INVESTIGATION OF ROADWAY GEOMETRIC AND TRAFFIC FLOW FACTORS FORVEHICLE CRASHES USING SPATIOTEMPORAL INTERACTION

Gurdiljot Singh Gill, Taha Sakrani, Wen Cheng, Jiandong Zhou2017 · DOI

TL;DR

This paper introduces a spatiotemporal Poisson lognormal model to better predict vehicle crashes on roadway segments by accounting for both spatial relationships between segments and temporal variations in influencing factors. This approach offers a more nuanced understanding of crash causes, moving beyond static models to capture the dynamic nature of traffic safety.

What Problem Does This Paper Address?

Traffic crashes impose significant economic and emotional burdens. While safety management processes aim to reduce these crashes, effectively identifying the root causes is challenging. Traditional crash prediction models often struggle to capture the complex interplay of factors that influence crash occurrences, particularly the spatial dependencies between adjacent road segments and the temporal changes in traffic conditions and geometric features over time. This paper addresses the need for a more sophisticated model that can simultaneously account for these spatial and temporal heterogeneities to improve crash prediction accuracy and inform targeted safety interventions.

Methods

The authors developed a spatiotemporal Poisson lognormal model with random effects. This approach builds upon existing overdispersed Poisson lognormal models by incorporating:

  • Spatial Random Effects: This component accounts for the spatial correlation between crash data on adjacent roadway segments. The idea is that crashes on one segment can influence the likelihood of crashes on nearby segments due to shared traffic patterns or geometric characteristics. They utilized a conditional auto-regressive (CAR) model for this, employing a binary adjacency-based weight matrix of the first order.
  • Time-Varying Coefficients: This allows the influence of explanatory variables (like lane width or speed limit) to change over time. This is crucial because the impact of these factors can evolve year-to-year due to changes in traffic behavior, infrastructure, or other unobserved factors.
  • Spatiotemporal Interaction: The model explicitly considers how the interaction between space and time affects crash occurrences, providing a more comprehensive understanding of crash dynamics.

The model was implemented using a Bayesian framework with Markov Chain Monte Carlo (MCMC) algorithms for parameter estimation.

Datasets

The study utilized a five-year dataset (2007-2011) from the Highway Safety Information System (HSIS), specifically focusing on freeway sections in California. This data was compiled from California's Traffic Accident Surveillance and Analysis System (TASAS) and the National Highway Traffic Safety Administration (NHTSA). The dataset included information on road characteristics (geometric features like lane width, median type), traffic flow (Average Annual Daily Traffic - AADT, design speed), and crash records for 279 roadway segments.

Key Insights and Ideas

  • Spatial dependencies matter: Crashes are not isolated events; their occurrence on one segment is influenced by neighboring segments. This highlights the importance of considering the "neighborhood" of a road segment in safety analysis.
  • Time is not static: The impact of factors like lane width or speed limit isn't constant. Capturing these temporal variations leads to more accurate predictions and a better understanding of evolving safety risks.
  • Combined spatial and temporal modeling is superior: By integrating both spatial and temporal dimensions, the model can better explain crash variability that simpler models might miss.
  • Road surface and speed are critical: The study reinforces the well-known impact of speed limits and road surface types (like bridge decks) on crash risk, but within a more sophisticated modeling framework.

Key Results

The study found significant correlations between several factors and vehicle crashes at roadway segments:

  • Lane Width: Generally, increased lane width was associated with a decreased risk of crashes. However, for a lane width of exactly 12 feet, the trend reversed, suggesting that drivers might drive faster or less cautiously with perceived extra space.
  • Speed Limit: A higher posted speed limit was positively correlated with an increased crash risk, as expected, due to reduced reaction times.
  • Road Surface Type: Driving on a bridge deck significantly increased the chance of a crash. Asphalt concrete pavement with less than seven inches of thickness also showed a heightened crash risk.
  • AADT: The Average Annual Daily Traffic showed a weak correlation with crashes, which the authors attribute to the rural nature of the analyzed road segments, implying lower overall traffic volumes.
  • The paper presents a table (Table 2) showing coefficients for various variables across different years, illustrating the time-varying nature of their influence. For instance, the coefficient for "Lane width 9, 10 feet" changes significantly from 0.18 in year 1 to -1.41 in year 3.

Limitations and Future Work

  • Limitations:
    • The study focused on rural freeway segments in California, so the findings might not be directly generalizable to urban areas or different road types.
  • The authors acknowledge that the "complex model formulation" might be computationally intensive.
  • The analysis of AADT's weak correlation might be specific to the rural context.
  • Future Work:
    • Explore different types of explanatory variables.
  • Incorporate different spatial levels of analysis (e.g., intersections, corridors).
  • Separate crash analysis by severity (injury, fatality) to identify factors specific to different crash outcomes.
  • The authors suggest that their complex model formulation can aid in "hotspot identification" by improving the precision and accuracy of crash count estimations.

Summary updated

Abstract

Traffic safety is a major concern in the transportation industry due to immense monetary and emotional burden caused by crashes of various severity levels, especially the injury and fatality ones. To reduce such crashes on all public roads, the safety management processes are commonly implemented which include network screening, problem diagnosis, countermeasure identification, and project prioritization. The selection of countermeasures for potential mitigation of crashes is governed by the influential factors which impact roadway crashes. Crash prediction model is the tool widely adopted by safety practitioners or researchers to link various influential factors to crash occurrences. Many different approaches have been used in the past studies to develop better fitting models which also exhibit prediction accuracy. In this study, a crash prediction model is developed to investigate the vehicular crashes occurring at roadway segments. The spatial and temporal nature of crash data is exploited to form a spatiotemporal model which accounts for the different types of heterogeneities among crash data and geometric or traffic flow variables. This study utilizes the Poisson lognormal model with random effects, which can accommodate the yearly variations in explanatory variables and the spatial correlations among segments. The dependency of different factors linked with roadway geometric, traffic flow, and road surface type on vehicular crashes occurring at segments was established as the width of lanes, posted speed limit, nature of pavement, and AADT were found to be correlated with vehicle crashes.
BibTeX
@inproceedings{gill2017investigation,
  author = {Gurdiljot Singh Gill and Taha Sakrani and Wen Cheng and Jiandong Zhou},
  title = {{INVESTIGATION OF ROADWAY GEOMETRIC AND TRAFFIC FLOW FACTORS FORVEHICLE CRASHES USING SPATIOTEMPORAL INTERACTION}},
  booktitle = {˜The},
  year = {2017},
  doi = {10.5194/isprs-archives-xlii-2-w7-1163-2017},
}
Figure