# Final Report Phase 1 - Temporary Wet-weather Pavement Markings for Mork Zones

## APPENDIX B: PRELIMINARY WORK PLAN FOR PHASE II

#### Phase I - Task 3: Coordination and logistical planning for field work-zone evaluation for Phase II

### Introduction

In Phase II, 3M plans to evaluate the effectiveness of the all-weather work zone pavement marking system (experimental treatment), as well as traditional paint and bead pavement markings (control treatment) at night during wet-weather in actual work zones. This technical memorandum contains the general experimental plan for these field studies. A more detailed experimental plan is to be developed in Phase II prior to the conduct of the field studies (as per the original proposal for this project).

### Desired Work Zone Characteristics

In work zones, unexpected changes in the roadway alignment, such as lane shifts and crossovers, may be challenging to drivers at night during wet-weather conditions since conventional pavement markings seem to disappear. Visible pavement markings under wet-weather conditions would continue to provide critical path guidance information and thus may reduce driver workload and in turn improve safety compared to conventional pavement markings. TTI researchers recommend that the two pavement marking treatments be applied in lane shifts since a minimal number of additional visual cues from other traffic control devices are typically present at these types of locations. Lane shifts that require greater changes in alignment (i.e., those with larger degrees of curvature and lateral offsets) are preferred. When existing pavement marking materials are removed in order to create a lane shift, the removal method typically leaves "ghost" markings which may be mistaken for actual pavement markings and thus cause confusion to drivers. Thus, lane shifts with "ghost" markings that do not align with the lane shift path are desired. TTI researchers believe that these types of conditions will maximize the potential to detect a benefit from the all-weather work zone pavement marking system.

Crossovers may also be used, if necessary. However, additional path guidance is typically provided by various other traffic control devices upstream and throughout the crossover (e.g., arrow panels at the upstream lane closure point, high target channelizing devices (drums), chevrons, etc.). Thus, the potential to detect a benefit from the all-weather markings is more likely to be diminished at these types of locations.

The field studies should be conducted on high-volume, multilane, divided roadways with a normal (non-work zone) posted speed limit ≥ 55 mph. Higher-volume roadways will result in greater numbers of vehicles traveling at night, which is the user group of primary interest in this type of study. Other variables that may impact the effectiveness of the all-weather work zone pavement marking system are type of roadway surface (asphalt versus concrete) and whether or not ambient lighting is present (existing high-mast lighting or temporary lighting for the work zone).

Table 1 contains the desired site matrix. At a minimum, TTI researchers recommend that the two pavement marking treatments be evaluated at four sites (i.e., one site per cell in Table 1). To increase the validity of the field studies, TTI suggests increasing the number of sites to eight (i.e., two sites per cell in Table 1).

Roadway Surface | Ambient Lighting | |
---|---|---|

Not Present | Present ^{b} | |

Asphalt | X |
X |

Concrete | X |
X |

^{a} Assuming all sites are on high-volume, multilane, divided roadways with a normal posted speed limit ≥ 55 mph.^{b} Could be existing high-mast lighting or temporary lighting for the work zone

### Experimental Plan

#### Study Design

At each work zone, researchers will evaluate the effectiveness of the all-weather work zone pavement marking system (experimental treatment) and traditional paint and bead pavement markings (control treatment) at night during wet-weather using operational measures as surrogates for safety. There are two types of study designs that could be used. The first is a before-and-after study in which both types of pavement markings are applied at different times at the same location in the work zone. While it will take longer to collect data at each site, using the before-and-after study approach significantly reduces the site-to-site variability that would be present in the data. Higher variability diminishes the ability to detect a difference between treatments. The second method would be to apply the two pavement marking treatments at the same time in different locations in the same work zone (e.g., one treatment located at the beginning of the lane shift and one treatment located at the end of the lane shift or if available, at the beginning of the lane shift in the opposite direction). While in theory this method reduces the amount of time needed to collect data at each site, the site differences themselves confound with the effects that the different pavement marking treatments may have, making it more difficult to detect actual differences between the pavement marking treatments in terms of their effect on the operational measure of effectiveness. In addition, if the sites are far apart the rainfall events at each location may vary considerably; and thus, actually increase the time needed to collect an adequate amount of data for each condition.

Based on the discussion above, TTI researchers recommend the use of the before-and-after study design. Typically, researchers will evaluate the control treatment in the before time period and the experimental treatment in the after time period. However, if more than one site per cell in Table 1 is used, at one site the control treatment (i.e., traditional paint and beads) should be evaluated first, and at the second site the experimental treatment (all-weather work zone pavement marking system) should be evaluated first. It should also be noted that since data will be collected at multiple work zones, the effect of the experimental treatment could still be confounded with the effect of uncontrollable extraneous variables that change between the before and after time periods. Comparison sites where no experimental treatment is applied are often used to help ensure the internal validity of the study by reducing confounding effects; however, finding work zones that are comparable to the study sites is almost impossible due to the vast assortment of work zone and roadway characteristic combinations. Thus, comparison sites are rarely used with this type of work zone study. Furthermore, given that the before and after time periods being proposed are likely to be fairly short and one right after the other, the potential for significant changes in the extraneous variables over both periods is likely to be minimal.

#### Data Collection

At a minimum, researchers should collect the following for each treatment at each site:

- the lateral placement of vehicles in the outside travel lane;
- rainfall data (e.g., amount and duration);
- pavement marking maintained presence;
- pavement marking retroreflectivity; and
- roadway and work zone characteristics.

The lateral placement data should be collected at three locations: immediately upstream of the lane shift (base location), at the midpoint of the first curve, and at the mid-point of the second curve. Since past studies (*1, 2*) have not shown a strong correlation between speed and delineation treatments, TTI does not believe it is worthwhile to collect vehicle speed data in the lane shift itself. However, vehicle speed data should be collected immediately upstream of the lane shift and compared amongst pavement marking treatments at each site and across sites to determine if the traffic characteristics were similar for all conditions or may have been affected by uncontrollable extraneous variables. Lateral placement and speed data should also be collected under dry pavement conditions (both at night and during the day) to help identify any unrelated conditions that might have affected the nighttime, wet-weather data.

#### Sample Size

TTI researchers determined the sample size (the number of vehicles) needed to detect a practically important minimum difference between pavement marking treatments and among the interaction effects between the pavement marking treatment and the weather factor (wet or dry) at each site by power analysis. The procedures given in Wheeler (*3*), Nelson (*4*), and Bratcher et al. (*5*) were used for the sample size calculation. Because the necessary sample size varies with the desired significance level a , the desired power, the standard deviation of the response variable, and the minimum difference of practical importance, those values were predetermined before the sample size calculation. By convention, the desired significance level and the desired power were set to 0.05 and 0.90, respectively. It was found from previous research, that the approximate standard deviation in lateral placement in the curves was 20 inches (*6*) and in work zones was 13 inches (*7*). To err on the conservative side, 20 inches was used. The minimum difference of interest before and after installation of the all-weather work zone pavement marking system was determined to be 6 inches for the mean lateral placements based on engineering judgment and previous research (*6*). It is believed that 6 inches is the minimum change in mean lateral placement that would be a practically significant change for at least two reasons: (1) field experience has shown that striping installations vary in width as much as ±0.5 inches and restriping can be misaligned by more than one inch, which may result in wide variability between pavement marking installations; and (2) previous research supported 6 inches (*6*).

The following equations were employed to determine the minimum sample size when the desired significance level is 0.05 and the desired power is 0.90. Equation 1 is used to detect main effects due to pavement marking treatment or the weather factor and Equation 2 is used to detect an interaction effect between pavement marking treatment and the weather factor.

From Equation 1, r is the number of levels of a factor,

From Equation 2, n is the number of interaction degrees of freedom, c is the number of factor-level combinations for the factors that are involved in the interaction, *k* is the number of factors involved in the interaction, and d is the minimum difference of interest among the interaction effects.

Based on Equation 1, the minimum sample size necessary for detecting a mean lateral placement difference of 6 inches before and after installation of the all-weather work zone pavement marking system at each site is

Based on Equation 2, the minimum sample size necessary for detecting a mean lateral placement difference of at least 6 inches in any two interaction means between pavement marking treatment and weather factor at each site is likewise

The sample size of 400 is selected to assure the power of the tests to be at least 0.90 for the mean lateral placement difference. Thus, at each site the desired number of vehicles to be observed at night for each pavement marking treatment (control and experimental) and weather condition (dry and wet) combination is at least 100. As mentioned previously, data should also be collected for each pavement marking treatment during daytime dry pavement conditions. At each site, the desired number of vehicles for each daytime treatment condition is also 100. It should be noted that these numbers are the minimum number of usable data points. More data will need to be collected in the field to ensure that at least 100 usable data points are obtained for each condition.

The above calculations were used to identify the minimum sample size needed for identifying meaningful differences in the mean lateral placement among treatments. However, previous research (*1, 2*) has also reported that the variance of vehicle lateral placement is strongly correlated with crash crates. Therefore, assuming a control treatment standard deviation equal to 20 inches, Table 2 shows the maximum experimental treatment standard deviations for which a difference could be detected for various sample sizes and confidence levels. For example, with a sample size of 100 vehicles and a 95 percent confidence level, a significant difference between the treatments' variances would only be detected when the experimental treatment standard deviation is equal to or less than 16 inches. In other words, it would be possible to detect a 20 percent reduction in lateral placement standard deviation with that sample size. Reducing the confidence level and/or increasing the sample size both increase the maximum experimental standard deviation for which a significant difference from the control condition can be detected. If it were desirable to detect as little as a 10 percent reduction in the lateral placement standard deviation (i.e., from 20 inches down to 18 inches), it would be necessary to collect data from 250 vehicles to retain a level of significance of 0.05. If, however, one accepted a higher level of significance, such as 0.15, then it would only be necessary to collect data from 150 vehicles to conclude that the standard deviations are different.

Sample Size (vehicles) | Level of Significance (alpha)^{b} | |||
---|---|---|---|---|

0.05 | 0.1 | 0.15 | 0.2 | |

100 | 16 | 17 | 17 | 18 |

150 | 16 | 17 | 18 | 18 |

200 | 17 | 18 | 18 | 18 |

250 | 18 | 18 | 18 | 18 |

300 | 18 | 18 | 18 | 19 |

^{a} The standard deviation is the square root of the variance. Based on a two-tailed F distribution. Assumes a control treatment standard deviation of 20 inches (variance of 400 inches^{2}).

^{b} The significance level is 1-alpha (e.g., alpha of 0.05 is a 95 percent confidence level).

Based on Table 2, TTI recommends a minimum sample size of 200 vehicles for each pavement marking treatment (control and experimental) and weather condition (dry and wet) combination; totaling 800 vehicles per site. This sample size will allow researchers to confidently identify differences in both the mean lateral placement and variance of vehicle lateral placement among treatments, without placing undue burden on the data collection effort (especially since all sites should be on high-volume roadways).

### Data Reduction and Analysis

At a minimum, the following measures of effectiveness (MOEs) should be utilized to determine the effectiveness of the all-weather work zone pavement marking system at night under wet-weather conditions:

- mean lateral placement of vehicle in the travel lane,
- variance in the lateral placement, and
- rate of inadvertent contact with edge line markings.

Comparison of the mean vehicle lateral placement data will allow researchers to determine if there are differences among the treatments in how drivers position their vehicles in the outside travel lane. It is believed that vehicle paths located near the center of the travel lane may result in higher levels of safety (*8*). Thus, the amount of deviation from the center of the travel lane provides an indication of crash potential.

As mentioned above, previous research has also shown a correlation between the variance of vehicle lateral placement and crash frequency. It is hypothesized that the installation of treatments that reduce the variance of lateral placement (indicating more uniform driving performance) will lead to a lower crash frequency.

The rate of inadvertent contact with the edge line marking is similar to lateral placement in that it indicates the potential of a crash resulting from inappropriate lateral position. Thus, a reduction in this rate would be considered a positive safety benefit.

Mean speed and variance at the upstream data collection location should be compared amongst pavement marking treatments at each site and across sites to determine if the traffic characteristics are similar or may have been affected by uncontrollable extraneous variables. Lateral placement and speed data collected under dry pavement conditions (both at night and during the day) should also be used to help identify any unrelated conditions that might have affected the nighttime, wet-weather data. In addition, these data may reveal additional benefits to using the all-weather work zone pavement marking system under dry pavement conditions (night or day) or during daytime wet-weather conditions.

### References

- Taylor, J.L., H.W. McGee, E.L. Sequin, and R.S. Hostetter.
*NCHRP Report 130 Roadway Delineation Systems*. Highway Research Board, National Research Council, Washington, D.C., 1972. - Stimpson, W.A., H.W. McGee, W.K. Kittelson, and R.H. Ruddy. Field Evaluation of Selected Delineation Treatments on Two-Lane Rural Highways. Report No. FHWA-RD-77-118, Federal Highway Administration, Washington, D.C., 1977.
- Wheeler, R. E. Portable Power.
*Technometrics*, 16, 1974, pp.193-201. - Nelson, L.S. Sample Size Tables for Analysis of Variance. In
*Journal of Quality Technology,*17, 1985, pp. 167-169. - Bratcher, T. L., M.A. Moran, and W.J. Zimmer. "Tables of Sample Sizes in the Analysis of Variances. In
*Journal of Quality Technology,*2, 1990, pp. 156-164. - Krammes, R.A., K.D. Tyer, D.R. Middleton, and S.A. Feldman.
*An Alternative to Post-Mounted Delineators at Horizontal Curves on Two-Lane Highways*. Report No. 1145-1F. Texas Transportation Institute, Texas A&M University, College Station, Texas, 1990. - Harvey, D.L, R. Mera, and S.R. Byington. Effect of Nonpermanent Pavement Markings on Driver Performance. In
*Transportation Research Record: Journal of the Transportation Research Board, No. 1409*, Transportation Research Board of the National Academies, Washington, D.C., 1993, pp. 52-61. - Thompson, H.T. and D.D. Perkins. Surrogate Measures for Accident Experience at Rural Isolated Horizontal Curves. In
*Transportation Research Record: Journal of the Transportation Research Board, No. 905*, Transportation Research Board of the National Academies, Washington, D.C., 1983, pp. 142-147.