Comparison of The Safety of Lighting Options on Urban Freeways

by Michael S. Griffith

Introduction

Nationwide accident statistics show that more than 50 percent of fatal accidents occur during the hours of darkness. Because only 25 percent of travel occurs during the same period, the fatality rate is about three times higher at night than during the day. The installation of overhead lighting is a potential countermeasure to this nighttime accident problem. However, this is expensive, and much of the research to date offers inconclusive results about its effect on highway safety.

Many previous studies have evaluated the relationship of urban freeway lighting and highway safety. However, the majority of these studies were conducted in the 1960s and early 1970s, and the results from these studies may be outdated. On our nation's highways, there have been many changes in traffic flow, vehicle fleet, and road-user demography in the past 20 to 30 years. Clearly, the volume of traffic on urban freeways is significantly higher today, and congestion is a greater problem.

This paper presents the results of a research study that integrated accident, roadway, and traffic volume data from many sources to compare the safety of continuously lighted urban freeways and urban freeways with interchange lighting only. A freeway section with continuous lighting has overhead lighting at the interchanges and between the interchanges, as opposed to overhead lighting at the interchanges only. The urban freeways evaluated in this study are located in the Minneapolis-St. Paul metropolitan area.

The primary data source for the study was the Highway Safety Information System (HSIS). HSIS uses data already collected by five states and annually acquired by the Federal Highway Administration (FHWA). It is a roadway-based system that provides quality data on a large number of accident variables, roadway characteristics, and traffic volumes for problem analyses. An examination of HSIS data files revealed that most of the desired accident and roadway variables to conduct this study were present.

However, none of the files has the complete lighting information or 24-hour traffic distributions needed to develop day versus night accident rates. Lighting information and automatic traffic recorder reports that provide summaries of 24-hour traffic distributions were obtained from Minnesota Department of Transportation (MnDOT). To classify accident and traffic volume data by day and night, sunrise and sunset information was acquired from the United States Naval Observatory.

Previous Studies

The International Commission on Illumination (CIE) Technical Report Road Lighting as an Accident Countermeasure provides a complete and detailed summary of all the studies that have investigated the relationship between urban freeway lighting and safety. (1) The summary includes before-and-after and cross-section studies that examined the safety effectiveness of continuous freeway lighting, interchange lighting, and the impact of reducing lighting levels on urban freeways. The before-and-after approach examines how safety has changed at sites where a treatment or modification was made. The cross-section approach is designed to compare the safety characteristics of two different groups.

In summary, it appears for the limited number of studies that obtained conclusive results, urban freeway lighting has a beneficial effect on safety. However, much of the research is 20 years old or older or was conducted in other countries, and it may not represent the current experience in the United States.

A cross-section study that is considered to be one of the most comprehensive studies of continuous freeway lighting was reported in 1972 by Box. (2) The study included about 320 kilometers (200 miles) of urban freeway on which more than 21,000 accidents occurred during the evaluation period of 1960 to 1968. The studied areas were Denver, Chicago, Atlanta, Dallas, Phoenix, Detroit, and Toronto. Box found that the average night/day accident-rate ratio was 66 percent greater on unlighted freeways than on lighted freeways, and he concluded that the illumination of an unlighted urban freeway could theoretically reduce night accidents by an average of 40 percent.

Three before-and-after studies were conducted in the 1970s. All three studies had fairly small sample sizes. Box evaluated approximately 8 km (5 mi) of a six-lane urban freeway with continuous lighting in Chicago. (2) He concluded that the installation of lighting possibly lowered the night accident rate; however, a complete statistical confirmation is lacking. The results of a French study by Onser showed that night accidents were reduced by 57 percent following installation of lighting on a section of the A1 Motorway. (3) A Japanese study by Nishimori examined a 12.9-km (8-mi) motorway section after overhead lighting was installed and found that the night accident rate was reduced by 56 percent. (4)

There are three studies that examined the impact of the switching off of an existing overhead lighting system on the safety of freeways. One study by Hilton examined the effects of shutting off the lighting system on a 10-lane freeway section located on I-95 in Virginia during the oil embargo period in 1973-74. The results indicate that switching off the lighting system had a negative impact on safety. The magnitude of the impact was found to vary with the season of the year. (5)

Approach

A cross-section approach was used in this research study to assess the safety effects of urban freeway lighting. The two groups in this study were urban freeway sections with continuous lighting and urban freeway sections with interchange lighting only. Unfortunately, in this case, there was not an opportunity to conduct a before-and-after study.

In identifying lighted urban freeway sections to be evaluated, only the five states within HSIS were considered. These five states are: Illinois, Maine, Michigan, Minnesota, and Utah. Since Illinois, Michigan, and Minnesota have more and larger metropolitan areas than Maine or Utah, it was decided to investigate the information available from these three states.

Each record on the HSIS roadlog files contains current characteristics of the road system, including shoulder and median information, lane information, surface type and width, etc. However, the roadlog files can't always provide all of the roadway information required for safety analyses.

Minnesota was selected for this study because MnDOT was able to provide the most comprehensive, supplementary information to support HSIS data. MnDOT provided a complete listing of where urban freeway lighting exists within the state and information about the type of lighting at each location. In addition, at the time of this study, Minnesota was the only HSIS state with a videodisc photolog system.

The videodisc photolog system provides additional roadway information that is not contained in the roadway files. Minnesota's videodisc photolog system can be accessed in the HSIS laboratory.

A photolog is a series of sequential images taken from a moving vehicle at approximately driver's eye level to provide a permanent record of the state-maintained roadway network. MnDOT uses an automated vehicle to annually film the entire state highway system in both directions with a videocamera. A photograph is taken every 0.016 km (0.01 mi) with the camera oriented slightly down and to the right for optimum coverage of the highway and roadside development. Video images of the highway are recorded and stored on laser videodiscs.

The videodisc photolog system is a key tool that allows users to have automatic computer access to all video images of the state-maintained roadway network. This system is used to collect supplementary data for studies and to verify existing data.

To enable the development of day versus night accident rates, Minnesota also provided 24-hour traffic distribution information from automatic traffic recorders (ATR), permanent count stations that collect 24-hour traffic data every day of the year.

The accident and traffic volume data were then classified into night and day periods based on sunrise/sunset information for Minneapolis, acquired from the United States Naval Observatory; the time of the accident recorded by the hour in HSIS accident files; and the Minnesota ATR summary reports.

Site Characteristics

A total of 87.9 km (54.6 mi) of urban freeway segments with continuous lighting and 57.1 km (35.5 mi) of urban freeway segments with interchange lighting only were used in the study. All of the study sections are located in the Minneapolis-St. Paul metropolitan area. A larger sample size (more miles) of urban freeway sections with interchange lighting only would have been useful in this study. Additional sections with interchange lighting only were considered, but they were judged to be inappropriate to include in the study. Information about the characteristics of the study sections were acquired from the HSIS, the Minnesota videodisc photolog system, and from field observations.

There is one major roadway difference between the urban freeway sections with continuous lighting and those with interchange lighting only. This difference pertains to the number of interchanges per mile. There are 1.2 interchanges per 1.6 km (1 mi) on the continuously lighted sections and 0.8 interchanges per 1.6 km (1 mi) on the sections with interchange lighting only. All of the interchanges are of the diamond type except three that are cloverleafs.

Interchange areas include the areas along the freeway mainline between and including acceleration lanes, deceleration lanes, and their respective ramps. (6) The accident experience in interchange and non-interchange areas is very different, and therefore, the location of interchange/non-interchange areas was considered important information to collect.

Using the videodisc photolog system, interchange areas and non-interchange areas were identified by beginning and ending mileposts for the study sections. About 59 percent of the total miles of continuously lighted sections were identified to be within interchange areas, and only 46 percent of the total miles of sections with interchange lighting only were within interchange areas.

In the field, each study section was examined during nighttime and daytime hours. Driving the sections during the day provided additional information about the roadway and operational characteristics. During night inspections, approximately 10 percent of the lights were observed to be off on the study sections. The lights were most likely off due to the malfunctioning of the system hardware and burned-out lamps. (7)

All of the continuously lighted sections had complete lighting at the interchanges. Complete interchange lighting is used to describe the process of applying lighting to the interchange in such a manner as to achieve illumination of all roadways in the interchanges. (8) Generally, complete interchange lighting is associated with freeways where the main lanes are lighted.

Partial interchange lighting only includes illumination of the parts of the interchange that are considered most critical. The parts most frequently lighted are the merge-diverge areas of the ramp connections, intersections, and critical roadway features.

With one exception, complete or partial lighting existed at the interchanges on the study sections with interchange lighting only. Partial lighting was the "rule" at these interchanges. One of the interchanges on the sections with interchange lighting only had high-mast lighting with the light sources mounted at 30.5 m (100 ft). High-mast lighting provides an uniform distribution of light over the entire area of an interchange. The lighting systems for the study sections have a combination of 200/400-watt, high-pressure sodium luminaires and 400-watt, mercury vapor luminaires. The majority of the luminaires are high-pressure sodium. The pole heights of the luminaires are 9.1 m (30 ft), 12.2 m (40 ft), and 15.2 m (50 ft).

The travel characteristics between the study sections were compared. The day/night distribution of the total vehicle miles traveled (VMT) for the continuously lighted sections and the sections with interchange lighting only was found to be identical. About 76 percent of the total travel occurred during day hours and 24 percent occurred during night hours. The traffic volumes for the freeway sections with continuous lighting and the freeway sections with interchange lighting only are virtually identical (within 5 percent). The average number of vehicles counted per year was 25.8 million for the sections with interchange lighting only and 27 million for the continuously lighted sections.

Results

Day and night accident rates were calculated for the study sections. Accident data from 1985-1990 was used. Accident rates corresponding to all accidents, serious injury accidents, injury accidents, property-damage-only accidents, interchange area accidents, and non-interchange area accidents were calculated.

Table 1 shows the total day accident rates, total night accident rates, and the total night/day accident-rate ratios. The total day accident rate for continuously lighted sections is three times higher than the total day accident rate for sections with interchange lighting only. This result appears to be reasonable given that there are more interchanges per mile on the freeway sections with continuous lighting than on the sections with interchange lighting only. The accident experience tends to be higher in interchange areas than in non-interchange areas. The total night accident rate for continuously lighted sections is also close to being three times higher than the total night accident rate for sections with interchange lighting only.

Table 1 - Accident Rates

To adjust for the differences between the freeway sections with continuous lighting and those with interchange lighting only, total night/day accident-rate ratios were calculated. The ratio is computed by dividing the total night accident rate by the total day accident rate. The total night/day accident-rate ratio for the sections with interchange lighting only is 12 percent higher than the total night/day accident-rate ratio for sections with continuous lighting (a larger night/day accident-rate ratio indicates a correspondingly more hazardous night condition). The validity of this result is dependent on the assumption that any significant changes (weather, vehicle fleet, road-user demography, etc.) that occurred over the study period, happened in the same manner at both the continuously lighted sections and sections with interchange lighting only. Since the study sections are adjacent to one another, this assumption is reasonable.

The null hypothesis that the total night/day accident-rate ratio for the sections with interchange lighting only is the same as the total night/day accident-rate ratio for the continuously lighted sections was evaluated with a Poisson statistical test at a level of significance equal to 0.05 and was rejected. So, from a statistical perspective, the total night/day accident-rate ratio for the sections with interchange lighting only is greater than the total night/day accident-rate ratio for the continuously lighted sections.

Figure 1 shows serious injury, injury, and PDO night/day accident-rate ratios. The only meaningful difference between the ratios is that the PDO night/day accident-rate ratio is 19 percent higher for the freeway sections with interchange lighting only than the PDO night/day accident-rate ratio for the continuously lighted sections.

Statistical models were fitted to the data to determine which roadway variables (number of lanes, recovery zone [median width + sum of left shoulder widths from both sides of the roadway], median type, and interchange/non-interchange area) are important in predicting accident rates. The best model obtained contains only one independent variable. This variable defines whether a roadway section is in an interchange area or a non-interchange area.

Figure 2 shows the accident-rate ratios for interchange and non-interchange areas.

For interchange areas, the night/day accident-rate ratios are statistically equal for continuously lighted sections and sections with interchange lighting only. One would expect these ratios to be similar since overhead lighting exists at the interchanges located on the sections with continuous lighting and the sections with interchange lighting only.

The night/day accident-rate ratio for non-interchange areas is 18 percent higher for sections with interchange lighting only than it is for continuously lighted sections. Statistically, the night/day accident-rate ratios for non-interchange areas are different. This result is based on a comparison of freeway sections between interchange areas that have overhead lighting to those without overhead lighting. Since it is of particular interest to compare freeway sections with and without overhead lighting, a more detailed analysis of the non-interchange sections was conducted.

Table 2 shows the non-interchange night/day accident-rate ratios for serious injury accidents, total injury accidents, and PDO accidents. The only meaningful difference between the ratios for the lighted sections and the unlighted sections is the one associated with PDO accidents. The PDO night/day accident-rate ratio for freeway sections between interchange areas without lighting is 32 percent higher than the PDO night/day accident-rate ratio for freeway sections with lighting between interchange areas. The differences in serious injury and total injury accident-rate ratios between the roadway sections with lighting and without lighting are statistically insignificant. The 29-percent difference in the serious injury accident-rate ratio between the unlighted and lighted sections is fairly large; however, the small sample size of serious injury accidents prevented this finding from being declared statistically significant using a Poisson statistical test.

Table 2 - Non-Interchange Night/Day Accident-Rate Ratios

A finding of this study and other traffic volume studies is that on average 25 percent of urban freeway traffic consistently occurs at night. This is an important finding for it allows the calculation of rate ratios in the absence of traffic counts. From the literature, it was found that the night/day accident-rate ratio is mathematically equal to three times the number of night accidents divided by the number of day accidents. (2)

This mathematical relationship was examined using the study data and was found to be fairly reasonable. For the lighted sections in non-interchange areas, three times the number of night accidents divided by the number of day accidents equals 1.28, which is practically equal to the computed night/day accident-rate ratio of 1.30. For the unlighted sections in non-interchange areas, the estimated rate of 1.47 is not much different than the computed night/day accident-rate ratio of 1.54. For the continuously lighted sections in interchange areas, the estimated rate of 1.33 is practically equal to the computed night/day accident-rate ratio of 1.30. For the sections with interchange lighting only in interchange areas, the estimated rate of 1.38 is not much different than the computed night/day accident-rate ratio of 1.33.

Given that the mathematical equation was found to be reasonable with data from this study, an analysis was conducted using a derivation of the equation to compute and compare the expected number of night accidents for the lighted and unlighted study sections in non-interchange areas. The expected number of night accidents can be calculated for any assumed night/day accident-rate ratio. Assume a freeway experiences 1,000 day accidents during a time period. For non-interchange areas, the lighted ratio of 1.30 from the study data produces 433 night accidents by use of the equation E_N = R * (A_D/3), where E_N = expected number of night accidents, R = night/day accident-rate ratio, and A_D = number of day accidents. (2) The unlighted ratio for non-interchange areas equal to 1.54 from the study data produces 513 night accidents. The difference of the expected number of night accidents between the lighted and unlighted sections is 80 (513 - 433), which represents 16 percent fewer night accidents under lighted conditions. If compared with 1,513 (1,000 + 513) total day and night accidents, the overall reduction would be 5 percent. This example indicates that the illumination of an unlighted urban freeway between interchange areas could theoretically reduce night accidents by 16 percent or total accidents by 5 percent.

Table 3 shows the crash costs for lighted sections between interchange areas classified by night and day. The costs are calculated for each accident severity by multiplying the accident frequency by the corresponding standard cost per crash. The total night and day crash costs are $37,617,605 and $46,270,320, respectively.

Table 3 -- Crash Costs for Lighted Sections between Interchange Areas

Note: The crash costs are in 1988 dollars. Source: The Costs of Highway Crashes, FHWA-RD-91-055.

Night/day crash-cost ratios were computed using the crash costs from tables 3 and 4. The night/day crash-cost ratios for freeway sections between interchange areas with lighting and without lighting are $37,617,605/$46,270,320 = 0.81 and $17,727,099/$17,228,489 = 1.03, respectively. The night/day crash-cost ratio for freeway sections between interchange areas without overhead lighting is 27 percent higher than the night/day crash-cost ratio for freeway sections between interchange areas with overhead lighting.

A benefit/cost analysis was conducted for the freeway sections between the interchange areas with and without lighting. The cost to install and provide electrical power for an urban freeway lighting system was obtained from a New York Department of Transportation (NYDOT) official. The NYDOT official could not provide maintenance costs. Based on the NYDOT cost information, it is estimated that the installation and power costs (in 1988 dollars) for the 35 km (22 mi) of continuously lighted freeways between the interchange areas is $4,528,560. The overhead lighting on the study sections was installed between 1960 and 1984.

For the benefit/cost analysis, the expected night crash costs (if roadway lighting did not exist at these locations) had to be calculated for the 35 km (22 mi) of continuously lighted freeways between the interchange areas. This was found by using the night/day crash-cost ratio for the unlighted sections between the interchange areas. One would expect if roadway lighting was not installed on the 35 km (22 mi) of continuously lighted freeways between interchange areas that the night/day crash-cost ratio would equal 1.03 (night/day crash-cost ratio for the unlighted sections) and not 0.81. The expected night crash costs are equal to $47,658,429 ($46,270,320 x 1.03).

The actual night crash costs are $37,617,605; this is $10,040,824 less than the expected night costs. Therefore, the estimated economic benefit of roadway lighting is $5,512,264 ($10,040,824 - $4,528,560 [estimated installation and power costs]) for the years 1985-1990. This is the amount saved by society. The estimate is conservative -- even if one considers that the costs to maintain the lighting systems have not been included in the calculations -- because the roadway lighting was installed at the study sections between 1960 and 1984 and installation costs are based on 1988 dollars. Therefore, the installation costs are inflated. In addition, the safety benefit of lighting before 1985 has been neglected since study data was obtained only for the years 1985-1990.

Conclusion

The results of this study and the 1972 cross-section study conducted by Box are different. There are several likely reasons why the results are different. These reasons are: (1) Box compared lighted against unlighted sections, and in this study continuously lighted sections were compared against sections with interchange lighting only. (2) About 25 years have passed between the studies. (3) Data from several cities were used in the Box study, and data from only one metropolitan area was used in this study.

Box found that the average night/day accident-rate ratio was 60 percent higher on unlighted freeways (no overhead lighting at the interchanges or between the interchanges) than on lighted freeways. Using accident data between 1985 and 1990, this study found that the total night/day accident-rate ratio for the sections with interchange lighting only is 12 percent higher than the total night/day accident-rate ratio for sections with continuous lighting. In the non-interchange areas, the night/day accident-rate ratio is 18 percent higher for sections with interchange lighting only than it is for continuously lighted sections. Box did not report separate ratios for the non-interchange and interchange areas.

A finding of the Box study indicates that the illumination of an unlighted freeway (installation of overhead lighting at the interchanges and between the interchanges) could theoretically reduce night accidents by an average of 40 percent. This study found that the illumination of an unlighted urban freeway between interchange areas (installation of overhead lighting between the interchanges only) could theoretically reduce night accidents by 16 percent.

Another conclusion of this study is that the relative benefit of overhead lighting for urban freeways between interchange areas is primarily associated with PDO accidents. The PDO night/day accident-rate ratio for unlighted freeway sections between interchange areas is 32 percent higher than the PDO night/day accident-rate ratio for lighted freeway sections between interchange areas. This difference is reflected in the cost analysis that found that the night/day crash-cost ratio for freeway sections between interchange areas without overhead lighting is 27 percent higher than the night/day crash-cost ratio for freeway sections between interchange areas with overhead lighting.

Jurisdictions considering the installation of a roadway lighting system for an urban freeway facility need to assess its potential economic impact. Benefit/cost analyses should be conducted for different types of freeway lighting systems. The estimated economic benefit of continuous lighting for the study's 22 miles of urban freeways between interchange areas is $5.5 million. This is the most important result of the study. As a minimum, the installation of roadway lighting at interchanges only on urban freeways should be considered by a jurisdiction as an intermediate step between no lighting and continuous lighting.

This study provided new information about urban freeway accidents and illumination and found a positive relationship between urban freeway lighting and highway safety. Additional research is needed to develop an even stronger knowledge base on urban freeway lighting. It would be desirable to conduct a before-and-after study with data from several states that have urban freeway lighting systems in service with a wide range of maintained illumination levels. MnDOT maintains an average horizontal illuminance in the range of 0.6 to 0.8 footcandle (6 to 9 lux) for their urban freeway lighting system, which meets the American Association of State and Highway Transportation Officials (AASHTO) recommendations. It is not known whether higher illumination levels would result in accident-reduction benefits.

References

(1) Road Lighting as an Accident Countermeasure, The International Commission on Illumination (CIE) Technical Report, First Edition, 1992.

(2) P.C. Box. "Freeway Accidents and Illumination," Highway Research Record 416, Highway Research Board, 1972.

(3) Onser. "Effect of Lighting on Motorways," National Organisation for Road Safety, 1973.

(4) S. Nishimori. "Effects of the Installation of Lighting on the Whole Length of the Motorway Upon Night Time Accidents," Japan Highway Public Corporation, paper presented at 11th Annual Conference of Japan Road Association, Tokyo, 1973.

(5) M.H. Hilton. "The Effectiveness of Freeway Lighting in Reducing Accidents," Compendium of Technical Papers ITE 49th Annual Meeting, Institute of Transportation Engineers, Toronto, Canada, 1979.

(6) J. Twomey, M. Heckman, and J. Hayward. Safety Effectiveness of Highway Design Features, Volume IV: Interchanges, Publication No. FHWA-RD-91-047, Federal Highway Administration, Washington, D.C., November 1992.

(7) "Maintenance Process Review of Highway Lighting," Federal Highway Administration, Minnesota Division, September 1988.

(8) Roadway Lighting Handbook, U.S. Department of Transportation, Federal Highway Administration, Washington, D.C., December 1978.

Michael S. Griffith is a mathematical statistician in FHWA's Office of Safety and Traffic Operations Research and Development. At the Turner-Fairbank Highway Research Center in McLean, Va., for the past five years, Griffith has conducted research and managed research contract activities. He earned his bachelor's degree in business management from Ithaca College and his master's degree in statistics from State University of New York at Buffalo.