Safety effects of reducing freeway illumination for energy conservation
Introduction
The addition of illumination where none was present is generally believed to have a positive effect on motor vehicle safety; reducing the frequency, as well as the severity of crashes. The operational cost of illumination, however, can make it a candidate for conservation during periods of high energy costs. In 2001, illegal manipulations of the energy markets in the Pacific Northwest and lower than average snowpack in the Cascade mountains created the perception that future energy shortages were likely. In response, Oregon's governor directed all state agencies to reduce power consumption by 10%. After review of power saving opportunities, the Oregon Department of Transportation (ODOT) elected to selectively reduce illumination on Oregon interstate highways as part of their energy-saving strategy. The illumination reductions occurred at both interchanges and along lineal freeway sections beginning in October 2001. The reductions occurred statewide, with a heavy focus on the Portland metropolitan area freeways (I-5, I-205, I-84, and US-26). An internal agency memorandum directed traffic engineers to select candidate locations for illumination modifications with above average conditions such as good striping, retroreflective signing, standard acceleration and deceleration lanes, typical geometry, and low crash history. Locations with adjacent pedestrian and bicycle facilities (e.g. paths) where highway illumination helped provide security were avoided. In a sense, only the safest locations were chosen for modification.
The reductions that were made fall into three general categories: (1) interchanges where lighting was reduced from a full lighting design to a partial design; (2) interchanges where lighting was from a partial plus design to partial lighting configuration; and (3) interstate freeway sections where mainline lineal lighting was reduced. Definitions of these categories are presented in a later section. A total of 44 interchanges and 5.5 miles of interstate freeway were modified. With the exception of 2.5 miles of the freeway sections, some level of illumination remained on at all locations. While the energy crisis did not materialize, the reductions were kept in place until an evaluation could be conducted since some of the locations were viewed to have excess illumination.
The objective of this paper is to quantify the safety effects of the reductions at these specific locations using an empirical-Bayes observational before–after methodology. Crash, geometry, weather, and volume information were collected for each of the modified locations as well as for a selected reference group. It is important to note that this research was not designed to study the safety effects of alternative lighting configurations at each interchange and no conclusions should be drawn about the safety effects of those designs. Changes in illumination were broadly classified and no before or after field measurements were taken of actual luminance values, lighting coverage, or other design specific values.
Section snippets
Previous evaluations
Previous work suggests that adding illumination where none was present has a positive effect on motor vehicle safety; reducing the frequency, as well as the severity of crashes on urban streets, highways, and at intersections (Elvik, 1995, Isebrands et al., 2006, Lamm et al., 1985, Walker and Roberts, 1976). Freeway-type facilities with their high speeds, volumes, and design standards are often strong candidates for illumination. There have been several cross-sectional comparisons of the safety
Research method
In the literature, a number of methods have been used for observational before–after studies of highway safety: (1) simple approach (naïve before–after); (2) yoked comparison site; (3) comparison group; and (4) empirical-Bayes (EB) methods. The primary challenge in highway safety evaluations is an accurate prediction of the expected number of crashes in the after period had the treatment not been implemented. After this prediction, a comparison can then be made with the estimated number of
Data
The locations where illumination was reduced were pooled into four groups based on the extent of lighting modifications and type of facility. Plan drawings showing lighting modifications were available to evaluate the before interchange lighting configuration. The groups are:
- •
Group A: full interchange lighting to partial interchange lighting (30 sites);
- •
Group B: lineal full lighting to no lineal lighting (2 sites, 2.5 miles);
- •
Group C: lineal full lighting to one direction lineal lighting (2 sites,
Developing safety performance functions
Predictive crash models are regression models that are used to estimate the frequency of crashes based on a set of explanatory variables. In most recent research, safety performance functions have been estimated with Poisson and negative binomial regression (NB) models (Harwood et al., 2002, Miaou and Lum, 1993, Poch and Mannering, 1996, Shankar et al., 1995, Vogt and Bared, 1998). As suggested in Lord et al. (2005), the Poisson and NB models are a theoretically appealing representation of the
Results
The analysis results from the pooled analysis for each crash severity and light condition and each lighting modification group are shown in Table 5. The estimated changes in safety for both day and night conditions are shown for the treatment locations. Because of the small sample size for the lineal modifications, the B and C groups were combined in the final safety evaluation.
While night crashes were the safety change of interest, day crashes were also evaluated for comparison. Day crash
Discussion
The evaluation of the effect of illumination reductions described in this study on crash performance was a challenging analysis for a number of reasons. First, the amount of illumination modified at the fully illuminated interchange locations was a partial reduction—lighting was still present in the after condition only at a reduced level. In the case of the partial plus to partial reductions the changes were even more subtle. For the lineal reductions, both total reductions and partial
Acknowledgements
The authors gratefully acknowledge the Oregon Department of Transportation for funding this research. The authors also thank the Technical Advisory Committee for guidance and suggestions that were valuable to completing this research. At Portland State University Michael Wolfe, Thareth Yin, Kartik Srinvas, Max Stephens and Peter Bosa contributed to this project. The authors wish to thank Craig Lyon of Ryerson University who provided assistance and guidance in developing the SAS models and the
References (31)
- et al.
Risk compensation—the case of road lighting
Accident Analysis & Prevention
(1999) - et al.
Drivers’ response to the installation of road lighting. An economic interpretation
Accident Analysis & Prevention
(2002) - et al.
Poisson, Poisson-Gamma and zero-inflated regression models of motor vehicle crashes: balancing statistical fit and theory
Accident Analysis and Prevention
(2005) - et al.
Modeling vehicle accidents and highway geometric design relationship
Accident Analysis and Prevention
(1993) - et al.
Effects of roadway geometrics and environmental factors on rural freeway accident frequencies
Accident Analysis and Prevention
(1995) Relieving the energy crisis—reduction in highway lighting
AASHTO Quarterly
(1977)- Bauer, K.M., Harwood, D.W., 1997. Statistical Models of Accidents on Interchange Ramps and Speed-Change Lanes....
Freeway accidents and illumination
Highway Research Record
(1972)Effect of lighting reduction on an urban major route
Traffic Engineering
(1976)- et al.
Implementing red light camera programs: guidance from economic analysis of safety benefits
Transportation Research Record: Journal of the Transportation Research Board
(2005)