Skip to contentUnited States Department of Transportation - Federal Highway Administration FHWA Home
Research Home
Report
This report is an archived publication and may contain dated technical, contact, and link information
Publication Number: FHWA-HRT-07-059
Date: October 2007

Updates to Research on Recommended Minimum Levels for Pavement Marking Retroreflectivity to Meet Driver Night Visibility Needs

PDF Version (660 KB)

PDF files can be viewed with the Acrobat® Reader®

3. RESEARCH METHODOLOGY

The scope of this research effort did not include the collection of additional human factors data from the field. The focus of this research was to examine the body of pavement marking visibility research in order to understand how different factors affect pavement marking visibility. Using the understanding gained from existing research, a visibility model (TARVIP) was employed to evaluate the effect that various pavement marking visibility parameters have upon the minimum retroreflectivity required by nighttime drivers. Some factors were analyzed in the TARVIP model while others were held constant to represent either reasonably worst-case visibility conditions (such as the absence of roadway lighting or the driver being of advanced age) or the conditions most likely faced by drivers on the roadway (such as conventional markings consisting of paint with beads). The results of the TARVIP analysis were then used, in conjunction with previous research findings reported herein, to provide recommended minimum pavement marking retroreflectivity levels.

3.1 Description of Tarvip

TARVIP is a deterministic model for evaluating the nighttime visibility of retroreflective objects from a driver’s perspective (the latest version of TARVIP may be obtained from the University of Iowa’s Operator Performance Laboratory website at http://opl.ecn.uiowa.edu/tarvip). TARVIP has a physical subsystem and a human factors subsystem. The physical subsystem determines the actual luminance contrasts of retroreflective objects on the roadway and is influenced by the following factors:

  • Three-dimensional spatial locations of the vehicle headlamps, driver eyes, and pavement markings.
  • Two-dimensional matrices of headlamp luminous intensity with respect to vertical and horizontal beam angle.
  • Two-dimensional matrices of pavement surface and pavement marking retroreflectivity coefficients (RL) with respect to entrance and observation angle.
  • Environmental data such as windshield transmission, atmospheric transmissivity, and ambient luminance.
The human factors subsystem uses Blackwell contrast threshold data to determine the average contrast detection ability of a human observer of a certain age under specific lighting and luminance contrast scenarios. The actual luminance contrast from the physical subsystem is compared with the average luminance contrast threshold from the human factors subsystem to determine how well the driver can see a pavement marking in the given geometric scenario, and what the driver’s needs are in terms of retroreflectivity for that scenario.(36)

3.2 Using the TARVIP Model

The process used in creating TARVIP models for the purpose of recommending minimum pavement marking retroreflectivity is detailed in the following paragraphs. In creating a visibility scenario, the first geometric component to create is the road itself. The user may select the width of roadway available on either side of the road center line and the geometry of the road in terms of plan and profile.

The user is then asked if the visibility of pavement markings, signs, or a diffuse (non-retroreflective) target is to be analyzed. Once pavement markings are selected, the user can choose which pavement marking configuration is to be analyzed. The three choices on the left are for center line configurations without edge lines. The “Combined” choice is for a single solid plus a single dashed center line (such as for a one-way no-passing zone). The final two options are for a single dash or double solid center line with solid white edge lines.

If the “Full + Dash” option is selected, the user is then prompted to define the widths of each pavement marking line and its lateral location on the roadway. The user is also asked for the longitudinal cycle length and the longitudinal gap length for the center line skip. If the “Full + Double Solid” option had been selected, the latter two would have been replaced with a prompt for the lateral separation between the two lines. The user is also asked to supply ASCII files containing tab-delimited text of the retroreflectivity matrices for the pavement surface and the pavement marking material. The efficiency of the material can be adjusted to represent old or worn marking materials.

The user may then define the length of the road and the location of a point of interest for use in tracking changes in photometric relationships as the vehicle moves along the roadway. Once this has been done, the vehicle and headlamp details can be selected. The user can select the speed of the vehicle, the ability of the windshield to allow light to pass, the location of the vehicle on the roadway, and the increments at which calculations should be performed. The user is also asked to supply ASCII files containing tab-delimited text of the luminous intensity matrices of the vehicle’s headlamps.

Once the file is supplied, the user can adjust the dimensions of the headlamp locations relative to a point halfway between the two headlamps on the pavement surface (termed the car origin), as well as adjusting the headlamp efficiency to represent deficient headlamps. TARVIP uses the headlamp luminous intensity matrices to project the headlamp beams onto the road surface for the visibility analysis (figure 4).

In the “Driver” window, the user may set the driver age, exposure time, minimum preview time, and the driver’s eye location relative to the car origin. If visibility under glare conditions is to be modeled, the user may enter the horizontal and vertical angle components of the driver’s glance away from the center of the traveled way.

Figure 4. Screen shot. Plot of iso-lux curves on road surface (TARVIP screen shot). This is a screen shot from the Target Visibility Predictor (TARVIP) software developed by the Operator Performance Laboratory at the University of Iowa. This dialog screen comes up when vehicle headlamp types are selected for visibility prediction. At the top right corner of the dialog the types of all headlamps used are shown. Here there are two headlamps, and both are “UMTRI-2004-50%-US-LB”, the 2004 UMTRI 50 th percentile market weighted low beam headlamp. The middle of the dialog is a plot of iso-lux curves on road surface from this particular headlamp configuration. Apart from headlamp types, the dimension of the vehicle and the positions of the headlamps also affect the curves, which would have been set up at an earlier step. The x-axis of the plot is labeled “Longitudinal Distance along Road Surface from Car Origin [m]”, with a range of 0 to 100, and the y-axis is labeled “Latitudinal Distance from Car Origin [m]”, with a range of -15 to 15. There are a total of 13 iso-lux curves visible in the plot, all in tear drop shape, with the tail pointing towards the +x direction, and roughly symmetric along the y = 0 line. The heads of the tear drops are all near x = 5, but the tails vary greatly. Only nine curves with the highest lux values are fully visible in the plot. The “eyes” of the curves are all centered around the point (8, 0). The inner most curve is the one with the highest lux value, and it spans an x range of about 6 to 15 and a y range of about -1.5 to 1.5. The next curve is about 5 to 19 and -2 to 2, and the next about 4.8 to 27 and -2.5 to 3, and so on. The ninth curve from inside, or the last one that is fully visible, has its tail reaching the point (-2, 84). At the bottom center of the dialog is a line of text that reads “Maximum Illuminance on Road Surface: 118 lux”, which refers to the illuminance value of the inner most iso-lux curve.

Figure 4. Plot of iso-lux curves on road surface (TARVIP screen shot).

The “Environment” window allows the user to adjust ambient luminance (cd/m²) and atmospheric transmissivity (km -1). The user can also enable fog luminance calculations and specify a fog droplet diameter if such analysis is to be undertaken.

Once the scenario is modeled in TARVIP to the user’s specifications, the visibility analysis may be performed. TARVIP provides a variety of data options for the output of the visibility analysis, including photometric angles, luminous intensity, illuminance, luminance, and contrast as the vehicle travels along the roadway toward the point of interest. TARVIP also includes a “Material Performance Summary” window that provides a minimum required RL for the scenario and allows for adjustments to be made to a few parameters without recalculating the entire model. The user can obtain the minimum required RL for any desired preview time and recalculate for changes in vehicle speed and the driver age. The user can also obtain the required preview time given a certain RL value as well as input a preview time to determine, on average, the oldest driver who would be able to see the pavement marking under the modeled conditions.

3.3 Validation of the TARVIP Model

The ability of TARVIP to generate reasonable measures of pavement marking visibility under various scenarios was ascertained by comparison of its outputs to data from various pavement marking visibility studies. The validation was performed using the study data from two Texas Transportation Institute (TTI) studies, reported in TTI Technical Report 5008-1 (2) and TTI Technical Report 4269-1. (15) In those two studies, dry, nighttime pavement marking detection distances were collected. The data were used to compare to TARVIP predictions. The comparison showed that the TARVIP curve fell within the 95 percent confidence range of the TTI study data under most conditions.

3.3.1 TTI Report 5008-1 Study Comparison

TTI Report 5008-1 was for a study evaluating the performance of several varieties of pavement markings in wet weather. However, as part of this research, dry detection distance data for each marking were also collected. Subjects drove a 2004 Ford Taurus with a researcher in the passenger seat. The research was conducted on an isolated test track. The subjects drove with the cruise control set at 48.3 km/h (30 mi/h) and told the researcher when they could see a pavement marking (white lines, yellow lines, or RRPMs). Distracter markings were located outside the travel lane to minimize the possibility of the subjects becoming accustomed to the pavement marking locations and guessing their location before actually seeing them. The markings of interest to the researchers were isolated skips located in the center of the travel lane. When the subject alerted the researcher when he/she could identify a pavement marking and its type, the researcher recorded the location values from a distance-measuring instrument.(2)

The research conditions used by Carlson et al. were duplicated as closely as possible in the TARVIP software; however, some assumptions were required. The pavement surfaces that were available in TARVIP when this analysis was conducted did not include an old or weathered asphalt pavement (after this phase of the research, the research team was able to add pavement- marking retroreflectivity files for old asphalt, which were ultimately used to generate the final recommendations in this report). However, the TTI study was conducted on old asphalt. Therefore, for the purposes of this validation effort, old concrete was used. Old concrete would most likely provide the nearest approximation of old asphalt in terms of contrast with pavement markings of any of the available pavement surfaces. Because old concrete has a lower retroreflectivity than old asphalt in the TARVIP pavement surface files, it will provide more luminance contrast with the pavement marking than old asphalt. Therefore, the resulting detection distances from the TARVIP analysis will be higher with old concrete than they would be with old asphalt, so the reported detection distances are more liberal than they would have been had an old asphalt pavement surface file been used in the analysis.

When comparing a tape product’s performance in TARVIP to its performance as reported by Carlson et al., (2) it was assumed that the product used in the TTI study had similar retroreflective characteristics to the pavement-marking file in TARVIP. To adjust for any differences, a material efficiency is used. This material efficiency is a ratio between the RL reading of the actual marking used and the 30-m (98-ft) geometry RL contained in the TARVIP pavement marking file. Such an adjustment assumes that this ratio is valid across the entire viewing matrix based on one RL value.

For the TARVIP thermoplastic marking files (as well as any of the other nonmanufactured markings), no information is available regarding the thickness, application method, binder, or bead gradation of these markings. Therefore, it was necessary to assume that the TARVIP thermoplastic marking file was similar to the material used in the TTI study.

It should be noted that the researchers used a 2003 Ford Taurus headlamp for the modeling, and the TTI study used a 2004 Ford Taurus. It was confirmed with a local Ford dealership that the headlamps of these two model years are compatible. The locations of the headlamp relative to the car origin and the driver’s eye relative to the headlamps were measured using the original vehicle from the study.

The TARVIP default values for windshield transmission, ambient luminance, and atmospheric transmissivity were retained since no data for these values are available from the TTI study.

The results of the comparison showed that the TARVIP curve fell within the 95 percent confidence range of the TTI study data (see figure 5 and figure 6).

Figure 5. Scatter diagram. Detection distance versus driver age--Structured tape. From a 2005 Texas Transportation Institute study (TTI Report 5008-1) by Carlson et al. The x-axis is labeled “Driver Age (yrs)”, with a range from 15 to 80, and the y-axis is labeled “Detection Distance (m)”, with a range from 0 to 250. Two types of points are present in the diagram—the green dots are marked as “5008-2 Study” and are data from the study, and the pink dots are marked as “TARVIP Output” and are prediction values from the TARVIP software performed afterwards. There are thirty-nine green dots and their x coordinates range from about 19 to 80. Twenty-two of them have a y coordinate between 50 and 100, twelve have a y coordinate between 100 and 150, four have a y coordinate between 150 and 200, and only one has a y coordinate between 200 and 250. The green dots with x coordinates smaller than 50 are scattered between the y = 50 and y = 210 lines, while the green dots with x coordinates greater than or equal to 50 are scattered between the y = 50 and y = 120 lines. This indicates the older drivers generally have a shorter detection distance. There are twenty pink dots and their x coordinates range from about 19 to 75. They fit closely to an imaginary line with one end at (19, 100) and the other end at (75, 70), approximately. There are also three lines in the diagram. The red line is marked as “Upper Bound” or the upper 95 percent bound line, the orange line as “Lower Bound” or the lower 95 percent bound line, and the green line as “5008-2 Regression Line”. The green regression line has an equation of “y = -0.3174x + 112.93”, and its coefficient of determination R 2 equals to 0.0296. The upper bound line has a slope of about 0.3 and intersects the y-axis (x = 15) at around (15, 140). The lower bound line has a slope of about -0.9 and intersects the y-axis at around (15, 70). There are 5 green dots above the upper bound line, 2 below the lower bound line, 12 between the upper bound line and the regression line, and 20 between the regression line and the lower bound line. The imaginary line formed by the pink dots is quite close to the 5008-2 regression line, about 5 m (16.4 ft) below at the left end (x = 19) and about 20 m (65.6 ft) below at the right end (x = 75). This means the TARVIP prediction fell well within the 95 percent confidence range of the TTI study data and very close to the regression line.

Figure 5. Detection distance versus driver age--Structured tape.

Figure 6. Scatter diagram. Detection distance versus driver age--Thermoplastic. From a 2005 Texas Transportation Institute study (TTI Report 5008-1) by Carlson et al. The x-axis is labeled “Driver Age (yrs)”, with a range from 15 to 80, and the y-axis is labeled “Detection Distance (m)”, with a range from 0 to 140. Two types of points are present in the diagram—the green dots are marked as “5008-2 Study” and are data from the study, and the pink dots are marked as “TARVIP Output” and are prediction values from the TARVIP software performed afterwards. There are eighteen green dots and their x coordinates range from about 19 to 80. Three of them have a y coordinate between 40 and 60, five have a y coordinate between 60 and 80, four have a y coordinate between 80 and 100, four have a y coordinate between 100 and 120, and only two have a y coordinate between 120 and 140. Nine green dots have an x coordinate between 19 and 25 and a y coordinate between 70 and 130, while seven green dots have an x coordinate between 67 and 77 and a y coordinate between 40 and 85. Only two green dots are at the middle of the x range, both at x = 44. There are nine pink dots and their x coordinates range from about 19 to 75. They fit closely to an imaginary line with one end at (19, 75) and the other end at (75, 54), approximately. There are also three lines in the diagram. The orange line is marked as “Upper 95%” or the upper 95 percent bound line, the red line as “Lower 95%” or the lower 95 percent bound line, and the green line as “5008-2 Regression Line”. The green regression line has an equation of “y = -0.5695x + 109.09”, and its coefficient of determination R 2 equals to 0.3871. The upper bound line has a slope of about -0.2 and intersects the y-axis (x = 15) at around (15, 126). The lower bound line has a slope of about -1.0 and intersects the y-axis at around (15, 77). There is one green dot above the upper bound line, none below the lower bound line, eight between the upper bound line and the regression line, and nine between the regression line and the lower bound line. The imaginary line formed by the pink dots is somewhere between the 5008-2 regression line and lower bound line. At the left end it almost touches the lower bound line, while at the right end it is about 15 m (49.2 ft) below the regression line. This means the TARVIP prediction fell within the 95 percent confidence range of the TTI study, but was not as close as the study data as in figure 5, the data for the structured tape.

Figure 6. Detection distance versus driver age--Thermoplastic.

3.3.2 TTI Report 4269-1 Study Comparison

TTI Report 4269-1 was for a study evaluating the visibility of signs and pavement markings from the perspective of passenger car drivers and commercial vehicle drivers (only the results from the passenger car drivers are reported herein). Two different types of pavement marking tape products were investigated, one of which was used both in new condition and with a clear mask applied. Thus, three different markings were tested, representing low, medium, and high retroreflectivity coefficients. Pavement marking detection distance data were collected in a 1998 Chevrolet Lumina traveling at 48.3 km/h (30 mi/h). Data were recorded similarly to that recorded by Carlson et al., except that participants were following a solid white right edge line and asked to state when they could clearly see the end of the pavement marking. The course also had signs that they were asked to identify during each run.(15)

The TTI study was performed on old concrete, so that pavement marking file was used. Assumptions regarding material efficiencies and TARVIP default values were the same as those made for the TTI Report 5008-1 study comparisons.

The most problematic assumption involved the vehicle headlamps, as TARVIP contains no Chevrolet Lumina headlamp file. TARVIP does have a 2001 UMTRI 50 percent low-beam headlamp file—the only headlamp that it has in common with the ERGO software used for sign visibility. Using ERGO, the luminance values recorded in the sign luminance portion of the study by Finley et al. and the output of scenarios modeled in ERGO were compared. From these two values, a headlamp efficiency value for each point was obtained. These efficiencies were then averaged to obtain one efficiency value. The two obvious limitations are

  • It is assumed that the average obtained is an effective method for comparing the Chevrolet Lumina headlamp to the UMTRI 50 percent low-beam headlamp.
  • The points of comparison are all located above the horizontal while the points of interest are below the horizontal.

The results of the comparison showed that the TARVIP curve fell within the 95 percent confidence interval for older drivers but slightly below the interval for younger driver s (approximately 50 years old and younger). See figure 7.

Figure 7. Scatter diagram. Detection distance versus driver age—Standard tape. From a 2002 Texas Transportation Institute study (TTI Report 4269-1) by Finley et al. The x-axis is labeled “Driver Age (yrs)”, with a range from 25 to 75, and the y-axis is labeled “Detection Distance (m)”, with a range from 0 to 250. Two types of points are present in the diagram—the green dots are marked as “4269 Study” and are data from the study, and the pink dots are marked as “TARVIP Output” and are prediction values from the TARVIP software performed afterwards. There are about one hundred and seven green dots and their x coordinates range from 25 to 73. Twenty-nine of them have a y coordinate between 50 and 100, fifty-six have a y coordinate between 100 and 150, nineteen have a y coordinate between 150 and 200, and only three have a y coordinate between 200 and 250. They are quite evenly distributed along the x-axis, with a bit more at the left side (younger age) than at the right. There are twenty-three pink dots and their x coordinates range from 25 to 73. They fit closely to an imaginary line with one end at (25, 88) and the other end at (73, 60), approximately. There are also three lines in the diagram. The orange line is marked as “Upper 95%” or the upper 95 percent bound line, the red line as “Lower 95%” or the lower 95 percent bound line, and the green line as “4269 Regression Line”. The green regression line has an equation of “y = -0.5201x + 144.64”, and its coefficient of determination R 2 equals to 0.0419. The upper bound line has a very slight negative slope and intersects the y-axis (x = 25) at around (25, 165). The lower bound line has a slope of about -1.0 and intersects the y-axis at around (25, 98). There are sixteen green dots above the upper bound line, fifteen below the lower bound line, thirty-seven between the upper bound line and the regression line, and thirty-nine between the regression line and the lower bound line. The imaginary line formed by the pink dots is quite far below the regression line. In fact, it is very close to the lower bound line, intersecting with it near the middle. This means the TARVIP prediction lie marginally within the 95 percent confidence interval for older drivers, but slightly below the interval for younger drivers (50 years old and younger).

Figure 7. Detection distance versus driver age—Standard tape.

Previous | Table of Contents | Next

ResearchFHWA
FHWA
United States Department of Transportation - Federal Highway Administration