Hotzone identification with GIS-based post-network screening analysis

This paper proposes a two-step method called post-network screening analysis for identifying collision hotzones (i.e., groups of neighboring hotspots) on a road network. The first step is the familiar safety network screening that uses AASHTO’s Highway Safety Manual (HSM) to identify and rank individual locations (i.e., hotspots) according to their potential for safety improvement. The second step is new and involves network-constrained kernel density estimation (KDE), a type of spatial analysis. KDE uses expected collision counts to estimate collision density, and outputs a graphical display that shows areas with a high density of expected collision counts. These are the hotzones. A particularly interesting area of application is the identification of high-collision corridors that may benefit from a program of systemic safety improvements. The proposed method was tested using five years of collision data (2005–2009) for the City of Regina, Saskatchewan. Three different network screening measures were compared: (1) observed collision counts, (2) observed severity-weighted collision counts, and (3) expected severity-weighted collision counts. The study found that observed severity-weighted collision counts produced a dramatic picture of the city’s hotzones, but this picture could be misleading as it could be heavily influenced by a small number of severe collisions. The results obtained from the expected severity-weighted collision counts smoothed the effects of the severity-weighting and successfully reduced regression-to-the-mean bias. A comparison is made between the proposed approach and the results of the HSM’s existing network screening method. As the proposed approach takes the spatial association of roadway segments into account, and is not limited to single roadway segments, the hotzones identified capture a higher number of expected equivalent-property-damage-only (EPDO) collisions than the existing HSM methodology. The study concludes that the proposed two-step method can help transportation safety professionals to prioritize hotzones within high-collision corridors more efficiently and scientifically.

[1]  D. Dai Identifying clusters and risk factors of injuries in pedestrian-vehicle crashes in a GIS environment , 2012 .

[2]  Ezra Hauer,et al.  Estimating Safety by the Empirical Bayes Method: A Tutorial , 2002 .

[3]  David O'Sullivan,et al.  Geographic Information Analysis , 2002 .

[4]  Craig Caulfield,et al.  Spatial and temporal visualisation techniques for crash analysis. , 2011, Accident; analysis and prevention.

[5]  A. Howard,et al.  Keeping children safe: rethinking how we design our surroundings , 2010, Canadian Medical Association Journal.

[6]  Ezra Hauer,et al.  How Best to Rank Sites with Promise , 2004 .

[7]  Bhagwant Persaud,et al.  Crash Cost Estimates by Maximum Police-Reported Injury Severity Within Selected Crash Geometries , 2005 .

[8]  Atsuyuki Okabe,et al.  A kernel density estimation method for networks, its computational method and a GIS‐based tool , 2009, Int. J. Geogr. Inf. Sci..

[9]  Michael J Demetsky,et al.  A Planning-Level Methodology for Identifying High-Crash Sections of Virginia’s Primary System , 2010 .

[10]  Isabelle Thomas,et al.  Spatial clustering of events on a network , 2010 .

[11]  Geert Wets,et al.  Identifying Hazardous Road Locations: Hot Spots versus Hot Zones , 2009, Trans. Comput. Sci..

[12]  Sekhar Somenahalli,et al.  Using GIS to Identify Pedestrian-Vehicle Crash Hot Spots and Unsafe Bus Stops , 2011 .

[13]  Darren J Torbic,et al.  SafetyAnalyst: Software Tools for Safety Management of Specific Highway Sites , 2007 .

[14]  Ezra Hauer Identification of Sites with Promise , 1996 .

[15]  G. Jenks The Data Model Concept in Statistical Mapping , 1967 .

[16]  June M. Tester,et al.  A matched case-control study evaluating the effectiveness of speed humps in reducing child pedestrian injuries. , 2004, American journal of public health.

[17]  Ibrahim Yilmaz,et al.  Geographical information systems aided traffic accident analysis system case study: city of Afyonkarahisar. , 2008, Accident; analysis and prevention.

[18]  Dominique Lord,et al.  Guidelines for Choosing Hot-Spot Analysis Tools Based on Data Characteristics, Network Restrictions, and Time Distributions , 2012 .

[19]  Simon Washington,et al.  Validation of Accident Models for Intersections , 2005 .

[20]  Geert Wets,et al.  The Magnitude of the Regression to the Mean Effect in Traffic Crashes , 2013 .

[21]  Jean-Claude Thill,et al.  Comparison of planar and network K-functions in traffic accident analysis , 2004 .

[22]  Peter Y. Park,et al.  Benefits of small municipalities using jurisdiction-specific safety performance functions rather than the Highway Safety Manual's calibrated or uncalibrated safety performance functions , 2013 .

[23]  Becky P.Y. Loo The Identification of Hazardous Road Locations: A Comparison of the Blacksite and Hot Zone Methodologies in Hong Kong , 2009 .

[24]  Michel Mouchart,et al.  The local spatial autocorrelation and the kernel method for identifying black zones. A comparative approach. , 2003, Accident; analysis and prevention.

[25]  Peter Y Park,et al.  Safety network screening for municipalities with incomplete traffic volume data. , 2013, Accident; analysis and prevention.

[26]  Tessa K Anderson,et al.  Kernel density estimation and K-means clustering to profile road accident hotspots. , 2009, Accident; analysis and prevention.

[27]  Alfonso Montella,et al.  A comparative analysis of hotspot identification methods. , 2010, Accident; analysis and prevention.

[28]  Jun Yan,et al.  Kernel Density Estimation of traffic accidents in a network space , 2008, Comput. Environ. Urban Syst..

[29]  Dominique Lord,et al.  Investigating Regression to the Mean in Before-and-After Speed Data Analysis , 2010 .