U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
202-366-4000


Skip to content U.S. Department of Transportation/Federal Highway AdministrationU.S. Department of Transportation/Federal Highway Administration

Accelerating Innovation

FHWA Home / Accelerating Innovation / HFL


SITE SELECTION

Site Selection Criteria

All aspects of this study took place in work zones. Work zones require drivers to make decisions that are not typical to normal driving operations, putting their cognitive processes under additional stress. For this project, crossovers were the chosen work zone maneuver for study, as opposed to lane drops or other maneuvers. Crossovers occur when traffic is diverted away from the existing roadway onto a temporary roadway section and then reintroduced to the existing roadway (this maneuver is typical when a bridge is being reconstructed and drivers are diverted onto a temporary bridge). Crossovers typically have less signage and information conveyed to the user, increasing the driver’s need for pavement markings or other devices installed along the roadway for navigation. For example, at crossovers through the construction site, information is conveyed to the user through special means such as cones, barrels, and temporary signage. To remove the effect that other devices may have on drivers’ ability to navigate a crossover, 3M provided certain site requirements and desiredcharacteristics of candidate work zones. With only minor deviations, these requirements were communicated to members of the research team before the project began, and included:

General Site Requirements

  • Work zone open for a minimum duration of 3 months.
  • Four or more lanes on the highway (two lanes per direction of travel).
  • No roadway lighting, if possible, especially construction lighting. If lighting is necessary, it should be approved and kept to a minimum

Desired Site Characteristics

  • Posted work zone speed limit of 45 miles per hour (72 kph) or higher.
  • State, US, or interstate highways.
  • Substantial lane shifts; for instance, no short crossovers to the shoulder (less than one lane width).
  • No temporary RPMs. These typically are installed on all work zone applications in North Carolina and as necessary in Ohio.
  • If jersey barriers are utilized at the site, no paddle reflectors installed on the top of the barriers.
  • A good pavement marking contractor, as indicated by the State department of transportation (DOT).
  • A history of multiple rainy days on record in the construction work zone season.
  • No traffic signals in or near the crossover being studied.

In addition, in Phase I of the present study, the researchers concluded that a total sample of 250 vehicles was necessary to determine a 0.05 level of significance in differences in lane line encroachments.(18) At this value, it should be possible to detect differences between the experimental and standard pavement markings and their interactions with wet conditions.

Using these criteria, the research team worked closely with DOTs and nearby municipalities to find sites that would be suitable for study. Sites that were considered were all under construction during the 2009-2011 construction seasons. Additional criteria the team used to choose sites included:

  • Two crossovers (entering and exiting the construction zone) so that one of the crossovers is utilized as a comparison site with standard pavement markings.
  • Moderate to high traffic volumes that would meet the sample size requirements.
  • Manageable implementation timeframes.
  • Proximity to research team base to get to a site quickly during potential rain events. Sites within a radius of approximately 1 to 1.5 hours were preferred; however, in special circumstances, sites within a 4-hour drive were considered feasible.

Site types of interest included bridge replacements, resurfacing and rehabilitation of roadway surfaces, and widening. The effect of the economy on road construction and the requirements necessary (especially number of lanes and time to site) posed a major challenge to members of the research team. Table 2 shows a list of sites. The Lanes column shows the number of lanes that remained open through the crossover in one direction of the highway; all highways had four lanes originally.

Table 2. Field test site descriptions.

Roadway Location Type Lanes Speed Limit
I-85 Henderson, NC Interstate 1 55 mph (89 kph)
US-15/501 Chapel Hill NC US Route 2 45 mph (72 kph)
US-421 Winston-Salem NC US Route 2 55 mph (89 kph)
US-32/33/50 Ashtabula, OH US/SR Routes 1 55 mph (89 kph)
I-90 Athens, OH Interstate 2 55 mph (89 kph)

Site Descriptions

I-85, Henderson, North Carolina

This project was a rehabilitation of several miles of I-85, designated North Carolina DOT Transportation Improvement Project (TIP) #I-2810. Henderson is located in the Northern Piedmont. During the data collection months of May and June 2009, the average monthly rainfall was above 3.6 inches (9.5 cm) and had a minimum of 9 hours and 22 minutes between sunset and sunrise.(22) During the period of construction, southbound I-85 was repaved, including some bridge rehabilitation. Median crossovers were utilized to bring traffic from southbound I-85 to the inside lane of northbound I-85. The north and south crossovers are approximately 6 miles (9.7 km) apart, shown in figure 6. Figure 7 shows daytime and nighttime views of the right curve into the treatment crossover on the south end of the work zone.

Figure 6. Photo. I-85 crossover locations.

Figure 6. Photo. I-85 crossover locations.

Figure 7. Photos. I-85 southern crossover during daytime and nighttime.

Figure 7. Photos. I-85 southern crossover during daytime and nighttime.

During the construction phase, traffic were reduced to one travel lane in each direction prior to entering the study area, using the original northbound travel lanes in a two-lane, two-way traffic pattern. At the time of the initial site selection process, the research team was not aware of the desire to study multiple lanes shifting into and out of the work zone. Because a sizeable effort was made to install data collection equipment, the site was kept for further data collection and analysis to note any differences between this site and other sites with multiple lane crossovers studied in this project. In addition, retroreflectivity studies were not affected by the need for more than one lane of traffic, so those findings are included in the summary findings. Figures 8 and 9 show aerial views of the southbound traffic shift at the northern and southern crossovers.

Figure 8. Photo. Aerial view of northern crossover. Southbound traffic shifts to northbound side divided by a jersey barrier.

Figure 8. Photo. Aerial view of northern crossover. Southbound traffic shifts to northbound side divided by a jersey barrier.

Figure 9. Photo. Aerial view of southern crossover. Southbound traffic shifts back to normal operation.

Figure 9. Photo. Aerial view of southern crossover. Southbound traffic shifts back to normal operation.

Standard and the AWP markings were installed at the crossovers located along southbound I-85. The northern crossover was marked by the standard pavement marking specified in the construction plans and included RPMs and traffic barrels. The southern crossover was marked by the AWP, traffic barrels on the passenger side, concrete barriers with reflectors on the driver side, and RPMs. Based on recommendations in the literature, a minimum 5-second preview time was utilized with the AWP on the southern crossover, corresponding to a 500-foot (152.4 m) application of the AWP along the tangent just prior to first radius of the crossover.(17) The speed limit was 65 miles per hour (105 kph), but a $250 speeding penalty was in effect and was affixed to speed limit signs throughout the site. Speeds were under enforcement along a tangent section in the northbound direction during many of the initial site visits the research team made early during the construction process, especially during daytime operations. No data were collected while enforcement was taking place.

Both the comparison and treatment sites had S-curve geometry. After some discussion with North Carolina DOT inspectors and the pavement marking contractor, it was decided that the last curve exiting the work zone on the southern site (treatment) could not be tied into the regular lanes with the AWP because there was not enough time to switch out the elements and properly calibrate the pavement marking thickness and element drop rate, all while slowly rolling traffic through the work zone. This meant the final curve (left turn of the treatment site) could not be studied. Thus, the right turns for the northern and southern crossovers would be the focus of study for this particular site for comparison of speeds and lane encroachments.

Another issue that the project team noted at the northern end of the test site was the atypical manner in which I-85 was reduced to one lane (see figure 7). At a single point, drivers are forced to choose between veering left onto I-85 southbound or staying straight to exit onto US-158. Drivers appeared confused by this arrangement; many cars were observed making last-second lane changes less than 200 feet (61 m) from the delineating barrels. Other drivers were seen coming to a complete stop on the highway to decide which way to go. Furthermore, some drivers would make the wrong choice, quickly pull off onto the shoulder, and either reverse down the highway or drive through the barrels separating the exit and I-85. Because the layout of the entry into the crossover was problematic, data were not collected during periods of obvious confusion. In addition, the first turn (left turn) into the comparison site crossover was not studied so that any other possible “confusion effect” was eliminated.

Last, it should be noted that the length of this site (approximately 6 miles [9.7 km]) could pose problems in the analysis. Rain duration and intensity could be very different at one crossover than at the other. To help alleviate any potential concerns, the research team only analyzed data 10 minutes after the obvious beginning of a rain event at each site and stopped 10 minutes before rain ceased.

In light of the various issues with this site and assumptions that had to be made, research team members recommend that any speed and lane encroachment effects should be evaluated carefully. This site should only be considered as potentially supplementing any findings from the other test sites.

US-15/501, Chapel Hill, North Carolina

This particular project was a bridge rehabilitation project approximately 0.9 miles (1.4 km) in length located on US-15/501 just north of Chapel Hill. Figure 10 shows an aerial photo of the entire construction site. The site is located near the Durham and Orange County borders, with US-15/501 intersecting I-40 approximately 1 mile (1.6 km) beyond the southern crossover point.

Figure 10. Photo. US-15/501 crossover locations.

Figure 10. Photo. US-15/501 crossover locations.

After the completion of the first of two bridges, southbound traffic was transitioned over to the northbound travel way to a temporary bridge built adjacent to the existing bridge. Two temporary lanes of traffic were constructed and separated from northbound traffic with a barrier. After crossing the temporary bridge, traffic was redirected back into the original southbound lanes. This second crossover was the focus of this study because of the nearby traffic signal at the northern crossover, especially the queues that spilled back into the crossover from the exit. Figure 11 is a photograph of the southern crossover. The research team installed a mast with an omnidirectional camera to observe this southern crossover point during both analysis periods (see figure 12) as both pavement markings were installed at this location in a staggered fashion. During the data collection months of May and July 2010, the average monthly rainfall was above 4.1 inches (10.4 cm), and there was a minimum of 13 hours and 38 minutes between sunset and sunrise.(22)

Figure 11. Photo. US-15/501 southern crossover – standard pavement marking and the AWP study site.

Figure 11. Photo. US-15/501 southern crossover – standard pavement marking and the AWP study site.

Figure 12. Photo. Nighttime camera views of US-15/501 northern crossover.

Figure 12. Photo. Nighttime camera views of US-15/501 northern crossover.

Both crossovers were two-lane movements. The comparison crossover consisted of standard pavement marking and a median barrier adjacent to the left lane. The treatment crossover was similar in that the AWP and median barrier were the only devices providing information to drivers regarding the configuration of the travel lanes. The speed limit was 55 miles per hour (89 kph). Both the comparison and the treatment sites had S-curve geometry similar to that of the Henderson site. However, the curves were slightly less abrupt than at the Henderson site, with larger radii and longer curve lengths.

US-421, Winston-Salem, North Carolina

The Winston-Salem site was similar to the Chapel Hill site in that it was also a bridge rehabilitation project. This site is located on US-421 approximately 2 miles (3.2 km) from I-40, Exit 188, between South Peace Haven Road and Lewisville Clemmons Road. An aerial photo of the site is shown in figure 13.

Figure 13. Photo. US-421 crossover locations.

Figure 13. Photo. US-421 crossover locations.

Temporary bridges were installed in each direction while new bridges were built in the existing locations. After the completion of these temporary bridges, traffic lanes were shifted onto them, with approximately 500 feet (152.4 m) of space between the bridge and each crossover. The research team monitored the eastern and western crossover points for both directions, as opposed to only one crossover point, which was the case at the US-15/501 site.

The lanes traveling eastbound were marked with standard pavement markings, designating it the comparison side. Consequently, the two westbound lanes were marked with the AWP, thus being the treatment side. Both sides had a barrier between the left travel lanes and the construction area that transitioned from barrels to concrete barrier and back to barrels. Additionally, during the course of the research, RPMs were installed on the comparison side, unbeknownst to the research team. Soon after finding out about the installation of the RPMs, a member of the team went to the site and, with the help of the North Carolina DOT, uninstalled them. This was done to obtain a fair and accurate comparison assessment of the visibility of the standard pavement markings and that of the AWP pavement markings.

The site was monitored using standard closed circuit television (CCTV) cameras, which were attached to two separate masts that were erected in the median a few hundred feet past the eastern and western crossover points looking directly down the lane lines. Figure 14 displays screen shots from the videos of the four curves. The use of two masts allowed the researchers to collect data for all four crossover points simultaneously. Because rain duration and intensity were the same on each of the crossovers, the simultaneous data collection was optimal for the data analysis. During the data collection months (May through October 2011), the average monthly rainfall was 3.5 inches (8.9 cm), and there were a minimum of 12 hours and 58 minutes between sunset and sunrise.(22)

Figure 14. Photos. Four curves at the US-421 test site.

Figure 14. Photos. Four curves at the US-421 test site.

US-32/33/50, Athens, Ohio

The US-32/33/50 project involved the reconstruction of the existing highway over a distance of approximately 2 miles. The Athens study site was located on US-32/33/50 between the interchanges with State Route 682 (SR 682) and Township Highway 60 (Blackburn Rd.). When US-33 and US-32/50 split near the end of the work zone, the study area continued on US-32/50 until just before Township Highway 60. The study area is shown in figure 15.

Figure 15. Photo. US-32/33/50 work zones.

Figure 15. Photo. US-32/33/50 work zones.

According to the Ohio DOT, the annual average daily traffic (AADT) in 2009 for the section of US-32/33/50 was 17,960 vehicles per day, with trucks representing 8 percent of the total traffic. For the section of US-32/50, the AADT was 13,050 vehicles per day, with trucks representing 7 percent of the total traffic.(23) During the data collection months from April through June 2010, the average monthly rainfall was above 3.4 inches (8.6 cm), and there were a minimum of 12 hours and 39 minutes between sunset and sunrise.(22) The AWP was used for the work zone located along the eastbound lanes of US-32/33/50, while standard pavement marking was used for the work zone located along the westbound lanes. For each type of pavement marking, an area before, during, and after the lane shift was studied for the work zones. Figure 16 shows the work zones during the daytime hours, and figure 17 shows the work zones during the nighttime hours.

Figure 16. Photos. The AWP (left) and standard pavement marking (right) work zones on eastbound US-32/33/50 during the day.

Figure 16. Photos. The AWP (left) and standard pavement marking (right) work zones on eastbound US-32/33/50 during the day.

Figure 17. Photos. The AWP (left) and standard pavement marking (right) work zones on eastbound US-32/33/50 at night.

Figure 17. Photos. The AWP (left) and standard pavement marking (right) work zones on eastbound US-32/33/50 at night.

While a site with a lane drop was not desirable, the site was selected due to proximity to the research office and capabilities of obtaining data during all weather conditions: daytime dry, nighttime dry, daytime wet, and nighttime wet. During construction, traffic was shifted over one lane to accommodate construction of half of the roadway at a time. The construction began during the spring of 2010 and continued through the fall of 2010. Data were only collected when the traffic was shifted to the inside shoulder of each direction of travel. During the construction, all exit ramps were open to traffic. Throughout each of the work zones the posted speed limit was 55 miles per hour (89 kph).

The data from this site were collected with video cameras mounted inside research vehicles that followed subject vehicles through the work zone. Data were collected for both directions of travel during the same time periods over several months. The amount of precipitation was noted during each rain event. As the data collection for the Athens site occurred over a several-week period, the average rainfall during the periods of data collection equaled 0.53 inches (1.3 cm) for the nighttime data collection and 0.54 inches (1.4 cm) for the daytime data collection.

I-90, Ashtabula, Ohio

The I-90 project involved the reconstruction of the existing concrete pavement over a distance of approximately 5.75 miles (9.3 km) between Paine Road in Lake County to the Ashtabula County Line. Specifically, the Ashtabula study site was located on I-90 between the interchanges with State Route 11 (SR-11) and State Routes 84 and 193 (SR-84 and SR-193), as shown in figure 18.

Figure 18. Photo. I-90 crossover locations.

Figure 18. Photo. I-90 crossover locations.

According to the Ohio DOT, the AADT in 2009 for this section of I-90 was 22,910 vehicles per day, with trucks representing 30 percent of the traffic.(24) During the data collection months of August and September 2011, the average monthly rainfall was above 3.8 inches (9.7 cm), and there were a minimum of 11 hours 48 minutes between sunset and sunrise.(22) The AWP was used for the double-lane crossover on the eastbound lanes of I-90, as shown in figure 19. Standard pavement marking was used for the double-lane crossover on the westbound lanes of I-90, as shown in figure 20. Views of each of the crossovers from within a vehicle can be seen in figure 21.

Figure 19. Photo. AWP treatment location on I-90.

Figure 19. Photo. AWP treatment location on I-90.

Figure 20. Photo. Standard pavement marking treatment location on I-90.

Figure 20. Photo. Standard pavement marking treatment location on I-90.

Figure 21. Photos. View of the AWP (left) and standard pavement marking (right) double-lane crossover.

Figure 21. Photos. View of the AWP (left) and standard pavement marking (right) double-lane crossover.

During the construction period, the two eastbound travel lanes were shifted to the two westbound travel lanes through a two-lane crossover. After crossing the construction site, the eastbound travel lanes were shifted back to their original travel path. The construction of the eastbound travel lanes began in late November of 2010. At the onset of the project, neither the AWP nor standard pavement marking were installed along the two-lane crossovers due to concerns with temperatures at the site, which were substantially below the desired 50-degree Fahrenheit (10 degrees Celsius) temperature for installation of the pavement markings. In the interim, a fast-dry temporary pavement marking was utilized for the pavement marking with raised pavement markers throughout the work zone until the temperatures in the spring (April) were adequate to install the AWP and standard pavement markings.

Both crossovers utilized similar geometric features, such as a reverse curve with a radius of 3,125 feet (952.5 m). The total length of the western crossover was 1,095 feet (333.8 m), and the total length of the eastern crossover was 1,085 feet (331 m). Throughout the work zone, the posted speed limit was 55 miles per hour (89 kph) for both directions of travel.

Data at the Ashtabula site were collected with sensors along the entire length of each crossover. Due to the number of sensors required for each crossover, the data were not collected simultaneously on both crossovers. Therefore, the amount of precipitation was noted for each period to assure similar precipitation events were compared in the analysis. The cumulative precipitation occurring during the 2-hour data collection for the AWP was 0.34 inches (.86 cm), while the precipitation for the standard pavement marking was 0.42 inches (1.1 cm). As the rainfall amounts were similar during the nighttime data collection periods, rainfall was not considered a confounding variable.

Discussion of Potential for Data Collection

The data collection for this project ultimately depended on the potential for rain events during nighttime conditions. The research team looked into the potential for data collection at each of the five sites by summarizing the average rainfall and nighttime hours during each month of the year, as shown in tables 3 and 4. Average rainfall totals are provided as the average monthly rainfall from 2003 to 2008. The available time for data collection during nighttime hours was calculated as the available time between ensuing darkness (30 minutes after sunset) and midnight, after which it was assumed that data collection probably was not feasible due to low traffic volumes and extended team travel time.

Table 3. Average monthly rainfall totals (inches).(22)

Location Jan Feb Mar Apr May June July Aug Sep Oct Nov Dec
Henderson, NC 3.6 3.5 3.8 3.0 3.9 3.8 4.3 4.7 3.7 3.4 3.3 3.3
Chapel Hill NC 3.7 3.9 4.3 3.3 4.5 4.4 4.1 4.4 3.3 3.5 3.5 3.5
Winston-Salem NC 3.3 3.3 3.7 2.8 4.0 3.8 4.5 3.9 3.5 3.5 3.0 3.4
Ashtabula, OH 2.7 2.2 2.8 3.6 4.0 4.5 4.7 3.8 4.3 4.0 3.7 3.5
Athens, OH 2.6 2.7 3.3 3.4 4.4 3.7 4.4 3.3 2.8 2.7 3.3 3.0

1 inch = 25.4 mm

Table 4. Average available time for nighttime data collection (hh:mm).(22)

Location Jan Feb Mar Apr May June July Aug Sep Oct Nov Dec
Henderson, NC 6:08 5:35 5:09 4:42 4:17 3:57 3:59 4:26 5:09 5:52 6:25 6:30
Chapel Hill NC 6:04 5:32 5:06 4:40 4:15 3:56 3:57 4:24 5:07 5:49 6:21 6:26
Winston-Salem NC 6:00 5:28 5:02 4:35 4:10 3:50 3:52 4:19 5:02 5:45 6:17 6:22
Ashtabula, OH 6:10 5:55 4:49 3:45 2:54 2:53 2:57 3:29 3:58 4:30 5:55 6:52
Athens, OH 5:58 5:46 4:44 3:44 2:55 2:36 2:39 3:09 3:33 5:00 5:23 6:20

The rainfall totals were interesting. Surprisingly, the average rainfall for each of the sites is fairly constant over time, with little variation between the seasons of the year. However, the available data collection time improves during winter months, as the winter solstice takes place in mid-December. The earlier sunset also affords the potential for higher traffic volumes during data collection, since the researchers would be collecting a portion of the peak hour traffic in some cases. Therefore, sites studied during winter months have the higher likelihood of collecting larger samples of data because the nighttime data collection time period is longer and includes potentially higher total traffic volumes by including a portion of the peak period travel.

Retroreflectivity

Retroreflectivity is a measure of the ability of a sign, pavement marking, or other countermeasure to reflect light back toward the source (em.e., the driver in the car). The research team took retroreflectivity measurements of standard pavement marking and AWP during the same day the markings were applied. The team formally documented the field installations of each pavement marking type from four different contractors in two different States, each given basic training and/or guidance from 3M on the installation of the new AWP prior to installation. No guidance was given on how to install the standard pavement marking, allowing the research team to document real-world field installations currently used in practice based on the specifications given by the respective DOTs.

Data Collection

Retroreflectometer

The retroreflectometer used in this study is in accordance with the standard test method for measurement of retroreflective pavement marking materials: it has an illumination angle of 1.24 degrees and an observation angle of 2.29 degrees, simulating an observation distance of 98.4 feet (30 m). The illumination and observation fields are shown in figure 22. Note how the illumination field, when measuring retroreflectivity of a wet continuous pavement, is not immediately below the device, as opposed to normal operation. This shift occurs when the wet night rails are attached, allowing the machine to measure pavement continually wetted by a water sprayer directly in front of it. All tests were to be performed with the wet night rails attached, per 3M instruction.

Figure 22. Diagram. Retroreflectance measurement fields.
1 mm = .04 inch

Figure 22. Diagram. Retroreflectance measurement fields.

Retroreflectivity values were collected approximately 30 minutes after the pavement markings were laid on the road surface. Before collecting retroreflectivity data, the section of pavement marking was swept to remove any loose optical elements and glass beads that did not bind to the pavement marking. This allowed for a more accurate retroreflective measurement because, over time, these elements would dislodge and scatter. One by one, three retroreflectivity tests were conducted on the test segment: the standard dry test (ASTM E1710-05), the wet continuous test (ASTM E2177-01), and the wet recovery test (ASTM E2176-01).

For the wetting of the pavement during the wet continuous test, the research team used a 4-gallon (15-liter) rechargeable sprayer. As suggested by 3M, the research team added a few drops of regular dishwashing soap to the water in the sprayer’s reservoir, about 1 tablespoon (14.8 ml) for 3 gallons (11.4 liters) of water, which is enough to give the solution a very light tint. Per the ASTM test specifications, the water was sprayed on the pavement marking at a rate of 9 inches per hour, and retroreflectivity readings were measured until the displayed values were constant with each recording. The water spraying then ceased, and the wet recovery retroreflectivity value was measured 45 seconds after.

Retroreflectivity Test and Samples

Following the aforementioned procedures, retroreflectivity tests were conducted at four of the five test sites: I-85 (Henderson, NC), US-15/501 (Chapel Hill, NC), US-421 (Winston-Salem, NC), and I-90 (Ashtabula, OH). Retroreflectivity measurements were taken at the US-32/33/50 site (Athens, OH); however, immediately after the markings were applied for both sections (the AWP and standard pavement marking), the contractor sprayed water and applied straw to the median, which affected the measurements. Therefore, these data were not included in the analyses.

Sample sizes were dependent on the team’s ability to collect data in various events. For instance, the Henderson site was a large facility that had a rolling road block. The first two rolling road blocks allowed each of the markings to dry. Then several more rolling road blocks were conducted based on the willingness of the contractor at the site. In all, a minimum of 10 retroreflectivity readings were conducted for each pavement marking type under the 3 simulated weather conditions (dry, wet continuous, and wet recovery). In addition, both yellow and white pavement markings were sampled. Table 5 shows the individual sample for each simulated weather condition for each pavement marking color and type. For example, in Henderson, 11 white treatment pavement marking readings was taken under dry conditions, followed by 11 separate wet and wet recovery readings.

Table 5. Retroreflectivity sample size obtained at each of the four sites studied.

Site Treatment (AWP) Comparison (Standard)
White Yellow White Yellow

I-85

Henderson, NC

11

11

11

11

US-15/501

Chapel Hill, NC

10

10

10

10

US-421

Winston-Salem, NC

22

26

20

16

I-90

Ashtabula, OH

9

6

11

6

Neither the I-85 nor the US-15/501 data sets needed any reduction in the measurements taken. As for the US-421 test site, three retroreflectometer readings were taken on the temporary bridge installments on both directions of the highway for both white and yellow striping under each simulated weather condition. The bridge deck striping required additional thermo-treatment prior to pavement marking installation. The inclusion of thermoplastic greatly enhanced the retroreflectivity and would bias the results; therefore, these measurements taken on the bridge decks were removed from the samples to be analyzed.

Analysis Methodology

At each test site, there were a variety of different retroreflectivity measurements. The three factors that distinguished results included (1) pavement marking treatment (the AWP or standard pavement marking), (2) pavement marking color (white or yellow), and (3) simulated weather conditions (dry, wet continuous, or wet recovery). Therefore, each site had 12 distinct retroreflectivity measurement sets. The measurement sets, defined by the combinations of factors, are listed in table 6.

Table 6. List of retroreflectivity samples at each site.

Retroreflectivity Samples
# Pavement Marking Treatment Pavement Marking Color Simulated Weather Condition

1

AWP

White

Dry

2

AWP

White

Wet Continuous

3

AWP

White

Wet Recovery

4

AWP

Yellow

Dry

5

AWP

Yellow

Wet Continuous

6

AWP

Yellow

Wet Recovery

7

Standard

White

Dry

8

Standard

White

Wet Continuous

9

Standard

White

Wet Recovery

10

Standard

Yellow

Dry

11

Standard

Yellow

Wet Continuous

12

Standard

Yellow

Wet Recovery

For each sample, the average and standard deviation were calculated. A standard two-sample t-test was conducted assuming unequal variances. Samples treated with the AWP pavement marking were compared to standard pavement marking treatment samples of the same striping color under the same simulated weather condition, and at the same locations for each weather condition. Finally, for each comparison, the percent difference between retroreflectivity averages was calculated.

Results

There were marked differences observed between retroreflectivity readings at each of the three sites even when treatments, striping colors, pavement type, and weather conditions were the same. Therefore, sites were rendered independent of one another, and the samples were analyzed separately.

For each site, the averages and standard deviations were calculated for the two types of pavement marking systems installed for both white and yellow striping under the three simulated weather conditions. The mean difference was calculated by subtracting each standard pavement marking average (baseline) from the AWP average. The percent difference was found by dividing the mean difference by the baseline standard pavement marking average. An unpaired two sample t-test was used to compare all the AWP samples with standard samples. The significance level was set to a 95 percent confidence interval. The null hypothesis assumed that the two pavement marking samples had similar retroreflectivity measurements. All four sites are summarized in table 7.

Table 7. Summary of statistics for retroreflectivity readings at all test sites.

Retroreflectivity White Yellow
(mcd/m2/lx)
Marking
Dry Wet Continuous Wet Recovery Dry Wet Continuous Wet Recovery

I-85 (Henderson, NC)

Standard

Avg.

344.6

220.0

315.7

155.5

101.4

137.7

Std. Dev.

24.6

12.8

21.9

17.3

15.3

17.7

AWP

Avg.

740.7

464.0

679.8

561.4

386.0

461.5

Std. Dev.

65.4

202.2

87.8

70.2

77.5

35.6

Mean Difference

+396.1

+244.0

+364.1

+405.9

+284.6

+323.8

Percent Difference

+114.9%

+110.9%

+115.3%

+261.0%

+280.7%

+235.1%

P(T<=t) two-tail

< 0.001*

0.0042*

< 0.001*

< 0.001*

< 0.001*

< 0.001*

US-15/501 (Chapel Hill, NC)

Standard

Avg.

166.2

10.2

9.3

122.6

8.8

37.9

Std. Dev.

40.7

1.7

9.8

20.9

1.0

9.2

AWP

Avg.

507.2

107.6

294.5

418.5

44.6

223.0

Std. Dev.

175.4

39.2

123.9

133.0

16.8

76.3

Mean Difference

+341.0

+97.4

+285.2

+295.9

+35.8

+185.1

Percent Difference

+205.2%

+954.9%

+3066.7%

+241.4%

+406.8%

+488.4%

P(T<=t) two-tail

< 0.001*

< 0.001*

< 0.001*

< 0.001*

< 0.001*

< 0.001*

US-421 (Winston-Salem, NC)

Standard

Avg.

274.9

37.9

54.7

194.0

22.5

26.7

Std. Dev.

33.1

10.0

13.4

12.9

8.0

3.6

AWP

Avg.

463.2

39.8

140.4

372.6

36.8

165.7

Std. Dev.

36.0

17.9

41.4

41.1

12.2

26.5

Mean Difference

+188.3

+1.9

+85.7

+178.6

+14.3

+139.0

Percent Difference

+68.5%

+5.0%

+156.7%

+92.1%

+63.6%

+520.6%

P(T<=t) two-tail

< 0.001*

0.69

< 0.001*

< 0.001*

< 0.001*

< 0.001*

I-90 (Ashtabula, OH)

Standard

Avg.

110.5

5.3

9.0

35.5

4.0

2.7

Std. Dev.

32.3

1.7

3.5

9.7

0.8

1.9

AWP

Avg.

536.1

17.2

230.3

263.2

17.4

166.2

Std. Dev.

167

6.3

88.7

44.4

5.7

68.9

Mean Difference

+425.6

+11.9

+221.3

+227.7

+13.4

+163.5

Percent Difference

+79.4%

+69.2%

+96.1%

+86.5%

+77.0%

+98.4%

P(T<=t) two-tail

< 0.001*

< 0.001*

< 0.001*

< 0.001*

0.002*

0.002*

* p-value is statistically significant (α = 0.05)

The p-values for the overwhelming majority of retroreflectivity tests were less than the significance level selected (α = 0.05); therefore, the null hypotheses that the two pavement markings had similar retroreflectivity measurements were rejected. The results of the t-tests imply that a statistically significant difference can be expected between the AWP and standard pavement markings’ retroreflectivity during all three weather scenarios and pavement marking colors at each individual site. Major improvements were noted across the board when the AWP was used; however, the most significant findings were in the increased retroreflectivity results for yellow pavement marking. In general, yellow pavement marking was less reflective than white pavement marking for standard marking and the AWP, so the notable increase in retroreflectivity could be very important for transportation agencies.

Research team members cannot readily explain the major differences in retroreflectivity between sites. Two possible theories may explain the differences. First, although a 3M representative was on-hand for the calibration of machinery and during the actual application of the pavement markings, the research team suspects there was some human error during the application process due to the differences in machinery used and experience and comfort of the team in actually applying different pavement marking mixtures. If this is true, pavement marking applications should be expected to have a lot of variability, which suggests that minimum pavement marking retroreflectivity standards should be considered by public entities (indeed, many agencies use such standards). Second, the team suspects that there may have been differences in results based on the ability of the pavement marking to shed water in some instances, possibly due to differences in pavement type, pavement roughness, or maybe even grade or crown in various roadways studied. Unfortunately, this possibility was not accounted for in the analysis.

Conclusions and Recommendations

As shown in table 7, based on the collection of retroreflectivity readings and the statistical analysis, it may be concluded that, when correctly installed, the AWP markings will provide a statistically significant increase in retroreflection when compared to standard pavement markings. In addition, the variability in the retroreflectivity readings for both pavement markings types being studied suggests that pavement marking contractors are not using consistent methods for applying their pavement markings.

In this study, the retroreflectivity readings were taken only after the initial installment of pavement markings. As the temporary work zone markings are expected to last at least 1 year, it is suggested that further investigation be conducted in examining how the AWP markings degrade over varying periods of time and under various traffic conditions.

Validating that the AWP markings were indeed more retroreflective, the research team further sought to examine whether these pavement markings would increase lane visibility and thereby improve navigation through work zone detours under nighttime, rainy conditions. The measures of effectiveness, including speeds, lane encroachments, and lane deviations, are discussed in the following chapters.

Speeds

AWP has been proven to be more retroreflective than the standard pavement marking. As a follow-up to that finding, the research team sought to discover how this new pavement marking affects driver behavior. The analysis of speed data was used as an indication of a motorist’s perceived risk while traveling through a work zone guided by the AWP or standard pavement marking through lane shifts and crossovers. Comparisons are made between entry and exit curves for both standard and AWP.

Data Collection Methods

LIDAR Speed Gun

Whenever possible, speeds were collected using a Class I laser speed gun. The speed gun uses light detection and ranging (LIDAR) technology, which uses a pulse of light that calculated the distance to the nearest vehicle 200 times per second, allowing for extremely accurate readings when calibrated. According to Laser Atlanta, the gun is capable of providing speeds within ± 1 mile per hour (± 1.16 kph). The speed gun is capable of acquiring its intended target in as little as 0.3 seconds, provided the user can acquire the target quickly and follow it with a steady hand. Speeds were recorded as vehicles were approaching or departing the location of the speed gun user, with no more than ± 5 degrees of angle between the user and vehicle being recorded to reduce any potential accuracy issues.

Two independent observers collected speed data concurrently in the crossover sections at the treatment and comparison locations when at the entry or exit in the same direction of travel. This meant that the population of drivers in the comparison and treatment speed bins was very similar as they entered and exited the work zone. In addition, the rain intensity at both segments was approximately the same throughout the entire data collection period. Of course, longer work zone segments likely had different rain intensities throughout the speed data collection time period; however, the differences were assumed to be negligible since data at longer sites were collected after a suitable period of time elapsed once rain was observed at both sites. If a platoon of vehicles was observed, the lead vehicle of the platoon was recorded as a single measurement and the following vehicles were ignored. This allowed the team to only collect speed data on vehicles that were maneuvering through the crossover at a safe speed as determined by the driver of that particular vehicle. Following vehicles would have likely been affected by the lead vehicle’s decision about what the safe traversing speed was for the curve and would not have been a true representation of what that driver’s speed may have been if not following a platoon.

Video Calculated Speeds

Speed data also were calculated from videos taken at the I-85, US-15/501, US-421, and US-32/33/50 test sites. The research team had three motivations for using videos to collect speed data. First, this method allowed speeds to be obtained during rain events at site locations farther from the research team’s offices (em.e., those requiring longer travel times), especially when the probability of a rain event was low. Second, speed samples could be collected for the exact same drivers through both crossovers at the same location every time, whereas the speed gun measurements included some human error in obtaining speeds at the same location in the transition every time. Third, video was required for lane encroachment analysis and was readily available for speed usage. Using the video taken to collect lane encroachments and volumes, speeds were collected from the same populations examined for lane encroachments rates.

The primary drawbacks to this method were two-fold. First, there is human error when using a stopwatch and known distance to collect speeds. To counter this, a time-stamp overlay was utilized to obtain accurate times within 1/30th of a frame. Second, the angle to the crossover where speeds were taken meant that some occlusion existed. Since occlusion was consistent at all known points, we assumed this to be consistent among all measurements.

Speeds were calculated in the following steps:

  1. Record for no less than 2 hours at each curve.
  2. Take physical distance measurements at each curve.
  3. Calibrate measurements with the speed gun.
  4. Watch the videos in the lab and record the times it took lead vehicles to traverse the measured distances.
  5. Calculate speeds using timing macros and functions in a computer-based database program.

Sensor Calculated Speeds

The research team used sensors to determine the speeds of vehicles traveling along the two-lane crossover on I-90. The sensors were located at 100-foot (30.5-m) intervals along the right and left travel lanes for the eastbound direction in which the AWP pavement markings were utilized. Along the westbound direction of travel, in which the standard pavement marking were utilized, the sensors were only located along the right travel lane due to the lack of an inside shoulder. The speeds were calculated by dividing the spacing of two sequential sensors by the time taken by a vehicle to traverse the distance between the sensors.

The sensors included a SHARP distance measuring sensor unit (Part No. GP2Y0A710K0F) and a MicroStrain SG-Link® Wireless Strain Node. The SHARP sensor was composed of a position-sensitive detector, infrared emitting diode, and a signal processing circuit, which provide a voltage output that corresponds to the detection distance (SHARP, 2006). The operating range for the SHARP sensors is 3.28 to 18.05 feet (1 to 5.5 m). The SG-Link® Wireless Strain Node was used to process the voltage output of the SHARP sensor and transmit it to a MicroStrain WSDA®-Base-mXRS™ Wireless Base Station connected to a laptop. The laptop connected with the Base Station had the MicroStrain Node Commander® software installed to conduct synchronous sampling for multiple SG-Link® Nodes. The SHARP sensor and the SG-Link® Node were installed in a plastic enclosure and were each connected to batteries for a power supply. The internal view of a sensor with the lid of the plastic enclosure removed is shown in figure 23. To securely place the sensors in the field, mounting bases were constructed using lumber, Velcro, and brackets. These mounting bases helped to elevate the sensors so they could measure the axle of the passing vehicle (opposed to the body of the vehicle) without being impeded by the cross-slope of the roadway. Also, the bases provided a means of securing the sensors to the graded shoulder. The sensor placed in the mounting base can be seen in figure 24.

Figure 23. Photo. Internal view of sensor.

Figure 23. Photo. Internal view of sensor.

Figure 24. Photo. Sensor on mounting base.

Figure 24. Photo. Sensor on mounting base.

Once the data were collected, the researchers needed to filter the data since there were continuous voltage readings for the entire data collection period. The team observed spikes in the voltage readings from a constant baseline level. This behavior is expected, as the output voltage increases as the detection distance decreases. Therefore, the baseline low voltage represents nothing present within the operating range, 3.28 to 18.05 feet (1 to 5.5 m). When the voltage increased above 1 volt, the sensor detected an object within the operating range. The researchers determined that to calculate speeds using the data collected, determining the initial voltage peak for each vehicle was necessary. This initial voltage peak would correspond to the front axle of the vehicle. As a result, a macro was created to filter through the data to find the initial voltage peaks. Once all of the readings were properly filtered, the sensor data, which produced a time-stamp for each collected reading, was utilized to determine vehicle speed. Theoretically, as a vehicle progressed through the crossover it would pass in front of each of the sensors and produce readings on each. Thus, the elapsed time from sensor to sensor can be determined by lining up the readings and, subsequently, the speed of the vehicle was calculated.

Analysis Methodology

This section outlines the differences in data collection at each of the sites, any particular problems with data collection, and sample sizes utilized during the data collection effort. Basic statistical analysis was conducted for the speed data collected at each test site after outliers were removed. First, averages and standard deviations were calculated from the filtered speed data. The samples’ averages and mean differences were then compared and calculated. Two different statistical tests were utilized. For North Carolina sites, the unpaired t-tests assuming unequal variances were used to test the significance levels in comparing speed samples. For Ohio sites, analysis of variance (ANOVA) was used. All sites were independent of one another and required separate analyses.

I-85

The team sought to obtain speeds during nighttime rain events through the northern and southern crossovers at I-85 using a LIDAR speed gun. However, due to the variable nature of rain events and the lack of operators and equipment, the team was only able to record a single speed sample at each crossover. The two speed samples were taken from vehicles well within their respective crossovers. Samples of only 15 reliable speeds were taken in the northern crossover (standard pavement marking), and 46 were taken in the southern crossover (the AWP), each sample in a platoon only including the lead vehicle. The samples were a bit small for any confidence in statistical analysis, but they are nonetheless available.

Upon further investigation in the office, camera angles at the northern crossover allowed speeds to be obtained at the entry and exit curves. (Collecting speeds at the southern crossover using video was not feasible because the relative camera angles presented too much occlusion from vehicles to consistently and accurately confirm distance.) Although not the primary goal, the team found it useful to examine the possibility of mean speed differences at the entry and exit of this single crossover. From the video, sample sizes of 208 and 211 were collected from the northern crossovers’ entry and exit curves, respectively. Outliers less or greater than 2 standard deviations were removed from the samples, reducing the sample sizes to 200 and 202, respectively.

Given the two speed samples, initially two separate statistical comparisons were conducted for data collected using both available samples. First, the mean difference was calculated for the northern and southern crossovers. As the sample sizes were small when comparing the AWP to standard pavement marking, the analysts decided against further analysis of this initial data set, since the findings could be incorrectly used with findings from other sites with similar analyses but much more data. Therefore, the only remaining speed analysis conducted at this particular site was from the northern crossover entry and exit curves. The two samples were compared using the unpaired two-sample t-test assuming unequal variances.

US-15/501

To obtain speeds at the US-15/501 test site, the team used the alternative video extraction method. As mentioned earlier, for this site only one curve was feasible for analysis due to the signal at the northern crossover. The research team worked with the pavement marking contractor and 3M to arrange for the AWP to be installed on top of the standard pavement marking after a series of rain event data had been collected on the standard pavement marking. After the AWP installation, video footage of nighttime rain conditions was again recorded from the same camera installation locations for future data extraction.

Although other samples of rainy nighttime data were available, only two samples were utilized: one sample during the time the lane shift was treated with standard pavement marking and the other after the AWP overlay. The primary reason for this decision was the team hypothesized that rain intensity would have a marked effect on drivers’ ability to safely traverse the work zone, so comparisons of greatly different rain events were not examined.

During the data collection for standard pavement marking, the rain intensity was approximately 0.50 inches (1.27 cm) per hour, while the approximate rain intensity for the AWP overlay was 0.40 inches (1.0 cm) per hour. A sample of 218 speeds was collected when the site was treated with standard pavement marking, and 275 speeds were collected with the AWP. The sample size was sufficiently large; therefore, the analyst removed outliers beyond 2 standard deviations from both samples, thereby reducing the samples sizes to 212 speeds and 262 speeds for standard pavement marking and the AWP, respectively. The two samples were compared using the unpaired two-sample t-test assuming unequal variances.

US-421

As described earlier, the field setup for the US-421 site was ideal for data collection because of the two bridges being reconstructed simultaneously. Therefore, both entry and exit crossovers along each direction of travel were analyzed independently, which is detailed previously in figure 14. This allowed the research team to utilize four synchronized videos of each crossover location. Speeds were taken for over 200 lead vehicles at each of the curves.

In all, 240 and 248 speeds were recorded at the AWP entry and exit curves, respectively, and 219 and 233 speeds were recorded for the standard pavement marking entry and exit curves, respectively. Again, any speeds less than or greater than 2 standard deviations of their curve speed averages were deemed outliers and filtered from the data. After the data filtration, there were 227, 239, 205, and 220 speeds for the AWP and standard pavement marking curves. The t-tests conducted compared the AWP entry speeds versus standard pavement marking entry speeds, the AWP exit speeds versus standard pavement marking exits speed, the AWP entry versus the AWP exit, and standard pavement marking entry versus standard pavement marking exit during nighttime rain conditions.

US-32/33/50

To obtain speeds at the US-32/33/50 test site, the video extraction method was utilized. Based on the placement of the video cameras, the speed data could be extracted by subdividing the work zone into three separate sections: at the beginning of the lane shift, in the midpoint of the lane shift, and at the end of the lane shift. Due to the location of an exit ramp near the lane shift, any vehicle exiting the highway via the ramp was excluded from the positional speed analysis. Similar vehicles were utilized for the speed data extraction as well as the lateral lane placement data. Data were collected during daytime and nighttime rain conditions as well as during nighttime dry conditions through the months of May, June, and July 2010. The sample sizes for the data collection ranged from 54 vehicles to 121 vehicles due to the elimination of vehicles that were not lead vehicles in a platoon and those that did not continue through the entire lane shift.

The samples at the US-32/33/50 test site included speeds along the travel lanes for entering, within, and exiting the lane shifts for both the AWP pavement markings and the standard pavement markings. A one-way analysis of variance was conducted to compare the speeds using several hypotheses, listed below:

  • Mean speeds were similar for entering, within, and exiting lane shifts for the AWP and standard pavement marking sites.
  • Mean entering speeds were similar for the AWP as compared to the standard pavement marking.
  • Mean speeds within the lane shift were similar for the AWP as compared to the standard pavement marking.
  • Mean exiting speeds were similar for the AWP as compared to the standard pavement marking.

I-90

Speeds at the I-90 test site were collected using sensors. Data were collected during daytime and nighttime dry conditions as well as during nighttime wet conditions through the months of August and September 2011. The sample sizes for the data collection ranged from 300 vehicles to 1,504 vehicles. The samples at the I-90 test site included speeds along the travel lanes for both the standard pavement marking and the AWP crossovers. A one-way analysis of variance was conducted to compare the speeds assuming a null hypothesis that stated that the mean speeds were similar for the AWP and the standard pavement marking sites. The analyses were conducted for daytime dry conditions, nighttime rain conditions, and nighttime dry conditions.

Results

From the speed data obtained by LIDAR, video extraction, and sensors, the research team compared various samples for each of the five test sites. Mean speeds were particularly dependent on site geometry, conditions, and characteristics; therefore, samples would only be compared with other samples from the same site. The mean speeds, standard deviations, and mean differences were first calculated, followed by a statistical analysis using either t-tests or ANOVA. Table 8 displays a summary table of the speed data collected and analyzed for all five test sites.

Table 8. Summary of statistics for speeds.

Site Light Weather Paint Location Mean speed (mph) Standard Deviation Mean Difference P(T<=t) two-tail
I-85 Night Rain Standard Entry Curve 44.9 5.2 1.4 0.009*
Standard Exit Curve 46.3 5.2
US-15-501 Night Rain Standard Entry Curve 38.7 6.2 3.5 <0.001*
AWP Entry Curve 42.2 6.4
US-421 Night Rain Standard Entry Curve 50.7 4.3 0.8 0.057
AWP Entry Curve 51.5 4.6
Standard Exit Curve 56.7 5.2 -1.3 0.039*
AWP Exit Curve 55.3 8.3
Standard Entry Curve 50.7 4.3 5.9 <0.001*
Standard Exit Curve 56.7 5.2
AWP Entry Curve 51.5 4.6 3.8 <0.001*
AWP Exit Curve 55.3 8.3
US-32/33/50 Day Rain Standard Entry Curve 35.7 3.6 1.6 0.526
AWP Entry Curve 37.3 6.4
Standard Within Workzone 36.4 5.2 1.5 0.476
AWP Within Workzone 37.9 5.1
Standard Exit Curve 36.5 5.9 1.6 0.59
AWP Exit Curve 38.1 5.8
Night Clear Standard Entry Curve 34.4 5.3 4.5 <0.001*
AWP Entry Curve 38.9 4.7
Standard Within Workzone 37.2 4.8 0.6 0.982
AWP Within Workzone 37.8 5.5
Standard Exit Curve 38.6 4.2 -2.3 0.057
AWP Exit Curve 36.3 5.0
Night Clear Standard Entry Curve 36.3 3.7 2.9 <0.001*
AWP Entry Curve 39.1 3.5
Standard Within Workzone 37.8 4.1 0.3 0.997
AWP Within Workzone 38.1 3.5
Standard Exit Curve 39.3 5.1 -1.8 0.102
AWP Exit Curve 37.6 3.5
I-90 Day Clear Standard Exit Curve 52.8 8.3 4.0 0.714
AWP Entry Curve 56.8 11.0
Night Clear Standard Exit Curve 55.9 9.8 -1.3 0.312
AWP Entry Curve 54.6 8.7
Night Rain Standard Exit Curve 57.0 7.8 -5.0 <0.001*
AWP Entry Curve 52.1 10.1

* p-value is statistically significant (α = 0.05)
1 mph = 1.6 kph

I-85

Speeds were compared between the entry and exit curves delineated with standard pavement marking at the northern crossover. The work zone speed limit was posted at 55 miles per hour (89 kph). Actual speeds at the northern crossover ranged from 44.9 to 46.3 miles per hour (72.3 to 74.5 kph). From the analysis, a higher mean speed was found for the exit curve. Furthermore, the p-value was statistically significant with 95 percent confidence. Speeds were likely lower than the posted speed limit because the site contained a single lane crossover with barriers on each side.

When comparing the speeds obtained from LIDAR speed guns, the northern crossover yielded a mean speed of 43.9 miles per hour (70.7 kph) based on a very limited sample size of 15, whereas the AWP (southern) crossover yielded a mean speed of 41.7 miles per hour (67.1 kph) using a sample size of 46. This resulted in a mean difference of 2.2 miles per hour (3.5 kph). From this limited analysis, drivers drove faster through the section with standard pavement marking than through the section with the AWP.

US-15/501

From the statistical analysis of the two samples obtained at US-15/501, mean speeds of 38.7 and 42.2 miles per hour (62.3 to 68 kph) were calculated for the standard pavement marking and the AWP entry curves, respectively. At a confidence interval of 95 percent, the mean speed of the AWP data was significantly greater (3.5 miles per hour, or 5.6 kph) than that of the standard pavement marking data. Both averages were noticeably lower than the posted speed limit of 45 miles per hour (72 kph).

US-421

The averages and standard deviations were calculated for the samples collected for each of the curves at the US-421 test site. The work zone speed limit was posted at 55 miles per hour (89 kph). Based on results from video data extraction, the mean speeds on each of the four crossovers ranged between 50.7 and 56.7 miles per hour (81.6 and 91.2 kph). Basic findings from entry and exit curves include the following:

  • For both standard pavement marking curves and the AWP curves, the average speeds at the entry curves were lower than those of the exit curves.
  • Average speeds on both entering curves were less than the posted speed limit.
  • Average speeds on both exiting curves were just above the posted speed limit.
  • Differences between the entering and exiting speeds were 3.0 to 6.0 miles per hour (4.8 to 9.7 kph).

More important are the results when comparing standard pavement marking and the AWP and the entry versus exit for each pavement marking type. The findings from this comparison at this site are the following:

  • The entry curve speeds in the sections with the standard pavement marking were compared to those with the AWP. The mean speed was slightly greater than 1 mile per hour (1.6 kph) faster for the AWP. A p-value of 0.057 indicates that there was no statistically significant difference between speeds at the entry curves of both pavement markings, although the difference was practically significant (90+ percent confidence).

  • In comparing the exit curves, the mean difference was just below 1.5 miles per hour (2.4 kph). Surprisingly, the exit curve with standard pavement marking had a higher mean speed than that with the AWP. The p-value calculated shows this finding to be significant with 95 percent confidence.

  • Finally, when comparing entry and exit curves of the same pavement marking, there was a mean difference of at least 3.5 miles per hour (5.6 kph), again indicating drivers exit the work zone faster than when entering. Both t-tests give p-values much less than the significance level.

The findings from the entry and exit curve comparison support the findings from the I-85 test site, in which higher speeds were observed in the exit curve. In addition, although not statistically significant, the findings between standard and the AWP pavement markings at similar crossovers were comparable to the results from US-15/501.

US-32/33/50

The averages and standard deviations were calculated for the samples collected for each of the crossovers at the US-32/33/50 test site. The posted work zone speed limit was set at 55 miles per hour (89 kph). Based on results from video data extraction, the mean speeds ranged between 34.4 to 39.3 miles per hour (55.2 to 63.2 kph) for the various conditions—much lower than the posted speed limit. Substantial deviations from the posted speed limit can likely be attributed to the barriers present along the work zone, the single lane, and even the high visitor population that was not familiar with the roadway geometry.

Statistical tests were used to determine if the mean speeds of the AWP crossovers were significantly different from those of the standard pavement marking crossovers. The one-way analysis of variance was utilized to compare the mean speeds. Due to heterogeneous variances, the Welch’s modification to the one-way analysis of variance was utilized, and the calculated F-value was based on an asymptotic distribution. In addition, due to the unequal variances and unequal sample sizes, the Games Howell post hoc test was utilized to examine specific differences within those samples tested in the one-way analysis of variance. Based on the one-way analysis of variance, it was determined that the mean speeds associated with the vehicles entering, within, and exiting the lane shift for the daytime rain, nighttime dry, and nighttime wet conditions were significantly different, as shown in table 9.

Table 9. US-32/33/50 speed ANOVA results.

Comparison Sum of Squares Degrees of Freedom Mean Squares F-Calc p-value

Daytime Rain Conditions

Between Groups

Within Groups

Total

328.57

14159.23

14487.79

5

183.62

65.713

27.44

2.843

0.17

Nighttime Dry Conditions

Between Groups

Within Groups

Total

555.89

7360.29

7916.17

5

200.07

111.18

15.93

7.171

0.000*

Nighttime Rain Conditions

Between Groups

Within Groups

Total

1044.29

9870.81

10915.10

5

411

208.86

24.02

8.696

0.000*

* p-value is statistically significant (α = 0.05)

For US-32/33/50, the ANOVA results indicated the following:

  • The only statistically significant differences between the AWP and standard pavement marking were during nighttime dry and rainy conditions.
  • Although not statistically significant, it is noteworthy that the AWP and standard pavement marking seem to have a large effect on speeds exiting the crossovers at nighttime dry and rainy conditions (p-values of 0.102 and 0.057, respectively).

Another trend is found when examining the data in a graphical visualization. As figure 25 shows, when plotting the data by location in the work zone (entry, within, and exiting) versus speed, it is evident that speeds generally increased through the work zone in all conditions except nighttime with the AWP.

Figure 25. Graphs. US-32/33/50 speed results by location in work zone.
1 mph = 1.6 kph

Figure 25. Graphs. US-32/33/50 speed results by location in work zone.

These findings suggest that the speed differential from entry to exit for the AWP is much more pronounced than with standard pavement marking, with a decreasing trend from the entrance through the exit. This also suggests that drivers are likely more confident entering the work zone with the AWP, as evidenced by the fact that speeds are higher during these two conditions than any other.

I-90

The averages and standard deviations were calculated for the samples collected for each of the two-lane crossovers at the I-90 test site. The posted work zone speed limit was set at 55 miles per hour (89 kph). Based on results from the sensor data extraction, the mean speeds ranged between 51.9 and 57.0 miles per hour (83.5 to 91.7 kph) for the various conditions. Average speeds on both two-lane crossovers were less than the posted speed limit for the AWP conditions, except for daytime, clear conditions. Average speeds during both the nighttime standard pavement marking conditions (clear and rain) exceeded the posted speed limit; however, the average speeds during the daytime clear conditions were less than the posted speed limit mainly due to the reduction of speeds through the work zone when workers were present, which continued through the end of the work zone.

Statistical tests were used to determine if the mean speed for the test site compared to the control site was statistically significant. Again, the one-way analysis of variance was utilized to compare the mean speeds. Due to heterogeneous variances, the Welch’s modification to the one-way analysis of variance was utilized, and the calculated F-value was based on an asymptotic distribution. In addition, due to the unequal variances and unequal sample sizes, the Games Howell post hoc test was utilized to examine specific differences within those samples tested in the one-way analysis of variance. Based on the one-way analysis of variance, it was determined that the mean speeds associated with the vehicles traveling along the two-lane crossover for the three conditions were significantly different, as shown in table 10.

Table 10. I-90 speed analysis ANOVA results.

Comparison Sum of Squares Degrees of Freedom Mean Squares F-Calc p-value

All Conditions

Between Groups

Within Groups

Total

21596.97

522466.36

544063.33

5

1849.71

4319.39

87.93

51.32

0.000

* p-value is statistically significant (α = 0.05)

For I-90, the ANOVA results provided earlier in table 8 indicated the following:

  • The only statistically significant differences between the AWP and standard pavement marking were during nighttime rainy conditions.

  • The mean speeds for those vehicles traveling along the two-lane crossover with the AWP were significantly lower than those vehicles traveling through the crossover with the standard pavement marking. However, unlike the US-32/33/50 site, the AWP was installed at an entry curve and the standard pavement marking was installed at an exit curve. Therefore, the difference in speeds at each crossover cannot be attributed solely to the pavement marking type, but could very likely be due to differences in entry versus exit curves. This is further validated by the findings at other sites with similar pavement markings at entry and exit curves, where major differences were noted between the two curves.

Conclusions and Recommendations

Vehicle speed was used as a surrogate measure used to evaluate the safety performance of more reflective lane markings. From the analysis of the speed data, it is not clear what effect the more reflective pavement marking has on driver behavior, or if an increase or decrease in speeds is a positive effect. Nonetheless, knowing the changes in speeds due to the new pavement marking should be instructive.

Speeds should be analyzed with some caution, for their variability and dependence on a number of factors. In the case of this study, major factors affecting speeds included (1) the geometry of highway crossover, particularly lane shifts, (2) the installation of highway detours and the nature of work zones, (3) the intensity of rain, and (4) the lighting conditions. The research team attempted to compensate for these factors by keeping them consistent (when possible) when comparing various samples.

At the I-85 test site, two issues arose. The geometry varied slightly, and more importantly, the intensity of the rain likely varied since the crossovers were so far apart (6 miles, or 9.7 km). This being the case, it was unreasonable to compare the AWP curves with the standard pavement marking curves. Consequently, the team was only able to compare entry with exit curves for the standard pavement marking installation. At a 95 percent confidence level, the researchers found a slight increase in speed at the exit curve. One could assume that drivers likely decelerate entering the crossover and accelerate slightly exiting the crossover.

Data collection at the US-15/501 site used the same curve for analysis, thus keeping the highway geometry and work zone layout consistent. This issue that remained was the rain intensity, since data were collected during different rain events. Standard pavement markings were initially installed and video was taken. When enough rainy nighttime footage was recorded, the standard markings were overlaid with the AWP pavement markings. Therefore, it was impossible to get identical rain events, even if the monthly rainfall was consistent during data collection periods. Every attempt was made to get similar rain intensities using rain data obtained from the National Oceanic and Atmospheric Administration (NOAA). From the footage used for analysis, the standard comparison sample was taken during a 2-hour period during which 1 inch (2.54 cm) of rain fell, whereas the AWP treatment sample saw just over 0.5 inches (1.27 cm) of rain for a 1 hour and 20 minute period, so the hourly rainfall intensities were very similar. Nonetheless, statistical analysis was used to compare the standard sample with the AWP sample. Speed was found to be higher on the AWP treatment sample. Even though average intensity was similar during both rain events, the researchers noted points of concern. First, the analysts observed consistent rainfall during the standard pavement marking data collection, while the rain intensity dramatically increased in the second half of the AWP data extraction. Second, possible changes in speed during phases of construction should be addressed. The standard comparison sample was collected just after the lane shift was constructed and opened as a detour. Even with familiar drivers, there can be an expected learning curve, thus decreasing maneuvering decision time and decreasing traveling speed until the mean speed could have returned to a steady state after the initial observation period.

The US-421 test site provided the most promising data from the North Carolina locations. Lane shift construction and opening was simultaneous as contractors worked to replace bridges on the four-lane divided highway. Both lane shifts had entry and exit crossovers side by side within a 0.5-mile (.8-km) stretch of highway. Therefore, the team was able to use synchronized video for data extraction and analysis. Some speed differences were noted when comparing the standard pavement marking curves with the AWP curves; however, the differences were small and conflicting (though statistically significant). For instance, when comparing pavement marking types at the entry curves, speeds were higher on the sections with AWP. This finding was not validated at the entry curves, where the standard pavement marking correlated with slightly higher speeds. Last, on both the AWP and standard pavement marking lane shifts, higher speeds were observed on exit curves than on entry curves.

The US-32/33/50 speed data analysis indicated that vehicles entering the lane shift utilizing the AWP during nighttime dry and rain conditions traveled at a higher rate of speed than on the lane shift utilizing the standard pavement marking. This is consistent with the theory that motorists will travel faster through a given length of highway due to the higher retroreflective pavement markings. On the other hand, the I-90 test site speed data analysis indicated that vehicles traveling through the two-lane crossover utilizing the AWP during the nighttime rain conditions traveled at a lower rate of speed as compared to the crossover utlizing the standard pavement marking. However, this finding at I-90 should be looked into further because both pavement markings were applied on different curves (entry and exit).

Overall, based on the comparisons between the mean speeds at the US-15/501, US-421, US-32/33/50, and I-90 test sites, the team finds no conclusive evidence that drivers travel faster through work zone detours delineated with the AWP than those with the standard pavement marking. There was no consistent positive or negative speed change between the two pavement marking types; in fact, the drivers on the AWP sections exhibited both higher and lower speeds than those on the standard pavement marking sections at each of the various crossovers studied.

In future studies, the research team recommends that speed data be collected by video extraction under the following circumstances:

  1. Using synchronized video for highway curves similar in geometry and work zone layout.
  2. Timing for speeds over a distance of at least 200 feet (61 m).
  3. Calibrating measurements with supplemental speed gun recordings.
  4. Using larger samples sizes.
  5. Removing outliers.
  6. Developing macro-enabled databases for analysis.


Page last modified on June 3, 2013.
Federal Highway Administration | 1200 New Jersey Avenue, SE | Washington, DC 20590 | 202-366-4000