To illuminate or not to illuminate: Roadway lighting as it affects traffic safety at intersections
Highlights
► Road intersection lighting is associated with an approximately 12% reduction in the night-to-day crash ratio. ► Visibility improvements from lighting are correlated with reductions in the night-to-day crash ratio. ► Such correlations yield a provisional transfer function with utility for safety analysis.
Introduction
A primary purpose for installing roadway lighting is to increase the visual range afforded by vehicle headlamps while driving at night (IES, 2000). It is generally believed that roadway lighting improves safety by reducing the frequency of crashes occurring at night through improvements in driver visibility. Some studies have examined statistical associations between roadway lighting presence and traffic safety, making no explicit assumptions about the role of visibility in mitigating nighttime crashes. Generally, the results from these studies (IES, 1989, Elvik, 1995) have led to the general conclusion that roadway lighting is associated with a reduction in the night-to-day crash ratio (CIE, 1992). A night-to-day crash ratio reduction of approximately 30% has been suggested for the overall crash safety effect from roadway lighting (CIE, 1992).
Most studies of roadway lighting and safety consider lighting as a binary variable (i.e., present or not), but several attempts to relate specific characteristics of roadway lighting and safety have been made, using various photometric measures as surrogates for visibility. One early study was undertaken by Box (1971) where he compared the ratio of night-to-day crash rates along 22 lighted and unlighted highway sections. The author grouped the highway sections into categories corresponding to mean horizontal illuminance levels between 3 and 6 lx, between 8 and 11 lx, and between 13 and 15 lx. Presumably, higher illuminances would result in greater visibility and might be expected to yield fewer nighttime crashes. The lighted sections had lower night-to-day crash rate ratios than the unlighted sections, but the lowest night-to-day crash rate ratio was found for the 3-to-6 lx category, with higher night-to-day crash rate ratios for the two higher illuminance categories. In a different study, Box (1976) evaluated the impact of reducing illuminance levels along a major highway from 14 lx to 9 lx and found that nighttime crash frequency increased by 10%, although daytime crash frequency also increased (by 4%). In comparison, Scott (1980) measured roadway luminance levels along 89 (each at least 1 km in length) two-lane roadway sections. Despite a great deal of variability in night-to-day crash ratios among all of the sites, a best-fitting exponential function to the data yielded a monotonically decreasing relationship between night-to-day crash ratios and luminance level, consistent with Box (1976) but inconsistent with Box (1971). The lack of agreement across these studies could be related to differences among the specific locations studied (e.g., roadway geometry or traffic control) or different evaluation paradigms (e.g., before/after or with/without comparisons). Even considering these differences, it is important to account for all the factors that could affect visibility. In particular, the impact of lighting on visual performance depends upon the contrast and the size of a hazard, not simply upon illuminance or luminance levels provided by the roadway lighting system.
In an attempt to relate specific measures of visibility associated with roadway lighting to safety, Janoff et al. (1978) studied nighttime crash frequency and measured photometric conditions along several roadways, and found nighttime crash frequency to have a weak inverse relationship with a visibility metric they derived, which was based on the ratio between an object's luminance contrast and the contrast it would require to be just visible. Also, Keck (2001) summarized a different study of roadway lighting, visibility and safety and reported that a similar visibility metric was not correlated with the frequency of nighttime crashes unless headlamp illumination was also considered. Even then, the statistical association was modest in magnitude (coefficient of determination r2 = 0.12).
In addition to the above issues related to study locations and evaluation methods, a factor that makes it difficult to assess the relationship between fixed roadway illumination systems and traffic safety is that lighting is installed for a variety of reasons (e.g., security, esthetics). Moreover, lighting is usually not the only roadway safety feature that is installed when a roadway is designed or improved (IES, 1989). For example, lighting may be installed when converting a stop-controlled to a signalized intersection, or may be installed when threshold levels of pedestrian or vehicular volumes, or annual crash frequencies, are exceeded. Thus, the estimates of association between roadway lighting presence and nighttime crashes in past studies may be confounded by other modifications to the study site.
In the present paper, an exploratory strategy is presented to test the theoretical links among roadway lighting, visibility, and safety while accounting for as many potential safety-influencing variables as possible. To overcome previous limitations in the literature, our goal was to determine if we could establish convergence between statistical and analytical approaches to crash safety. Thus, both statistical and analytical approaches were undertaken to relate crash frequencies to visual performance levels for the same set of roadway lighting conditions. Donnell et al. (2010) merged roadway lighting presence and roadway geometric and traffic volume and control data with daytime and nighttime crash data to assess the statistical association between roadway intersection lighting presence and the night-to-day crash ratio in Minnesota. In their study, many potential safety-influencing variables that were not considered in previous lighting-safety research were included in the statistical model estimation. In a parallel but independent investigative domain, Rea et al. (2010) developed photometric simulations of a large variety of roadway intersection lighting configurations to make context-specific predictions of visual performance levels provided by roadway lighting systems varying in light level, spatial configuration, and ambient characteristics for drivers of different ages.
Here we present a methodology to link roadway lighting characteristics to visual performance levels and, therefore, to traffic crashes. Importantly, our objective was to test whether there is convergence between statistical and analytical approaches, which would bolster the expected relationships among lighting presence, improved visibility, and improved safety. Each approach reduces the inherent uncertainty associated with the other by providing an independent basis to explore the theoretical relationship between roadway intersection lighting and reductions in nighttime crashes through improvements in driver visual performance.
The present paper summarizes the parallel statistical and analytical approaches, developed by Donnell et al. (2010) and by Rea et al. (2010), respectively, used to probe the theoretical relationship described above. First, the background, methods and results of each independent approach are provided; next, the evidence for the convergence of these approaches is described. As stated above, the approaches used in this paper are described in previous publications (Donnell et al., 2010, Rea et al., 2010); the present paper describes how these approaches were adapted for different roadway intersection types in Minnesota.
If the necessary links between roadway lighting, visual performance and traffic safety can in fact be forged through converging approaches, it could then be possible for traffic engineers to improve traffic safety by considering the visual performance levels and the costs of proposed roadway illumination systems through a provisional proposed transfer function relating visibility from lighting to crash safety.
Section snippets
Description of parallel approaches
In the present section, the statistical modeling and the visual performance modeling are described together with the findings from these two approaches.
Integration of results from the statistical and analytical approaches
As described above, the visibility coverage area analysis developed by Rea et al. (2010) was used here together with the statistical associations between lighting and nighttime crashes based on Donnell et al. (2010) to examine the theoretical relationships among lighting, visibility and traffic safety. Since the statistical associations between crash frequency and roadway lighting for different types of intersections had been modeled by Donnell et al. (2009a) and since the visibility coverage
Discussion
The present study formally examined the theoretical relationships among lighting, visibility and safety using two independent approaches, a statistical and an analytical approach, for the same lighting context. A key to the present study was that the roadway lighting conditions associated with crashes in Minnesota could be modeled in terms of the levels of visibility those same roadway lighting conditions provide. The close correspondence between the statistical and the analytical results (Fig.
Acknowledgments
We gratefully acknowledge the input and technical contributions of our colleagues Peter Morante and Leora Radetsky from the Lighting Research Center, Rensselaer Polytechnic Institute.
References (71)
- et al.
A note on modeling vehicle-accident frequencies with random-parameters count models
Accident Analysis and Prevention
(2009) - et al.
Applying the random effect negative binomial model to examine traffic accident occurrence at signalized intersections
Accident Analysis and Prevention
(2003) - et al.
A framework for estimating the safety effects of roadway lighting at intersections
Safety Science
(2010) - et al.
Case–control and cross-sectional methods for estimating crash modification factors: comparisons from roadway lighting and lane and shoulder width safety effect studies
Journal of Safety Research
(2011) - et al.
The statistical analysis of crash-frequency data: a review and assessment of methodological alternatives
Transportation Research Part A
(2010) - et al.
Application of propensity scores and potential outcomes to estimate effectiveness of traffic safety countermeasures: exploratory analysis using intersection lighting data
Accident Analysis and Prevention
(2013) Visibility of targets: model for calculation
Lighting Research and Technology
(1989)Highway Safety Manual
(2010)- et al.
Size as a determinant of reading speed
Journal of the Illuminating Engineering Society
(1993) - Bauer, K.M., Harwood, D.W., 1999. Statistical Models of At-Grade Intersection Accidents—Addendum. Report No....