- Email: [email protected]
Night-time driving visibility associated with LED streetlight dimming
Accident Analysis and Prevention 121 (2018) 295–300
Contents lists available at ScienceDirect
Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap
Night-time driving visibility associated with LED streetlight dimming a,⁎
b
a
c
Joanne M. Wood , Gillian Isoardi , Alex Black , Ian Cowling
T
a
School of Optometry and Vision Science, Institute of Health and Biomedical Innovation, Queensland University of Technology, Brisbane, Queensland, Australia Light Naturally, Brisbane, Queensland, Australia School of Chemistry, Physics and Mechanical Engineering, Faculty of Science and Engineering Faculty, Queensland University of Technology, Brisbane, Queensland, Australia
b c
ARTICLE INFO
ABSTRACT
Keywords: LED streetlights Dimming Night-time visibility Driving performance
New LED streetlighting designs and dimming are being introduced worldwide, however, while their cost savings are well established, their impact on driving performance has received little attention. This study investigated the effect of streetlight dimming on night-time driving performance. Participants included 14 licensed drivers (mean age 34.2 ± 4.9 years, range 27–40 years) who drove an instrumented vehicle around a closed circuit at night. Six LED streetlights were positioned along a 250 m, straight section and their light output varied between laps (dimming levels of 25%, 50%, 75% and 100% of maximum output; L25, L50, L75 and L100 respectively; at 100% average road surface luminance of 1.14 cd/m2). Driving tasks involved recognition distances and reaction times to a low contrast, moving target and a pedestrian walking at the roadside. Participants drove at an average driving speed of 55 km/hr in the streetlight zone. Streetlight dimming significantly delayed driver reaction times to the moving target (F3,13.06 = 6.404; p = 0.007); with an average 0.4 s delay in reaction times under L25 compared to L100, (estimated reduction in recognition distances of 6 m). Pedestrian recognition distances were significantly shorter under dimmed streetlight levels (F3,12.75 = 8.27; p = 0.003); average pedestrian recognition distances were 15 m shorter under L25 compared to L100, and 11 m shorter under L50 compared to L100. These data suggest that streetlight dimming impacts on driver visibility but it is unclear how these differences impact on safety; future studies are required to inform decisions on safe dimming levels for road networks.
1. Introduction
were published prior to 1990, with the review authors stating that the risk of bias and susceptibility to confounders was considered high, given that none of the studies involved randomized control trials. In addition, none of the studies considered the safety aspects of newer streetlighting technologies, such as Light Emitting Diode (LED) luminaires. LED streetlighting is being introduced worldwide to take advantage of the high efficacy, long lifetime and cost effectiveness of LED light sources. It also provides the opportunity for further lighting infrastructure savings through dimming and adaptive light controls. However, while important data is accumulating regarding the energy savings associated with LED streetlights (Clinton Climate Initiative, 2009; Lockwood et al., 2011; Huang et al., 2012; Pipattanasomporn et al., 2014), only limited research has been undertaken on the drivingrelated human factors and safety aspects, particularly how dimming might impact on road user performance (Fotios and Gibbons, 2018). Two meta-analyses of crash data and associated streetlighting levels in the US (Gibbons et al., 2014) and UK (Steinbach et al., 2015), demonstrated the known benefits of streetlighting for reducing night-time
Night-time driving is dangerous, with fatality rates being up to three times higher than for daytime driving when adjusted for distances driven (NHTSA, 2007). These effects are even more pronounced for fatal crashes involving pedestrians, where night-time pedestrian fatality rates are up to seven times higher than those in the day (Sullivan and Flannagan, 2007) and of greater severity (Mohamed et al., 2013). Analyses of crash statistics indicate that reduced lighting and poor visibility are the primary factors associated with these high crash rates, rather than other factors that vary between day and night-time driving, such as driver fatigue and alcohol consumption (Owens and Sivak, 1996; Sullivan and Flannagan, 2002). This suggests that drivers are often unable to recognise and respond to pedestrians and cyclists at night until it is too late to avoid a collision (Rumar, 1990). Streetlighting is a cost-effective night-time road safety intervention. A Cochrane review involving seventeen studies demonstrated the capacity of streetlighting to reduce night-time crash risk (Beyer and Ker, 2009). Importantly, the majority of studies considered in this review
⁎
Corresponding author: School of Optometry and Vision Science, Queensland University of Technology, Victoria Park Rd, Kelvin Grove Q 4059, Australia. E-mail address: [email protected] (J.M. Wood).
https://doi.org/10.1016/j.aap.2018.08.023 Received 13 April 2018; Received in revised form 4 July 2018; Accepted 21 August 2018 0001-4575/ © 2018 Elsevier Ltd. All rights reserved.
Accident Analysis and Prevention 121 (2018) 295–300
J.M. Wood et al.
crash risk, but also highlighted that reducing the level of traditional streetlighting may not adversely impact safety. These studies suggest the potential for dimming and adaptive strategies, where lighting levels are adjusted based on the needs of roadway users. However, there is still a critical lack of evidence regarding the impact of LED streetlight configurations, dimming and adaptive capacity on both visual performance and road safety (Bullough and Radetsky, 2013). This information is essential to provide policy makers and road safety authorities with the information necessary to make evidence-based decisions on the levels of dimmed LED streetlighting that balance the potential for cost savings against the need to maintain road safety at night. In the current study we investigated the effect of four different levels of LED streetlighting (maximum output, and three levels of dimming) on the roadway lighting characteristics and light levels at a driver’s eye. We also investigated how different levels of LED streetlighting affected driver visibility performance, in terms of driver reaction times and pedestrian recognition distances, in a group of licensed drivers with normal vision.
been used in previous studies (Wood and Troutbeck, 1994; Tyrrell et al., 2004; Wood et al., 2005; Tyrrell et al., 2009; Wood et al., 2012). All testing occurred at least one hour after sunset and on nights during which there was no active precipitation and all road surfaces were dry. The circuit is representative of a rural road, and includes a series of hills, bends, curves, intersections, lengthy straight sections and standard road signs and lane markings; a 1.8 km section of the outside loop of the circuit that does not involve negotiating any intersections was used for this study. The use of a closed road circuit is considered a critical aspect of the validity of the study from a vision perspective, since the broad dynamic range of luminous conditions at the eye during night-time driving cannot be well reproduced in driving simulators (Wood and Chaparro, 2011). There are no existing streetlights on the circuit making it ideal for evaluating the effects of different streetlighting configurations, which were introduced onto the circuit specifically for this project. A schematic representation of the driving circuit as used during testing is shown in Fig. 1. Six LED streetlights (LED Roadway Lighting NXT 78 M), mounted on trailers, were erected on the left-hand lane of the long straight stretch of the circuit at 50 m intervals, leaving the two other lanes as a typical carriageway (7.5 m wide in total). The total distance between the first and last streetlight was 250 ms. The streetlights were mounted at a height of 11 m, with an outreach of 2 m. Measures of the luminance of the roadway, as well as the stationary and moving targets were made with a calibrated LMK 5 luminance measuring video photometer and a Konica Minolta LS-100 luminance meter. This demonstrated that when operated at maximum output (158 W at 100%), the streetlighting arrangement achieved a mean road luminance of 1.14 cd/m2 through the lit section of the circuit. The streetlight dimming levels were managed using wireless control of a single application network established through nodes connected to each streetlight. The network could be controlled and monitored through a digital interface on site, allowing all of the streetlights to be dimmed simultaneously. The light levels selected were (by power consumption and light output), 25%, 50%, 75% and 100% of the maximum output; identified as L25, L50, L75 and L100 respectively
2. Materials and methods 2.1. Participants Fourteen drivers (mean age 34.2 ± 4.9 years, range 27–40 years; 7 males, 7 females), were recruited via a number of different methods, including notices placed on university noticeboards (both electronic and physical), participation in previous studies and graduate students. All participants were licensed drivers and satisfied the minimum Australian drivers’ licensing criteria for binocular visual acuity of 6/12 (+0.30 logMAR) or better when wearing their habitual optical correction (if any). The study followed the tenets of the Declaration of Helsinki and was approved by the QUT Human Research Ethics Committee. All participants were given a full explanation of the nature and possible consequences of the study, and written informed consent was obtained with the option to withdraw from the study at any time. Participants completed a pre-experimental questionnaire that included questions on their general driving habits, as used in previous vision and driving studies (Owsley et al., 1999; Wood et al., 2005; Owens et al., 2007).
2.3.1. Target systems A low contrast (57% Weber contrast at L100) grey rectangular moving target (75 cm long and 25 cm high) was presented at 1.8 km/hr, with the movement of the target triggered by the vehicle passing through an infrared laser photosensor positioned 75 m before the target (Light Barrier 1; Fig. 1). To reduce the drivers’ expectation that this target would always move, a stationary target of the same configuration was positioned either before (run A) or after (run B) the moving target (Fig. 1), so that for each run the driver encountered two low contrast targets, of which only one moved. Two experimenters acted as pedestrians, walking in place on the opposite side of the road, to provide naturalistic pedestrian motion and ensure the safety of the pedestrian (as they were at known locations). Both pedestrians wore plain street clothing, which consisted of a light grey long-sleeve t-shirt along with a pair of light grey track pants. One pedestrian was located along the 250 m lit straight section of roadway (primary pedestrian). To reduce the drivers’ expectation that a person would always be in a single location, a second experimenter walked in place at a corner at the opposite side of the circuit (secondary pedestrian; Fig. 1). Recognition distances were not measured for the secondary pedestrian due to the limited sight distance available and because there were no streetlights at this location on the circuit. The participants drove around the circuit twice for each streetlight level. To further minimize driver expectations, the pedestrians were located in two different positions A or B for each of these laps, as indicated in Fig. 1.
2.2. Vision assessment Participants completed a short battery of standard vision tests, which were conducted binocularly while wearing their habitual driving optical correction. 2.2.1. Visual acuity Distance high contrast visual acuity was measured with the Early Treatment for Diabetic Retinopathy Study (ETDRS) chart at 5 m, at a luminance of 100 cd/m2, using the letter-by-letter scoring method. 2.2.2. Contrast sensitivity Letter contrast sensitivity was measured with the Pelli-Robson Contrast Sensitivity chart at 1 m at a luminance of 110 cd/m2, using the letter-by-letter scoring method (Elliott et al., 1991). A + 1.00 DS lens was used to compensate for the working distance. 2.2.3. Visual fields Binocular visual fields were measured using the Binocular Esterman program on the Humphrey Field Analyser (HFA), a computerized visual field testing instrument; the percentage of points seen out of 120 was recorded.
2.3.2. Experimental vehicle and light monitoring systems The experimental vehicle was an automatic transmission sedan (2015 Toyota Camry) with the halogen headlights set to low-beam for
2.3. Closed road circuit and test vehicle Testing was conducted at night on a closed-road circuit that has 296
Accident Analysis and Prevention 121 (2018) 295–300
J.M. Wood et al.
Fig. 1. Schematic and images of the driving circuit: The six streetlights on the long straight section are represented by open circles, the light barriers by open triangles, and the physical targets are represented by rectangles. Pedestrian positions around the circuit are also shown. Arrows indicate the direction of travel.
failed to read road signs at any point on the circuit. At the end of each lap, participants were asked to rate their level of agreement with the following statement: “The streetlighting on my last lap was effective in providing good driving visibility.” A five-option response scale was used: 1. Strongly Agree; 2. Agree; 3. Neither Agree or Disagree; 4. Disagree; 5. Strongly Disagree. Two driving visibility outcome measures are reported. The first is the distance on each lap at which the driver first recognised that a moving target or pedestrian was present. Response distances were coded as zero for trials where the driver failed to respond to the pedestrian or moving target or had passed the pedestrian or moving target before pressing the touchpad. The second variable is the driving reaction time to the detection of the moving target.
testing, to reflect general driving conditions. The vehicle was instrumented with four Konica Minolta illuminance sensors and a GPS logger. Two of the illuminance sensors were mounted on the roof of the vehicle to measure horizontal and vertical illuminance at the roof of the car. One small, high-resolution sensor was attached to a (lens free) spectacle frame and worn by the driver to measure illuminance at the eye and one was placed alongside the driver’s head to measure the vertical illuminance in the vehicle. The in-vehicle GPS logger allowed the location of the vehicle and vehicle speed to be recorded when a driver indicated that they detected the moving target. GPS position loggers were also stationed at the trigger and moving target positions. The exact locations by GPS coordinates of all pedestrian and stationary targets were also recorded prior to testing. The combination of the active logging GPS sensors and the known positions of all targets and pedestrians enabled the data logging system to identify the exact moment (and distance at which) drivers first recognised pedestrians, and the moving targets as they drove around the circuit; the accuracy of the system was within 1.2–1.8 m.
2.5. Statistical analysis For each of the driving outcome measures (driving speed, reaction time to the moving target, and recognition distances for the moving target and the primary pedestrian in locations A and B), a separate linear mixed model with repeated measures (using an unstructured covariance structure) was used to test for the effects of streetlight levels (4 levels) as a fixed-effect factor. Each model was adjusted for testing order as a fixed effect, to account for any potential learning effects. Post-hoc pairwise comparisons were conducted using a Bonferroni adjusted method. Statistical analyses were performed using SPSS version 23.0 (SPSS, Chicago, IL), and the level of significance was set at p < 0.05.
2.4. Procedures Each participant completed nine laps of the test circuit, including one practice lap and eight data collection laps. The purpose of the first lap (practice lap) was to familiarise the driver with both the test vehicle and the driving circuit. The data collection laps consisted of two laps for each of the four streetlight levels (L25, L50, L75, L100), with different locations of the pedestrian and stationary targets for each (lap A and B). The order of the streetlight levels was counterbalanced, as was the order of testing laps A and B. At the start of each data collection lap, the participant was instructed to drive along the specified route at a comfortable speed and press a large dash-mounted touch pad (and concurrently announce “person!”) as soon as they recognised that a person was present in the road scene ahead. They were also informed that there would be a number of other grey targets located around the circuit, and instructed to press the response pad as soon as they saw any of the targets move, as for the pedestrians. In an effort to increase driver workload, participants were also instructed to verbally report all road signs encountered around the circuit. Performance on this task was not recorded or analyzed, although participants were prompted if they
3. Results The mean habitual visual acuity of the 14 drivers was -0.14 ± 0.06 logMAR (more than one line better than 6/6 or 20/20 vision), and all had normal levels of Pelli-Robson contrast sensitivity (mean 1.97 ± 0.03 logCS). All participants had binocular visual fields that were within normal limits, which satisfied the visual licensing requirements for a private vehicle in Australia of a binocular field of 110 degrees horizontally, 10˚ above and below the horizontal. Drivers reported a group mean of 16.3 ± 5.0 years of driving experience and, in a typical week, drove an average of 63 ± 48 kms 297
Accident Analysis and Prevention 121 (2018) 295–300
J.M. Wood et al.
Table 1 Measured photometric data (luminance and illuminance) under the four streetlight levels. STREETLIGHT LEVEL L25 2
L50
L75
L100
Mean road luminance (cd/m )
0.25
0.59
0.83
1.14
% of L100 - road luminance
22%
52%
73%
100%
Comparable Australian Road Lighting Category by average carriageway luminance (standard required average luminance, cd/m2)*
< V5 (0.35)
V4 (0.5)
V3 (0.75)
V2 (1.0)
Illuminance at the eye, mean (SD) of all drivers through lighted section of road (lux)**
0.31 (0.13)
0.55 (0.23)
0.83 (0.34)
0.95 (0.42)
* From AS/NZS 1158.1.1:2005 - Lighting for roads and public spaces. ** The average illuminance at the eye recorded when driving through the lit segment of the road was calculated separately for each driver. The average value across all drivers is presented.
during the day and 27 ± 20 kms at night. Self-reported driving ability was rated to be “average” or better, and nearly all (91%) reported “a little” or “no difficulty” when driving at night. The lighting levels at the road surface and at the drivers’ eyes, provided by the six trailer-mounted LED lights, across the four streetlight conditions, are given in Table 1. The average speed at Light Barrier 1 (located 75 m before the moving target) was 55 km/hr (Table 2). There were no significant differences in driving speeds recorded along this section of the circuit across the four streetlighting conditions (F3,15.04 = 0.21; p = 0.89). There was a significant effect of streetlight dimming level on driver reaction times to the moving target (F3,13.06 = 6.404; p = 0.007, Table 2), where reaction times were faster at the higher light levels than at the lower levels. Post-hoc pairwise comparisons indicated that the differences in driving reaction times were only significant for the comparison between the highest L100 and the lowest light level L25 (0.4 s; p = 0.008), which represents a 21% increase in reaction time (2.21 s vs 1.82 s). Drivers were also able to detect the movement of the target at significantly longer distances with increasing streetlight levels (F3,13.48 = 16.1; p < 0.001). Post-hoc analysis indicated that the differences were significant between the L100 and L25 (7 m; p = 0.006) and L50 (5.6 m; p < 0.001) For the primary pedestrian at position A, higher streetlight levels were associated with the presence of the pedestrian being detected at significantly longer distances (F3,12.75 = 8.27; p = 0.003, Table 2; Fig. 2). Post-hoc analysis indicated that these differences reached significance for the comparison between the L100 and the two lowest light levels L25 (14.7 m; p = 0.002) and L50 (11.2 m; p = 0.015). For the primary pedestrian at position B, who was located outside the streetlight illumination, there were no significant differences between the pedestrian recognition distances across the four streetlight levels (F3,12.98 = 1.68; p = 0.22). When participants were asked about their perceptions of the
streetlight levels, it was clear that there were considerable differences between L25 and L50, and L75 and L100 (Fig. 3). Nearly 80% of drivers Agreed or Strongly Agreed with the statement that “The streetlighting on my last lap was effective in providing good driving visibility” for the two highest light levels (L75 and L100). In contrast, the percentage of participants who Agreed or Strongly Agreed with this statement was less than half for L50 (43%), and less than a quarter (21%) for the lowest light level L25. 4. Discussion In this study, we investigated the effect of four different dimming levels of LED streetlighting on the ability of drivers with normal vision to detect the movement of a target and the presence of a pedestrian walking at the roadside under night-time conditions. Overall, our findings demonstrate a small but significant delay in driver reaction times to the moving target and a reduction in pedestrian recognition distances under the lower levels of LED streetlighting. Overall, there were significant differences in the distances at which drivers could first recognise a moving, low contrast target and a pedestrian wearing non-retroreflective clothing between the highest level of LED streetlighting and dimming to 25% and 50%. There was also a 0.4 s increase in response time to the low contrast target (for streetlight dimming of 25% relative to 100%), which represents a 9 m increase in stopping distance for a vehicle travelling at 80 km/hr. Importantly, there were no significant differences in any of the driver performance measures when LED streetlighting was dimmed from 100% to 75% and no differences between the 25%, 50% and 75% levels. However, it is difficult to directly compare these reductions in driving visibility performance with previous studies, given the lack of research in this area. One recent study reported the detection and recognition distances of night-time bicyclists under full streetlight and no street illumination (Costa et al., 2017) but did not specifically explore
Table 2 Group mean (SD) of driving outcome measures across the four streetlight levels. STREETLIGHT LEVEL L25
L50
L75
L100
Test Statistic; p-value
Post-hoc Pairwise Comparisons (p < 0.05, Bonferroni)
Driving speed at Light Barrier 1 (km/hr)
55.1 (4.8)
55.1 (4.5)
54.6 (3.9)
54.9 (3.2)
F3,15.04 = 0.21; p = 0.89
–
Reaction Time to the moving target (s)
2.21 (0.55)
2.08 (0.56)
2.05 (0.63)
1.82 (0.45)
F3,13.06 = 6.40; p = 0.007
L100–L25
Moving Target recognition distance (m)
55.0 (10.1)
L100–L50, L100–L25
56.4 (6.3)
58.0 (11.0)
62.0 (5.0)
F3,13.48 = 16.1; p < 0.001
Primary Pedestrian (Location A) recognition distance (m) 46.1 (14.3)
49.6 (17.1)
52.7 (27.2)
60.8 (24.1)
F3,12.75 = 8.27; p = 0.003
L100–L50, L100–L25
Primary Pedestrian (Location B) recognition distance (m) 39.1 (21.5)
43.9 (24.5)
37.0 (23.0)
49.1 (22.7)
F3,12.98 = 1.68; p = 0.22
–
298
Accident Analysis and Prevention 121 (2018) 295–300
J.M. Wood et al.
Fig. 2. Group mean (SE) distances at which participants first recognised the presence of the primary pedestrian located at position A on the circuit as a function of streetlight level. Significant differences in the post-hoc pairwise comparisons (p < 0.05) indicated by the asterisks.
Fig. 3. Drivers’ responses to the statement “The streetlighting on my last lap was effective in providing good driving visibility”.
the effects of streetlight dimming. To some extent, our data can be compared to findings of a government report by Gibbons et al. (2015), which also showed that recognition distances for pedestrians wearing non-reflective clothing were shorter when LED streetlighting was dimmed by a factor of four (1.25 vs 5 lx); interestingly, these differences were more marked for LED compared to HPS streetlight dimming. An important consideration is whether these reductions in driving reaction times and recognition distances under the lower streetlighting conditions are likely to have a negative impact on night-time road safety. While the two meta-analyses of crash data and road lighting suggest that some reduction in streetlighting levels may not be problematic in terms of the number of collisions (Gibbons et al., 2014; Steinbach et al., 2015), they provide no information on associated changes in driving exposure for the different streetlight levels which is important to calculate collision rate data. In addition, evidence over a number of decades suggests that drivers don’t tend to reduce their speeds in response to the reduced visibility conditions of night-time driving (Leibowitz and Owens, 1977; Leibowitz et al., 1998; Jagerbrand
and Sjobergh, 2016) and if they do it is not sufficient to compensate for the reduced visibility (Owens et al., 2007). Our findings also support this suggestion, where there was no significant change in driving speeds, regardless of the level of streetlight dimming. Thus while reducing light levels to 50% output or lower may affect driving performance, this requires further exploration to determine the extent to which these differences impact on road safety at night. Interestingly, the differences in lighting levels afforded by the different streetlight outputs were recognised to some extent by the participants, where a quarter judged the lowest level to provide good driving visibility, with this percentage doubling to under half of participants for streetlight L50; almost 80% of all participants judged that both streetlight L75 and L100 provided good night-time visibility. These data suggest that while to some extent drivers’ perceptions of their own visibility ability reflects that of their actual ability, the differences in magnitude are different and they fail to recognise any change in performance between L100 and L75, as evidenced in Fig. 3. It is thus clear that drivers are aware to some extent of the impact of changing 299
Accident Analysis and Prevention 121 (2018) 295–300
J.M. Wood et al.
streetlight levels on their driving visibility in an experimental manipulation like this, however, whether that would be the case under normal road conditions at night is unclear. The impact of dimmed levels of streetlighting on night-time driving is particularly significant as more LED luminaires are being installed throughout the road networks, and decisions on energy-saving, ‘smart’ control systems and scheduled dimming are being made. The long lifetime of LED technologies means that they can gradually dim below acceptable light levels long before they ever noticeably fail completely. This will also inevitably lead to reductions in light output over extended periods of time with increasing maintenance cycles. Therefore it is important that efforts are made to determine the lowest light levels (that might occur because of dimming or lumen depreciation for these streetlight types) that still allow safe driving at night. While the size and demographic of the cohort reported here is too limited to allow major conclusions to be drawn on safe dimming levels for general road networks, the fact that we have been able to demonstrate statistically significant differences, indicates that additional work with a wider population of drivers is required to inform decisions regarding dimming levels. One particular group of interest are older drivers, given that they self-report night-time driving difficulties, particularly problems with glare and visibility under low light levels (Kimlin et al., 2016). In conclusion, this study demonstrated that there are small but significant reductions in both driver reaction times and pedestrian recognition distances when LED streetlighting operates at dimmed levels. Future research should explore these effects in a larger sample of drivers, including those of different ages and with a range of levels of visual performance that would be more representative of the true driving population. Other factors that need to be explored include the effects of driver speed on these relationships, as well as the effects of different LED spectral power distributions, colour temperatures and different weather conditions. These studies are critical in order to provide sufficient evidence to inform decisions on safe dimming levels for road networks.
59, 485–500. Elliott, D., Bullimore, M., Bailey, I., 1991. Improving the reliability of the Pelli-Robson contrast sensitivity test. Clin. Vision Sci. 6, 471–475. Fotios, S., Gibbons, R., 2018. Road lighting research for drivers and pedestrians: the basis of luminance and illuminance recommendations. Light. Res. Technol. 50, 154–186. Gibbons, R., Guo, F., Medina, A., Terry, T., Du, J., Lutkevich, P., Li, Q., 2014. Design criteria for adaptive roadway lighting. In: FHWA-HRT-14-051 ed. US Department of Transportation, Federal Highway Administration. Gibbons, R., Meyer, J., Terry, T., Bhagavathula, R., Lewis, A., Flanagan, M., Connell, C., 2015. Evaluation of the impact of spectral power distribution on driver performance. In: FHWA-HRT-15-047 ed. US Department of Transportation, Federal Highway Administration. Huang, S.C., Lee, L.L., Jeng, M.S., Hsieh, Y.C., 2012. Assessment of energy-efficient LED street lighting through large-scale demonstration. In: 2012 International Conference on Renewable Energy Research and Applications (ICRERA). Nagasaki. pp. 1–5. Jagerbrand, A.K., Sjobergh, J., 2016. Effects of weather conditions, light conditions, and road lighting on vehicle speed. Springerplus 5, 505. Kimlin, J.A., Black, A.A., Djaja, N., Wood, J.M., 2016. Development and validation of a vision and night driving questionnaire. Ophthalmic Physiol. Opt. 36 (4), 465–476. Leibowitz, H.W., Owens, D.A., 1977. Nighttime driving accidents and selective visual degradation. Science 197 (4302), 422–423. Leibowitz, H.W., Owens, D.A., Tyrrell, R.A., 1998. The assured clear distance ahead rule: implications for nighttime traffic safety and the law. Accid. Anal. Prev. 30 (1), 93–99. Lockwood, R., Selwyn, T., Morgan-Taylor, M., 2011. A review of local authority road lighting initiatives aimed at reducing costs, carbon emissions and light pollution: report for Defra. Temple Group, London. Mohamed, M.G., Saunier, N., Miranda-Moreno, L.F., Ukkusuri, S.V., 2013. A clustering regression approach: A comprehensive injury severity analysis of pedestrian-vehicle crashes in New York, US and Montreal, Canada. Saf. Sci. 54, 27–37. NHTSA, 2007. Traffic safety facts passenger vehicle occupant fatalities by day and night: A contrast (DOT HS 810 637). In: National Center for Statistics and Analysis, NHTSA ed. DOT HS 809923 ed. National Highway Traffic Safety Administration, US Department of Transportation, Washington, D.C. Owens, D.A., Sivak, M., 1996. Differentiation of visibility and alcohol as contributors to twilight road fatalities. Hum. Factors 38 (4), 680–689. Owens, D.A., Wood, J.M., Owens, J.M., 2007. Effects of age and illumination on night driving: a road test. Hum. Factors 49, 1115–1131. Owsley, C., Stalvey, B., Wells, J., Sloane, M.E., 1999. Older drivers and cataract: driving habits and crash risk. J. Gerontol.: Biol. Sci. Med. Sci. 54 (4), M203–M211. Pipattanasomporn, M., Rahman, S., Flory, I., Teklu, Y., 2014. Engineering design and assessment of a demand-sensitive LED streetlighting system. Sustain. Energy Technol. Assess. 7, 136–146. Rumar, K., 1990. The basic driver error: late detection. Ergonomics 32, 1281–1290. Steinbach, R., Perkins, C., Tompson, L., Johnson, S., Armstrong, B., Green, J., Grundy, C., Wilkinson, P., Edwards, P., 2015. The effect of reduced street lighting on road casualties and crime in England and Wales: Controlled interrupted time series analysis. J. Epidemiol. Commun. Health 69 (11), 1118–1124. Sullivan, J.M., Flannagan, M.J., 2002. The role of ambient light level in fatal crashes: inferences from daylight saving time transitions. Accid. Anal. Prev. 34 (4), 487–498. Sullivan, J.M., Flannagan, M.J., 2007. Determining the potential safety benefit of improved lighting in three pedestrian crash scenarios. Accid. Anal. Prev. 39 (3), 638–647. Tyrrell, R.A., Wood, J.M., Carberry, T.P., 2004. On-road measures of pedestrians’ estimates of their own nighttime conspicuity. J. Safety Res. 35 (5), 483–490. Tyrrell, R.A., Wood, J.M., Chaparro, A., Carberry, T.P., Chu, B.S., Marszalek, R.P., 2009. Seeing pedestrians at night: visual clutter does not mask biological motion. Accid. Anal. Prev. 41 (3), 506–512. Wood, J.M., Troutbeck, R., 1994. Effect of visual impairment on driving. Hum. Factors 36 (3), 476–487. Wood, J.M., Tyrrell, R.A., Carberry, T.P., 2005. Limitations in drivers’ ability to recognize pedestrians at night. Hum. Factors 47 (3), 644–653. Wood, J.M., Tyrrell, R.A., Chaparro, A., Marszalek, R.P., Carberry, T.P., Chu, B.S., 2012. Even moderate visual impairments degrade drivers’ ability to see pedestrians at night. Invest. Ophthalmol. Vis. Sci. 53 (6), 2586–2592. Wood, J.M., Chaparro, A., 2011. Night driving: How low illumination affects driving and the challenges of simulation. In: Fisher DL, Rizzo M., Caird JK, Lee JD ed. Handbook of Driving Simulation for Engineering, Medicine and Psychology. CRC Press/Taylor & Francis, pp. 28.1-28.12.
Acknowledgements The authors acknowledge the funding support of Transport and Main Roads and the cooperation of the staff of the Mt Cotton Driving Circuit in facilitating this research. We would also like to acknowledge the contribution of LED Roadway Lighting, in providing the luminaires and associated lighting control systems. The authors would also like to thank the QUT research teams and all of the participants who gave so generously of their time. References Beyer, F.R., Ker, K., 2009. Street lighting for preventing road traffic injuries. Cochrane Database Syst. Rev.(1) CD004728. Bullough, J.D., Radetsky, L.C., 2013. Analysis of New Highway Lighting Technologies. Lighting Research Center, Rensselaer Polytechnic Institute, Troy, NY. Clinton Climate Initiative, 2009. City of Los Angeles LED Street-lighting Case Study. Clinton Climate Initiative, Los Angeles. Costa, M., Bonetti, L., Bellelli, M., Lantieri, C., Vignali, V., Simone, A., 2017. Reflective tape applied to bicycle frame and conspicuity enhancement at night. Hum. Factors
300
Contents lists available at ScienceDirect
Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap
Night-time driving visibility associated with LED streetlight dimming a,⁎
b
a
c
Joanne M. Wood , Gillian Isoardi , Alex Black , Ian Cowling
T
a
School of Optometry and Vision Science, Institute of Health and Biomedical Innovation, Queensland University of Technology, Brisbane, Queensland, Australia Light Naturally, Brisbane, Queensland, Australia School of Chemistry, Physics and Mechanical Engineering, Faculty of Science and Engineering Faculty, Queensland University of Technology, Brisbane, Queensland, Australia
b c
ARTICLE INFO
ABSTRACT
Keywords: LED streetlights Dimming Night-time visibility Driving performance
New LED streetlighting designs and dimming are being introduced worldwide, however, while their cost savings are well established, their impact on driving performance has received little attention. This study investigated the effect of streetlight dimming on night-time driving performance. Participants included 14 licensed drivers (mean age 34.2 ± 4.9 years, range 27–40 years) who drove an instrumented vehicle around a closed circuit at night. Six LED streetlights were positioned along a 250 m, straight section and their light output varied between laps (dimming levels of 25%, 50%, 75% and 100% of maximum output; L25, L50, L75 and L100 respectively; at 100% average road surface luminance of 1.14 cd/m2). Driving tasks involved recognition distances and reaction times to a low contrast, moving target and a pedestrian walking at the roadside. Participants drove at an average driving speed of 55 km/hr in the streetlight zone. Streetlight dimming significantly delayed driver reaction times to the moving target (F3,13.06 = 6.404; p = 0.007); with an average 0.4 s delay in reaction times under L25 compared to L100, (estimated reduction in recognition distances of 6 m). Pedestrian recognition distances were significantly shorter under dimmed streetlight levels (F3,12.75 = 8.27; p = 0.003); average pedestrian recognition distances were 15 m shorter under L25 compared to L100, and 11 m shorter under L50 compared to L100. These data suggest that streetlight dimming impacts on driver visibility but it is unclear how these differences impact on safety; future studies are required to inform decisions on safe dimming levels for road networks.
1. Introduction
were published prior to 1990, with the review authors stating that the risk of bias and susceptibility to confounders was considered high, given that none of the studies involved randomized control trials. In addition, none of the studies considered the safety aspects of newer streetlighting technologies, such as Light Emitting Diode (LED) luminaires. LED streetlighting is being introduced worldwide to take advantage of the high efficacy, long lifetime and cost effectiveness of LED light sources. It also provides the opportunity for further lighting infrastructure savings through dimming and adaptive light controls. However, while important data is accumulating regarding the energy savings associated with LED streetlights (Clinton Climate Initiative, 2009; Lockwood et al., 2011; Huang et al., 2012; Pipattanasomporn et al., 2014), only limited research has been undertaken on the drivingrelated human factors and safety aspects, particularly how dimming might impact on road user performance (Fotios and Gibbons, 2018). Two meta-analyses of crash data and associated streetlighting levels in the US (Gibbons et al., 2014) and UK (Steinbach et al., 2015), demonstrated the known benefits of streetlighting for reducing night-time
Night-time driving is dangerous, with fatality rates being up to three times higher than for daytime driving when adjusted for distances driven (NHTSA, 2007). These effects are even more pronounced for fatal crashes involving pedestrians, where night-time pedestrian fatality rates are up to seven times higher than those in the day (Sullivan and Flannagan, 2007) and of greater severity (Mohamed et al., 2013). Analyses of crash statistics indicate that reduced lighting and poor visibility are the primary factors associated with these high crash rates, rather than other factors that vary between day and night-time driving, such as driver fatigue and alcohol consumption (Owens and Sivak, 1996; Sullivan and Flannagan, 2002). This suggests that drivers are often unable to recognise and respond to pedestrians and cyclists at night until it is too late to avoid a collision (Rumar, 1990). Streetlighting is a cost-effective night-time road safety intervention. A Cochrane review involving seventeen studies demonstrated the capacity of streetlighting to reduce night-time crash risk (Beyer and Ker, 2009). Importantly, the majority of studies considered in this review
⁎
Corresponding author: School of Optometry and Vision Science, Queensland University of Technology, Victoria Park Rd, Kelvin Grove Q 4059, Australia. E-mail address: [email protected] (J.M. Wood).
https://doi.org/10.1016/j.aap.2018.08.023 Received 13 April 2018; Received in revised form 4 July 2018; Accepted 21 August 2018 0001-4575/ © 2018 Elsevier Ltd. All rights reserved.
Accident Analysis and Prevention 121 (2018) 295–300
J.M. Wood et al.
crash risk, but also highlighted that reducing the level of traditional streetlighting may not adversely impact safety. These studies suggest the potential for dimming and adaptive strategies, where lighting levels are adjusted based on the needs of roadway users. However, there is still a critical lack of evidence regarding the impact of LED streetlight configurations, dimming and adaptive capacity on both visual performance and road safety (Bullough and Radetsky, 2013). This information is essential to provide policy makers and road safety authorities with the information necessary to make evidence-based decisions on the levels of dimmed LED streetlighting that balance the potential for cost savings against the need to maintain road safety at night. In the current study we investigated the effect of four different levels of LED streetlighting (maximum output, and three levels of dimming) on the roadway lighting characteristics and light levels at a driver’s eye. We also investigated how different levels of LED streetlighting affected driver visibility performance, in terms of driver reaction times and pedestrian recognition distances, in a group of licensed drivers with normal vision.
been used in previous studies (Wood and Troutbeck, 1994; Tyrrell et al., 2004; Wood et al., 2005; Tyrrell et al., 2009; Wood et al., 2012). All testing occurred at least one hour after sunset and on nights during which there was no active precipitation and all road surfaces were dry. The circuit is representative of a rural road, and includes a series of hills, bends, curves, intersections, lengthy straight sections and standard road signs and lane markings; a 1.8 km section of the outside loop of the circuit that does not involve negotiating any intersections was used for this study. The use of a closed road circuit is considered a critical aspect of the validity of the study from a vision perspective, since the broad dynamic range of luminous conditions at the eye during night-time driving cannot be well reproduced in driving simulators (Wood and Chaparro, 2011). There are no existing streetlights on the circuit making it ideal for evaluating the effects of different streetlighting configurations, which were introduced onto the circuit specifically for this project. A schematic representation of the driving circuit as used during testing is shown in Fig. 1. Six LED streetlights (LED Roadway Lighting NXT 78 M), mounted on trailers, were erected on the left-hand lane of the long straight stretch of the circuit at 50 m intervals, leaving the two other lanes as a typical carriageway (7.5 m wide in total). The total distance between the first and last streetlight was 250 ms. The streetlights were mounted at a height of 11 m, with an outreach of 2 m. Measures of the luminance of the roadway, as well as the stationary and moving targets were made with a calibrated LMK 5 luminance measuring video photometer and a Konica Minolta LS-100 luminance meter. This demonstrated that when operated at maximum output (158 W at 100%), the streetlighting arrangement achieved a mean road luminance of 1.14 cd/m2 through the lit section of the circuit. The streetlight dimming levels were managed using wireless control of a single application network established through nodes connected to each streetlight. The network could be controlled and monitored through a digital interface on site, allowing all of the streetlights to be dimmed simultaneously. The light levels selected were (by power consumption and light output), 25%, 50%, 75% and 100% of the maximum output; identified as L25, L50, L75 and L100 respectively
2. Materials and methods 2.1. Participants Fourteen drivers (mean age 34.2 ± 4.9 years, range 27–40 years; 7 males, 7 females), were recruited via a number of different methods, including notices placed on university noticeboards (both electronic and physical), participation in previous studies and graduate students. All participants were licensed drivers and satisfied the minimum Australian drivers’ licensing criteria for binocular visual acuity of 6/12 (+0.30 logMAR) or better when wearing their habitual optical correction (if any). The study followed the tenets of the Declaration of Helsinki and was approved by the QUT Human Research Ethics Committee. All participants were given a full explanation of the nature and possible consequences of the study, and written informed consent was obtained with the option to withdraw from the study at any time. Participants completed a pre-experimental questionnaire that included questions on their general driving habits, as used in previous vision and driving studies (Owsley et al., 1999; Wood et al., 2005; Owens et al., 2007).
2.3.1. Target systems A low contrast (57% Weber contrast at L100) grey rectangular moving target (75 cm long and 25 cm high) was presented at 1.8 km/hr, with the movement of the target triggered by the vehicle passing through an infrared laser photosensor positioned 75 m before the target (Light Barrier 1; Fig. 1). To reduce the drivers’ expectation that this target would always move, a stationary target of the same configuration was positioned either before (run A) or after (run B) the moving target (Fig. 1), so that for each run the driver encountered two low contrast targets, of which only one moved. Two experimenters acted as pedestrians, walking in place on the opposite side of the road, to provide naturalistic pedestrian motion and ensure the safety of the pedestrian (as they were at known locations). Both pedestrians wore plain street clothing, which consisted of a light grey long-sleeve t-shirt along with a pair of light grey track pants. One pedestrian was located along the 250 m lit straight section of roadway (primary pedestrian). To reduce the drivers’ expectation that a person would always be in a single location, a second experimenter walked in place at a corner at the opposite side of the circuit (secondary pedestrian; Fig. 1). Recognition distances were not measured for the secondary pedestrian due to the limited sight distance available and because there were no streetlights at this location on the circuit. The participants drove around the circuit twice for each streetlight level. To further minimize driver expectations, the pedestrians were located in two different positions A or B for each of these laps, as indicated in Fig. 1.
2.2. Vision assessment Participants completed a short battery of standard vision tests, which were conducted binocularly while wearing their habitual driving optical correction. 2.2.1. Visual acuity Distance high contrast visual acuity was measured with the Early Treatment for Diabetic Retinopathy Study (ETDRS) chart at 5 m, at a luminance of 100 cd/m2, using the letter-by-letter scoring method. 2.2.2. Contrast sensitivity Letter contrast sensitivity was measured with the Pelli-Robson Contrast Sensitivity chart at 1 m at a luminance of 110 cd/m2, using the letter-by-letter scoring method (Elliott et al., 1991). A + 1.00 DS lens was used to compensate for the working distance. 2.2.3. Visual fields Binocular visual fields were measured using the Binocular Esterman program on the Humphrey Field Analyser (HFA), a computerized visual field testing instrument; the percentage of points seen out of 120 was recorded.
2.3.2. Experimental vehicle and light monitoring systems The experimental vehicle was an automatic transmission sedan (2015 Toyota Camry) with the halogen headlights set to low-beam for
2.3. Closed road circuit and test vehicle Testing was conducted at night on a closed-road circuit that has 296
Accident Analysis and Prevention 121 (2018) 295–300
J.M. Wood et al.
Fig. 1. Schematic and images of the driving circuit: The six streetlights on the long straight section are represented by open circles, the light barriers by open triangles, and the physical targets are represented by rectangles. Pedestrian positions around the circuit are also shown. Arrows indicate the direction of travel.
failed to read road signs at any point on the circuit. At the end of each lap, participants were asked to rate their level of agreement with the following statement: “The streetlighting on my last lap was effective in providing good driving visibility.” A five-option response scale was used: 1. Strongly Agree; 2. Agree; 3. Neither Agree or Disagree; 4. Disagree; 5. Strongly Disagree. Two driving visibility outcome measures are reported. The first is the distance on each lap at which the driver first recognised that a moving target or pedestrian was present. Response distances were coded as zero for trials where the driver failed to respond to the pedestrian or moving target or had passed the pedestrian or moving target before pressing the touchpad. The second variable is the driving reaction time to the detection of the moving target.
testing, to reflect general driving conditions. The vehicle was instrumented with four Konica Minolta illuminance sensors and a GPS logger. Two of the illuminance sensors were mounted on the roof of the vehicle to measure horizontal and vertical illuminance at the roof of the car. One small, high-resolution sensor was attached to a (lens free) spectacle frame and worn by the driver to measure illuminance at the eye and one was placed alongside the driver’s head to measure the vertical illuminance in the vehicle. The in-vehicle GPS logger allowed the location of the vehicle and vehicle speed to be recorded when a driver indicated that they detected the moving target. GPS position loggers were also stationed at the trigger and moving target positions. The exact locations by GPS coordinates of all pedestrian and stationary targets were also recorded prior to testing. The combination of the active logging GPS sensors and the known positions of all targets and pedestrians enabled the data logging system to identify the exact moment (and distance at which) drivers first recognised pedestrians, and the moving targets as they drove around the circuit; the accuracy of the system was within 1.2–1.8 m.
2.5. Statistical analysis For each of the driving outcome measures (driving speed, reaction time to the moving target, and recognition distances for the moving target and the primary pedestrian in locations A and B), a separate linear mixed model with repeated measures (using an unstructured covariance structure) was used to test for the effects of streetlight levels (4 levels) as a fixed-effect factor. Each model was adjusted for testing order as a fixed effect, to account for any potential learning effects. Post-hoc pairwise comparisons were conducted using a Bonferroni adjusted method. Statistical analyses were performed using SPSS version 23.0 (SPSS, Chicago, IL), and the level of significance was set at p < 0.05.
2.4. Procedures Each participant completed nine laps of the test circuit, including one practice lap and eight data collection laps. The purpose of the first lap (practice lap) was to familiarise the driver with both the test vehicle and the driving circuit. The data collection laps consisted of two laps for each of the four streetlight levels (L25, L50, L75, L100), with different locations of the pedestrian and stationary targets for each (lap A and B). The order of the streetlight levels was counterbalanced, as was the order of testing laps A and B. At the start of each data collection lap, the participant was instructed to drive along the specified route at a comfortable speed and press a large dash-mounted touch pad (and concurrently announce “person!”) as soon as they recognised that a person was present in the road scene ahead. They were also informed that there would be a number of other grey targets located around the circuit, and instructed to press the response pad as soon as they saw any of the targets move, as for the pedestrians. In an effort to increase driver workload, participants were also instructed to verbally report all road signs encountered around the circuit. Performance on this task was not recorded or analyzed, although participants were prompted if they
3. Results The mean habitual visual acuity of the 14 drivers was -0.14 ± 0.06 logMAR (more than one line better than 6/6 or 20/20 vision), and all had normal levels of Pelli-Robson contrast sensitivity (mean 1.97 ± 0.03 logCS). All participants had binocular visual fields that were within normal limits, which satisfied the visual licensing requirements for a private vehicle in Australia of a binocular field of 110 degrees horizontally, 10˚ above and below the horizontal. Drivers reported a group mean of 16.3 ± 5.0 years of driving experience and, in a typical week, drove an average of 63 ± 48 kms 297
Accident Analysis and Prevention 121 (2018) 295–300
J.M. Wood et al.
Table 1 Measured photometric data (luminance and illuminance) under the four streetlight levels. STREETLIGHT LEVEL L25 2
L50
L75
L100
Mean road luminance (cd/m )
0.25
0.59
0.83
1.14
% of L100 - road luminance
22%
52%
73%
100%
Comparable Australian Road Lighting Category by average carriageway luminance (standard required average luminance, cd/m2)*
< V5 (0.35)
V4 (0.5)
V3 (0.75)
V2 (1.0)
Illuminance at the eye, mean (SD) of all drivers through lighted section of road (lux)**
0.31 (0.13)
0.55 (0.23)
0.83 (0.34)
0.95 (0.42)
* From AS/NZS 1158.1.1:2005 - Lighting for roads and public spaces. ** The average illuminance at the eye recorded when driving through the lit segment of the road was calculated separately for each driver. The average value across all drivers is presented.
during the day and 27 ± 20 kms at night. Self-reported driving ability was rated to be “average” or better, and nearly all (91%) reported “a little” or “no difficulty” when driving at night. The lighting levels at the road surface and at the drivers’ eyes, provided by the six trailer-mounted LED lights, across the four streetlight conditions, are given in Table 1. The average speed at Light Barrier 1 (located 75 m before the moving target) was 55 km/hr (Table 2). There were no significant differences in driving speeds recorded along this section of the circuit across the four streetlighting conditions (F3,15.04 = 0.21; p = 0.89). There was a significant effect of streetlight dimming level on driver reaction times to the moving target (F3,13.06 = 6.404; p = 0.007, Table 2), where reaction times were faster at the higher light levels than at the lower levels. Post-hoc pairwise comparisons indicated that the differences in driving reaction times were only significant for the comparison between the highest L100 and the lowest light level L25 (0.4 s; p = 0.008), which represents a 21% increase in reaction time (2.21 s vs 1.82 s). Drivers were also able to detect the movement of the target at significantly longer distances with increasing streetlight levels (F3,13.48 = 16.1; p < 0.001). Post-hoc analysis indicated that the differences were significant between the L100 and L25 (7 m; p = 0.006) and L50 (5.6 m; p < 0.001) For the primary pedestrian at position A, higher streetlight levels were associated with the presence of the pedestrian being detected at significantly longer distances (F3,12.75 = 8.27; p = 0.003, Table 2; Fig. 2). Post-hoc analysis indicated that these differences reached significance for the comparison between the L100 and the two lowest light levels L25 (14.7 m; p = 0.002) and L50 (11.2 m; p = 0.015). For the primary pedestrian at position B, who was located outside the streetlight illumination, there were no significant differences between the pedestrian recognition distances across the four streetlight levels (F3,12.98 = 1.68; p = 0.22). When participants were asked about their perceptions of the
streetlight levels, it was clear that there were considerable differences between L25 and L50, and L75 and L100 (Fig. 3). Nearly 80% of drivers Agreed or Strongly Agreed with the statement that “The streetlighting on my last lap was effective in providing good driving visibility” for the two highest light levels (L75 and L100). In contrast, the percentage of participants who Agreed or Strongly Agreed with this statement was less than half for L50 (43%), and less than a quarter (21%) for the lowest light level L25. 4. Discussion In this study, we investigated the effect of four different dimming levels of LED streetlighting on the ability of drivers with normal vision to detect the movement of a target and the presence of a pedestrian walking at the roadside under night-time conditions. Overall, our findings demonstrate a small but significant delay in driver reaction times to the moving target and a reduction in pedestrian recognition distances under the lower levels of LED streetlighting. Overall, there were significant differences in the distances at which drivers could first recognise a moving, low contrast target and a pedestrian wearing non-retroreflective clothing between the highest level of LED streetlighting and dimming to 25% and 50%. There was also a 0.4 s increase in response time to the low contrast target (for streetlight dimming of 25% relative to 100%), which represents a 9 m increase in stopping distance for a vehicle travelling at 80 km/hr. Importantly, there were no significant differences in any of the driver performance measures when LED streetlighting was dimmed from 100% to 75% and no differences between the 25%, 50% and 75% levels. However, it is difficult to directly compare these reductions in driving visibility performance with previous studies, given the lack of research in this area. One recent study reported the detection and recognition distances of night-time bicyclists under full streetlight and no street illumination (Costa et al., 2017) but did not specifically explore
Table 2 Group mean (SD) of driving outcome measures across the four streetlight levels. STREETLIGHT LEVEL L25
L50
L75
L100
Test Statistic; p-value
Post-hoc Pairwise Comparisons (p < 0.05, Bonferroni)
Driving speed at Light Barrier 1 (km/hr)
55.1 (4.8)
55.1 (4.5)
54.6 (3.9)
54.9 (3.2)
F3,15.04 = 0.21; p = 0.89
–
Reaction Time to the moving target (s)
2.21 (0.55)
2.08 (0.56)
2.05 (0.63)
1.82 (0.45)
F3,13.06 = 6.40; p = 0.007
L100–L25
Moving Target recognition distance (m)
55.0 (10.1)
L100–L50, L100–L25
56.4 (6.3)
58.0 (11.0)
62.0 (5.0)
F3,13.48 = 16.1; p < 0.001
Primary Pedestrian (Location A) recognition distance (m) 46.1 (14.3)
49.6 (17.1)
52.7 (27.2)
60.8 (24.1)
F3,12.75 = 8.27; p = 0.003
L100–L50, L100–L25
Primary Pedestrian (Location B) recognition distance (m) 39.1 (21.5)
43.9 (24.5)
37.0 (23.0)
49.1 (22.7)
F3,12.98 = 1.68; p = 0.22
–
298
Accident Analysis and Prevention 121 (2018) 295–300
J.M. Wood et al.
Fig. 2. Group mean (SE) distances at which participants first recognised the presence of the primary pedestrian located at position A on the circuit as a function of streetlight level. Significant differences in the post-hoc pairwise comparisons (p < 0.05) indicated by the asterisks.
Fig. 3. Drivers’ responses to the statement “The streetlighting on my last lap was effective in providing good driving visibility”.
the effects of streetlight dimming. To some extent, our data can be compared to findings of a government report by Gibbons et al. (2015), which also showed that recognition distances for pedestrians wearing non-reflective clothing were shorter when LED streetlighting was dimmed by a factor of four (1.25 vs 5 lx); interestingly, these differences were more marked for LED compared to HPS streetlight dimming. An important consideration is whether these reductions in driving reaction times and recognition distances under the lower streetlighting conditions are likely to have a negative impact on night-time road safety. While the two meta-analyses of crash data and road lighting suggest that some reduction in streetlighting levels may not be problematic in terms of the number of collisions (Gibbons et al., 2014; Steinbach et al., 2015), they provide no information on associated changes in driving exposure for the different streetlight levels which is important to calculate collision rate data. In addition, evidence over a number of decades suggests that drivers don’t tend to reduce their speeds in response to the reduced visibility conditions of night-time driving (Leibowitz and Owens, 1977; Leibowitz et al., 1998; Jagerbrand
and Sjobergh, 2016) and if they do it is not sufficient to compensate for the reduced visibility (Owens et al., 2007). Our findings also support this suggestion, where there was no significant change in driving speeds, regardless of the level of streetlight dimming. Thus while reducing light levels to 50% output or lower may affect driving performance, this requires further exploration to determine the extent to which these differences impact on road safety at night. Interestingly, the differences in lighting levels afforded by the different streetlight outputs were recognised to some extent by the participants, where a quarter judged the lowest level to provide good driving visibility, with this percentage doubling to under half of participants for streetlight L50; almost 80% of all participants judged that both streetlight L75 and L100 provided good night-time visibility. These data suggest that while to some extent drivers’ perceptions of their own visibility ability reflects that of their actual ability, the differences in magnitude are different and they fail to recognise any change in performance between L100 and L75, as evidenced in Fig. 3. It is thus clear that drivers are aware to some extent of the impact of changing 299
Accident Analysis and Prevention 121 (2018) 295–300
J.M. Wood et al.
streetlight levels on their driving visibility in an experimental manipulation like this, however, whether that would be the case under normal road conditions at night is unclear. The impact of dimmed levels of streetlighting on night-time driving is particularly significant as more LED luminaires are being installed throughout the road networks, and decisions on energy-saving, ‘smart’ control systems and scheduled dimming are being made. The long lifetime of LED technologies means that they can gradually dim below acceptable light levels long before they ever noticeably fail completely. This will also inevitably lead to reductions in light output over extended periods of time with increasing maintenance cycles. Therefore it is important that efforts are made to determine the lowest light levels (that might occur because of dimming or lumen depreciation for these streetlight types) that still allow safe driving at night. While the size and demographic of the cohort reported here is too limited to allow major conclusions to be drawn on safe dimming levels for general road networks, the fact that we have been able to demonstrate statistically significant differences, indicates that additional work with a wider population of drivers is required to inform decisions regarding dimming levels. One particular group of interest are older drivers, given that they self-report night-time driving difficulties, particularly problems with glare and visibility under low light levels (Kimlin et al., 2016). In conclusion, this study demonstrated that there are small but significant reductions in both driver reaction times and pedestrian recognition distances when LED streetlighting operates at dimmed levels. Future research should explore these effects in a larger sample of drivers, including those of different ages and with a range of levels of visual performance that would be more representative of the true driving population. Other factors that need to be explored include the effects of driver speed on these relationships, as well as the effects of different LED spectral power distributions, colour temperatures and different weather conditions. These studies are critical in order to provide sufficient evidence to inform decisions on safe dimming levels for road networks.
59, 485–500. Elliott, D., Bullimore, M., Bailey, I., 1991. Improving the reliability of the Pelli-Robson contrast sensitivity test. Clin. Vision Sci. 6, 471–475. Fotios, S., Gibbons, R., 2018. Road lighting research for drivers and pedestrians: the basis of luminance and illuminance recommendations. Light. Res. Technol. 50, 154–186. Gibbons, R., Guo, F., Medina, A., Terry, T., Du, J., Lutkevich, P., Li, Q., 2014. Design criteria for adaptive roadway lighting. In: FHWA-HRT-14-051 ed. US Department of Transportation, Federal Highway Administration. Gibbons, R., Meyer, J., Terry, T., Bhagavathula, R., Lewis, A., Flanagan, M., Connell, C., 2015. Evaluation of the impact of spectral power distribution on driver performance. In: FHWA-HRT-15-047 ed. US Department of Transportation, Federal Highway Administration. Huang, S.C., Lee, L.L., Jeng, M.S., Hsieh, Y.C., 2012. Assessment of energy-efficient LED street lighting through large-scale demonstration. In: 2012 International Conference on Renewable Energy Research and Applications (ICRERA). Nagasaki. pp. 1–5. Jagerbrand, A.K., Sjobergh, J., 2016. Effects of weather conditions, light conditions, and road lighting on vehicle speed. Springerplus 5, 505. Kimlin, J.A., Black, A.A., Djaja, N., Wood, J.M., 2016. Development and validation of a vision and night driving questionnaire. Ophthalmic Physiol. Opt. 36 (4), 465–476. Leibowitz, H.W., Owens, D.A., 1977. Nighttime driving accidents and selective visual degradation. Science 197 (4302), 422–423. Leibowitz, H.W., Owens, D.A., Tyrrell, R.A., 1998. The assured clear distance ahead rule: implications for nighttime traffic safety and the law. Accid. Anal. Prev. 30 (1), 93–99. Lockwood, R., Selwyn, T., Morgan-Taylor, M., 2011. A review of local authority road lighting initiatives aimed at reducing costs, carbon emissions and light pollution: report for Defra. Temple Group, London. Mohamed, M.G., Saunier, N., Miranda-Moreno, L.F., Ukkusuri, S.V., 2013. A clustering regression approach: A comprehensive injury severity analysis of pedestrian-vehicle crashes in New York, US and Montreal, Canada. Saf. Sci. 54, 27–37. NHTSA, 2007. Traffic safety facts passenger vehicle occupant fatalities by day and night: A contrast (DOT HS 810 637). In: National Center for Statistics and Analysis, NHTSA ed. DOT HS 809923 ed. National Highway Traffic Safety Administration, US Department of Transportation, Washington, D.C. Owens, D.A., Sivak, M., 1996. Differentiation of visibility and alcohol as contributors to twilight road fatalities. Hum. Factors 38 (4), 680–689. Owens, D.A., Wood, J.M., Owens, J.M., 2007. Effects of age and illumination on night driving: a road test. Hum. Factors 49, 1115–1131. Owsley, C., Stalvey, B., Wells, J., Sloane, M.E., 1999. Older drivers and cataract: driving habits and crash risk. J. Gerontol.: Biol. Sci. Med. Sci. 54 (4), M203–M211. Pipattanasomporn, M., Rahman, S., Flory, I., Teklu, Y., 2014. Engineering design and assessment of a demand-sensitive LED streetlighting system. Sustain. Energy Technol. Assess. 7, 136–146. Rumar, K., 1990. The basic driver error: late detection. Ergonomics 32, 1281–1290. Steinbach, R., Perkins, C., Tompson, L., Johnson, S., Armstrong, B., Green, J., Grundy, C., Wilkinson, P., Edwards, P., 2015. The effect of reduced street lighting on road casualties and crime in England and Wales: Controlled interrupted time series analysis. J. Epidemiol. Commun. Health 69 (11), 1118–1124. Sullivan, J.M., Flannagan, M.J., 2002. The role of ambient light level in fatal crashes: inferences from daylight saving time transitions. Accid. Anal. Prev. 34 (4), 487–498. Sullivan, J.M., Flannagan, M.J., 2007. Determining the potential safety benefit of improved lighting in three pedestrian crash scenarios. Accid. Anal. Prev. 39 (3), 638–647. Tyrrell, R.A., Wood, J.M., Carberry, T.P., 2004. On-road measures of pedestrians’ estimates of their own nighttime conspicuity. J. Safety Res. 35 (5), 483–490. Tyrrell, R.A., Wood, J.M., Chaparro, A., Carberry, T.P., Chu, B.S., Marszalek, R.P., 2009. Seeing pedestrians at night: visual clutter does not mask biological motion. Accid. Anal. Prev. 41 (3), 506–512. Wood, J.M., Troutbeck, R., 1994. Effect of visual impairment on driving. Hum. Factors 36 (3), 476–487. Wood, J.M., Tyrrell, R.A., Carberry, T.P., 2005. Limitations in drivers’ ability to recognize pedestrians at night. Hum. Factors 47 (3), 644–653. Wood, J.M., Tyrrell, R.A., Chaparro, A., Marszalek, R.P., Carberry, T.P., Chu, B.S., 2012. Even moderate visual impairments degrade drivers’ ability to see pedestrians at night. Invest. Ophthalmol. Vis. Sci. 53 (6), 2586–2592. Wood, J.M., Chaparro, A., 2011. Night driving: How low illumination affects driving and the challenges of simulation. In: Fisher DL, Rizzo M., Caird JK, Lee JD ed. Handbook of Driving Simulation for Engineering, Medicine and Psychology. CRC Press/Taylor & Francis, pp. 28.1-28.12.
Acknowledgements The authors acknowledge the funding support of Transport and Main Roads and the cooperation of the staff of the Mt Cotton Driving Circuit in facilitating this research. We would also like to acknowledge the contribution of LED Roadway Lighting, in providing the luminaires and associated lighting control systems. The authors would also like to thank the QUT research teams and all of the participants who gave so generously of their time. References Beyer, F.R., Ker, K., 2009. Street lighting for preventing road traffic injuries. Cochrane Database Syst. Rev.(1) CD004728. Bullough, J.D., Radetsky, L.C., 2013. Analysis of New Highway Lighting Technologies. Lighting Research Center, Rensselaer Polytechnic Institute, Troy, NY. Clinton Climate Initiative, 2009. City of Los Angeles LED Street-lighting Case Study. Clinton Climate Initiative, Los Angeles. Costa, M., Bonetti, L., Bellelli, M., Lantieri, C., Vignali, V., Simone, A., 2017. Reflective tape applied to bicycle frame and conspicuity enhancement at night. Hum. Factors
300