Guidelines for Selection of Speed Reduction Treatments at High-Speed Intersections: Supplement to NCHRP Report 613 (2008)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-1 5 Testing Results 5.1. EXECUTIVE SUMMARY Three speed-reduction treatments were tested at 10 intersection sites for a total of 19 intersection approaches. The testing was conducted to determine the effectiveness of each treatment at reducing speeds on high-speed intersection approaches. The treatments included peripheral transverse pavement markings, rumble strips, and dynamic warning signs activated by speed. Speed data were collected at four locations along each of the 19 intersection approaches before and after treatment installation. The speeds measured at these four locations established a speed profile that could be used to estimate the speed of a vehicle at any point on the intersection approach. The sensors were placed at the same locations before and after a treatment was installed so that direct comparisons could be made between vehicle speeds on each intersection approach. Three types of analyses were conducted: • Analysis by Location – to determine the speed reduction observed at each data collection location (i.e., at each of the four locations along an intersection approach) • Analysis by Intersection Approach – to determine the overall speed reduction observed along an intersection approach • Analysis by Treatment Type – to quantify the effect of each treatment across all intersection approaches at which it was installed The analysis by location showed that installing treatments generally reduced speeds by as much as 3.2 mph. At a few locations, increases in speed of as much as 1.8 mph were observed. The analysis by intersection approach showed that at most intersection approaches, the treatment installation resulted in a statistically significant speed reduction. At three approaches, there was an increase in speed; however, only one of these was statistically significant. All three intersection approaches on which dynamic warning signs were installed experienced a statistically significant speed reduction. The analysis by treatment type showed that dynamic warning signs and peripheral transverse pavement marking had a statistically significant overall effect in reducing speeds. Dynamic warning signs had a mean speed reduction of 1.7 mph, while peripheral transverse pavement marking had a mean speed reduction of 0.6 mph. Dynamic warning signs resulted in the largest speed reduction compared to the other treatments. Rumble strips did not have a statistically significant overall effect in reducing speeds; however, the results of the analysis by treatment type were based on only three intersection approaches where rumble strips were installed. All three treatment types may reduce speeds on high-speed intersection approaches; however, that speed reduction is likely to be minimal (i.e., less than 3 mph). Of the three treatment types, dynamic warning signs may be the most effective at reducing speeds and may result in the

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-2 greatest speed reduction. However, this conclusion is based on testing results at three intersection approaches. 5.2. INTRODUCTION Chapter 5 presents the results of data collection and analyses conducted to evaluate the effectiveness of various treatments intended to reduce vehicle speeds on high-speed intersection approaches. The research team coordinated with the Oregon Department of Transportation (ODOT), Washington State Department of Transportation (WSDOT), Texas Department of Transportation (TXDOT), and Clackamas County, Oregon, to install treatments at 10 intersections. These treatments included peripheral transverse pavement markings, rumble strips, and dynamic warning signs. This chapter is organized as follows: • Data Collection Approach • Data Analysis Approach • Data Analysis Results o Results of Analysis by Location (Step 1 Analysis) o Results of Analysis by Intersection Approach (Step 2 Analysis) o Results of Analysis by Treatment Type (Step 3 Analysis) o Results of 85th Percentile Speeds o Testing Results Summary • Conclusions 5.3. DATA COLLECTION APPROACH A total of 19 intersection approaches from 10 intersection sites in Oregon, Texas, and Washington were tested. Exhibit 5-1 presents basic site information, including the type of treatment evaluated, for each intersection site. The techniques used to collect field data included: • Automated traffic recorders to count traffic volumes and collect data on the speeds, headways, and lengths of individual vehicles. • Video recorders to film traffic events at the intersection for later review and data reduction, as needed.

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-3 • The automated traffic recorders were Unicorn traffic classifiers manufactured by Diamond Traffic Products. Vehicle axle passages to determine speeds, headways, and vehicle lengths were detected with piezoelectric cables taped to the pavement surface. Speed data were collected before and after treatment installation to determine the effectiveness of each treatment at reducing speeds on a high-speed intersection approach. After treatment installation, a three-month acclimation period occurred prior to collecting speed data to allow for the novelty of the treatment to subside and for motorists to adjust to the presence of the treatment. Data were collected under daylight, off-peak, and free-flow conditions in dry weather. With the exception of one intersection approach, speed data were collected at four locations along each intersection approach to obtain the following speeds: • Free-Flow Speed (Location D) - A set of sensors was placed upstream of the treatment in a location where vehicle speeds would not be affected by the presence of the intersection (before treatment installation) or the treatment itself (after treatment installation). • Intersection Perception/Response Speed (Location C) - A set of sensors was placed at the point where the intersection or the treatment (after treatment installation) would first become visible to the driver, i.e., the point at which the driver might first react to the presence of the intersection or treatment.

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-4 Exhibit 5-1 Summary of Speed Data Collection Sites Site No. Intersection Site Intersection Approach Treatment Through Lanes STOP Control Speed Limit (mph) NB Peripheral transverse pavement markings 1 Y 55 OR1 Meridian Rd / Whiskey Hill Rd EB Peripheral transverse pavement markings 1 N 55 NB Peripheral transverse pavement markings 1 N 55 OR2 Canby-Marquam Hwy / Lone Elder Rd SB Peripheral transverse pavement markings 1 N 55 EB Peripheral transverse pavement markings 1 N 55 OR3 Redland Rd / Ferguson Rd WB Peripheral transverse pavement markings 1 N 50 EB Peripheral transverse pavement markings 1 N 55 OR4 Redland Rd / Bradley Rd WB Peripheral transverse pavement markings 1 N 55 EB Peripheral transverse pavement markings 1 N 55 OR6 SR 6 / Wilson River Loop Rd WB Peripheral transverse pavement markings 1 N 55 TX1 US 271 / FM 726 SB Rumble strips 2 Na 70 NB Rumble strips 2 Na 70 TX2 US 271 / FM 2088 SB Rumble strips 2 Na 70 NB Rumble strips 2 Na 70 TX3 US 271 / FM 593 SB Rumble strips 2 Na 70 EB Dynamic warning sign 1 Na 55 TX4 US 82 / SH 98 WB Dynamic warning sign 1 Na 55 EB Dynamic warning sign 1 N 50 WA1 SR 26 / S 1st Ave WB Dynamic warning sign 1 N 50b a The intersection has a flashing-yellow signal on the major road and a flashing-red STOP-control signal on the minor road. b At 1,850 feet from the intersection, the speed limit is reduced from 60 to 50 mph.

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-5 • Accident Avoidance Speed (Location B) - A set of sensors was placed at the point that represents a driver’s “last chance” (or emergency opportunity) to avoid an accident at the intersection, in response to a surprise or unexpected hazard. Such a maneuver would require an instant perception-reaction-time response. • Intersection Entry Speed (Location A) - A set of sensors was placed close (e.g., at or near the stop bar) to the intersection to measure the speed at which vehicles enter the intersection. The speeds measured at these four locations established a speed profile that could be used to estimate the speed of a vehicle at any point on the intersection approach. The sensors were placed at the same locations before and after a treatment was installed so that direct comparisons could be made between vehicle speeds on each intersection approach. All distances were measured from the point where the roadway approach entered the intersection proper. As discussed in the following section, three primary types of analyses were conducted in the study: (1) the speed reduction observed at each traffic classifier location; (2) the speed reduction observed along each intersection approach; and (3) the speed reduction effectiveness for each treatment type. A secondary analysis (of 85th-percentile speeds) was conducted for each intersection approach and each treatment type; however, less emphasis is placed on this analysis as explained later. To clarify the terminology used throughout the analysis discussion, a summary of terms is presented below: • Location – A data collection location refers to a single point along an intersection approach at which a traffic classifier and two accompanying sensors were placed to measure speeds before and after treatment installation. Typically, traffic classifiers were placed at four such locations (e.g., Locations A, B, C, and D) along each intersection approach. • Intersection Approach – An intersection approach refers to a single intersection leg (i.e., EB, WB, NB, or SB). As many as four traffic classifiers were placed on an intersection approach to establish a speed profile as drivers approached the intersections. • Intersection Site – An intersection site refers to an entire intersection, including all approaches. At all but one of the intersection sites, speeds were measured at two of the intersection approaches (see Table 5-1). 5.4. DATA ANALYSIS APPROACH The objective of the speed data analysis was to evaluate the effectiveness of three treatments (peripheral transverse pavement markings, rumble strips, and dynamic warning signs activated by speed) at reducing vehicle speeds on high-speed intersection approaches. The primary analysis consisted of three steps, with each step building upon the previous one. The three steps were: • Determine the speed reduction at each data collection location (i.e., A, B, C, and D) – The speed reduction at an individual location was determined by comparing the speed data at an individual location before and after treatment installation. This approach is perhaps the least meaningful because it estimates treatment effectiveness only at a specific point along the

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-6 intersection approach. The results do not provide information about speed reduction at any other point along that intersection approach; thus, no information is provided about the treatment’s effect on the overall speed profile. However, because individual location speed data might be of interest to the reader, they have been summarized and are presented in Appendix G. • Determine the speed reduction along each intersection approach – The effect of a treatment on the speed profile of an intersection approach is determined by comparing speed data, before and after treatment installation, at all locations while accounting for base speed changes along the intersection approach before treatment installation. This analysis should identify, for example, if a treatment causes drivers to begin their deceleration farther upstream of the intersection than before the treatment was installed. • Determine the overall effectiveness of a particular treatment type – The effectiveness of a particular treatment type (i.e., peripheral transverse pavement markings, rumble strips, dynamic warning signs) is determined by comparing speed data, before and after treatment installation, across all intersection approaches at which the treatment was installed while accounting for site-to-site differences. Each analysis step is described in greater detail below. 5.4.1. Analysis by Location (Step 1 Analysis) The purpose of the analysis by location is to determine the speed reduction observed at each data collection location on an intersection approach as a result of the treatment installed on that approach. The observed speed reduction provides an estimate of the treatment effectiveness at a given location. This analysis is perhaps the least meaningful of the three analyses because it estimates the effectiveness of the treatment only at a specific point along the intersection approach. The effectiveness of a treatment at reducing the mean or median speed and the speed variability at a specific location were evaluated by comparing the speed data measured at an individual location before and after treatment installation. The first step of the analysis was to verify that the speed data were normally distributed as many statistical tests are based on this assumption. To accomplish this, histograms of speed data were plotted for each location. A visual inspection of the histograms identified traffic classifiers from which speed data did not appear to be normally distributed. The most common location for this to occur, on a given intersection approach, was the location closest to the intersection (Location A). It appears that turning movements taking place at the intersection created a sufficient number of slower vehicles to result in either skewed or bimodal data and, thus, violate the assumption of normality at that location. Histograms of the speed profile data at all ten sites are shown in Appendix F. The next step of the analysis was to compare speed variability between before and after treatment periods. This was accomplished using the modified Levene’s test of homogeneity of variance. This test is less sensitive than the standard Levene’s test to the assumption of normally distributed data. In essence, the test is constructed by calculating the absolute deviation from the sample median for each speed measurement, separately for the before and after speed groups,

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-7 and then using analysis of variance (ANOVA) to test that the means of this quantity are the same for the before and after speed groups. This ANOVA provides an F-test and associated p-value. Small p-values (≤ 0.05) are an indication that the before and after speed variances are unequal at the 95-percent confidence level. Based on the results of the check of normality and the homogeneity of variance test (i.e., modified Levene’s test), either mean or median speeds were compared to determine if the difference between the before and after speeds were significant. The statistical approach was as follows: • Normally distributed speed data and equal variances: a two-sample t-test with pooled variance was used to compare mean speeds before and after treatment • Normally distributed speed data and unequal variances: a two-sample t-test with weighted variance (with Satterthwaite’s approximation of degrees of freedom) was used to compare mean speeds before and after treatment • Non-normally distributed (skewed or bimodal) speed data (no test of equal variance is needed): a Wilcoxon rank-sum test was used to compare median speeds before and after treatment 5.4.2. Analysis by Intersection Approach (Step 2 Analysis) The purpose of the analysis by intersection approach was to determine the overall speed reduction observed along an intersection approach as a result of the treatment installed on that approach. The observed speed reduction provided an estimate of the overall treatment effectiveness along the approach. The overall effect of the treatment at a particular intersection approach, i.e., the average speed reduction across all data collection locations, was evaluated using a two-way analysis of variance (ANOVA). Two factors were included in the ANOVA model: • Distance from intersection to data collection location (i.e., A, B, C, or D) • Data collection period (before or after). An interaction between data collection location and period is also included to account for differences in the treatment effect by location. The assumptions made in the ANOVA include: • The variability in speed measurements was allowed to differ across locations and collection period. • All data (regardless of data collection location and period) was normally distributed. Although this assumption may not hold for a few of the locations, speed reduction estimates were unaffected by this assumption, and the effect on the overall conclusions was believed minor given the degree of non-normality.

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-8 • While speed measurements across data collection locations were likely correlated, due to the same vehicle being measured at each location as it approached the intersection, an assumption of independent measurements across locations was made. The implication of this assumption was considered when evaluating the statistical significance of the overall treatment effect. For each intersection approach, the overall speed difference (across Locations A, B, C, and D) was estimated and an F-test used to determine if the difference in mean speeds before and after treatment installation was significant. An interaction between treatment and location was also evaluated with an F-test. 5.4.3. Analysis by Treatment Type (Step 3 Analysis) The purpose of the analysis by treatment type is to determine the overall effectiveness of a particular treatment type (i.e., peripheral transverse pavement markings, rumble strips, dynamic warning signs). The effect of the three treatment types across all sites was evaluated through ANOVA with random effects. The fixed effects considered in the ANOVA are the data collection period (before/after) and the interaction between data collection period and location. The random effects included in the ANOVA are: • Intersection approach • Interaction between data collection period and intersection approach • Interaction between intersection approach and data collection location • Three-way interaction between intersection approach, data collection location, and data collection period The assumptions made in the ANOVA include: • Equal variance in speed measurements at each site approach and location • Independence of measurements across locations for a site-approach • Normal distribution for all random effects and speed measurements. For each treatment, the overall treatment effect on speeds across intersection approaches and data collection locations was estimated and an F-test used to determine if the difference in mean speeds before and after treatment installation was significant. The data collection period effect (treatment effect) was tested in comparison to the collection period by intersection approach interaction. Thus, the treatment effect was evaluated based on the number of intersection approaches where the treatment was applied. The treatment by location interaction was similarly tested. The collection period by location interaction was tested in comparison to the interaction between intersection approach, data collection location, and data collection period. Thus, the

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-9 treatment by location was compared to the number of locations for all intersection approaches with the same treatment. 5.4.4. Secondary Analysis of 85th-Percentile Speeds The statistical analyses discussed in the analyses by location, intersection approach, and treatment type (Step 1, Step 2, and Step 3 analyses) clearly focus on the comparison of speed distributions and their changes due installation of a speed-reduction treatment. In each of these analyses, the entire speed distribution at each individual data collection location is considered when making final estimates of speed differences between the before and after periods due to treatment installation. Although the emphasis in these analyses is placed on differences in mean speeds, the results of these analyses are applicable to the entire speed distribution; that is, to any quantile or percentile of that distribution. As such, the results of analyses of mean speeds may be applied to 85th-percentile speeds as well, assuming that speed distributions are normal. A rigorous statistical procedure, similar to the comparison of mean speeds as discussed in the Step 1, Step 2, and Step 3 analyses, has not yet been developed for 85th-percentiles (or any other quantile). However, an attempt was made to statistically compare 85th-percentile speeds before and after treatment installation for the analysis by interaction approach (Step 2 analysis) and the analysis by treatment type (Step 3 analysis). In each case, the 85th-percentile speeds were compared using a paired t-test on the differences between before and after 85th-percentile speeds obtained at each site-location with the following caveats: • The 85th-percentile speed was estimated from the speed distribution at each site and location individually; as a result, the variability in the 85th-percentile speed at each site and location is ignored. This, is turn, reduces the variability in the test statistic and increases the significance of the results (i.e., a difference in 85th-percentile speeds may be significant when it is not, simply because extra variability is ignored.) • The t-test assumes normal distribution of the 85th-percentile speeds; this assumption could not be checked due to small sample sizes (e.g., between 2 and 4 locations at the intersection approach level); normality was simply assumed in all the t-tests. • When an intersection has two lanes per approach with unequal traffic volumes (e.g., Texas sites with rumble strips), no weighting by traffic volume could be performed in either the before or after period and across periods. • The different sample sizes between the before and after treatment periods could not be accounted for in the t-tests; this was done in the more rigorous average speed comparisons since properly weighting by the number of speed measurements affects the data analysis. 5.4.5. Data Preparation Several steps were taken to prepare the speed data for analysis, including general data cleanup and data selection for a particular analysis. 5.4.5.1. General Data Cleanup

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-10 Speed data were carefully reviewed to determine which data to include in the analyses. For example, speed data of the following vehicles were not included in the analyses: • Vehicles traveling in the opposite direction but traveling over the sensors (assumed to be a vehicle involved in a passing maneuver) • Vehicles following another vehicle, with a headway less than 3 seconds between the two vehicles (i.e., the speed of the first vehicle was retained but the speed of the closely following second vehicle was not) In addition, plots of vehicle speed vs. time of day (by traffic classifier) and traffic flow rate vs. time of day (by traffic classifier) were reviewed to identify potential problems such as: • Inconsistencies in traffic flow – Major disruptions in traffic flow could indicate an accident on the roadway. While no major disruptions in traffic flow were identified, the traffic flow plots revealed at least two occasions when one classifier recorded substantially fewer vehicles than the other classifiers on the same approach (and there was no logical explanation, such as a major intersection or driveway between the classifiers). • Occasional malfunctioning of traffic classifier or roadway sensors – Periods of time during which no speeds were recorded or traffic flow rates appeared to be erratic usually indicated that either the traffic classifier was malfunctioning or the roadway sensors had come loose from the roadway. When a particular classifier recorded substantially fewer vehicles than the other classifiers on the approach, speed data from that classifier were not included in the analysis. When a particular classifier appeared to be malfunctioning or its roadway sensors came loose from the roadway, speed data from that classifier during the time period in question were not included in the analysis. Generally, equipment failure was an infrequent occurrence (site-specific problems with individual classifiers are addressed in the section “Results of Analysis by Intersection Approach”). However, data collection efforts at TX4 experienced substantial equipment problems. As a result, no speed data from TX4 could be included in an analysis. 5.4.5.2. Data Selection for Analyses Two data selection procedures were implemented to generate datasets that were most appropriate for a particular analysis approach (i.e., analysis by location, intersection approach, or treatment type): • When conducting the analysis by data collection location (Step 1 above), the most encompassing set of speed data was used. As such, no restrictions were imposed on matching time of day across locations at a particular intersection approach. • Evaluating the effectiveness of a treatment along an intersection approach (Step 2) or comparing the effectiveness of one treatment type to another (Step 3) involved developing speed profiles for each intersection approach. In this case, great care was taken to consider only those data collection periods during which all traffic classifiers on an intersection approach (Locations A, B, C, and D) were collecting speed data. This permitted the general

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-11 assumption that the same vehicles traversed all four data collection locations and that the population of vehicles at Location D were generally the same as the population of vehicles at the other locations. 5.5. DATA ANALYSIS RESULTS 5.5.1 Results of Analysis by Location (Step 1 Analysis) As discussed previously, the analysis of the speed data collected at each individual location (A, B, C, and D) is of minor importance in assessing overall treatment effectiveness. Nonetheless, descriptive summaries and the results from a before-after comparison of either mean or median speeds at each individual location are provided in Appendix G. General observations from the results of the analysis by location are: • Before-after speed comparisons were made for 51 individual traffic classifier locations (Locations A, B, C, and D) • At approximately 75 percent of the locations, speeds decreased; the decrease in speeds ranged from 0.1 to 8.7 mph. (It should be noted that the decrease of 8.7 mph – at Location A at WA1 eastbound – is an anomaly. The next largest decrease in speed is 3.2 mph.) • At approximately 25 percent of the locations, speeds increased; the increase ranged from 0.1 to 1.8 mph. • Of the 51 before-after speed differences, 34 were statistically significant at the 5-percent level (including a p-value of 0.051). The distribution of the 34 statistically significant speed differences by traffic classifier location was: o Location A (8) o Location B (10) o Location C (10) o Location D (6)

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-12 5.5.2 Results of Analysis by Intersection Approach (Step 2 Analysis) This section is organized by site in the order listed in Exhibit 5-1. The discussion of each site begins with a general site description, a summary of the data collection activities conducted at that site, and a site diagram. Next, description summaries of the speed data collected at each site are presented, including separately for the before-and-after treatment periods: • number of measurements • speed range (i.e., minimum and maximum speeds) • mean, median, and 85th-percentile speeds • speed standard deviation Next, the ANOVA results from the before-after statistical comparison of speed profile data at a particular intersection approach are presented in a table and subsequently discussed. • The column “Main Effect or Interaction” in that table refers to either the main treatment effect or the treatment by location interaction (abbreviated as “Treatment x Location”). • Columns 3 and 4 present the estimated speed difference (between the before and after periods), its standard error, and an estimate of the precision of the speed difference. A negative estimated speed difference indicates a decrease in mean speeds from before to after treatment installation. • The last three columns show the results of the F-test, with the last column indicating statistical significance of the treatment or the treatment by location interaction: a p-value of 0.05 or less (highlighted in yellow) will indicate a significant effect at that intersection approach at the 5-percent significance level. Thus, a p-value of 0.05 or less in a “Treatment” row will indicate a statistically significant treatment effect on speed profile at that approach. A p-value of 0.05 or less in a “Treatment x Location” row will indicate that the treatment effect on speed profile varies by location. • To conclude the discussion of each site, speed profile plots are shown separately for each approach. Each figure presents the mean speed profile, separately for the before period (solid blue line) and after period (dashed red line). The vertical dashed lines on each figure represent the locations (A, B, C, or D) on the approach where the speed-reduction treatment was installed. The interval around each estimated mean speed represents the 95-percent confidence bound of the mean. A summary discussion across all sites follows the presentation of the individual site results. 5.5.2.1 OR1 – Meridian Road/Whiskey Hill Road The Meridian Road/Whiskey Hill Road intersection is located in Clackamas County, Oregon. Peripheral transverse pavement markings were installed on the northbound approach of Meridian

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-13 Road and the eastbound approach of Whiskey Hill Road. The intersection is unsignalized with no control on Whiskey Hill Road and STOP-control on Meridian Road. The speed data collection sensors on the northbound approach of Meridian Road were placed 165, 265, 600, and 1,800 feet from the intersection. Peripheral transverse pavement markings were installed at two locations on the northbound intersection approach: 265 and 795 feet from the intersection. The speed data collection sensors on Whiskey Hill Road’s eastbound approach were placed 100, 250, 600, and 1,300 feet from the intersection. Peripheral transverse pavement markings were installed at two locations on the eastbound intersection approach: 420 and 728 feet from the intersection. Exhibit 5-2 presents a site diagram of the intersection and illustrates the treatment and speed data collection locations. WHISKEY HILL ROAD M E R ID IA N R O A D A 100' B 250' C 600' D 1300' A 165' B 265' C 600' D 1800' TPM 420' TPM 728' TPM 450' TPM 795' Exhibit 5-2 Site Diagram for OR1 ↑ N

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-14 Basic descriptive summaries for the before and after speed data at each location are presented in Exhibit 5-3. The comparison of the before and after mean speeds at all locations, while accounting for base speed changes (mostly deceleration) along the intersection approach, is summarized in Exhibit 5-4, separately for each intersection approach. Accounting for changes between segment speeds and intersection speeds before a treatment is implemented is important so as not to confound any potential treatment effect on speed with naturally occurring speed changes as drivers approach an intersection. Exhibit 5-3 Descriptive Summaries for OR1 Speed Statistics (mph) Location Number of Measurements Speed Range (Minimum - Maximum) Mean Median 85 th Percentile Standard Deviation Before Treatment A 417 8.3 - 41.5 29.3 29.8 34.2 5.2 B 410 9.9 - 48.8 35.0 35.2 40.3 5.7 C 410 8.2 - 64.9 44.7 45.3 52.3 8.0 NB D 385 16.4 - 77.8 52.4 53.1 61.2 8.6 A 180 10.7 - 50.4 30.4 29.2 38.8 7.0 B 177 19.5 - 53.3 37.0 37.3 43.0 5.9 C 177 27.1 - 59.2 44.3 45.2 50.1 6.0 EB D 159 27.5 - 72.2 49.2 49.4 55.5 6.7 After Treatment A 235 10.4 - 40.2 27.6 28.1 32.9 5.4 B 234 4.6 - 50.0 33.5 34.4 39.4 6.3 C 230 8.6 - 68.3 43.8 44.5 52.1 9.2 NB D 234 12.9 - 83.4 51.4 51.9 59.5 9.7 A 437 7.8 - 56.5 28.3 27.4 34.5 6.4 B 440 6.1 - 58.3 35.0 35.5 40.8 6.6 C 429 7.2 - 59.0 41.7 42.4 48.1 7.1 EB D 432 14.1 - 72.0 47.4 48.5 54.1 8.1

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-15 Exhibit 5-4 Treatment Effect on Speed Profile at OR1 Approaches Approach Main Effect or Interaction Estimated Speed Difference (mph) Standard Error (mph) F-Statistic Degrees of Freedom P-Value Treatment -1.2 0.3 15.6 (1, 2547) <.001 Treatment x Location 0.35 (3, 2547) 0.786 Location A -1.6 0.4 Location B -1.4 0.5 Location C -0.9 0.7 NB Location D -1.0 0.8 Treatment -2.1 0.3 50.87 (1, 2423) <.001 Treatment x Location 0.32 (3, 2423) 0.811 Location A -2.1 0.6 Location B -2.0 0.5 Location C -2.6 0.6 EB Location D -1.8 0.7 Exhibits 5-3 and 5-4 present mean speed profiles for the northbound and eastbound approach, respectively. Both intersection approaches experienced a slight reduction in mean, median, and 85th-percentile speeds after installing the peripheral transverse pavement markings (Exhibit 5-3). The speed reduction appears to be greater on the eastbound approach than on the northbound approach (Exhibit 5-4). The reduction in mean speeds on both intersection approaches (1.2 mph and 2.1 mph) is statistically significant as indicated by the small p-values for treatment in Exhibit 5-4.

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-16 Transverse Pavem ent M arkings T ransverse Pavem ent M arkings A B C D Col lection Period Before A fter S p e e d ( m p h ) 20 30 40 50 60 Distance from In tersection (ft) 1,800 1,600 1,400 1,200 1,000 800 600 400 200 0 Exhibit 5-5 Mean Speed Profile for NB Traffic at OR1 T ransverse Pavement M arkings Transverse Pavem ent M arkings A B C D Col lection Period Before A fter S p e e d ( m p h ) 20 30 40 50 60 Distance from In tersection (ft) 1,800 1,600 1,400 1,200 1,000 800 600 400 200 0 Exhibit 5-6 Mean Speed Profile for EB Traffic at OR1 5.5.2.2 OR2 – Canby-Marquam Highway/Lone Elder Road The Canby-Marquam Highway/Lone Elder Road intersection is in Clackamas County, Oregon.

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-17 Peripheral transverse pavement markings were installed on the northbound and southbound approaches. The intersection is unsignalized with no control on Canby-Marquam Highway and STOP-control on Lone Elder Road. The speed data collection sensors on the northbound approach of Canby-Marquam Highway were placed 100, 250, 700, and 1,500 feet from the intersection. Peripheral transverse pavement markings were installed at two locations on the intersection approach: 417 and 860 feet from the intersection. The speed data collection sensors on the southbound approach of Canby-Marquam Highway were placed 175, 250, 625, and 1,300 feet from the intersection. Peripheral transverse pavement markings were installed at two locations on the intersection approach: 350 and 854 feet from the intersection. Exhibit 5-7 presents an intersection site diagram and illustrates the treatment and speed data collection locations. LONE ELDER ROAD C A N B Y -M A R Q U A M H W Y A 175' TPM 417' B 250' C 625' TPM 860' D 1300' A 100' B 250' C 700' D 1500' TPM 350' TPM 854' Exhibit 5-7 Site Diagram for OR2 ↑ N

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-18 Basic descriptive summaries for the before and after speed data at each location are presented in Exhibit 5-8. The comparison of the before and after mean speeds at all locations, while accounting for base speed changes along the intersection approach before treatment implementation, is summarized in Exhibit 5-9, separately for each intersection approach. Exhibit 5-8 Descriptive Summaries for OR2 Speed Statistics (mph) Location Number of Measurements Speed Range (Minimum - Maximum) Mean Median 85th Percentile Standard Deviation Before Treatment A 1,457 4.3 – 65.2 37.6 38.8 49.6 11.4 C 1,399 16.7 - 66.0 47.4 47.7 53.7 6.3 NB D 1,278 10.4 - 65.5 46.8 47.5 52.4 6.3 B 1,307 12.3 - 69.5 42.6 43.5 50.1 7.8 C 1,302 14.1 - 73.2 48.5 48.8 54.0 6.0 SB D 1,264 15.5 - 75.8 49.8 50.0 55.1 5.7 After Treatment A 1,023 11.4 - 67.8 38.0 39.5 49.2 10.5 C 971 10.8 - 72.8 47.2 47.5 53.0 6.4 NB D 850 15.3 - 65.1 47.1 47.5 52.1 5.7 B 571 16.2 - 60.6 42.2 42.7 49.1 7.0 C 576 4.7 – 63.9 47.5 48.1 53.4 6.4 SB D 581 21.3 - 66.1 48.7 49.1 54.5 5.9 The traffic classifier at Location B on the northbound approach recorded substantially fewer vehicles in the before period than the other traffic classifiers on the approach. Since there was no logical explanation for this (e.g., major intersection or driveway onto which vehicles would be turning), it was assumed that the classifier was not functioning properly and, therefore, no data from this classifier were included in the analysis of the northbound approach. Also, on the southbound approach, before-period speed data were collected on a different day at Location A than at Locations B, C, and D. Therefore, no data from the traffic classifier at Location A were included in the southbound approach analysis.

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-19 Exhibit 5-9 Treatment Effect on Speed Profile at OR2 Approaches Approach Main Effect or Interaction Estimated Speed Difference (mph) Standard Error (mph) F-Statistic Degrees of Freedom P-Value Treatment 0.15 0.19 0.59 (1, 6972) 0.443 Treatment x Location 1.07 (2, 6972) 0.343 Location A 0.42 0.44 Location C -0.21 0.27 NB Location D 0.24 0.26 Treatment -0.83 0.19 19.55 (1, 5595) <.001 Treatment x Location 1.31 (2, 5595) 0.269 Location B -0.38 0.36 Location C -1.03 0.31 SB Location D -1.08 0.29 Exhibits 5-10 and 5-11 present mean speed profiles for the northbound and southbound approaches, respectively. The speed differential on the northbound approach after installing the peripheral transverse pavement markings appears to be negligible; mean and median speeds either slightly increased or decreased depending on the location along the intersection; and all 85th -percentile speeds slightly decreased (Exhibit 5-8). The overall increase of 0.15 mph in mean speeds is not statistically significant as indicated by the large p-value for treatment in Exhibit 5-9. The southbound approach appears to have experienced a slight reduction in mean, median, and 85th-percentile speeds after installing the peripheral transverse pavement markings (Exhibit 5-8). The reduction of 0.83 mph in mean speeds on the southbound approach is statistically significant as shown by the small treatment p-value in Exhibit 5-9.

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-20 Transverse Pavem ent M arkings T ransverse Pavem ent M arkings A CD Col lection Period Before A fter S p e e d ( m p h ) 20 30 40 50 60 Distance from In tersection (ft) 1,800 1,600 1,400 1,200 1,000 800 600 400 200 0 Exhibit 5-10 Mean Speed Profile for NB Traffic at OR2 T ransverse Pavement M arkings T ransverse Pavem ent M arkings B C D Col lection Period Before A fter S p e e d ( m p h ) 20 30 40 50 60 Distance from In tersection (ft) 1,800 1,600 1,400 1,200 1,000 800 600 400 200 0 Exhibit 5-11 Mean Speed Profile for SB Traffic at OR2

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-21 5.5.2.3 OR3 – Redland Road/Ferguson Road The Redland Road/Ferguson Road intersection is in Clackamas County, Oregon. Peripheral transverse pavement markings were installed on the eastbound and westbound approaches. The intersection immediately east of this site (Redland and Bradley roads) was also a study site (OR4), and speed data were collected at these locations simultaneously. Exhibit 5-12 presents a site diagram of Sites OR3 and OR4 and illustrates the locations of treatment installation and speed data collection for both sites. The close proximity of these intersections did not allow for multiple data collection points between the intersections; only one set of sensors was placed on each approach between the intersections. Therefore, data collection was focused on the eastbound approach at OR3 and the westbound approach at OR4. The intersection of Redland and Ferguson roads is unsignalized with no control on Redland Road and STOP-control on Ferguson Road. The speed data collection sensors on the eastbound approach were placed 100, 250, 500, and 1,200 feet from the intersection. Peripheral transverse pavement markings were installed at two locations on the eastbound intersection approach: 318 and 550 feet from the intersection. The speed data collection sensors on the westbound approach were placed 275 feet from the intersection and peripheral transverse pavement markings were installed 660 feet from the intersection. Speed data from the traffic classifier on the westbound approach were not included in the analysis since there was only one data collection point and, therefore, no speed profile could be developed. Basic descriptive summaries for the before and after speed data at each location are presented in Exhibit 5-13. The comparison of the before and after mean speeds at all locations, while accounting for base deceleration along the intersection approach before treatment implementation, is summarized in Exhibit 5-14, separately for each intersection approach. F E R G U S O N R O A D REDLAND ROAD D 1200' B R A D LE Y R O A D C 500' B 250' A 100' TPM 550' TPM 318' B 250' C 1200' D 2000' TPM 508' TPM 1488' TPM 660' TPM 485' B 275' B 275' Exhibit 5-12 Site Diagram for OR3 and OR4 ↑ N

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-22 Exhibit 5-13 Descriptive Summaries for OR3 Speed Statistics (mph) Location Number of Measurements Speed Range (Minimum - Maximum) Mean Median 85th Percentile Standard Deviation Before Treatment A 1,381 7.0 - 68.4 45.3 47.7 53.1 9.1 B 1,351 8.3 - 90.2 47.0 47.9 53.0 7.3 C 1,356 14.5 - 71.2 48.9 49.1 54.0 5.5 EB D 1,336 18.0 - 71.4 49.6 49.8 54.6 5.6 After Treatment A 780 8.1 - 67.5 44.9 46.9 52.3 8.5 B 802 11.6 - 67.0 46.5 47.1 52.4 6.3 C 777 23.4 - 67.3 48.3 48.5 53.3 5.4 EB D 785 5.4 - 74.0 49.2 50.0 54.5 6.8 Exhibit 5-14 Treatment Effect on Speed Profile at OR3 Approach Approach Main Effect or Interaction Estimated Speed Difference (mph) Standard Error (mph) F-Statistic Degrees of Freedom P-Value Treatment -0.44 0.15 8.03 (1, 8560) 0.005 Treatment x Location 0.15 (3, 8560) 0.932 Location A -0.46 0.39 Location B -0.41 0.3 Location C -0.56 0.24 EB Location D -0.32 0.29 Exhibit 5-15 presents mean speed profiles for the eastbound approach. The eastbound approach experienced a slight reduction in mean, median, and 85th-percentile speeds after installing the peripheral transverse pavement markings (Exhibit 5-13). The reduction of 0.44 mph in mean speeds is statistically significant as shown by the small treatment p-value in Exhibit 5-14.

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-23 T ransverse Pavement M arkings T ransverse Pavem ent Markings A B CD Col lection Period Before After S p e e d ( m p h ) 20 30 40 50 60 Distance from In tersection (ft) 1,800 1,600 1,400 1,200 1,000 800 600 400 200 0 Exhibit 5-15 Mean Speed Profile for EB Traffic at OR3 5.5.2.4 OR4 – Redland Road and Bradley Road The Redland and Bradley road intersection is in Clackamas County, Oregon. Peripheral transverse pavement markings were installed on the eastbound and westbound approaches. The intersection immediately to the east of this site (Redland and Ferguson roads) was also a study site (OR3) and, as mentioned previously, speed data were collected at both locations simultaneously. The intersection of Redland and Bradley roads is unsignalized with no control on Redland Road and STOP-control on Bradley Road. The speed data collection sensors on the eastbound approach were placed 275 feet from the intersection and peripheral transverse pavement markings were installed 485 feet from the intersection. Speed data from the traffic classifier on the eastbound approach were not included in the analysis since there was only one data collection point and, therefore, no speed profile could be developed. The speed data collection sensors on the westbound approach were placed 250, 1,200, and 2,000 feet from the intersection. Peripheral transverse pavement markings were installed at two locations on the westbound intersection approach: 508 and 1,488 feet from the intersection. Basic descriptive summaries for the before and after speed data at each location are presented in Exhibit 5-16. The comparison of the before and after mean speeds at all locations, while accounting for base deceleration along the intersection approach before treatment implementation, is summarized in Exhibit 5-17, separately for each intersection approach.

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-24 Exhibit 5-16 Descriptive Summaries for OR4 Speed Statistics (mph) Location Number of Measurements Speed Range (Minimum - Maximum) Mean Median 85th Percentile Standard Deviation Before Treatment B 1,349 9.6 - 74.3 52.6 53.3 58.2 6.5 C 1,352 16.0 - 77.4 52.0 52.4 57.5 6.7 WB D 1,366 5.0 - 78.5 54.8 55.4 60.5 7.0 After Treatment B 1,072 19.6 - 73.0 51.2 51.8 56.7 6.3 C 1,029 3.9 - 105.0 50.0 51.0 56.7 8.6 WB D 1,070 4.6 - 78.5 53.3 54.1 59.2 7.6 Exhibit 5-17 Treatment Effect on Speed Profile at OR4 Approach Approach Main Effect or Interaction Estimated Speed Difference (mph) Standard Error (mph) F-Statistic Degrees of Freedom P-Value Treatment -1.6 0.2 87.56 (1, 7232) <.001 Treatment x Location 0.99 (2, 7232) 0.372 Location B -1.4 0.3 Location C -2.0 0.3 WB Location D -1.4 0.3 Exhibit 5-18 presents mean speed profiles for the westbound approach. The westbound approach experienced a slight reduction in mean, median, and 85th-percentile speeds after installing the peripheral transverse pavement markings (Exhibit 5-16). The reduction of 1.6 mph in mean speeds is statistically significant as indicated by the small p-value for treatment shown in Exhibit 5-17.

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-25 T ransverse Pavem ent Marki ngs T ransverse Pavement M arkings BC D Collection Period Before A fter S p e e d ( m p h ) 20 30 40 50 60 Distance from In tersecti on (ft) 2,000 1,800 1,600 1,400 1,200 1,000 800 600 400 200 0 Exhibit 5-18 Mean Speed Profile for WB Traffic at OR4 5.5.2.5 OR6 – OR 6/Wilson River Loop Road The OR 6/Wilson River Loop Road intersection is in Tillamook County, Oregon. Peripheral transverse pavement markings were installed on the eastbound and westbound approaches. The intersection is unsignalized with no control on OR 6 and STOP-control on Wilson River Loop Road. The speed data collection sensors on the eastbound approach of OR 6 were placed 100, 250, 900, and 1,700 feet from the intersection. Peripheral transverse pavement markings were installed at approximately 530 and 1,320 feet from the intersection on the approach. The speed data collection sensors on the westbound approach of OR 6 were placed 280, 440, 980, and 1,580 feet from the intersection. Peripheral transverse pavement markings were installed at approximately 630 and 1,320 feet from the intersection on the approach. Exhibit 5-19 presents a site diagram of the intersection and illustrates the projected locations of treatment installation and actual speed data collection locations. The traffic classifier at Location D on the eastbound approach never functioned properly during the before- and after-period data collection efforts. In fact, from the traffic flow plots that were reviewed during data cleanup, it appeared that this classifier did not record any speeds. Therefore, no speed data from this classifier were available for analysis. Also, it should be noted that there were no data collection time periods at Location D that were common with the other three locations, so data from the classifier at Location D were not included in the analysis.

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-26 SR6 WILSON RIVER LOOP ROAD A 100' TPM 530' B 250'C 900' TPM 1320' A 280' B 440' C 980' TPM 630' D 1700' D 1580' TPM 1320' Exhibit 5-19 Site Diagram for OR6 Basic descriptive summaries for the before and after speed data at each location are presented in Exhibit 5-20. The comparison of the before and after mean speeds at all locations, while accounting for base speed changes along the intersection approach before treatment implementation, is summarized in Exhibit 5-21, separately for each intersection approach. ↑ N

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-27 Exhibit 5-20 Descriptive Summaries for OR6 Speed Statistics (mph) Location Number of Measurements Speed Range (Minimum - Maximum) Mean Median 85th Percentile Standard Deviation Before Treatment A 822 13.5 – 66.6 51.5 54.0 58.5 9.4 B 852 23.2 – 66.4 52.7 54.1 58.6 6.8 EB C 918 8.9 – 71.6 55.9 56.2 60.4 4.9 A 1,139 4.7 – 69.2 51.0 52.9 58.6 8.3 B 1,079 13.5 – 74.2 53.3 54.0 59.4 6.4 C 1,212 7.2 – 73.3 55.5 55.9 60.4 5.7 WB D 1,232 15.3 – 76.2 54.5 56.6 61.4 9.6 After Treatment A 1,029 15.0 - 74.4 53.1 54.7 59.0 8.2 B 1,065 5.4 - 76.7 53.6 54.8 59.3 6.9 EB C 1,125 14.8 - 76.7 56.3 56.4 60.6 5.0 A 668 12.6 - 68.7 51.8 53.6 59.8 8.0 B 685 33.6 - 66.7 52.2 53.0 58.2 5.6 C 687 37.7 - 71.9 55.6 55.8 60.2 5.0 WB D 683 17.6 - 76.9 53.3 55.4 59.8 9.7 Exhibit 5-21 Treatment Effect on Speed Profile at OR6 Approaches Approach Main Effect or Interaction Estimated Speed Difference (mph) Standard Error (mph) F-Statistic Degrees of Freedom P-Value Treatment 1.0 0.2 27.01 (1, 5805) <.001 Treatment x Location 3.22 (2, 5805) 0.04 Location A 1.6 0.4 Location B 0.9 0.3 EB Location C 0.4 0.2 Treatment -0.4 0.2 4.03 (1, 7377) 0.045 Treatment x Location 7.02 (3, 7377) <.001 Location A 0.8 0.4 Location B -1.1 0.3 Location C 0.1 0.3 WB Location D -1.2 0.5

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-28 Exhibits 5-22 and 5-23 present mean speed profiles for the eastbound and westbound approaches, respectively. The eastbound approach experienced a slight and unexpected increase in mean, median, and 85th-percentile speeds after installing the peripheral transverse pavement markings (Exhibit 5-20). The overall increase of 1.0 mph in mean speeds was statistically significant as indicated by the small treatment p-value in Exhibit 5-21. In addition, the significant interaction between treatment and location along the intersection approach as shown by the small p-value for interaction in Exhibit 5-21 indicates that the speed increase due to treatment varies from location to location (1.6 mph at Location A; 0.9 mph at Location B; and 0.4 mph at Location C). No explanation for the increase in mean speeds or the varying effect across locations was identified. In the westbound direction, there was a slight change (increase or decrease) in mean, median, and 85th-percentile speeds after installing treatments, depending on the location along the intersection (Exhibit 5-20). The overall mean speed reduction of 0.4 mph on the westbound approach was statistically significant as shown by the small treatment p-value in Exhibit 5-21. However, it should be noted that the interaction between treatment and location (i.e., increase of 0.8 mph at Location A; decrease of 1.1 mph at Location B; increase of 0.1 mph at Location C; and decrease of 1.2 mph at Location D) was also statistically significant as indicated by the small p-value for interaction in Exhibit 5-21, indicating that the degree of treatment effectiveness was not consistent along the intersection approach. T ransverse Pavem ent M arkings T ransverse Pavem ent M arkings AB C Col lection Peri od Before After S p e e d ( m p h ) 20 30 40 50 60 Distance from In tersection (ft) 1 ,800 1,600 1,400 1,200 1,000 800 600 400 200 0 Exhibit 5-22 Mean Speed Profile for EB Traffic at OR6

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-29 Tran sverse Pave m ent M a rkin gs T ra nsverse Pa vem e nt M arking s A B C D Co llection Pe riod Before After S p e e d ( m p h ) 20 30 40 50 60 Distan ce fro m Intersectio n (ft) 1,800 1,6 00 1,400 1,20 0 1 ,0 00 800 6 00 400 200 0 Exhibit 5-23 Mean Speed Profile for WB Traffic at OR6 5.5.2.6 TX1 – US 271 and FM 726 The US 271/FM 726 intersection is in Upshur Country, Texas. Rumble strips were installed on the southbound approach (rumble strips were already present on the northbound approach). The intersection has a flashing-yellow signal on US 271 and a flashing-red STOP-control signal on FM 726. The speed data collection sensors on US 271’s southbound approach were placed 100, 200, 600, and 1,800 feet from the intersection. Rumble strips were installed at three locations on the intersection approach: 330, 935, and 1,710 feet from the intersection. Exhibit 5-24 presents a site diagram of the intersection and illustrates the treatment installation and speed data collection locations.

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-30 FM 726 A 100' B 200' C 600' D 1800' H IG H W A Y 2 71 RS 935' RS 330' RS 1710' Exhibit 5-24 Site Diagram for TX1 Basic descriptive summaries for the before and after speed data at each location are presented in Exhibit 5-25. The comparison of the before and after mean speeds at all locations, while accounting for base speed changes along the intersection approach before treatment implementation, is summarized in Exhibit 5-26, separately for each intersection approach. ↑ N

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-31 Exhibit 5-25 Descriptive Summaries for TX1 Speed Statistics (mph) Location Number of Measurements Speed Range (Minimum – Maximum) Mean Median 85th Percentile Standard Deviation Before Treatment A 1,343 19.4 - 78.8 57.1 59.0 67.6 11.1 B 1,329 16.5 - 79.1 59.3 60.4 68.3 8.7 C 1,357 20.3 - 80.0 63.4 63.9 69.8 6.5 SB D 1,363 35.7 - 85.2 66.0 66.7 71.8 6.2 After Treatment A 1,274 24.3 - 80.7 57.2 58.6 67.5 10.7 B 1,296 35.1 - 81.4 59.0 59.7 68.0 8.5 C 1,284 30.9 - 89.5 62.7 63.1 70.2 7.1 SB D 1,260 22.6 - 110.0 65.9 66.9 71.9 6.5 Exhibit 5-26 Treatment Effect on Speed Profile at TX1 Approach Approach Main Effect or Interaction Estimated Speed Difference (mph) Standard Error (mph) F-Statistic Degrees of Freedom P-Value Treatment -0.3 0.2 4.33 (1, 7883) 0.037 Treatment x Location 1.55 (2, 7883) 0.212 Location B -0.4 0.3 Location C -0.7 0.3 SB Location D 0.0 0.3

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-32 Exhibit 5-27 presents mean speed profiles for the southbound approach. There was a slight, and statistically significant, speed reduction of 0.3 mph on the southbound intersection approach after installing rumble strips. Rum ble Strips Rumble Strips Rumble Strips B C D Collection Period Before After S p e e d ( m p h ) 20 30 40 50 60 70 80 Distance from Intersection (ft) 1,800 1,600 1,400 1,200 1,000 800 600 400 200 0 Exhibit 5-27 Mean Speed Profile for SB Traffic at TX1

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-33 5.5.2.7 TX2 – US 271/FM 2088 The US 271/FM 2088 intersection is in Upshur County, Texas. Rumble strips were installed on the northbound and southbound approaches. The intersection has a flashing-yellow signal on US 271 and a flashing-red STOP-control signal on FM 2088. The speed data collection sensors on the northbound approach of US 271 were placed 175, 300, 750, and 1,800 feet from the intersection. Rumble strips were installed at two locations on the intersection approach: 700 and 1,080 feet from the intersection. The speed data collection sensors on the southbound approach of US 271 were placed 100, 250, 750, and 1,615 feet from the intersection. Rumble strips were installed at two locations on the intersection approach: 500 and 900 feet from the intersection. Exhibit 5-28 presents a site diagram of the intersection and illustrates the treatment installation and speed data collection locations. A 100' B 250' C 750' D 1615' H IG H W A Y 2 71 RS 500' RS 900' A 175' B 300' C 750' D 1800' RS 700' RS 1080' FM 2088 Exhibit 5-28 Site Diagram for TX2 ↑ N

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-34 Basic descriptive summaries for the before and after speed data at each location are presented in Exhibit 5-29. The comparison of the before and after mean speeds at all locations, while accounting for base speed changes along the intersection approach before treatment implementation, is summarized in Exhibit 5-30, separately for each intersection approach. Exhibit 5-29 Descriptive Summaries for TX2 Speed Statistics (mph) Location Number of Measurements Speed Range (Minimum - Maximum) Mean Median 85th Percentile Standard Deviation Before Treatment A 353 18.4 - 81.4 56.5 57.3 65.1 8.8 B 294 12.2 - 81.6 55.5 57.2 63.4 10.4 C 413 38.1 - 84.5 61.8 62.7 68.2 7.0 NB D 100 23.8 - 77.6 65.0 66.8 71.9 8.3 A 48 14.8 - 77.4 56.0 56.7 64.7 10.7 B 47 26.5 - 78.5 57.8 57.7 65.1 8.9 C 47 50.7 - 79.5 61.5 60.3 67.0 7.0 SB D 43 48.8 - 85.2 67.2 67.5 71.5 6.5 After Treatment A 1,172 10.4 - 85.6 56.2 57.0 64.1 9.1 B 320 18.9 - 74.8 56.3 56.7 63.4 7.8 C 1,268 16.7 - 87.6 60.4 61.5 67.9 8.5 NB D 998 20.9 - 87.4 66.3 67.0 72.1 6.3 A 264 21.6 - 75.6 54.8 55.5 61.3 8.2 B 230 24.6 - 75.7 55.4 56.4 61.8 8.2 C 257 20.3 - 76.2 59.1 59.3 65.5 6.7 SB D 245 43.0 - 80.4 66.7 67.2 71.6 5.3

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-35 Exhibit 5-30 Treatment Effect on Speed Profile at TX2 Approaches Approach Main Effect or Interaction Estimated Speed Difference (mph) Standard Error (mph) F-Statistic Agrees of Freedom P-Value Treatment 0.1 0.3 0.05 (1, 4910) 0.83 Treatment x Location 3.71 (3, 4910) 0.011 Location A -0.3 0.5 Location B 0.7 0.8 Location C -1.4 0.4 NB Location D 1.2 0.9 Treatment -1.6 0.7 5.86 (1, 1173) 0.016 Treatment x Location 0.63 (3, 1173) 0.595 Location A -1.2 1.6 Location B -2.3 1.4 Location C -2.4 1.1 SB Location D -0.5 1.0 Exhibits 5-31 and 5-32 present mean speed profiles for the northbound and southbound approaches, respectively. The northbound approach experienced a slight reduction in mean, median, and 85th-percentile speed at all locations except Location D (all three speed statistics) and Location B (mean speed only) (Exhibit 5-29). The overall increase of 0.1 mph in mean speed is not statistically significant as indicated by the large treatment p-value in Exhibit 5-30. However, the extent to which mean speed reduction varied across data collection locations was statistically significant (small interaction p-value in Exhibit 5-30). Looking at Exhibit 5-31, it is apparent that the speed profiles from the before and after data cross each other twice, indicating that the treatment effectiveness on this intersection approach is inconsistent. The southbound approach overall experienced a reduction in mean, median, and 85th-percentile speeds (Exhibit 5-29). The overall reduction of 1.6 mph in mean speeds was statistically significant (small treatment p-value in Table Exhibit 5-30) and is consistent across locations along the intersection approach (large interaction p-value in Exhibit 5-30).

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-36 Rum ble S trips Rum ble S tri ps AB C D Col lection P eriod Before A fter S p e e d ( m p h ) 20 30 40 50 60 70 80 Distance from In tersection (ft) 1 ,800 1,600 1,400 1,200 1,000 800 600 400 200 0 Exhibit 5-31 Mean Speed Profile for NB Traffic at TX2 Rumble Strips Rum ble Stri ps A B C D Collection Period Before A fter S p e e d ( m p h ) 20 30 40 50 60 70 80 Distance from In tersecti on (ft) 1,800 1,600 1,400 1,200 1,000 800 600 400 200 0 Exhibit 5-32 Mean Speed Profile for SB Traffic at TX2

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-37 5.5.2.8 TX3 – US 271/FM 593 The US 271/FM 593 intersection is in Upshur County, Texas. Rumble strips were installed on the northbound and southbound approaches. The intersection has a flashing-yellow signal on US 271 and a flashing-red STOP-control signal on FM 593. The speed data collection sensors on US 271’s northbound approach were placed 125, 275, 600, and 1,800 feet from the intersection. Rumble strips were installed at three locations on the intersection approach: 330, 900 and 1,700 feet from the intersection. The speed data collection sensors on US 271’s southbound approach were placed 125, 250, 600, and 1800 feet from the intersection. Rumble strips were installed at three locations on the intersection approach: 330, 940, and 1,730 feet from the intersection. Exhibit 5-33 presents a site diagram of the intersection and illustrates the treatment installation and speed data collection locations. Data collection efforts at this site experienced substantial equipment problems. As a result, no data collection time periods were common across traffic classifiers. Thus, no analysis could be conducted. A 125' B 250' C 600' D 1800' H IG H W A Y 2 71 RS 330' RS 940' A 125' B 275' C 600' D 1800' RS 330' RS 900' FM 593 RS 1730' RS 1700' Exhibit 5-33 Site Diagram for TX3 ↑ N

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-38 5.5.2.9 TX4 – US 82/SH 98 The US 82/SH 98 intersection is in Bowie County, Texas. Dynamic warning signs activated by speed were installed on the eastbound and westbound approaches. The intersection has a flashing-yellow signal on US 82 and a flashing-red STOP-control signal on SH 98. The speed data collection sensors on US 82’s eastbound approach were placed 100, 250, 600, and 1,300 feet from the intersection. The dynamic warning sign activated by speed was installed 650 feet from the intersection on the approach. The speed data collection sensors on US 82’s westbound approach were placed 100, 300, 600, and 1,800 feet from the intersection. The dynamic warning sign was installed 630 feet from the intersection on the approach. Exhibit 5-34 presents an intersection site diagram and illustrates the treatment installation and speed data collection locations. Basic descriptive summaries for the before and after speed data at each location are presented in Exhibit 5-35. The comparison of the before and after mean speeds at all locations, while accounting for base deceleration along the intersection approach before treatment implementation, is summarized in Exhibit 5-36, separately for each intersection approach. S H 9 8 A 100' DWS 630' US 82 B 300' C 600' D 1800' A 100' B 250' C 600' D 1300' DWS 650' Exhibit 5-34 Site Diagram for TX4 ↑ N

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-39 Exhibit 5-35 Descriptive Summaries for TX4 Speed Statistics (mph) Location Number of Measurements Speed Range (Minimum - Maximum) Mean Median 85th Percentile Standard Deviation Before Period B 389 24.5 - 84.1 52.0 52.4 58.4 7.2 C 452 23.7 - 67.4 53.7 53.8 58.5 5.2 WB D 439 14.7 - 73.9 55.1 55.3 60.4 5.5 After Period B 816 12.0 - 67.4 49.1 50.5 55.9 7.1 C 796 18.0 - 80.7 51.6 52.4 56.9 6.0 WB D 781 19.6 - 78.2 54.5 54.7 60.3 6.1 Exhibit 5-36 Treatment Effect on Speed Profile at TX4 Approach Approach Main Effect or Interaction Estimated Speed Difference (mph) Standard Error (mph) F-Statistic Degrees of Freedom P-Value Treatment -1.9 0.2 75.08 (1, 3667) <.001 Treatment x Location 9.92 (2, 3667) <.001 Location B -2.9 0.4 Location C -2.1 0.3 WB Location D -0.6 0.3 Exhibit 5-37 presents mean speed profiles for the westbound approach. The westbound approach experienced a reduction in mean, median, and 85th-percentile speeds at all locations (Exhibit 5- 35). The overall reduction of 1.9 mph in mean speeds was statistically significant (small treatment p-value shown in Exhibit 5-36). The treatment effect in reducing speeds was not consistent along locations as shown by the significant interaction p-value in Exhibit 5-36 (2.9- mph speed reduction at Location B; 2.1-mph speed reduction at Location C; and 0.6-mph speed reduction at Location D). No analysis was conducted for the eastbound approach due to equipment problems (and resulting inconsistent data collection periods across locations). Also, no data were collected at Location A on the westbound approach due to a malfunctioning traffic classifier.

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-40 Dynam ic Warn ing Sign B C D Coll ection Period Before A fter S p e e d ( m p h ) 20 30 40 50 60 Distance from In tersecti on (ft) 1,800 1,600 1,400 1,200 1,000 800 600 400 200 0 Exhibit 5-37 Mean Speed Profile for WB Traffic at TX4 5.5.2.10 WA1 – SR 26/S 1st Ave The SR 26/S 1st Ave intersection is in Adams County, Washington. Dynamic warning signs activated by speed were installed on the eastbound and westbound approaches. The intersection is unsignalized with no control on SR 26 and STOP-control on S 1st Ave. The speed data collection sensors on SR 26’s eastbound approach were placed 250, 850, 1,120, and 1,850 feet from the intersection. A dynamic warning sign activated by speed was installed 850 feet from the intersection. The speed data collection sensors on the westbound approach of SR 26 were placed 375, 850, 1,150, and 1,850 feet from the intersection. A dynamic warning sign activated by speed was installed 850 feet from the intersection. Exhibit 5-38 presents an intersection site diagram and illustrates the treatment installation and speed data collection locations. Basic descriptive summaries for the before and after speed data at each location are presented in Exhibit 5-39. The comparison of the before and after mean speeds at all locations, while accounting for base deceleration along the intersection approach before treatment implementation, is summarized in Exhibit 5-40, separately for each intersection approach. On the eastbound approach, the traffic classifier at Location A recorded substantially fewer speeds than the other three classifiers during the before period; in the after period, it did not function properly until the last hour of the data collection period. Therefore, no speed data from Location A were included in the analysis. The traffic classifier at Location D did not record any speeds in the before and after periods. Therefore, no speed data from Location D were available for analysis.

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-41 SR26 D 1850'1S T A V E A 250' C 1150' B 850' A 375' B 850' C 1120' D 1850' DWS 850' DWS 850' Exhibit 5-38 Site Diagram for WA1 Exhibit 5-39 Descriptive Summaries for WA1 Speed Statistics (mph) Location Number of Measurements Speed Range (Minimum - Maximum) Mean Median 85th Percentile Standard Deviation Before Treatment B 1,036 8.5 - 65.9 47.4 47.8 53.2 6.0 EB C 959 16.0 - 84.5 48.2 48.4 53.7 6.4 A 390 5.8 - 61.6 43.3 43.6 50.1 7.2 B 427 10.1 - 85.8 47.4 47.7 52.9 7.2 C 456 10.3 - 69.8 49.5 49.7 54.4 6.1 WB D 478 11.5 - 70.8 51.7 51.8 56.7 6.5 After Treatment B 352 9.1 - 64.9 44.2 44.9 51.0 7.1 EB C 352 4.5 - 89.6 46.2 46.8 52.3 7.1 A 533 3.4 - 95.4 42.2 42.4 48.7 7.4 B 480 4.2 - 98.7 45.4 46.0 50.9 6.7 C 582 4.0 - 67.7 46.7 47.7 52.5 7.4 WB D 563 14.4 - 70.5 50.6 50.6 56.0 6.1 ↑ N

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-42 Exhibit 5-40 Treatment Effect on Speed Profile at WA1 Approaches Approach Main Effect or Interaction Estimated Speed Difference (mph) Standard Error (mph) F-Statistic Agrees of Freedom P-Value Treatment -2.6 0.3 74.34 (1, 2695) <.001 Treatment x Location 3.81 (1, 2695) 0.051 Location B -3.2 0.4 EB Location C -2.0 0.4 Treatment -1.8 0.2 62.67 (1, 3901) <.001 Treatment x Location 3.51 (3, 3901) 0.015 Location A -1.1 0.5 Location B -2.0 0.5 Location C -2.8 0.4 WB Location D -1.2 0.4 Exhibits 5-41 and 5-42 present mean speed profiles for the eastbound and westbound approaches, respectively. Both the eastbound and westbound intersection approaches experienced a reduction in mean, median, and 85th-percentile speeds (Exhibit 5-39). The reductions in mean speeds of 2.6 mph at the eastbound approach and 1.8 mph at the westbound approach are both statistically significant as indicated by the significant treatment p-values in Exhibit 5-40. The difference in effectiveness across locations was also statistically significant in both directions as shown by the significant interaction p-values in Exhibit 5-40.

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-43 Dynam ic Warni ng Sign B C Collection Pe riod Before After S p e e d ( m p h ) 2 0 30 40 50 60 Distance from In tersection (ft) 1,800 1,600 1,400 1 ,200 1,000 800 600 400 200 0 Exhibit 5-41 Speed Profile for EB Traffic at WA1 Dynam ic Warn ing Sign A B C D Collection Pe riod Before After S p e e d ( m p h ) 2 0 30 40 50 60 Distance from In tersection (ft) 2,000 1,800 1,600 1,400 1,200 1,000 800 600 400 200 0 Exhibit 5-42 Speed Profile for WB Traffic at WA1

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-44 5.5.2.11 Summary of Results from Analysis by Intersection Approach ( Step 2 Analysis) The purpose of the analysis by intersection approach was to determine the overall speed reduction observed along an intersection approach as a result of the treatment installed on that approach. The analysis by intersection approach included 14 intersection approaches, with the three treatment types distributed as follows: • Peripheral transverse pavement markings – 8 approaches • Rumble strips – 3 approaches • Dynamic warning signs activated by speed – 3 approaches Exhibit 5-43 presents a summary of the results. On all but three of the intersection approaches, the estimated speed difference from before to after treatment installation was negative; that is, the speed decreased after treatment installation. Furthermore, all of the estimated speed reductions were statistically significant. Of the three cases where the estimated speed increased, only one was statistically significant. Another interesting finding is that all three intersection approaches on which dynamic warning signs were installed experienced a statistically significant speed reduction. Also, it appears that speed reductions were greatest on those approaches where dynamic warning signs were installed. The results of the analysis by treatment type are presented in the following section.

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-45 Exhibit 5-43 Summary of Treatment Effect on Speed Profile at Individual Data Collection Sites Site No. Intersection Site Intersection Approach Treatment Estimated Speed Difference (mph) Standard Error (mph) Statistically Significant? NB Peripheral transverse pavement markings -1.2 0.3 Yes OR1 Meridian Rd / Whiskey Hill Rd EB Peripheral transverse pavement markings -2.1 0.3 Yes NB Peripheral transverse pavement markings 0.2 0.2 No OR2 Canby-Marquam Hwy / Lone Elder Rd SB Peripheral transverse pavement markings -0.8 0.2 Yes OR3 Redland Rd / Ferguson Rd EB Peripheral transverse pavement markings -0.4 0.2 Yes OR4 Redland Rd / Bradley Rd WB Peripheral transverse pavement markings -1.6 0.2 Yes EB Peripheral transverse pavement markings 1.0 0.2 Yes OR6 SR 6 / Wilson River Loop Rd WB Peripheral transverse pavement markings -0.4 0.2 Yes TX1 US 271 / FM 726 SB Rumble strips -0.3 0.2 Yes NB Rumble strips 0.1 0.3 No TX2 US 271 / FM 2088 SB Rumble strips -1.6 0.7 Yes TX4 US 82 / SH 98 WB Dynamic warning sign -1.9 0.2 Yes EB Dynamic warning sign -2.6 0.3 Yes WA1 SR 26 / S 1st Ave WB Dynamic warning sign -1.8 0.2 Yes

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-46 5.5.3 Results of Analysis by Treatment Type (Step 3 Analysis) The previous section discussed potential treatment effects in reducing speeds on an intersection approach. The next logical step is to quantify the effect of a specific speed-reduction treatment across all intersection approaches at which it was installed. Transverse pavement markings were installed at 5 intersection sites for a total of 10 intersection approaches; rumble strips were installed at 2 intersection sites for a total of 3 intersection approaches; and dynamic warning signs were installed at 2 intersection sites for a total of 4 intersection approaches. The estimated effectiveness of each of the three treatments across all intersection approaches is presented in Exhibit 5-44. The “Main Effect or Interaction” column refers to either the overall treatment effect (Treatment), the interaction effect between treatment and data collection location (abbreviated as “Treatment x Location”), or the treatment effect at the individual locations (A, B, C, or D). The statistical significance of an effect is based on the p-value in the last column. Rows with p-values highlighted in yellow and bold indicate statistical significance based on a p- value cut-off of 0.05 (i.e., 5-percent significance level). The two footnotes pertain to the type of test used to test for significance: an F-test to test for significant treatment or treatment-by- location interaction; a t-test to test for significant treatment effect at a particular data collection location. The treatment effect is the estimated speed reduction across all intersection approaches. The treatment-by-location interaction tests whether the treatment effect is the same across data collection locations (note that in this analysis, differences between intersection sites are controlled for in the model). The location effects correspond to the estimated speed reduction across sites for each data collection location (i.e., A, B, C, or D).

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-47 Exhibit 5-44 Overall Treatment Effect on Speed Profile Across All Intersection Approaches Treatment Type Main Effect or Interaction Estimated Speed Difference (mph) Standard Error (mph) Test Statistic Degrees of Freedom P-Value Treatment a -0.6 0.3 3.68 (1,7) 0.097 Treatment x Location a 6.04 (6,17) 0.002 Location A b -0.3 0.4 -0.65 17 0.522 Location B b -0.7 0.4 -1.89 17 0.076 Location C b -0.9 0.4 -2.33 17 0.032 Peripheral transverse Pavement Markings Location D b -0.8 0.4 -2.04 17 0.058 Treatment a -0.6 0.4 1.83 (1,3) 0.269 Treatment x Location a 23.70 (6,6) 0.001 Location A b -0.7 0.7 -1.00 6 0.354 Location B b -0.4 0.5 -0.69 6 0.516 Location C b -1.3 0.5 -2.45 6 0.050 Rumble Strips Location D b 0.2 0.6 0.29 6 0.781 Treatment a -1.7 0.2 81.18 (1,2) 0.012 Treatment x Location a 71.06 (6,3) 0.003 Location A a -1.1 0.5 -1.98 3 0.141 Location B a -2.8 0.3 -9.25 3 0.003 Location C a -2.3 0.3 -7.70 3 0.005 Dynamic Warning Sign Location D a -0.9 0.4 -2.44 3 0.093 a F-Test a T- Test Exhibit 5-44 shows the following: • All treatment installations resulted in speed reductions from before to after installation at given locations, with the exception of rumble strips at Location D (speed increase of 0.2 mph with a standard error of 0.6 mph). • Only a few speed reductions (either overall or on a location basis) were statistically significant at the 5-percent level. • Only dynamic warning signs activated by speed have a significant overall effect (at the 5- percent significance level) in reducing speeds from before to after installation, with a mean speed reduction of 1.7 mph (standard error or precision of 0.2 mph).

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-48 • Peripheral transverse pavement markings have a significant overall effect (at the 10-percent significance level) in reducing speeds from before to after installation, with a mean speed reduction of 0.6 mph (standard error or precision of 0.3 mph). • Rumble strips did not have an overall effect in reducing speeds from before to after installation. • Treatment by location interactions was significant at the 5-percent level for all three treatment types; that is, the effectiveness of each treatment, whether overall statistically significant or not, is not consistent across location along an approach. • All three treatment types were effective in reducing speeds at Location C (intersection perception/response speed location) with mean speed reduction (standard error) of 0.9 mph (0.4 mph) for peripheral transverse pavement markings; 1.3 mph (0.5 mph) for rumble strips; and 2.3 mph (0.3 mph) for dynamic warning signs activated by speed. • Dynamic warning signs activated by speed were effective in reducing speeds at Location B (accident avoidance speed location), with a mean speed reduction (standard error) of 2.8 mph (0.3 mph). • Dynamic warning signs activated by speed showed the largest significant speed reduction as compared to all other reductions, with a mean speed decrease of 2.8 mph (standard error of 0.3 mph) at Location B. • None of the treatments significantly reduced speeds at Location D (farthest away from intersection), as expected. 5.5.4 Results of Secondary Analysis of 85th-Percentile Speeds As previously discussed in Section 5.4.4, a secondary analysis of 85th-percentile speeds was conducted for the analysis by intersection approach (Step 2 analysis) and the analysis by treatment type (Step 3 analysis). The results of the analyses, based on a paired t-test of the differences between before and after 85th-percentile speeds, are presented in Appendix H and discuss the following: • Treatment effect on 85th-percentile speed by intersection approach (to match level 2 analysis of mean speeds) • Overall treatment effect on 85th-percentile speed across all intersection approaches (to match level 3 analysis of mean speeds). The estimated speed differences in 85th percentiles shown in Exhibit H-1 may be compared to those shown in the treatment effect tables shown in Exhibits 5-4 through 5-40; while the differences are mostly comparable both in magnitude and direction (sign of difference), the significance levels (p-values) are often smaller; this is due to the fact that the variance in the 85th-percentile speed at a given location is ignored in this approach.

NCHRP 3-74 Chapter 5 Selection of Speed Reduction Treatments at High-Speed Intersections Testing Results Page 5-49 Similarly, the estimated speed differences in 85th percentiles shown in Exhibit H-2 may be compared to the estimated speed differences shown in Exhibit 5-44; again, while the differences are mostly comparable both in magnitude and direction (sign of difference), the significance levels (p-values) are often smaller; this is due to the fact that the variance in the 85th-percentile speed at a given location is ignored in this approach. 5.5.5. Testing Results Summary An analysis by location was conducted to determine the speed reduction observed at each data collection location (Location A, B, C, and D) on an intersection approach as a result of the treatment installed on that approach. Before-after speed comparisons were made for 51 individual locations. Key findings from this analysis include: • At approximately 75 percent of the locations, speed reductions of as much as 3.2 mph were observed (with one anomaly speed reduction of 8.7 mph). • At approximately 25 percent of the locations, speed increases of as much as 1.8 mph were observed. • Of the 51 before-after speed differences, 34 were statistically significant. An analysis by intersection approach was conducted to determine the overall speed reduction observed along an intersection approach as a result of the treatment installed on that approach. Fourteen intersection approaches – 8 with peripheral transverse pavement markings, 3 with rumble strips, and 3 with dynamic warning signs activated by speed – were included in this analysis. Key findings include: • At 11 of the 14 intersection approaches, the estimated speed difference from before to after treatment installation was negative; that is, the speed decreased after treatment installation. The estimated speed reductions were statistically significant. • At 3 of the 14 intersection approaches, the estimated speed difference was positive (the speed increased); however, only one of these speed differences was statistically significant. • All three intersection approaches on which dynamic warning signs were installed experienced a statistically significant speed reduction. • The speed reductions appeared to be greatest where dynamic warning signs were installed. An analysis by treatment type was conducted to quantify the effect of a specific speed-reduction treatment across all intersection approaches at which it was installed. Nineteen intersection approaches – 10 with peripheral transverse pavement markings, 5 with rumble strips, and 4 with dynamic warning signs activated by speed – were included in this analysis. Key findings include:

Chapter 5 NCHRP 3-74 Testing Results Selection of Speed Reduction Treatments at High-Speed Intersections Page 5-50 • Dynamic warning signs activated by speed were the only treatment to have a statistically significant overall effect at the 5-percent significance level in reducing speeds from before to after installation, with a mean speed reduction of 1.7 mph. • Peripheral transverse pavement markings had a statistically significant overall effect at the 10-percent significance level in reducing speeds from before to after installation, with a mean speed reduction of 0.6 mph. • Rumble strips did not have a statistically significant overall effect in reducing speeds from before to after installation • All three treatment types were effective in reducing speeds at Location C (intersection perception/response speed location), with mean speed reduction of 0.9 mph for peripheral transverse pavement markings; 1.3 mph for rumble strips; and 2.3 mph for dynamic warning signs activated by speed. • Dynamic warning signs activated by speed were effective in reducing speeds at Location B (accident avoidance speed location), with a mean speed reduction of 2.8 mph. • Dynamic warning signs activated by speed resulted in the largest speed reduction as compared to all other treatments. 5.6. CONCLUSIONS Overall conclusions from the testing results include: • All three treatment types may reduce speeds on high-speed intersection approaches; however, speed reduction is likely to be minimal (i.e., less than 3 mph). • The three treatment types appear to be most effective at reducing speeds at Location C (intersection perception/response speed location), which is the point where the intersection would first become visible to the driver or where the driver might first react to the intersection. • Of the three treatment types tested, dynamic warning signs activated by speed may be the most effective at reducing speeds. However, this conclusion is based on only three intersection approaches. • Peripheral transverse pavement marking also appear potentially effective at reducing speeds. • Based on a limited number of sites, rumble strips do not appear to be as effective at reducing speeds as dynamic warning signs or transverse pavement markings.