Network-Based Highway Crash Prediction Using Geographic Information Systems
(2007)- Abstract
- The objectives of this project were to estimate network-based crash prediction models that will predict the expected crash experience in any given geographic area as a function of the highway link, intersection and land use features observed in the area. The result is a system of GIS programs that permit a polygon to be drawn on a map, or a set of links and intersections to be selected, and then predict the number of crashes expected to occur on the selected traffic facilities. These expected values can then be compared with observed values to identify locations with higher than usual crash incidence and may require attention to improve the safety of the location. Alternatively, this tool could be used to estimate the safety impacts of... (More)
- The objectives of this project were to estimate network-based crash prediction models that will predict the expected crash experience in any given geographic area as a function of the highway link, intersection and land use features observed in the area. The result is a system of GIS programs that permit a polygon to be drawn on a map, or a set of links and intersections to be selected, and then predict the number of crashes expected to occur on the selected traffic facilities. These expected values can then be compared with observed values to identify locations with higher than usual crash incidence and may require attention to improve the safety of the location. Alternatively, this tool could be used to estimate the safety impacts of proposed changes in highway facilities or in different land development scenarios.
A network approach was chosen to solve this problem, in which separate models were estimated for crashes at major intersections, and intersection-related and segment-related crashes on road segments. All three sets of models can then be used to predict the number of crashes for an entire highway facility delineated as the user desires – including all intersections. These models also consider all relevant road features, in particular the intensity of traffic at intersections and driveways resulting from the surrounding land use. Gathering traffic volumes at every intersection and driveway on the road network would preclude the feasibility of such an approach, both for estimation and in practice. Instead, the link between land development and trip generation was exploited to estimate the driveway and minor road volumes. Land development intensity variables were generated from land use inventories organized using Geographic Information Systems (GIS), permitting virtually automatic preparation of the required data sets for model estimation and application and prediction of crash counts on roads. Specifically, population and retail and non-retail employment counts were associated with each analysis segment to represent vehicle exposure to intersection-related crashes.
GIS was used for two purposes in this project: 1) distributing population and employment counts in a traffic analysis zone (TAZ) among all the links in that zone. 2) Visually comparing the predicted and observed accident counts in order to identify higher than usual crash locations. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/838867
- author
- Ivan, John N. ; Gårder, Per ; Bindra, Sumit ; Jonsson, Thomas LU ; Shin, Hyeon-Shic and Deng, Zuxuan
- organization
- publishing date
- 2007
- type
- Book/Report
- publication status
- published
- subject
- keywords
- prediction, demographics, land use, safety, GIS, accidents, blackspots
- pages
- 82 pages
- publisher
- New England Transportation Consortium
- report number
- NETCR67
- language
- English
- LU publication?
- yes
- id
- 4102eecb-f755-4e92-a209-5f6c5378797d (old id 838867)
- alternative location
- http://www.netc.umassd.edu/
- date added to LUP
- 2016-04-04 10:46:09
- date last changed
- 2018-11-21 21:00:40
@techreport{4102eecb-f755-4e92-a209-5f6c5378797d,
abstract = {{The objectives of this project were to estimate network-based crash prediction models that will predict the expected crash experience in any given geographic area as a function of the highway link, intersection and land use features observed in the area. The result is a system of GIS programs that permit a polygon to be drawn on a map, or a set of links and intersections to be selected, and then predict the number of crashes expected to occur on the selected traffic facilities. These expected values can then be compared with observed values to identify locations with higher than usual crash incidence and may require attention to improve the safety of the location. Alternatively, this tool could be used to estimate the safety impacts of proposed changes in highway facilities or in different land development scenarios.<br/><br>
A network approach was chosen to solve this problem, in which separate models were estimated for crashes at major intersections, and intersection-related and segment-related crashes on road segments. All three sets of models can then be used to predict the number of crashes for an entire highway facility delineated as the user desires – including all intersections. These models also consider all relevant road features, in particular the intensity of traffic at intersections and driveways resulting from the surrounding land use. Gathering traffic volumes at every intersection and driveway on the road network would preclude the feasibility of such an approach, both for estimation and in practice. Instead, the link between land development and trip generation was exploited to estimate the driveway and minor road volumes. Land development intensity variables were generated from land use inventories organized using Geographic Information Systems (GIS), permitting virtually automatic preparation of the required data sets for model estimation and application and prediction of crash counts on roads. Specifically, population and retail and non-retail employment counts were associated with each analysis segment to represent vehicle exposure to intersection-related crashes.<br/><br>
GIS was used for two purposes in this project: 1) distributing population and employment counts in a traffic analysis zone (TAZ) among all the links in that zone. 2) Visually comparing the predicted and observed accident counts in order to identify higher than usual crash locations.}},
author = {{Ivan, John N. and Gårder, Per and Bindra, Sumit and Jonsson, Thomas and Shin, Hyeon-Shic and Deng, Zuxuan}},
institution = {{New England Transportation Consortium}},
keywords = {{prediction; demographics; land use; safety; GIS; accidents; blackspots}},
language = {{eng}},
number = {{NETCR67}},
title = {{Network-Based Highway Crash Prediction Using Geographic Information Systems}},
url = {{https://lup.lub.lu.se/search/files/5616827/4157319.pdf}},
year = {{2007}},
}