Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities (2011)

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.

24 This chapter describes the field data collection methodol- ogy used in this research. It discusses the experimental design and field methodology that were used to collect data with blind study volunteers at the test locations. It further discusses the process of identifying and selecting treatments, which ulti- mately led to the decision of which treatments should be tested in the field. The chapter then presents a summary of the site selection process and describes the treatment and sites used in the NCHRP Project 3-78A studies. Experimental Design and Field Methodology The general procedure used for crossing trials replicated field methodologies from previous work by research team members. To determine the effect of treatments, a pretest– posttest, within-participant design was used. In other words, participants were recruited to complete pretest crossings, and then returned for a second study (posttest) after one or more treatments were installed at the site. This experimental design allowed the team to control for confounding factors. By using the same participants for baseline and after-treatment data collection, each participant served as his or her own control, which provided more statistical power when using a relatively small number of participants. Therefore, all participants in the posttest had also participated in the pretest to allow for a direct within-participant comparison, although not all participants returned for the second round of testing. The same basic pro- cedure was used in all studies for this project. In each session, the participant made crossings back and forth on each experimental crosswalk, using a white cane and determining when to begin crossing independently. Partic- ipants were assisted with aligning to cross or maintaining alignment during crossings, as needed. Only one individual participated at a time, completing a number of crossings in each session. As in previous studies, participants were allowed to cross at their own pace and using their own judgment of when to begin crossing. Participants were accompanied at all times by a certified O&M specialist. If a participant decided to begin crossing at a time that the specialist judged to be unsafe based on vehicle positions and speeds, the O&M specialist stopped the participant, usually by grasping the participant’s arm or shoulder. This is referred to as an O&M intervention and is later used as a safety performance measure in the analysis. The same O&M specialist was used in all experimental trials performed in this research to ensure uniformity of instruc- tions and consistency in behavior, particularly as related to O&M interventions. After all participants had completed the pretest, treatments were installed. The posttest was conducted after a driver adap- tation or acclimation period, which allowed drivers to become familiar with the treatments. The posttest procedure was iden- tical to the pretest procedure, except for changes related to the treatments (for example, pushing the push button on the pedestrian hybrid beacon). Participants Participants were individuals who were blind with light- perception or less vision and who had no ability to visually detect crosswalk lines, poles, objects, or vehicles. Participants were individuals who traveled independently using a long cane or guide dog and reported that they typically crossed streets independently. This was done in an attempt to draw a sample of participants from an appropriate target population of blind pedestrians who might cross at the experimental locations outside of the studies. A local O&M specialist in each city assisted in recruitment of participants based on these criteria. Participants who were guide dog users were also proficient in long cane use. All participants used a long cane during the crossings because repetitious street crossings are confusing for guide dogs. Participants were identified by recruitment coordinators and screened via a phone interview to determine if they met the criteria for the research. Written consent was C H A P T E R 3 Methodology

obtained before any data was collected, using consent mate- rials approved by the IRB at Western Michigan University and NAS (Appendix N). The participants received an honorar- ium and were provided transportation assistance as needed. Participation was strictly voluntary, and participants were allowed to withdraw from the experiment at any time or for any reason. The honorarium and transportation assistance were provided independent of whether a participant com- pleted all trials. Sixteen to 18 participants were recruited for the pretest in each location, with approximately 12 individu- als completing both pretest and posttest at each site. Orientation After consent was obtained, participants were oriented to the roundabout or CTL by the O&M specialist. Orientation included using tactile maps; walking around the facility with the experimenter, who described features; walking across the crosswalk while guided by an experimenter; and then cross- ing independently using the experimental procedure. Partic- ipants were encouraged to ask questions about the layout of the intersection and crosswalks, the traffic movement, the pedestrian facilities available, and any other features of inter- est. During the posttest, orientation included a description and demonstration of the treatments and their operation. Procedure Participants were instructed to cross whenever they believed it was appropriate to do so, using the cues that were available (traffic in pretest, and combination of traffic and treatment- related cues in posttest). Participants crossed a specified num- ber of times at each site. The number of crossings was deter- mined through pilot testing at each location to fit within the approximately 90 min experimental timeframe and to not result in inordinate fatigue for participants. Each partici- pant made six round trips (entry, exit, exit, entry or exit, entry, entry, exit), or 24 crossings, at the Charlotte, NC, single-lane roundabout; 20 round trips or 40 crossings at the CTLs; and 12 round trips, or 48 crossings, at the Golden, CO, sites (four round trips at the single-lane roundabout and four round trips each at both the two-lane roundabout crosswalks). These sites are described further later in this chapter. A trial was defined as a crossing of one approach, which could entail one or two lanes of traffic depending on the site (e.g., single-lane or two-lane roundabout). Trials were blocked by crosswalk (for sites with multiple test approaches) to save time and to avoid confusing the participants. The starting location (e.g., entry or exit leg at a roundabout) was systematically varied to control for order effects. Posttesting was conducted at the same crosswalks as pretesting. For each trial, participants were guided to the middle of the curb ramp and were aligned to face across the crosswalk. While approaching the crossing location, participants were told which lane of the roundabout they were crossing, which direction traffic would approach from, and whether they were crossing from the island or curb. For example: “You are crossing the entry lane of Davidson from the curb, with traf- fic coming from your left [touch left shoulder]. Cross when- ever you’re ready.” Participants were reminded that the exper- imenter merely informed them when the trial began, not that it was a safe time to begin crossing. Before beginning trials, they were told that after the experimenter said “cross when- ever you’re ready,” they should identify a safe time to begin crossing and then cross the street. The experimenter stopped each participant on the opposite side of the street (or on the island) at the end of that trial. After at least one vehicle had crossed the crosswalk (or 30 s, approximately, if no vehicles), the experimenter guided the participant to the starting point for the next crossing and began another trial. Participants were allowed to take breaks as needed, and refreshments were provided. After all crossings were completed, each participant completed a short debriefing questionnaire. Participant Questionnaires After each testing session, each participant completed a debriefing questionnaire. The questionnaire was intended to ask the participant about the crossings just completed and to learn about their confidence in crossing at that location. Questions included, for example: • “How would you rate your confidence in your ability to cross here safely on a scale of 1 to 5, with 1 being not at all and 5 very confident?” • “Would you use this crossing if it was on the most direct route to and from work?” Other debriefing questions focused on crossing strategies, the perceived difference between entry and exit lanes at round- abouts, the difference of crossing from the curb or the split- ter island, and information about the treatments. Questions included, for example: • “What cues did you use to decide when to cross?” • “Did it matter whether you were crossing to or from the island?” • “Were you using the sound from the strips to help you decide?” • “Did you think the flashing yellow beacon made a differ- ence in driver behavior?” While the responses to these survey questions are subjec- tive, they add an important feature to the analysis. The results are presented in Chapter 5 along with more objective perfor- mance assessment, such as the average delay time. Appendix G 25

shows the questionnaires used during the debriefing process for different sites. Identification and Selection of Treatments This section describes the process used by the research team to collectively arrive at a set of treatments to be experi- mentally applied at selected single- and two-lane roundabouts and CTLs. The objective of the treatments was to improve access for blind pedestrians. The definition for what consti- tutes an accessible crossing and the performance evaluation framework used in the analysis is presented in Chapter 4. The treatment selection process combined information on treat- ment effectiveness available in the literature and the applied research and practical experience of those on the project team. Given the operational similarities of roundabouts and CTLs discussed in Chapter 1, comparable treatment strategies were hypothesized to enhance accessibility at both facility types. Accessibility for blind pedestrians at these types of facilities is a function of (1) traffic conditions associated with a low occurrence of naturally occurring crossable gaps, (2) blind pedestrians’ ability to detect the presence of naturally occur- ring gaps, (3) the likelihood of motorists yielding to pedes- trians, and (4) the ability of blind pedestrians to reliably detect those yield events. Given the dynamic nature of the pedestrian–vehicle interaction, the temporal efficiency is another critical aspect since gap duration and driver patience (in yielding) are limited. In addition to the above conditions for crossing, equivalent access to these types of facilities by blind and sighted pedestri- ans will also be a function of how effectively the blind pedes- trian can (1) locate the crosswalk, (2) correctly align for cross- ing, and (3) maintain alignment while crossing. Failures of facilities to support these requirements can result in (a) an increase in the total pedestrian travel time associated with crossing, (b) exposure to risk by crossing at an inappropriate (i.e., unmarked) crossing location, or (c) both an increase in crossing time and exposure to risk associated with veering during the crossing. While the focus of the experimental tri- als in this research was on the actual crossing task, these other aspects are discussed in general terms. The long list of treatments given in Chapter 2 represents all treatments that were deemed to have some potential in improving the accessibility of roundabouts and CTLs. It was beyond the scope of this project to test all of these treatments at multiple facility types. Consequently, a process had to be devised to arrive at a short list of treatments to propose for further testing. Having completed the literature review on treatment types and effectiveness and having extensive prior research and field knowledge in the engineering and accessibil- ity fields, the research team represented a diverse and qualified group for this task. Therefore, an independent internal team survey was conducted to weigh different treatment options in terms of their perceived effectiveness, cost, and applicability to the different facility types. Internal Team Survey The internal team survey was intended to reduce the long list of treatments to a short list to move forward in the field- testing stage of the project. A survey tool was developed to gather the input of all members of the research team. The survey tool and team results are provided as Appendix C. Each member of the research team evaluated each of the potential treatments in terms of (1) the extent to which each of the treatments would have an impact on the likelihood of gap detection (estimated separately for blind and sighted pedestrians), (2) the likelihood of having an impact on yield detection, and (3) the likelihood of drivers yielding to pedes- trians. These are referred to in the survey as behavioral attri- butes associated with treatments. Estimates were also provided for the extent to which each potential treatment was believed to have an impact on performance (i.e., both the delay and risk experienced by blind and sighted pedestrians), as well the effect on vehicular traffic. Estimates were also provided for the perceived applicability of treatments for implementation at single- and two-lane roundabouts and at CTLs. Treatments were divided into six categories following the discussion in Chapter 2: driver information treatments, traf- fic calming treatments, pedestrian information treatments, crosswalk geometry modification, signalization with APS, and grade separation. The survey results suggested that most driver informa- tion treatments and traffic calming treatments were judged to have no substantial impact on blind pedestrians’ ability to detect gaps and/or yielding vehicles. However, these treat- ments were generally believed to be beneficial to sighted pedes- trians and were believed to be applicable to single-lane round- about and CTL crossings, but to a lesser extent to two-lane roundabouts. While they weren’t believed to affect detection of crossing opportunities, their perceived impact on increasing frequency of crossing opportunities (more yields) was deemed beneficial. At a relatively low cost and perceived low impact to vehicular operations, one treatment in each of these categories was considered for further testing: a pedestrian-actuated flash- ing beacon and a raised crosswalk. For pedestrian information treatments, members of the research team hypothesized mixed effects on gap and yield detection, although generally higher effectiveness than treat- ments in the previous category. Estimated effectiveness for blind pedestrians was moderate for surface treatments intended to generate additional acoustic cues and higher for more 26

automated methods of gap and yield detection. It should be noted that, although considered in the internal survey, auto- mated gap and yield detection capabilities of the type that have undergone preliminary field evaluation in the NIH/NEI proj- ect (NIH 2010) are beyond the scope of what are consid- ered off-the-shelf treatments available for evaluation by NCHRP Project 3-78A. Such treatments, while promising, are too experimental at this point to be considered as avail- able treatment options for state DOTs. The concept of a sur- face sound-strip treatment showed potential in prior research (Inman et al. 2005) and was considered for further study at single-lane roundabouts and CTL crosswalks. Signalization treatment options generally showed mod- erate to high potential for reducing delay and risk for blind pedestrians. As might have been expected, vehicle delay was judged to be negatively affected for signalized treatments. Of special interest was the PHB, an innovative signalization strategy that reduces vehicle delay through unconventional phasing while still providing a red vehicle display and associ- ated walk phase for pedestrians. Initial estimates of potential treatment effectiveness were also obtained for crosswalk geometry modifications and showed moderate potential in the ability of blind pedestrians to detect gaps and yields. Lastly, grade-separation treatments were rated as having the highest potential impact on behavioral param- eters and performance measures. However, the applicability of modified crosswalk geometry and grade separations as treat- ment options to be considered under this research effort was low due to the high costs and permanent nature associated with the treatments. Treatments Recommended for Installation The results of the team treatment survey were used to develop recommended treatments or combinations of treat- ments for installation at the test locations under consideration of the project scope and budget. Each installation required cooperation from a municipality, was associated with instal- lation cost, and required substantial team resources for data- collection planning, execution, and analysis. Treatment rec- ommendations were chosen based on the perceived level of effectiveness in producing a measurable impact on the occur- rence of yielding events or utilized gaps. Recommendations also considered the level of cost associated with installing the treatment and the expected benefit to travelers. Due to the relatively low number of available sites willing to install treat- ments, as well as budget constraints that limited the number of sites for which data could be collected, combinations of treatments were considered a plus. Based on these factors, the team recommended installing two treatment installations. The team recommended testing these treatment combina- tions concurrently at two CTLs of the same intersection in order to make most efficient use of project resources and to allow for a direct comparison using the same participants. • Channelized turn lane: The team recommended instal- lation of sound strips and lane delineators to provide addi- tional auditory information to blind pedestrians about traffic patterns in the CTL and to distinguish that traffic from adjacent through movements. The team further recommended testing the sound strip and lane delineator treatment in isolation and supplemented with a pedestrian- actuated flashing beacon. The latter was intended to test the effect of enhancing pedestrian visibility to the driver and the effect of the treatment on increased yielding behavior. This proposed approach provides a measurable effect of three possible CTL treatments: the effect of sound strips, the effect of sound strips in combination with a pedestrian-actuated flashing beacon, and the flashing beacon alone (derived from the net difference between the two installations). The use of lane delineators was a fixed addition, assuming that they would be necessary to ensure proper functionality of the sound strips. • Single-lane roundabout: The team recommended the same treatments used for the CTL, but for obvious reasons with- out the lane delineators. Proposed treatments were sound strips in isolation and sound strips supplemented with a pedestrian-actuated flashing beacon. The same assumptions and control factors applied since two different approaches of the roundabout were recommended with similar geome- tries and traffic volumes. • Two-lane roundabout: The team recommended installing two different treatments at two approaches of the same two- lane roundabout: A raised crosswalk (RCW) intended to slow traffic and increase yielding, and a PHB intended to stop traffic at a red signal display and to provide additional audible information to the blind pedestrian. The team con- cluded that installing these two significant treatments was reasonable given the perceived challenges at two-lane cross- ings and the draft PROWAG language. Additional details on the specific treatment installations are provided after the discussion of site selection and along with the overview of the test sites. Site Selection The research team evaluated a list of potential study sites and selected those that were deemed suitable for further field inves- tigation of the proposed treatments. Criteria for site selection included: • Feasibility of implementing one or more of the desired treat- ments at a given site within project schedule and budget; 27

• Level of federal, state, and local support and cost sharing in implementing the proposed treatments; • Sufficient vehicle and pedestrian demand to enable a mean- ingful evaluation of the treatment’s impact on the system performance; • Availability of adequate numbers of potential research par- ticipants who are blind and are in reasonable proximity to the sites identified for data collection tasks; • Proximity of the sites to the data collection team and to one another to maximize use of limited budget resources; and • Adequate representation of the various geometric condi- tions to be considered. Identification of Treatment Site Alternatives The research team used three methods to identify candi- date sites. First, a solicitation for sites was posted on e-mail list serves. Appendix D contains aerial and site photographs of the sites identified in the responses to the request. In the second method of site identification, agencies and practicing engineers active in the planning, design, and construction of round- abouts were contacted. While roundabouts were not the only focus of the study, it was hypothesized that CTL sites would be readily identified in proximity to (more rare) roundabout sites. Agencies and companies contacted included Maryland State Highway Administration; Kansas DOT; Washington State DOT; New York State DOT; North Carolina DOT; California DOT; Florida DOT; City of Clearwater, FL; City of Kennewick, WA; City of Modesto, CA; City of Bend, OR; City of Portland, OR; City of Tucson, AZ; City of Golden, CO; Town of Vail, CO; MTJ Engineering; Roundabouts, USA; Ourston Round- about Engineering; and Alternate Street Design. The team had follow-up meetings and/or telephone conversations with the Maryland State Highway Administration, Washington State Department of Transportation, New York State Depart- ment of Transportation, and Ourston Roundabout Engineer- ing. The third method for site selection consisted of reviewing sites studied under NCHRP Project 3-72, “Lane Widths, Chan- nelized Right Turns, and Right-Turn Deceleration Lanes in Urban and Suburban Areas” and NCHRP Report 572: Round- abouts in the United States. Selection of Treatment Sites Channelized Turn Lanes Providence Road at Pineville-Matthews Road – Charlotte, NC. The selection of a channelized turn lane facility was tied to the location of roundabout sites since CTL locations are common to most jurisdictions. The research team considered four candidate sites, two of which were located in Charlotte, NC, and the other two in Towson, MD. Given that no sites were chosen for roundabout treatments in Towson, the deci- sion was made to focus on turn-lane treatments in the Char- lotte region. Ultimately, the intersection of Providence Road at Pineville-Matthews Road was selected for treatment evaluation. The main intersection is signal controlled with an eight- phase actuated-coordinated signalization scheme and pro- tected dual-left turns on all four approaches. The through movements on Providence Road and Pineville-Matthews Road have two and three lanes per direction, respectively. The site has CTLs on all four approaches, each yield controlled at the downstream merge and with deceleration lanes present. No acceleration lanes were present on any of the approaches. The posted speed limit on all approaches is 45 mph. Traffic volumes were high during the peak hour periods. Land uses near the intersection include office buildings, retail strip malls, and residential. Two of the four CTLs were selected for this study. Both served the right-turn movements from Providence Road onto Pineville-Matthews Road. Figure 5 shows the loca- tions of the studied crosswalks at the southeast and northwest corner of the intersection. Traffic volumes (in 2005) at the tested crosswalk in the southeast corner show 24-hour turning flows of 3,200 vehicles per day (vpd) and adjacent through traffic of 12,000 vpd. The downstream conflicting movement is composed of the oppos- ing through traffic (10,600 vpd) and the opposing left turn traffic (3,000 vpd). Volumes at the northwest corner crosswalk are generally higher at 6,000 vpd in the CTL, and adjacent through traffic of 18,400 vpd. The downstream conflicting through flow was 22,400 vpd, and the opposing left turn was 2,700 vpd. Low pedestrian activity was observed at the site, 28 This figure shows an aerial photograph of the CTL treatment site at the intersection of Providence Road and Pineville-Matthews Road in Charlotte, NC. The studied crosswalks are highlighted and located in the southeast and northwest corners of the intersection. Ph ot o by G oo gl e Figure 5. Aerial view of CTL site.

on the order of 20 pedestrians per day, primarily during the midday off-peak period. However, all main approaches of the intersection had marked crosswalks and pedestrian sig- nals. Drivers were therefore believed to expect (occasional) pedestrian activity. In the pretest, crosswalks at the CTLs were unsignalized and outfitted with standard marking and pedestrian signage. Figure 6 shows street-level photographs of each of the studied approaches and the installed treatment. The treat- ments were sound strips that were intended to increase the awareness of pedestrians of approaching vehicles at the northwest corner (SS-ONLY) and sound strips in combi- nation with a pedestrian-actuated flashing beacon that was intended to increase driver yielding behavior at the south- east corner (SS+FB). Both approaches further had lane delin- eators installed to prevent late merges into the turn lane. This was necessary to ensure that the sound strips in fact picked up all right-turning vehicles. The sound-strip treatment tested was an off-the-shelf, tem- porary, and self-adhesive rumble-strip material. The sound strips were white in color and had a vertical elevation of 0.25 in. The sound strips were evenly spaced over approximately 50 ft, and a total of 5 strips were installed at each turn lane. The first sound strip was installed at a distance of 10 ft upstream of the crosswalk. At a constant speed of 50 ft/s (34.1mph), this cor- responds to a temporal spacing of the “clack” sounds of 1 s. A faster speed results in a shorter rate of clack sounds. A vehi- cle that slows down while traversing the sequence of sound strips will generate an audible pattern where the rate of clack sounds decreases as the vehicle slows (longer time between clacks). The sound strips therefore aim to accomplish two goals: (1) an audible distinction between right-turning (con- flicting) and adjacent through traffic, and (2) an audible pat- tern that helps identify slowing or yielding vehicles. The flashing beacons were installed on both sides of the street at the SS+FB corner and had a pole-mounted dual- head signal display. The beacons rested in “Dark” mode and flashed in wig-wag patterns (a 1-s flash frequency) for 20 s after a pedestrian pushed the button. The flashing beacon was further outfitted with a push-button integrated locator tone and an audible speech message, “cross with caution; cars may not stop,” that was repeated concurrent with the flash- ing display. Single-Lane Roundabouts The research team considered four candidate sites: the Pullen-Stinson roundabout in Raleigh, NC, the two single-lane approaches to the Towson, MD, roundabout, a single-lane roundabout located in Voorheesville, NY, and a single-lane roundabout in Charlotte, NC. The Pullen- Stinson roundabout has been studied by research team mem- bers under other research grants. Based on the fact that other treatments have been used and evaluated at this site, the high proximity of the roundabout to the campus of North Carolina State University where pedestrian activity is extremely high, and the fact that other research trials as part of the NIH proj- ect were planned with blind participants during the project timeframe, the decision was made to eliminate this round- about for study of further treatments. However, the analysis ended up including some prior data collected at the site. The Towson roundabout was intriguing; however, previous treat- ments installed at the roundabout were removed because of their negative effect on vehicular movements, which may be 29 This figure shows two photographs of the treatments installed at the CTL study site. The left photograph shows a view from the crosswalk splitter island down the deceleration lane with the sound-strip and lane-delineator treatments. The right photograph shows the view of a driver approaching the crosswalk with the sound strips, lane delineators, and flashing beacons installed. Ph ot o by B as tia n Sc hr oe de r Ph ot o by B as tia n Sc hr oe de r Figure 6. Photo of sound strip and flashing beacon treatments.

problematic from the standpoint of installing new treatments for evaluation. The Voorheesville roundabout was consid- ered problematic in terms of proximity to the research team. Ninth Street at Davidson Street – Charlotte, NC. The site chosen for single-lane roundabout data collection was located at the intersection of 9th and Davidson streets in Charlotte. The urban roundabout site was close to the research team members, and the team was in contact with city personnel willing and able to install the necessary treatments. The site is located in a primarily residential neighborhood just northeast of uptown Charlotte, the city center. The inscribed diameter of the roundabout is approximately 140 ft (42.7 m), with approach speed limits of 25 mph (40 km/h). The crossing distance is approximately 16 ft (4.9 m) for each leg of the crossing. The major approach had average annual daily traffic (AADT) of 9,900 vpd, which is well within the capacity limits of a single-lane roundabout. Peak hour traffic counts showed total roundabout entering volumes of 830, 530, and 780 vehi- cles per hour (vph) for a.m., lunch, and p.m. peak periods, respectively. About 80% to 85% of these flows were on the major approaches on Davidson Street. Volumes on 9th Street were low at or below 100 vph. The team therefore concluded to focus the study on the two approaches along Davidson Street. Figure 7 shows the location of the two studied cross- walks, as well as a street level view of the southeast entry approach. No posttest treatment evaluation was performed at this single-lane roundabout. The recommendation and direction not to proceed with treatment installation was made after the pretest was completed because it was decided that project resources were better spent elsewhere. Pullen Road at Stinson Drive – Raleigh, NC. Although not officially part of the NCHRP Project 3-78A data collection effort, further analysis of an alternate single-lane roundabout was conducted to test the hypothesis that pedestrian cross- ing performance was affected by higher traffic volumes. The roundabout at the intersection of Pullen Road and Stinson Drive located in Raleigh was one studied previously by the research team. For this research, a supplemental evaluation was performed using pre-existing video of crossing trials prior to any treatment installation and using the same experimen- tal protocol used in NCHRP Project 3-78A. The roundabout is located in close proximity to the cam- pus of North Carolina State University and has a high volume of pedestrian traffic from students walking to and from class. The roundabout has an inscribed diameter of 88 ft (26.8 m) and approach speeds of 25 mph (40 km/h). The crossing dis- tance is approximately 13 ft (4.0 m). The AADT at this site is higher than at the Charlotte site at 15,000 vpd along the major approach, Pullen Road. Peak hour traffic counts showed total roundabout entering volumes on the order of 1,300 vph for a.m. and lunch peaks, and 1,500 vph in the p.m. peak. About 90% of these flows were on the major approaches on Pullen Road. Figure 8 shows the location of the studied crosswalk, as well as a view of the southern approach from an adjacent building rooftop. Since this site was originally studied under 30 This figure shows two photographs of the studied single-lane roundabout at the intersection of Davidson Street and 9th Street in Charlotte, NC. The left image shows an aerial photograph with the studied crosswalks highlighted on the southwest and northeast approaches of Davidson Street. The right image shows a pedestrian-level photograph of the roundabout entry leg. Ph ot o by G oo gl e Ph ot o by B as tia n Sc hr oe de r Figure 7. Aerial and pedestrian-level photographs of Davidson Street at 9th Street in Charlotte, NC – single- lane roundabout.

a different research project, no NCHRP Project 3-78A treat- ments were installed. Golden Road at Ulysses Drive – Golden, CO. One addi- tional single-lane roundabout site was studied by the research team. This site, located in Golden, CO, was intended to pro- vide a control condition for a nearby two-lane roundabout. The single-lane site was studied concurrently with the pretest and posttest treatment conditions at the two-lane site. Both roundabouts were on the same corridor (Golden Road). The purpose of this additional analysis was to test for a learning effect in the same participants returning to the same round- about without treatment installation. It further allowed a direct within-participant comparison of the single-lane and two-lane roundabout crossing ability of the same participants. The roundabout has a central island diameter of 100 ft (30.5 m), including a truck apron of 10 ft (3.1 m). The lanes at the studied crosswalk are 20 ft wide, partly to accommodate nearby roadside bus stops. The crosswalk is located approxi- mately 60 ft (18.2 m) from the circulating lane measured at the exit side, and approximately 50 ft (15.2 m) from the round- about yield line at the entry. The two-stage crossing is divided by a 6-ft, raised splitter island, but the crossing itself is at pavement elevation. No pedestrian-detectable warning sur- faces were installed on the splitter island, so the study partic- ipants were instructed by the O&M specialist when they com- pleted the first half of the crossing. No detectable warning surfaces were installed on the crosswalks at this roundabout. The crosswalk was marked and outfitted with standard pedes- trian signage. Traffic volumes on Golden Road indicated an AADT of 15,000 vpd. Figure 9 shows the location of the studied crosswalk as well as a street view of the eastern approach. Two-Lane Roundabout The research team identified three candidate two-lane roundabout sites, in Gatineau, Quebec, Canada; Towson, MD; and Golden, CO. The Gatineau site already had a sig- nal installed and would have therefore prevented the team from performing a pretest and posttest comparison. The Towson site has a very good mix of pedestrian and vehicular traffic, with heaviest pedestrian traffic during the middle of the day. This site also offers the advantage of providing a mix of single-lane and two-lane entries/exits. However, the geometry is very different from most single- or two-lane roundabouts, which was an important consideration in rejecting the site. Further, the local agency was only supportive of treatment installations short of signalization. Golden Road at Johnson Road – Golden, CO. The two- lane roundabout site finally selected for installation of treat- ments was located in Golden. The site provided two good approaches for treatment installations, and the team enjoyed significant local support for the research effort and treatment installation. Figure 10 shows an aerial view of the site and the locations of the studied crosswalks. The PHB treatment was 31 This figure shows two photographs of the studied single-lane roundabout at the intersection of Pullen Road and Stinson Drive in Raleigh, NC. The left image shows an aerial photograph with the studied crosswalk highlighted on the southwest approach of Pullen Road. The right image shows a photograph of the studied crosswalk taken from a nearby building. Ph ot o by G oo gl e Ph ot o by B as tia n Sc hr oe de r Figure 8. Aerial and pedestrian-level photographs of Pullen Road at Stinson Road in Raleigh, NC – single- lane roundabout.

installed on the northwest corner, and the raised crosswalk was installed on the southeast corner. The roundabout has a central island diameter of 90 ft (27.4 m), including a truck apron of 10 ft (3.1 m). The lanes at the studied crosswalk are 11 ft wide (3.4 m), resulting in a total crossing distance of 22 ft (6.8 m). The crosswalk is located approximately 30 ft (9.1 m) from the circulating lane at the northwest corner, and approximately 20 ft (6.1 m) at the southeast corner. The two-stage crossings are divided by 15-ft raised and landscaped splitter islands. Detectable warn- ing surfaces were installed on the splitter islands and on the curbs at both crossings. The crosswalks were marked and out- fitted with standard pedestrian signage. Traffic volumes on Golden Road indicated an AADT of 15,000 vpd. Peak hour total entering flows at the roundabout were 1,900 vph in the a.m. peak and 2,100 vph in the p.m. peak. Traffic volumes were highest on the two Golden Road approaches (approxi- mately 70% to 75% of entering traffic). Figure 11 shows a photograph of the RCW installed at the southeast approach of the roundabout on Golden Road. The RCW was installed at a vertical elevation of 3 in. from pavement surface and a transitional slope of 1:15. This means that the transition to the 3-in. elevation occurred over approx- imately 45 in., or just short of 4 ft. This installation resulted 32 This figure shows two photographs of the studied single-lane roundabout at the intersection of Golden Road and Ulysses Street in Golden, CO. The left image shows an aerial photograph with the studied crosswalks highlighted on the east approaches of Golden Road. The right image shows a pedestrian-level photograph of the roundabout entry leg. Ph ot o by B as tia n Sc hr oe de r Figure 9. Aerial and pedestrian-level photographs of Golden Road at Ulysses Street in Golden, CO – single- lane roundabout. This figure shows an aerial photograph of the two-lane roundabout treatment site at the intersection of Golden Road and Johnson Road in Golden, CO. The studied crosswalks are highlighted and located in the southeast and northwest approaches of Golden Road. Ph ot o by G oo gl e Figure 10. Aerial view of Golden Road at Johnson Road in Golden, CO – two-lane roundabout. This figure shows a pedestrian-view photograph of the RCW treatment installed at the southeastern approach of the Golden Road two-lane roundabout. Ph ot o by Ja ne t B ar lo w Figure 11. Raised crosswalk treatment – Golden, CO.

in a less severe impact on vehicle traffic compared to other installations visited by the research team. Other installations of RCWs commonly feature vertical elevations up to 5 in. and/or transitional slopes as steep as 1:10. Clearly, the speed reduction effect is greater with higher vertical elevation and a steeper slope. The RCW was intended as a temporary installation. It was therefore simply installed on top of existing pavement and used the existing curb and gutter system for drainage. This resulted in the RCW being sloped downward on each side. A permanent installation would likely be installed flush with the sidewalk and incorporate drainage into the RCW design via piping. As a result of this installation, pedestrians first walked down the curb ramp from sidewalk level and then back up onto the crosswalk. Numerous participants noted that this unexpected down/up transition was uncomfort- able to walk on. The RCW installation therefore satisfied the intended treatment effect (on traffic), but the pedestrian aspects of the design should be modified for a permanent installation. Pavement markings were as shown in Figure 11 and emphasized the upslope of the RCW. Figure 12 shows a photograph of the PHB installation at the northwest approach of the roundabout on Golden Road. The PHB was installed on roadside poles following MUTCD requirements for vertical height and additional signage. The PHB installation featured a total of four pole-mounted devices, with two each at the entry and exit legs. All four devices were outfitted with push-button integrated APS devices. The signal timing was such that the two sides (entry and exit) oper- ated fully independently of each other. The vehicle display rested in “Dark” mode; the pedestrian display rested in “Don’t Walk.” The phasing sequence of the PHB after pedestrian acti- vation and the times for each phase are shown in Figure 13. After pedestrian push-button activation, the PHB entered a “Flashing Yellow” phase for vehicles (3 s), followed by a “Steady Yellow” phase (3 s). The pedestrian signal during both of these phases remained as “Don’t Walk.” Next, the vehicle display changed to a “Steady Red” phase, while the pedestrian displays showed “Walk” (4 s). The PHB then showed a “Flash- ing Red” for vehicles concurrent with the pedestrian “Flash- ing Don’t Walk” (10 s). Finally, the vehicle display returned to “Dark” and the pedestrian display to “Don’t Walk.” The PHB was timed to rest in this phase for a minimum of 15 s before cycling back through the phases. Expected pedestrian behavior at the PHB was as it would be at a standard pedestrian signal. Pedestrians were expected to push the button (since they observed a “Don’t Walk” phase) and cross during the “Walk” interval. Drivers were expected to be stopped during the “Red” phase. The “Flashing Yellow” was intended to alert the driver of the impending signal phase change. The “Solid Red” offered drivers dilemma-zone pro- tection in case they were too fast and too close to the crosswalk to come to a stop. Drivers were allowed to proceed with cau- tion after having stopped during the “Flashing Red” phase, provided the pedestrian had cleared the crosswalk. This feature reduced the required stop time for drivers from 14 s to 4 s. Appendix F gives additional details on the PHB installation, including timing parameters and signal plans. 33 This figure shows a pedestrian-view photograph of the PHB treatment installed at the northwestern approach of the Golden Road two-lane roundabout. Ph ot o by Le e R od eg er ts Figure 12. PHB treatment – Golden, CO. This figure shows a graphic of the PHB phasing scheme. The sequence of vehicle/pedestrian displays is “Flashing Yellow/Don't Walk,” “Steady Yellow/Don't Walk,” “Steady Red/Walk,” “Flashing Red/Flashing Don't Walk,” and “Dark/Don’t Walk.” Phase durations for these phases in sequence are 3 s, 3 s, 4 s, 10 s, and 15 s, where the last represents a minimum time for the “Dark/Don’t Walk” and the PHB rests in this phase if no further pedestrian calls are placed. 10 seconds 3 seconds 3 seconds 4 seconds 15 seconds minimum signal rests in this phase Figure 13. PHB phases and signal timing.