequency and severity of RTA. Extensive research has been going on to develop methods to analyse historical RTA data using GIS tools to identify and rank RTA hotspots. This helps in prioritising expenditure distribution on road maintenance at road sections identified as RTA hotspots. The aim of this paper is to critically review some of the planar and network spatial analyses in RTA hotspot identification. The planar spatial analyses reviewed in this paper are one global statistical approach - Global Moran's I and three local statistical approaches - Local Anselin Moran's I, planar Kernel Density Estimation and Getis-Ord GI*. These geostatistical approaches are embedded in ESRI ArcMAP analyst extension tools. The network spatial analyses reviewed in this paper are two tools based on network Kernel Density Estimation - SANET and KDE+. A new addition is a network-based approach that accounts for RTA frequency, severity and socioeconomic costs - Spatial Traffic Accident Analysis (STAA). After review, it is found that network spatial analyses are more preferred to planar spatial analyses for RTA hotspot analysis (HSA) on or along a road network.