You are on page 1of 34

Marine Corps Base Hawaii: Alternative Lighting

Evaluation

Final Report
Submitted by
Virginia Tech Transportation Institute
3500 Transportation Research Plaza
Blacksburg, VA 24061

December 2009
Marine Corps Base Hawaii: Alternative Lighting Evaluation

REPORT

Dr. Ronald B. Gibbons


Christopher J. Edwards
Derek Viita
and
Jason Meyer
VTTI

December 2009

ii
TABLE OF CONTENTS
EXECUTIVE SUMMARY ................................................................................................... VII
INTRODUCTION .....................................................................................................................1
ASSESSMENT OF ALTERNATE LIGHTING ....................................................................................1

RESEARCH OBJECTIVES ............................................................................................................1

ASSESSMENT OF ALTERNATIVE LIGHTING TECHNOLOGIES .................................2


EXPERIMENTAL DESIGN............................................................................................................2
PARTICIPANTS ..........................................................................................................................2

EQUIPMENT ..............................................................................................................................2
Roadway Lighting Mobile Measurement System ...................................................................3
PROCEDURE .............................................................................................................................4
RESULTS ..................................................................................................................................7
Detection Distance ...............................................................................................................7
Recognition Distance ......................................................................................................... 12

Horizontal Roadway Illuminance ....................................................................................... 16


DISCUSSION ........................................................................................................................... 20
Detection Distance ............................................................................................................. 20
Recognition Distance ......................................................................................................... 23
Horizontal Roadway illuminance ....................................................................................... 23
CONCLUSIONS ........................................................................................................................ 25

LIMITATIONS .......................................................................................................................... 26
FUTURE CONSIDERATIONS AND IMPROVEMENTS...................................................................... 26
REFERENCES .......................................................................................................................... 27

iii
LIST OF FIGURES
Figure 1. Experimental vehicle with RLMMS components. .........................................................4
Figure 2. RLMMS components mounted on and inside a vehicle.................................................4
Figure 3. Overhead View of the route with Luminaires designated. .............................................5
Figure 4. Examples of large and small colored targets placed along the experimental route. ........6
Figure 5. Significant 3-way interaction among target color, target size, and target illuminance
level. ....................................................................................................................................8
Figure 6. Mean detection distances for target color and size. ..................................................... 10
Figure 7. Mean detection distances for each target color and target illuminance level. ............... 11
Figure 8. Mean detection distances for each target size and target illuminance level. ................. 12
Figure 9. Significant 3-way interaction among target color, target size, and target illuminance
level. .................................................................................................................................. 13
Figure 10. Mean recognition distances for target color and size. ................................................ 14
Figure 11. Mean recognition distances for each target color and target illuminance level. .......... 15
Figure 12. Mean recognition distances for each target size and target illuminance level............. 16
Figure 13. Sensor locations for Horizontal Roadway illuminance analysis. ................................ 17
Figure 15. Glare levels for entire road length. ............................................................................ 20
Figure 16. Detection distance comparison for identical small target colors at different sites....... 22
Figure 17. Detection distance comparisons between locations and alternative lighting............... 23
Figure 18. Mean Illuminance comparisons across testing sites. .................................................. 24
Figure 19. Mean glare comparisons across testing sites. ............................................................ 25

iv
LIST OF TABLES
Table 1. Experimental Variables and Descriptions.......................................................................2
Table 2. Target Characteristics of ANOVA Results (Detection Distances) ..................................7
Table 3. Target Characteristics of ANOVA Results (Recognition Distances)............................. 12
Table 4. Horizontal roadway illuminance Metrics ..................................................................... 18

v
EXECUTIVE SUMMARY
Marine Corps Base Hawaii (MCBH) in Kaneohe Bay, Hawaii recently retrofitted existing street
lighting with a new efficient fluorescent technology in an effort to reduce power consumption
and save on energy and maintenance costs. The U.S. Navy had previously explored other
alternative lighting technologies such as light-emitting diodes (LEDs), induction, and ceramic
metal halide lighting to provide solutions to energy consumption. The final choice by the MCBH
was the fluorescent, and retrofitting was completed in July 2009. The lighting choice achieved
the energy and maintenance costs as vetted in previous testing; however, the street lighting
impact on drivers on the base premises was not known. Magnaray® (the seller of the final
product) contacted the Virginia Tech Transportation Institute (VTTI) in order to evaluate the
street lighting impact on drivers. VTTI measured driver visibility through target detection and
recognition tasks. Photometric measurements were also collected during each run. The data were
gathered through a roadway lighting mobile measurement system (RLMMS) that collected
visibility detection data in addition to illuminance data. The data were then cleaned and
analyzed. The analysis compared the detection distance of target types (large, small), target color
(four colors), lighting level (high/low), speed, and recognition. It also compared each section of
the alternate lighting to one another and also to the control section for target detection distances.
In addition, the photometric analysis compared illuminance measurements throughout each
section and compared them across sections. The final detection distance results suggest that the
alternate lighting did not perform as well as the control section; however, additional research is
required in order to justify lowering the light level from the control installation. The illuminance
results follow a similar pattern as reflected in the detection distance data gathered. These
differences and the influence of alternate lighting are discussed in greater detail in this report.

vii
INTRODUCTION
Developments in roadway lighting have produced a number of lighting alternatives that provide
cost savings and energy efficiency when compared to roadway lighting currently in use. These
technologies may provide substantial cost savings in the form of reduced energy consumption;
however, it is not known what effect they have on driver safety in terms of visibility at night.
Military establishments tested a number of alternative lighting technologies that included light-
emitting diodes (LED), induction, and ceramic metal halide lighting. Specifically, Marine Corps
Base Hawaii (MCBH) in Kaneohe Bay, Hawaii implemented an alternative lighting technology
that utilized fluorescent street lighting which was installed on both active roadways and streets
within the marine base (Leetzow, personal communication, August 11 th 2009).The test
installation was completed in 2006 with a retrofit completed in mid 2009. In an effort to gauge
both energy savings and nighttime safety, the sponsor of the project requested an evaluation of
the alternative lighting technology. The evaluation required the collection of driver visibility data
that included detection and recognition distance data and photometric characteristics such as
illuminance and luminance. The specifics for each alternative lighting technology were then
compared to each other and to current roadway lighting technologies. A detailed review of the
methodology, analysis techniques, and final results are presented in this report.

Assessment of Alternate Lighting


The goal of the lighting retrofit was to increase cost savings for the military base with respect to
the outdoor roadway lighting. This retrofit also allowed a opportunity to evaluate the impact of
the lighting on driver visibility. For this aspect of the retrofit, the research team was contracted
by Magnaray International to evaluate an alternative roadway lighting technology at MCBH. The
project required a complete evaluation of the alternative lighting technology in terms of
photometric characteristics and visibility metrics for drivers.

Research Objectives
The key research objectives for this project are to:
1) Collect visibility detection data from Marine personnel on base,
2) Collect photometric data for the lighting technology,
3) Analyze and review the target detection data for differences in target detection,
recognition, and the impact of color, target size, and speed of approach,
4) Analyze and review the photometric data, and
5) Provide methodology, results, and discussion points in a report format.

1
ASSESSMENT OF ALTERNATIVE LIGHTING TECHNOLOGIES
The assessment of the MCBH alternative street lighting involved a human factors and lighting
metric approach. In order to accomplish these tasks, an experiment was designed that
incorporated a human visibility metric in addition to luminance metrics collected before, during,
and after the visibility measurements. To collect the required data, the Virginia Tech
Transportation Institute (VTTI) Lighting and Infrastructure Technology (LIT) group utilized a
state-of-the-art roadway lighting mobile measurement system (RLMMS). The system combines
visibility metrics (e.g., detection distance) and photometric measurements and collects them
dynamically during the experiment.

Experimental Design
The experimental design involved testing a fluorescent street lighting system retrofitted on a
Marine Corps base located in Kaneohe Bay on the island of O‟ahu, Hawaii. Target type size also
varied, with Small Target Visibility (7"x7") targets and larger targets (14"x14") used within the
testing area. Additional experimental variables included target illuminance level (which was
obtained by manipulating target position under the lighting section), a color comparison using
four specific colors of targets, two travel speeds of the experimental vehicle, and target
recognition (e.g., tab direction on the target). Details of each variable are shown in Table 1.

Table 1. Experimental Variables and Descriptions


Variable Description
Lighting Type Magnaray Luminaire with Philips 50 Watt Twin T5 Fluorescent
Target Illuminance Level High (10 lux) and low (3 lux)
Color Gray (18% reflectance), blue (7% reflectance), red (18%
reflectance), or green (18% reflectance) target
Target Type Small target (7"x7") and large target (14"x14")
Target Direction Position of target tab (pointing left or right)
Speed 25 mph and 45 mph

Participants
A total of 11 participants volunteered to be drivers in the data collection vehicle. The participants
were recruited from the Marine Corp base. The participant pool contained both males and
females aged 18 and older. It should be noted that gender and age were not controlled for this
study, thus were not analyzed. Each participant drove the specified route for six complete laps.
Participants drove four laps at 25 mph and two laps at 45 mph.

Equipment
The researchers used a suite of measurement equipment to gather visibility metrics and
photometric data. The lighting section and overall test area was defined by MCBH and
Magnaray® such that it allowed a relatively straight section of road to be tested uninterrupted by

2
base traffic/pedestrians. Magnaray® sold MCBH the required lighting and the test section was
retrofitted with the lighting.

Roadway Lighting Mobile Measurement System


The VTTI LIT group used a suite of data collection equipment during the experiment which
captured a variety of variables such as: illuminance, luminance, color, and participant response
data. The RLMMS was created by the LIT group at VTTI as a method for collecting roadway
lighting data in addition to participant response data. Illuminance data were collected by five
Minolta T10 detector heads, each mounted in a specific location on the vehicle. Specifically, four
waterproof Minolta detector heads were placed within a “Spider” apparatus that was mounted
directly on the roof of the experimental vehicle. The heads were positioned horizontally on the
vehicle roof such that two illuminance meters were positioned over the right and left wheel paths
and the other two meters were placed along the center line of the vehicle. The fifth Minolta
illuminance meter was positioned on the vehicle windshield and collected glare data from the
lighting installation. Each of the Minolta heads was connected to a single Minolta T10 body.
Data from these units were then sent via Ethernet to the data collection Personal Computer (PC)
located in the cargo area of the Sport Utility Vehicle (SUV). A NovaTel global positioning
system (GPS) device was positioned at the center of the four roof-mounted illuminance meters
and attached to the “Spider” apparatus. The GPS device was connected to the data collection box
and the vehicle latitude and longitude position data were incorporated into the overall data file.

Two separate video cameras were mounted on the vehicle windshield; one collected color
images of the forward driving scene and the second camera collected calibrated luminance
images of the forward driving scene. Each camera connected to a stand-alone PC computer that
was then connected to the data collection PC. The data collection PC was responsible for
collecting illuminance and GPS data and also synchronized the camera PC images with a
common time stamp. Additional equipment inside the vehicle consisted of button boxes for
participant responses and a keyboard for input by the in-vehicle experimenter.

Each component of the RLMMS was controlled by a specialized software program created in
LabVIEW™. The entire hardware suite was synchronized through the software program and data
collection rates were set at 20Hz. The image capture rate for each camera was set at 3.75 frames
per second (fps). The final output file used during the analysis contained a synchronization index
stamp, GPS latitude and longitude data, button/keyboard presses, individual images from each of
the cameras inside the vehicle, and also the illuminance meter data from each of the Minolta T-
10s. Figure 1 below shows the test vehicle used for this study. Figure 2 shows the “Spider”
apparatus with incorporated Minolta waterproof heads in addition to the GPS unit and cameras
mounted inside the vehicle.

3
Figure 1. Experimental vehicle with RLMMS components.

Figure 2. RLMMS components mounted on and inside a vehicle.

Procedure
Participants were asked to meet the research team at the start of the route. The route contained a
section of roadway used as an alternate entrance to the Marine base during daytime hours. The
gate closed at 10pm and thus traffic was non-existent during the experiment. The route is shown
in Figure 3 (from Google Maps). The numbers associated with the luminaires are the mounting
heights. The starting point (close to the motorpool) and the end point are shown with Red lines.

Upon arrival participants were asked to sit in the driver seat of the experimental vehicle and
adjust the seat so that they were comfortable to drive. An in-vehicle experimenter was present at
all times and sat in the passenger seat of the vehicle. When the participant was comfortable, the
in-vehicle experimenter reviewed the tasks involved. Each participant was asked to drive a
specific speed along the route and, while driving, to identify different-sized targets placed along
the roadway (see Figure 4).

4
Figure 3. Overhead View of the route with Luminaires designated.

Two specific tasks for target identification were explained to each participant. The first task was
to search for or identify targets along the route. Participants were asked to tell the in-vehicle
experimenter when they were confident they saw a target. The second task required participants
to indicate the color and tab direction of the target (e.g., target tab pointing towards or away from
the roadway – in or out). The two tasks resulted in two separate methods of identification. While
the first task was simply to detect a target without specific details, the second task was to
recognize the color and tab direction of the target, such that increased attention was oriented
towards the target in an effort to describe target characteristics. Target colors were switched half
way through each participant run in order for participants to view large and small targets of each
color type. The targets were also moved between target stations so that no learning effects were
evident in the experimental data. Each color was displayed for both large and small targets for
both high and low target illuminance levels.

5
Figure 4. Examples of large and small colored targets placed along the experimental route.

After reviewing the target detection tasks, the in-vehicle experimenter then explained the speed
requirements during the target identification task. The speed chosen was dependent on the order
of participation by the participant. A simple split-half arrangement was devised to randomize the
speed in which participants viewed the objects. For example, the first participant drove at 25
mph for the first lap, 45 mph for the second lap, and 25 mph for the third lap. The subsequent
fourth lap was travelled at 45 mph, followed by the fifth and sixth laps at 25mph. The second
participant began the first lap at 45 mph, followed by two laps at 25 mph. The second
participants‟ fourth lap was then 25 mph, followed by 45 mph and 25 mph for the final lap. The
third participant followed the same speed order as the first and the remaining participants
alternated depending on the previous participant‟s speed order. Alternating the initial speed in
which participants viewed the objects on the first lap allowed distribution of effects from the two
speed levels chosen. The in-vehicle experimenter asked if there were any further questions
regarding what was being asked of the participant. If there were no additional questions about
the task or speed requirement, the participant was then asked to put the vehicle in drive and
proceed to drive the route.

While the participant was maintaining the set speed limits driving the route, the in-vehicle
experimenter was tasked with marking the detection, recognition, and target location data. This
was achieved by a keyboard press when participants first detected a target, followed by a
different keyboard press upon recognizing the target color and tab direction. Finally, upon
passing the target location, the in-vehicle experimenter pressed a space bar to mark location of
the target object. This information was then logged in a data file for later analysis.

6
At the end of the six laps (approximately 30 minutes of driving), the participant was asked to
park the vehicle at the start of the testing route so that it was ready for the next participant.
Before exiting the vehicle the participant was asked if there were any additional questions.
Participants were then thanked for their time and dismissed. The in-vehicle experimenter then
greeted the next participant and continued with the experimentation. At the end of the testing
period the experimental team packed up the targets and transferred the data to a secure storage
drive.

Results
The entire data file including the button and space bar presses, latitude and longitude
information, and the respective image names from the color and luminance cameras were
imported into Statistical Analysis Software (SAS) for review and analysis. For example, the
distance calculation was derived from using latitude and longitude coordinates for each button
and space bar press and passing that information through a coordinate equation to derive distance
in meters. When the distance calculations were completed the data set underwent an additional
data check where outliers were analyzed and corrections were made (e.g., either deletions for
false button presses or frame corrections).

The illuminance data for the lighting collected by the illuminance meter network were also
imported into SAS for review and analysis. The illuminance data gave an approximation of the
light intensity reaching the road surface, which gave a further understanding of the performance
of the lighting. These results, in addition to the detection distance information, are presented in
greater detail in the following results sections.

Detection Distance
An ANalysis Of VAriance (ANOVA) was conducted on the detection distance data to identify if
differences occurred among a variety of target characteristics. Follow-up Student Newman-Keuls
(SNK) tests were also used to identify where these differences occurred. The results of the
ANOVAs for detection distances are shown in Table 2. Comparisons which proved insignificant
included speed condition (25 and 45 mph) and lap number (exposure order).
Table 2. Target Characteristics of ANOVA Results (Detection Distances)
Source F value P value Significant
Target Size 100.67 <0.0001 *
Target Color 77.65 <0.0001 *
Target Illuminance Level 8.06 0.0047 *
Target Size * Target Color 2.79 0.0403 *
Target Color * Target Illuminance Level 17.58 <0.0001 *
Target Size * Target Illuminance Level 19.99 <0.0001 *
Size * Color * Target Illuminance Level 16.01 <0.0001 *

For detection distances, the primary factors of Target Size (Large or Small), Target Color (Blue,
Gray, Green, or Red), Target illuminance Level (High or Low), Target Size and Target Color
interaction, Target Color and Target illuminance Level interaction, Target Size and Target

7
illuminance Level interaction, and the 3-way interaction among Target Size, Target Color, and
Target illuminance Level were all significant or showed differences.

A 3-way interaction occurred among high and low target illuminance conditions, large and small
targets, and the four target color conditions. The graphed interaction is shown in Figure 4. The
general trend for the results showed that target size, color, and target illuminance influenced the
detection distances of participants. Of little surprise, larger targets were detected sooner than
smaller ones and higher illumination provided longer detection distances; however, the role of
target color appears to be influenced both by target size and target illuminance level.

In an effort to better understand what role target color played with respect to lighting used, a
second set of analyses was performed. The first was to compare target colors across the specific
target size range and the second was to compare the impact of target illuminance (e.g., High and
Low) on the target color chosen.

Interaction of Color, Size, and Illuminance


140

120
Mean Detection Distance (m)

100

80

60

40

20

0
High Low High Low High Low High Low High Low High Low High Low High Low

Blue Gray Green Red Blue Gray Green Red

Large Small

Figure 5. Significant 3-way interaction among target color, target size, and target
illuminance level.

The first analysis comparing target size and color is shown in Figure 6 below. Pairwise
comparisons of the results were performed using an SNK analysis. The results are shown in the
figure by the letters at the top of the bars. Significant differences between target colors are
indicated by different letters. For example, green targets elicited much greater detection

8
distances, and this is indicated by the letter „A‟. Those columns containing the same letters are
not significantly different from one another.

When comparing within target sizes and specifically reviewing the large target group,
significantly longer detection distances were seen for the red and green targets (approximately
99 m) compared to the remaining two target colors. Furthermore, red and green detection
distances were significantly longer than gray (83 m), which was also significantly longer than the
blue target (52 m). The blue target performance was likely due to the dark blue shade chosen for
the experiment. The dark blue likely was low in contrast, resulting in significantly reduced
detection distance performance when contrasted against the grass background. Similar to the
large target detection distances, green (78 m), red (65 m), and gray (71 m) had significantly
longer detection distances than the blue target (33 m). Unlike the large target performance, the
small green target had significantly longer detection distances than red, but did not have
significantly longer detection distances than gray, perhaps due to the Color Rending Index (CRI)
of the lamp. No significant differences occurred between the gray and red targets.

Overall, large targets had longer detection distances than small targets, with green and red having
comparable distances for large targets. This relationship changed for the small targets used with
higher detection distances for green targets but there were no differences between gray and red.
The issue with the target color red-green relationship change with target size is likely due to a
change in the contrast with the background. The targets were located along the side of roadway
on a mowed shoulder. The small targets primarily appeared against the grass and the mowed area
along the shoulder. The larger targets appeared against the mowed area and an un-mowed area
outside of the mowed area.

Both target types (large and small) showed a significantly shorter detection distance for the blue
target.

9
Detection Distance by Target Color & Size
120
A A
Mean Detection Distance (m)

100
B
A
80 A, B
B

60 C
Blue
Gray
40 C Green
Red
20

0
Large Small
Target Size

Figure 6. Mean detection distances for target color and size.

The second analysis compared high and low target illuminance conditions across the four target
colors used with respect to detection distances. The results are presented in Figure 6. In general,
the high target illuminance group had longer detection distances than the low target illuminance
level group, as expected. However, looking within groups, subtle differences occurred. For the
high target illuminance group the green outperformed all other target colors with an average
detection distance of 104 m. This detection distance was significantly longer than the red target
(89 m), the gray target (76 m), and the blue target (38 m). The benefit of a white light source
appears to have favored most of the colors used in the experiment, except for blue, which may
have suffered from the type of shade used for the target color. For the low target illuminance
condition detection distances were similar across red and green (~ 77m), however gray and
significantly longer detection distances (83m) than either red or green. Again, blue had the
lowest detection distances (50m). It is important to remember that these are large and small
target colors combined and thus gray may have benefited by increased contrast against the
background of the road type used in the experimentation. This did not happen with the red and
green targets where the detection distance for these targets was reduced with the vertical target
illuminance level. The low target illuminance conditions did not benefit green or red. It is
thought that the size of target moderated the likelihood of target detection rather than color alone.
The other critical aspect here is the contrast. As contrast is the metric by which we perceive
objects, the target contrast may be actually increased by a lower target illuminance level as the
target may be darker than its background. This effect may not be consistent across the color of
the target as differences in color also contribute to evaluation of the contrast. These results

10
further confirm that target illuminance, target color, and size of target, with addition of lighting
type, influence the likelihood of target detection.

Target Color & Illuminance - Detection Distances


120
A
Mean Detection Distance (m)

100 B
A
C B B
80

60 Blue
C
Gray
D
40 Green
20 Red

0
High Low
Illuminance Level

Figure 7. Mean detection distances for each target color and target illuminance level.

As a final check for effects, the researchers performed a final analysis to check the influence of
target size and target illuminance levels, combining the target colors into one group. Figure 8
below shows the differences between target size based on the target illuminance level. As can be
seen, large targets benefit in both high and low target illuminance conditions, although this
benefit is not equal across all target illuminance conditions. Target size and target illuminance
definitely play a large role in target detection; however, introducing different color types also
influences target detection under the fluorescent lighting. Greens and reds appear to benefit the
most from the broad source spectrum lighting. The blue appears to be the most impacted;
however, the chosen color content of the blue in this experiment may have influenced when it
was detected.

Overall, when testing broad spectrum light sources, there is a strong benefit from using multiple
colored targets. The impact of spectral implications is based on target size, and target
illuminance levels impact the likelihood of target detection. There is a strong benefit from using
multiple colored targets in an effort to measure driving behavioral effects.

11
Target Size & Illuminance - Detection Distance
100
90
Mean Detection Distance (m)

80
70
60
50
40 Large
30 Small
20
10
0
High Low
Illuminance Level

Figure 8. Mean detection distances for each target size and target illuminance level.

Recognition Distance
In addition to the detection distance measurements, the research team also introduced, for this
research effort, a task that required participants to recognize object color and also specify in
which direction the tab on the target was pointing. The culmination of these combined tasks was
thought to take more attentional effort requiring participants to identify and interpret the target
information. These recognition measurements underwent a similar but separate analysis as
detection distances. An ANOVA was conducted on the recognition distance data to identify if
differences occurred among a variety of target characteristics. Similar to the detection distances,
follow-up SNK tests were also used to identify where these differences occurred. The results of
the significant ANOVAs for recognition distances are shown in Table 3. Again, similar to the
detection distance results, comparisons between speed, tab direction (correct and incorrect), and
lap number (exposure order effects) were not significantly different.
Table 3. Target Characteristics of ANOVA Results (Recognition Distances)
Source F value P value Significant
Target Size 78.50 <0.0001 *
Target Color 22.84 <0.0001 *
Target illuminance Level 10.89 0.0010 *
Target Size * Target Color 3.34 0.0191 *
Target Color * Target illuminance Level 6.07 0.0005 *
Target Size * Target illuminance Level 12.37 0.0005 *
Size * Color * Target illuminance 8.79 0.0002 *

12
Similar to the detection distance results, the variables for recognition distances of Target Size
(Large or Small), Target Color (Blue, Gray, Green, or Red), Target illuminance Level (High or
Low), Target Size and Target Color interaction, Target Color and Target illuminance Level
interaction, Target Size and Target illuminance Level interaction, and the 3-way interaction
among Target Size, Target Color, and Target illuminance Level were all significant or showed
differences.

Again, a 3-way interaction occurred among target illuminance level, target size, and target
colors. The interaction of these variables is presented in Figure 9 below. Similar trends to
detection distances were seen with larger targets being recognized sooner, and higher target
illumination provided earlier recognition. Again, target color interacted with both target size and
target illuminance, requiring an in-depth review of the color variable.

Similar to the detection distance analysis, recognition distances were also subjected to additional
analyses. The first analysis examined target size and color interaction. The second analysis
examined target illuminance and color in an effort to identify how these factors interacted.

Interaction of Color, Size, and Illuminance


90

80

70
Mean Recognition Distance (m)

60

50

40

30

20

10

0
High Low High Low High Low High Low High Low High Low High Low High Low

Blue Gray Green Red Blue Gray Green Red

Large Small

Figure 9. Significant 3-way interaction among target color, target size, and target
illuminance level.
The results of analysis of the interaction between target color and size are shown in Figure 9.
Pairwise comparisons of the results were performed using an SNK analysis. Similar to the

13
findings for detection distance, the blue target color had significantly shorter recognition
distances than the other colors used during the experiment, irrespective of target size.
Participants likely had to detect and then concentrate on the target itself to identify the color
type. This was likely influenced by the dark shade of blue used for the experimentation. What‟s
interesting in the recognition comparisons–and opposite of the detection distance comparisons–is
the lack of significant differences between the remaining color types. Interaction effects are
likely driven by the blue target alone, with significantly shorter recognition distances for both
target size groups. Participants had slightly longer recognition distances for green and red targets
(60 m for both) compared to gray (57 m) for large targets, although they were not significantly
different. A similar outcome is seen for the small targets with little variation in recognition
distance between green, red, and gray (39 m, 36 m, 40 m, respectively). From these results it is
evident that there is a benefit in target size for recognizing target characteristics sooner or farther
away; however, within group comparisons it appears that the lighting provides ample recognition
for a variety of color types around the same distance. These results likely represent threshold
contrast issues. Further inspection of how target illuminance factors into these results is pursued
in the next analysis.

Recognition Distance by Target Color & Size


120
Mean Recognition Distance (m)

100

80
A A
60 A Blue
Gray
A A A
40 Green
B
Red
B
20

0
Large Small
Target Size

Figure 10. Mean recognition distances for target color and size.

The next analysis compared high and low target illuminance conditions, the four target colors,
and corresponding recognition distances. Analysis results are illustrated in Figure 11. Target
illuminance levels affected recognition distances based on the type of target color. Differences in

14
recognition distances in the high target illuminance level were only seen for the blue target;
again, with the target color having significantly lower detection distances than the remaining
colors. Interestingly, the low target illuminance condition showed differences between the
specific colors, with gray having the highest recognition distance (50 m) which was significantly
longer when compared to red (44 m), green (39 m), and blue (26 m). Red and green were
significantly longer than blue, but the two colors had similar recognition distances. The similar
distances across target illuminance levels for the blue target suggest that the color chosen
delayed recognition regardless of target illuminance level. For the remaining colors, increased
target illuminance does increase a person‟s ability to recognize the target sooner, as expected.

Target Color & Illuminance - Recognition Distances


120
Mean Recognition Distance (m)

100

80
A
A A Blue
60 A
B Gray
B
40 Green
B C
Red
20

0
High Low
Illuminance Level

Figure 11. Mean recognition distances for each target color and target illuminance level.

As a final analysis for recognition distance, the differences across target illuminance levels were
examined when collapsed across color type. Figure 12 shows recognition distances for each of
the target sizes and target illuminance levels. As expected, large targets outperform small targets
in both high and low target illuminance conditions, which can be attributed to the surface area of
the object. An interesting result is that the comparison of small targets across target illuminance
levels (with target color combined) appears to not fluctuate dramatically, with 36 m for high
condition and 33 m for the low condition. This result suggests that, irrespective of target
illuminance level, recognition performance for the small targets appears to be comparable,
perhaps indicating that small targets are necessarily a sufficient standard to base both detection
and recognition performance on, with respect to broad spectrum lighting.

15
Target Size & Illuminance - Recognition Distance
100
Mean Recognition Distance (m)

90
80
70
60
50
40 Large
30 Small
20
10
0
High Low
Illuminance Level

Figure 12. Mean recognition distances for each target size and target illuminance level.

Horizontal Roadway Illuminance


The results from the horizontal roadway illuminance data collected are presented below.
Horizontal roadway illuminance measurements were taken via four illuminance meters mounted
on the vehicle roof (see equipment section). For this assessment north and southbound travel lane
sensors were combined to provide an entire roadway illuminance picture. For example, Figure 12
illustrates the method by which roadway illuminance was evaluated. Illuminance data from
Minolta T10 waterproof sensors–representing six points across the roadway–were collected.
Sensor points 1 and 6 were collected from the right side of the vehicle, and sensor points 3 and 4
were collected from the left side of the vehicle. Sensor points 2 and 5 represent an average of
data collected from the front and rear sensors. A final in-vehicle sensor attached to the
windshield collected glare data while traveling the route and is presented separately.

16
N
Street Northbound E
lighting on to
W
west motorpool
shoulder S
1 2 3 4 5 6
Southbound
to
gatehouse

Figure 13. Sensor locations for Horizontal Roadway illuminance analysis.

The horizontal roadway illuminance analysis combined data for the six sensors previously
described and is presented based on the distance of the experimental route. The base motorpool,
or participant starting location, was used as a starting point for the horizontal roadway
illuminance measurements. The research team then processed the metrics for the sensors for the
entire experimental route length (~ 1100 m). It should be noted that these sensors are located on
the vehicle roof which was approxmiately 72 inches above the roadway surface and these values
do not represent overall roadway illuminance values which are typically at the road surface. As
shown in Figure 13 there is a noticeable difference among horizontal roadway illuminance levels
across the roadway as well as a gap where a luminaire was not functioning. Sensor 1, which was
closest to the light sources (tower side), showed the highest horizontal roadway illuminance
levels and, as the metering progressed over the roadway, the illuminance levels decreased as
expected. The lowest horizontal roadway illuminance levels were for sensor 6, which was also
the furthest away from the light source. Horizontal roadway illuminance levels for the tower side
ranged between 22 lux and 47 lux. The higher levels were likely achieved during the slight curve
in the roadway test route. The non-tower side achieved the highest levels for sensor 5 (front and
rear sensors combined) which ranged between 8-17 lux depending on the location within the
route.

17
Illuminance Meter Levels Over Entire Roadway Length
50

45

40

35

30
Illuminance (lux)

Sensor 1

25 Sensor 2
Sensor 3
Sensor 4
20
Sensor 5
Sensor 6
15

10

0
0 200 400 600 800 1000
Test Route Length (m)

Figure 14. Horizontal roadway illuminance levels of six sensors over entire road length.
Table 4 below provides additional metrics collected by the RLMMS. The unusually high
uniformity ratios are likely the result of the retrofit on the current poles. As the pole spacing‟s
were not changed, the retrofit did not provide an optimal design for the luminaires. This resulted
in the RLMMS capturing extremely low (minimum) illuminance levels between poles (e.g., see
Figure 14). This did not influence the detection and recognition performance as the target
illuminance values were measured and standardized for all of the target locations. It should also
be noted that the last 5 luminaires in the series were at a lower mounting height (24ft rather than
30ft) which is seen in the slightly higher illuminance values for these luminaires as seen in
Figure 14.
Table 4. Horizontal roadway illuminance Metrics
Uniformity Uniformity
Measurement Sensor Minimum Average Maximum Ratio (Avg Ratio (Max
Location Location Illuminance Illuminance Illuminance to Min) to Min)
Tower side 1 0.1 5.31 57.17 53.1 571.7
(Southbound) 2 0.1 4.02 35.83 40.2 358.3
3 0.2 4.30 41.14 21.5 205.7
Non-Tower 4 0.2 2.32 19.9 11.6 99.5
side 5 0.1 2.86 21.36 28.6 213.6
(Northbound) 6 0.1 2.21 28.79 22.1 287.9

18
In an effort to validate some of the onsite calculations obtained during the research effort, design
simulations were conducted using AGI-32 design software. Based on the current design, pole
spacing, luminaire type, arm length and other metrics was entered into AGI-32 design software.
The results for current spacing resembled the measurements gathered on-site. For example,
average illuminance was calculated at 4.52 lux, with maximum illuminance of 48.6 lux, and a
minimum of 0.1 lux. In addition, the average to minimum ratio obtained through the software
was 45.2.

The same metrics were taken into account when calculating the recommended RP-8 values for
both the illuminance and luminance methods. Using the illuminance method for an R2 pavement,
a level of 7 lux is recommended. That information, along with the other metrics, was entered into
AGI-32. Based on the illuminance method, the design software recommended a spacing of 232ft.
The luminance methods was also calcuated based on a 0.5 cd/m2 RP-8 value. The spacing
recommendation based on this design was 180ft. Based on this spacing, the Uniformity ratio of 5
was calculated which is within the specifications recommended in RP-8. This would be a more
typical lighting design for street than the 360‟ spacings provided through the retrofit. However,
given the present studies results of target detecton, the current lighting design does not appear to
be detrimental for driver visibility. The calculations to meet recommended IESNA values were
not attainable in this retrofit situation and yet the visual performance remains the same or
improved based on the detection distance measurements performed for this research.

Figure 15 illustrates that glare levels collected over the entire length of the roadway show similar
effects as the horizontal roadway illuminance levels. The largest impact occurred on the tower
side, where glare levels ranged from 5 to 9 lux, again dependening on the area of roadway.
Similar to the previous illuminance measurements, a gap was seen near the end of the route
where a luminaire was not functioning. Travel in the opposite direction, on the other side of the
towers, resulted in glare ratings that maxed out at 5.5 lux and had a minimum of just over 2 lux.
The largest impact was, as expected, on the tower side where the glare meter picked up highs of
9 lux.

19
Glare Levels for Entire Roadway Length
10

7
Illuminance Level (Lux)

Glare:
5 Tower Side

Glare: Non-
3 Tower Side

0
0 200 400 600 800 1000
Test Route Length (m)

Figure 15. Glare levels for entire road length.

Discussion
A comparison of detection and recognition distances for large and small targets of different
colors was collected at MCBH. It is important to note that, compared to other testing sites (e.g.,
see Gibbons & Edwards, 2009; Mutmansky et al., 2009) , this test location provided only
consisted of one alternative lighting type. Furthermore the test site had a total of 10 luminaires,
comparable with one previous test locations (see Gibbons & Edwards, 2009). Participants ranged
in age and gender similar to previous studies. Detection distance was collected based on target
color and target illuminance level; however, an additional variable–target size–was also
included. Finally, an additional behavioral metric was collected–recognition–in an effort to
understand how target color, size, and target illuminance levels influenced correct recognition.

Detection Distance
Detection distances were affected by the color, target illuminance, and size of the target. As
expected, larger targets had longer detection distances than smaller targets. A larger object in this
instance gave participants a larger surface area to detect and thus resulted in longer detection
distances. However, color of the target also affected detection distance, with large red and green
targets having significantly longer detection distances than blue or gray. In turn, large gray
targets had significantly longer detection distances than did large blue targets. An interesting
result occurred for the smaller targets, where green had significantly longer detection distances
than red, gray, or blue. The spectral properties of the fluorescent lighting favored green targets
for detection, followed by no differences between red and gray target detection. Again, blue had

20
significantly shorter detection distances than any of the other target colors used. In addition to
size of the target, target illuminance level (high or low) impacted the detection of different
colored targets. As expected, higher target illuminance levels allowed participants to detect the
targets sooner, which again was impacted by target size. However, the results showed that target
size and target illuminance level interact and influence moderate when targets are detected. For
example, large green and red targets in the high target illuminance level were detected at around
120 m, but in the low target illuminance condition, these detection distances were reduced to less
than 80 m. Having sufficient target illuminance levels to maximize the benefit of color is
required. These results suggest that the role of an alternate light source appears to benefit
detection distances which is moderated by target illuminance and size of the target. The results
also support the use of different target color types when evaluating alternative light sources.

The impact of color on detection distances was also compared to previous studies. Figure 16
presents detection distance data collected from Anchorage and San Diego (see Gibbons &
Edwards, 2009; Mutmansky et al., 2009). Comparisons were made between detection distances
for small targets of varying colors under High Pressure Sodium (HPS) lighting. Immediate
benefits (i.e., higher detection distances) of broad spectrum lighting tested at MCBH can be seen
for the green, gray, and red colored targets when compared to identical targets under the HPS
condition. The blue target had higher detection distances in Anchorage under HPS compared to
MCBH or San Diego. This result was likely due to the contrast or environment that the target
was presented in. The targets‟ locations in Anchorage were on top of a snow bank, thus
providing increased contrast and higher detection distances. However, when MCBH data are
compared to San Diego HPS data, the broad spectrum benefit can be seen, despite the blue target
color disadvantages (e.g., dark). Comparisons across test locations appear to support the benefit
of using newer broad spectrum light sources compared to current HPS roadway lighting. It
should be noted that blue target benefits in overall detection distance observed in Anchorage are
likely due to the snow background for target contrast.

21
Detection Distance by Site and Small Target Color
100

90

80
Detection distance (m)

70

60
Blue
50 Gray
40 Green
Red
30

20

10

0
Hawaii Anchorage San Diego
Site

Figure 16. Detection distance comparison for identical small target colors at different sites.

From the initial analysis in Figure 16 above it appears that a valid comparison, based on
environmental influences is between the Hawaii and San Diego sites. An additional comparison,
Figure 17, was made with respect to blue and gray targets; red and green targets were not utilized
in San Diego. When compared directly to other broad spectrum sources, the fluorescent appears
to provide improved detection performance for the gray target. Comparisons across the blue
target appear to be within the same range as other alternative lighting sources. The lower
detection distances for the blue target again reflect the target color used, not necessarily the
performance of the luminaires. The increased performance on the gray target may also be due, in
part, to the amount of data available based on the number of luminaires at each testing site. It
should be noted that San Diego used a maximum of four luminaires per alternative lighting
technology and the results may be a ceiling effect or maximum detection distance based on the
sections being close to one another.

22
Detection Distance Comparisons
80

70

60
Detection Distance (m)

50

40
Blue
Gray
30

20

10

0
Powerlux US Light Tech Fiti LED 165w Leotek LED 94w Existing HPS Magnaray (Philips
Induction 165w 150W Street 50 Watts x2)

San Diego Hawaii

Figure 17. Detection distance comparisons between locations and alternative lighting.

Recognition Distance
Complementary to the detection distance, the current study also collected recognition distances.
Recognition distance provides an additional metric to gauge not only how a person identifies a
target object, but also how he/she had to interpret the information from the target. Again, similar
to the detection distances obtained that were described previously, the recognition distances were
influenced not only by target size, but also by target color and target illuminance levels. A
noticeable difference when comparing detection distance and recognition distance is that
participants in some instances detected objects at ~120 m but recognized target characteristics at
~ 80 m. For that example, the 40-meter difference translates into an additional (approximately) 2
- 3 seconds (depending on speed, 25 or 45 mph) to identify, process, and respond to target
characteristics. While detecting the object is an important first step that may influence how you
react to it, the use of recognition distance is important for understanding when someone can
interpret that information. This is the first study to incorporate recognition distance when
reviewing broad spectrum lighting and it is hoped that this metric will be incorporated into future
research to understand the impact of recognition under current and alternative light sources.

Horizontal Roadway illuminance


A roadway illuminance profile was established for the lighting on the entire roadway, which
included both tower side (high illuminance) and non-tower side measurements (lower

23
illuminance). High values were obtained when directly under the fluorescent which reached
values between 22-47 lux. Lower levels obtained for the non-tower which included values
between 8-17 lux. In addition to the profile, Figure 18 below compares mean roadway
illuminance levels collected from other broad spectrum lighting sources, in addition to HPS, with
the current technology under investigation. As can be seen from the figure, the fluorescent
average was around 3.5 lux, but provided noticeably higher detection distances than HPS and
comparable distances to other broad spectrum sources obtained in previous studies (see Figure
17). Similar to other findings of broad spectrum sources, the fluorescent appears to provide
adequate detection levels despite having a lower average roadway illuminance than current HPS
roadway lighting.

Mean Illuminance Comparisons


25

20
Illuminance (lux)

15

10

0
Magnaray Powerlux US Light Tech Fiti LED 165w Leotek LED Existing HPS 400W HPS
Flourescent Induction 150W 94w Street
(Philips 50 165w
Watts x2)

Hawaii San Diego Anchorage


Data Collection Locations

Figure 18. Mean Illuminance comparisons across testing sites.


A glare analysis was also conducted on the glare using a surrogate glare measure, namely an
illuminance meter mounted on the windshield. A glare profile was also produced for the entire
roadway that showed both tower and no-tower glare experienced by the driver. As expected,
higher glare was experienced when the participant was on the tower side of the roadway when
compared directly to the non-tower side. To understand the impact of the glare, comparisons
were also made to previous studies which are presented in Figure 19. From the comparisons
across locations, the Magnaray® setup in Hawaii appears to have a similar mean glare impact as
the Fiti LED used in San Diego. The mean glare rating appears to be within the same level at
other alternative technologies and lower than standard HPS street lighting. Again, these results

24
and cross location comparisons should be treated as a first attempt. It should be noted that the
amount of data collected was greater in the Hawaii location compared to the San Diego location,
which may in turn influence the mean values obtained.

Mean Glare Levels Across Locations


8

6
Illuminance (lux)

0
Powerlux Induction US Light Tech 150W Fiti LED 165 Watts Leotek LED 94 Existing HPS Street Magnaray
165 Watts Watts Fluorescent (Philips
50 Watts x2)

San Diego Hawaii


Data Collection Locations

Figure 19. Mean glare comparisons across testing sites.

Conclusions
The results of the visibility analysis at MCBH indicated that reduced levels of roadway
illuminance appear to provide adequate detection and recognition performance when targets are
the object of interest. The result is similar to other research conducted in different areas of the
country. The Magnaray® product used in Hawaii also helps with recognition visibility, an
additional variable collected for the first time during this research effort. Recognition is another
important tool when trying to assess the impact on driver safety. Not only do drivers have to
detect the object but they must also recognize what type of object may or may not impact their
general driving safety. The broad spectrum source studied in Hawaii also appears to be
comparable to other broad spectrum sources examined in previous research studies. When
comparisons were made to targets in Anchorage and San Diego, the Magnaray® fluorescent
performed as well as, and in some cases better than, other broad spectrum sources.

The illuminance levels obtained during this research effort also appear to be comparable with
alternative products previously investigated. However, it should be noted that the current project
had a closed roadway environment with 10 luminaires, whereas previous San Diego research had
25
a limited number of luminaires or a publicly open roadway collection environment. Additional
variables (i.e., other traffic) may have influenced the quality of data collected in previous studies
and may not reflect consistent results. Additional research is warranted where adequate lighting
setups can be achieved in order to compare not only across alternative and current lighting
technologies, but also within alternative lighting technologies currently being considered by
counties and municipalities.

Limitations
The results of this study were likely impacted based on a number of factors outside the
constraints or the control of the project. The participants used for this research were all younger
(e.g., 18-30 years old and perhaps provide the optimal detection and recognition performance
outcomes. The impact of broad spectrum lighting and detection distances should be explored for
a wider age range in an effort to account for the aging eye.

The participants were also explicitly asked to look for targets along the side or edge of the
roadway. Though no significant differences were found between the first lap (e.g., initial
exposure) and other laps, there was little reason for participants to modify their visual scanning
behavior. It is likely that people intentionally focused on the side of the roadway in anticipation
of a target. The results for this detection task may therefore reflect optimal conditions where
participants were primed to look for an object. Alternating target types in addition to color will
provide a better understanding of detection and recognition for real-world objects (e.g.,
pedestrians).

Overall, despite some limitations to the current study, the results add to the body of knowledge
for alternative lighting configurations. These limitations may have influenced the analysis
outcomes to some degree; however, some level of control is surrendered given the partially
naturalistic data collection environment. These limitations provide valuable lessons for future
research.

Future Considerations and Improvements


More research is required to assess the impacts of lighting type and driver safety. Despite the
availability of alternative light source comparisons from previous research, a baseline
measurement for direct comparisons should be obtained. For example, previous studies had
different luminaire numbers and mounting heights, which may impact the benefits and
comparisons that can be made within and across products. It is preferable that the same number
of luminaires and similar roadway types are used for comparisons across studies. The impact of
color and the spectral implications from this research project provide additional support to the
benefits of broad spectrum lighting. Detection and recognition of targets is impacted by the color
of the target and the ability of a participant to see and interpret that information. Thus, broad
spectrum lighting as used in the present study provides benefits for participants to see and
interpret information under lower lighting levels. Additional research in a controlled
environment using broad spectrum and current road lighting sources is required in an effort to
fully understand the potential benefits to driver safety.

26
References

Gibbons, R.B., Edwards, C.J., and Meyer, J.E. (2009). Municipality of Anchorage: Alternative
Lighting Evaluations. Final Report. Virginia Tech Transportation Institute and Clanton &
Associates.

Mutmansky, M., Givler, T., Garcia, J., Clanton, N., Gibbons, R., Edwards, C., DeJean, D.,
Cartier, T., Hudson, C. (2009). Advanced Street Lighting Technologies Assessment Project –
City of San Diego. Final Report. Clanton & Associates.

27

You might also like