National Academies Press: OpenBook
« Previous: References
Page 99
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 99
Page 100
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 100
Page 101
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 101
Page 102
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 102
Page 103
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 103
Page 104
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 104
Page 105
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 105
Page 106
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 106
Page 107
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 107
Page 108
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 108
Page 109
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 109
Page 110
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 110
Page 111
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 111
Page 112
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 112
Page 113
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 113
Page 114
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 114
Page 115
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 115
Page 116
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 116
Page 117
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 117
Page 118
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 118
Page 119
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 119
Page 120
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 120
Page 121
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 121
Page 122
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 122
Page 123
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 123
Page 124
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 124
Page 125
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 125
Page 126
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 126
Page 127
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 127
Page 128
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 128
Page 129
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 129
Page 130
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 130
Page 131
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 131
Page 132
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 132
Page 133
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 133
Page 134
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 134
Page 135
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 135
Page 136
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 136
Page 137
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 137
Page 138
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 138
Page 139
Suggested Citation:"Appendix A - Detailed Results." National Academies of Sciences, Engineering, and Medicine. 2011. Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities. Washington, DC: The National Academies Press. doi: 10.17226/14473.
×
Page 139

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.

99 This appendix contains five parts detailing analysis results for the different studies performed under NCHRP Project 3-78A: Part 1: Detailed Channelized Turn Lane Results Part 2: Detailed Single-Lane Roundabout: Golden, CO Part 3: Detailed Single-Lane Roundabout: North Carolina Part 4: Detailed Two-Lane Roundabout: Golden, CO – RCW Part 5: Detailed Two-Lane Roundabout: Golden, CO – PHB A P P E N D I X A Detailed Results

C O N T E N T S 101 Part 1 Detailed Channelized Turn Lane Results 101 Introduction 101 CTL Analysis Results 101 Site Description 101 Crossing Statistics for CTL Site and Treatments 108 Channelized Turn Lane Results Summary 110 Part 2 Detailed Single-Lane Roundabout: Golden, CO 110 Introduction 110 Golden Single-Lane Analysis Results 110 Site Description 110 Crossing Statistics for Crosswalk 115 Golden, CO, Single-Lane Roundabout Summary 116 Part 3 Detailed Single-Lane Roundabout: North Carolina 116 Introduction 116 North Carolina Single-Lane Analysis Results 116 Site Description 118 Crossing Statistics 120 Discussion 121 Part 4 Detailed Two-Lane Roundabout: Golden, CO – RCW 121 Introduction 121 Raised Crosswalk Evaluation 121 Raised Crosswalk Treatment Overview 121 Pretest Pedestrian Behavior at the RCW 123 Posttest Pedestrian Behavior at the RCW 124 Performance Statistics for RCW 127 RCW Summary 128 Part 5 Detailed Two-Lane Roundabout: Golden, CO – PHB 128 Introduction 128 Pedestrian Hybrid Beacon Evaluation 128 Pedestrian Hybrid Beacon Treatment Overview 128 Pretest Pedestrian Behavior at the PHB Crosswalk 130 Posttest Blind Pedestrian Behavior at the PHB Crosswalk 133 Performance Statistics at the PHB Crosswalk 136 Driver Behavior at the PHB 138 PHB Crosswalk Summary 100

101 Introduction This section describes detailed analysis results of data collected at the channelized right turn lane in Charlotte, NC, at the intersection of Providence Road and Pineville-Matthews Road (Exhibit 1). The focus is on pedestrian-related measures, including the availability and utilization of yield and gaps, as well as pedestrian delay and O&M interventions. Two turn lane crosswalks at the CTL sites were studied in this project: the crosswalks in the southeast (SE) and northwest (NW) corner. Both turn lanes served the right-turn movements from Providence Road onto Pineville-Matthews Road. Similar to other data collection sites, the CTL was studied in a pre and post study design with treatment installation. The treatments were (1) sound strips that were intended to increase the aware- ness of pedestrians of approaching vehicles at the NW corner and (2) sound strips in combination with a pedestrian-actuated flashing beacon that was intended to increase driver yielding behavior at the SE corner. In the following discussion, the cross- walks will be identified by treatments installed as SS-ONLY and SS+FB, respectively. Both turn lanes were further supplemented with lane delineators that were intended to prevent late merges into the turn lane. All treatments, including the lane delin- eators, were installed between pre and post studies. The pre study was completed in May 2008; a total of 16 blind travelers participated. Fourteen of the original 16 participants returned for the post experiment in November 2008. The treat- ments were installed in early October 2008, allowing six weeks for driver adaptation. CTL Analysis Results Site Description The CTL site is located at the intersection of two major arterial streets in southeast Charlotte. Providence Road has a four-lane cross-section, and Pineville-Matthews Road has a six-lane cross-section in the vicinity of the intersection. All four left-turn movements have dual left-turn lanes and are thus controlled by protected signal phases. In the PRE condition, all right-turn lanes were free flowing (no signal) and were con- trolled only by a downstream yield sign. As a result, vehicle speeds through the turn lane were relatively fast. Vehicle movements were entirely uninhibited and free flowing during signal phases where the downstream lanes were clear, which were the signal phases serving the adjacent through movements on Providence Road and the opposing left turns on Pineville- Matthews Road. The treatments tested at the CTL site were intended to provide a relatively low-cost solution to make the site accessible to and usable by pedestrians who are blind. The high-end treat- ment would have been a pedestrian signal, which was not tested because its effects on crossing performance are predictable. The treatment tested was sound strips that enhance the auditory sound patterns of approaching vehicles. The hypothesis was that sound strips enhance the rate of opportunity utilization of pedestrians and therefore reduce delay. Presumably, sound strips would also help reduce the rate of O&M interventions if the subjects have increased awareness of the presence of a vehicle in the turn lane. The sound strips were tested in isolation at the SS-ONLY corner; Exhibit 2 shows a photo of the installation along with the mentioned lane delineators. At the SS+FB corner, the sound strip and lane delineator treatment was supplemented with a pedestrian-actuated flash- ing beacon (Exhibit 3). The FB was intended to increase driver awareness of the pedestrian’s intent to cross and thus increase the rate of yielding. When activated, the FB would transmit an audible speech message saying “Flashing Beacon Is On” for the duration of the flashing mode. The flashing mode would terminate after 20 s. Crossing Statistics for CTL Site and Treatments The analysis of crossing performance focuses on aspects of pedestrian–vehicle interaction following the NCHRP Project 3-78A analysis framework. The first analysis component describes the availability and utilization of yields in both the P A R T 1 Detailed Channelized Turn Lane Results

pre and post treatment conditions. Two yield measures were used in the analysis: • P(Y_ENC): The probability of encountering a yield event, defined as the number of yields divided by the total of all events encountered by the pedestrian until he/she completes the crossing. • P(GO|Y): The probability of yield utilization, defined by the number of crossings in a yield divided by total number of yields encountered by the pedestrian. The P(Y_ENC) measure is somewhat different from the traditionally used probability of yielding since it is calculated on the basis of all pedestrian–vehicle events and not just poten- tial yielders. Chapter 3 provides additional discussion on these and other performance measures, including examples on the difference between the yielding measures. Treatments at the SS-ONLY crosswalks were sound strips and lane delineators only; the SS+FB corner was further sup- plemented with a pedestrian-actuated flashing beacon. The fig- ures shown represent the mean results considering all subjects. Each subject completed ten crossing trials at the roundabout, with each trial consisting of two lane crossings (e.g., curb to splitter island and island to curb). For example, a subject in the pre condition would have crossed each crosswalk 20 times (twice in each of 10 trials) and would have performed a total of 40 crossings at the site. The average performance for each crosswalk in the pre condition was then calculated from the mean of these 20 crossing for all 16 subjects. In the post con- ditions, 14 subjects participated in the experiment. In total, 30 subjects were included in the study (16 pre, 14 post) and each performed 40 lane crossings, resulting in a theoretical total of 1,200 crossing attempts at this site. However, several subjects appeared to struggle with too many crossing attempts, and the number of trials per crosswalk was therefore capped at six for some participants. Overall, 993 crossings were com- pleted by the participants. Exhibit 4 shows the statistics for the studied crosswalks pre and post treatment installation. Exhibit 4 shows that the probability of encountering a yield, P(Y_ENC), was not significantly different at the two studied crosswalks in the pre condition (p = 0.2728).The installation of the sound strips and lane delineators at the SS-ONLY corner did not result in a notable change in yield encounters. The added installation of the flashing beacon increased the likelihood of encountering a yield from 15.2% to 22.0%, which is small but significant at p = 0.0363. From a driver perspective 102 Ph ot o by G oo gl e Exhibit 1. Aerial view of CTL site. Ph ot o by B as tia n Sc hr oe de r Exhibit 2. Sound strip installation at CTL. Ph ot o by B as tia n Sc hr oe de r Exhibit 3. FB installation at CTL.

[P(Yield), which is not shown in Exhibit 4], the rate of yield- ing increased from 24.1% to 43.1% (p = 0.0123). This means that with the installation of the FB, 43% of drivers stopped at the beacon, but these events still only represented 22.0% of the encountered vehicle events. The remaining events are in the form of gaps. The exhibit further shows the rates of yield utilization, P(GO|Y), defined as the rate of yields that resulted in a pedes- trian crossing the roadway. The yield utilization rates at the SS-ONLY corner actually appeared to decrease from 50.8% to 40.5%, although that change is not statistically significant (p = 0.2878) due to a high standard deviation. Similarly, the appar- ent increase in P(GO|Y) at the SS+FB corner from 53.1% to 64.4% is not significant at the given sample size (p = 0.2769). The high standard deviations in the yield utilization measure suggest great inter-subject variability. In the range of P(GO|Y), it is evident that some pedestrians had perfect yield utilization, while others utilized only 8% to 10% of yield opportunities. A fraction of yields further fell into the “forced yield” cat- egory, which is defined as the pedestrian stepping out into the roadway before the vehicle initiated the yielding process. The degree of risk associated with these events depends on the rel- ative position and speed of the vehicle at the time of crossing initiation. Forced yield events should therefore not necessar- ily be interpreted as poor or risky decisions. In the pre condi- tion, 11.3% and 11.5% of yields were forced at the SS-ONLY and SS+FB crossings, respectively. In the post condition, the corresponding forced yield percentages were reduced to 6.3% and 5.9%. This reduction in the percentage of forced yields was not statistically significant due to high standard devia- tions (p = 0.3520 and p = 0.1902 for SS-ONLY and SS+FB, respectively). The analysis next considered the availability and utilization of crossable gaps. For the purpose of this analysis, a crossable gap was defined as a gap greater than 6.5 s, which was sufficient to cross the 16-ft crosswalk at a walking speed of 3.5 ft/s while allowing for a 2-s buffer. These 2 seconds allowed for some pedestrian reaction time before initiating the crossing, as well as a safety buffer between a completed crossing and the next vehicle arrival. Similar to the yield statistics, two gap-related parameters are defined: • P(CG_ENC): The probability of encountering a CG event (gap greater than 6.5 s), defined as the number of crossable gaps divided by the total of all events encountered by the pedestrian. • P(GO|CG): The probability of crossable gap utilization, defined by the number of crossings in a CG divided by total number of CGs encountered by the pedestrian. Exhibit 5 shows the statistics for the studied crosswalk. The results in Exhibit 5 show a slightly higher P(CG_ENC) at the SS+FB crosswalk, which is significant at p = 0.0554. With the installation of the treatments, the rate of crossable gap encounter increases for both the SS-ONLY and SS+FB cross- walks, but neither increase is significant given the high standard deviations across subjects (p = 0.1666 and 0.4440, respectively). The rates of gap utilization are again comparable between SS+FB and SS-ONLY crosswalks in the pre condition. While the SS-ONLY treatments did not significantly affect gap utiliza- tion (p = 0.4238), the added installation of the flashing bea- con increased P(GO|CG) from 63.2% to 89.3% at the SS+FB crosswalk (p = 0.0011). The effect may be attributable to an increased level of confidence resulting from the speech message emitted from the beacon. 103 a) P(Y_ENC) Pre Avg. Min. Max. Std. Dev. SS-ONLY (n = 16) 18.4% 4.2% 37.5% 8.3% SS+FB (n = 16) 15.2% 6.0% 36.4% 7.9% Post SS-ONLY (n = 14) 18.6% 10.0% 75.0% 20.6% SS+FB (n = 14) 22.0% 0.0% 35.7% 8.9% b) P(GO|Y) Pre Avg. Min. Max. Std. Dev. SS-ONLY (n = 16) 50.8% 0.0% 100.0% 31.0% SS+FB (n = 16) 53.1% 8.0% 100.0% 28.5% Post SS-ONLY (n = 14) 40.5% 10.0% 75.0% 20.6% SS+FB (n = 14) 64.6% 20.0% 100.0% 28.2% Exhibit 4. Yield availability and utilization statistics for CTL crosswalks.

The combined effect of gap and yield availability and uti- lization is reflected in the delay experienced by pedestrians. Delay statistics in Exhibit 6 are provided for two delay measures: • Observed Delay per Leg (s): The average pedestrian delay in seconds, defined as the time difference between when the trial started and when the pedestrian initiated the crossing. • Delay>Min (s): The delay beyond the first opportunity (Delay>Min), defined as the time difference between first yield or crossable gap encountered by the pedestrian and the actual crossing initiation. Statistics for all measures are for crossing one lane of channelized turn lane. Exhibit 6 shows similar delays and Delay>Min at both turn lanes in the pre condition (p = 0.6972). As was the case for the roundabout sites, the observed ranges and standard deviations of the delay estimates are large, suggesting great variability across subjects. The highest average delay in the pre study was 80.6 s for one subject, while another one had an average delay of only 3.7 s across all trials. With installation of the treatments, the SS-ONLY delay dropped from 26.2 s to 18.5 s, which is not statistically signifi- cant at p = 0.1898. However, it is evident that both the range of observed delays and the standard deviation of the estimate showed corresponding reductions, suggesting at least some impact from the sound strip installation. The SS+FB corner saw a higher delay reduction, from 23.4 s to 12.2 s, which is significant at p = 0.0453. Again, both the range and standard deviation of the delay estimate show a reduction, suggesting more consistent behavior across subjects after treatment installation. The single highest delays for any subject in a trial were 119.0 and 113.1 s in the pre and post conditions, respectively, excluding events that were capped at the 2-min time-out limit. Overall, at the SS-ONLY crosswalk the 2-min limit was reached 19 times in the pre and 16 times in the post study (with two fewer subjects participating). At the SS+FB crosswalk the time-out was reached 10 and 2 times in the pre and post, respectively. So despite average delay improvements, isolated trials still performed very poorly after treatment installation. The results for Delay>Min also show corresponding trends. There was no significant difference between SS+FB and SS-ONLY in the pre condition (p = 0.9089). A small but statistically insignificant drop was evident for the SS-ONLY corner (15.6 to 11.7 s, p = 0.4224), while the SS+FB corner saw a significant reduction from 14.9 to 4.9 s (p = 0.0342). The difference between delay and Delay>Min suggests that quite a few participants missed crossing opportunities, especially 104 a) P(CG_ENC) Pre Avg. Min. Max. Std. Dev. SS-ONLY (n = 16) 34.9% 16.9% 64.7% 11.3% SS+FB (n = 16) 44.7% 27.9% 84.6% 16.1% Post SS-ONLY (n = 14) 41.2% 24.7% 62.1% 12.8% SS+FB (n = 14) 49.2% 29.1% 75.0% 15.6% b) P(GO|Gap>Min) Pre Avg. Min. Max. Std. Dev. SS-ONLY (n = 16) 60.3% 4.0% 100.0% 28.9% SS+FB (n = 16) 63.2% 10.0% 100.0% 24.9% Post SS-ONLY (n = 14) 68.2% 5.8% 100.0% 24.4% SS+FB (n = 14) 89.3% 58.3% 100.0% 13.4% Exhibit 5. Crossable gap availability and utilization statistics for CTL crosswalks. a) Observed Delay per Leg (s) Pre Avg. Min. Max. Std. Dev. SS-ONLY (n = 16) 26.2 3.7 80.6 20.7 SS+FB (n = 16) 23.4 4.1 75.7 19.6 Post SS-ONLY (n = 14) 18.5 5.3 34.5 9.2 SS+FB (n = 14) 12.2 3.2 36.0 8.0 b) Delay>Min (s) Pre Avg. Min. Max. Std. Dev. SS-ONLY (n = 16) 15.6 0.3 65.4 17.4 SS+FB (n = 16) 14.9 0.6 63.4 16.9 Post SS-ONLY (n = 14) 11.7 2.1 27.9 7.5 SS+FB (n = 14) 4.9 0.0 20.4 5.7 Exhibit 6. Average pedestrian delay statistics for studied crosswalk.

when considering the range and standard deviation of the estimates. Similar to delay, the Delay>Min parameter saw some tightening in these variability measures after treatment installation. Exhibit 7 shows the cumulative distribution of delay for all subjects in the pre and post conditions for both SS-ONLY (a) and SS+FB (b) crosswalks. The figures show a relative shift of the pre and post curves for both crosswalks, with a bigger effect expectedly at the SS+FB corner. The 85th percentile overall delay was reduced from 40.9 to 32.7 s at the SS-ONLY corner, and from 38.6 s to 17.9 s at the SS+FB crosswalk. Exhibit 8 shows the 85th percentile delay estimate by subject for SS-ONLY (a) and SS+FB (b) crosswalks. The exhibit makes evident that there is a lot of inter-subject variability. 105 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 0 10 20 30 40 50 60 70 80 90 100 110 120 Pe rc en til e Delay (sec.) a) SS-ONLY PRE POST 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 0 10 20 30 40 50 60 70 80 90 100 110 120 Pe rc e n til e Delay (sec.) b) SS+FB PRE POST 85%ILE DELAY POST PRE Exhibit 7. Cumulative delay distribution all subjects – channelized turn lane.

The results of the SS-ONLY crosswalk show that some partic- ipants seemed to benefit from the sound strips (subjects 1, 4, 7, 9, 11, 12, 13, and 15), while others experienced no difference (subjects 2, 6, 8, 10, and 16). One subject experienced greater delay after treatment installation (subject 3). Note that sub- jects 5 and 14 did not participate in the post study. For the SS+FB crosswalk, Exhibit 8 shows that several participants experienced reductions in 85th percentile delay after installation of the treatments (subjects 2, 4, 7, 8, 9, 10, 11, 12, 13, 15, and 16), while others stayed approximately constant (subjects 1 and 3) and one actually experienced slightly higher delay after the treatment was installed (subject 6). Subjects 5 and 14 did not participate in the post study. The results are further arranged by time of day. It appears that participants in the 8:30 a.m. time slot experienced lower delay than those participating later in the day, when traffic volumes were higher. However, given the low sample of obser- vations in each category, no effect can be isolated. Most par- ticipants were in the 10:30 a.m. time slot, during which a range of delay times was observed. 106 0 20 40 60 80 100 120 10 13 14 1 3 5 8 1 3 5 8 11 15 12 16 6 2 8:30am 10:30am 1:30pm 3:30pm 85 th P er c en til e D el ay (s ec .) Subject a) SS-ONLY PRE POST 0 20 40 60 80 100 120 10 13 14 11 15 8:30am 10:30am 1:30pm 3:30pm 85 th Pe rc en til e D e la y (se c. ) Subject b) SS+FB PRE POST 4 7 9 12 16 6 2 4 7 9 Exhibit 8. 85th percentile delay by subject – channelized turn lane.

The analysis further investigated two parameters that were intended to describe the efficiency with which crossing opportunities were utilized: • Latency (s): Latency is defined as the time between when the previous vehicle went through the crosswalk and the time the pedestrian initiated the crossing. • Yield Lost Time (s): The YLT is defined as the time between when a driver first yields and the time the crossing is initiated. Note that in some cases, pedestrians may prefer to cross only after a car has come to a full stop (stopped yield), and so some inherent yield utilization time is expected. Exhibit 9 shows statistics for both measures. The latency results in Exhibit 9 suggest that on average pedestrians waited 8 to 10 s into a crossable gap before ini- tiating the crossing in the pre study. This suggests a lot of inefficiency in decision-making and likely contributes to the low overall rate of gap utilization at the site. Individual sub- jects even experienced average latency times up to 32 s. With installation of the treatments, the average latency dropped at the SS+FB crosswalk from 10.0 to 6.8 s (p = 0.0986). It appears that the combination of sound strips and flashing beacon either gave pedestrians more confidence in their actions or that the sound strips helped with identifying gap crossing opportunities (in this case through the absence of sound). For the yield lost time measure, there was no measurable difference between SS-ONLY and SS+FB corners and no detectable impact with treatment installation. The average YLT was in the range of 3 to 4 s, which again points to inefficiencies in the utilization of yields. Similar to observations at round- abouts, isolated YLTs reached a maximum average of 18.3 s. It is expected that few drivers are willing to wait that long to let a pedestrian cross, unless they experience a downstream conflict and therefore do not incur any additional delay. At the CTL, there may be a downstream queue resulting from vehicles yielding to cross-street traffic. At the same time, these high YLTs prove that there are some determined yielders that disrupt traffic operations similar to the way a signalized cross- ing would, or even more so. For the 16-ft crossing, a pedestrian signal would likely be timed as 4 s of “Walk” followed by 5 s of “Flashing Don’t Walk” (16 ft/3.5 fps = 4.6 s), which is significantly less than the time some drivers yielded. Pre- sumably, a PHB or HAWK phasing scheme would further reduce the signal impact. Finally, the analysis evaluates the rate of O&M interven- tions, a measure of pedestrian risk during the crossings. The study participants were at all times accompanied by a certi- fied O&M specialist, who was directed to stop the participants if the crossing decision would have resulted in undue risk to pedestrian and/or driver. The resulting rate of O&M inter- vention is defined as follows: • Intervention Rate (%): The intervention rate is defined by the number of times the O&M specialist intervened for a particular subject divided by the total number of lanes crossed for a particular condition. For example, one inter- vention over a set of 20 lane crossings at one turn lane cor- responds to an intervention rate of 5%. The summary statistics for O&M interventions are given in Exhibit 10. The results show that a total of 44 O&M interventions were observed across all 16 participants in the pre case, 30 at the SS-ONLY, and 14 at the SS+FB crosswalk. On average, each participant experienced 1.9 interventions at the SS_ONLY 107 a) Latency (s) Pre Avg. Min. Max. Std. Dev. SS-ONLY (n = 16) 8.2 2.1 15.9 3.8 SS+FB (n = 16) 10.0 3.1 32.0 7.0 Post SS-ONLY (n = 14) 7.6 2.2 12.5 2.9 SS+FB (n = 14) 6.8 1.7 11.8 2.5 b) Yield Lost Time (s) Pre Avg. Min. Max. Std. Dev. SS-ONLY (n = 16) 3.6 –0.9 13.7 3.7 SS+FB (n = 16) 3.4 –5.1 18.3 5.1 Post SS-ONLY (n = 14) 4.1 0.5 13.0 3.9 SS+FB (n = 14) 3.8 1.1 9.6 2.3 Exhibit 9. Latency and yield lost time statistics for CTL crosswalks.

crosswalk in the pre condition, which corresponds to 9.4% of crossing attempts. At the SS+FB crosswalk in the pre con- dition, the average rate of interventions was 0.9 interventions per participant, which equates to 5.6% of crossing attempts, as shown in Exhibit 10. These rates are much higher than for other observed sites and indicate a lot of risk at this crossing. While some pedestrians didn’t experience any interventions, others had up to six interventions out of 20 crossings at the SS-ONLY crosswalk, resulting in an intervention rate of 30%. While fewer average interventions were observed at the SS+FB crosswalk compared to SS-ONLY, the difference is not statistically significant (p = 0.1614). The installation of treatments reduced interventions sig- nificantly at both crosswalks to 2.9% and 1.4% at the SS-ONLY and SS+FB corners, respectively (p = 0.0204 and p = 0.0112). This suggests that while the treatments didn’t have a huge effect on crossing performance in terms of opportunity utilization and mixed results on delay, the impact on interventions seems clear and noteworthy for both crosswalks. While the effect of the treatments on the reduction of inter- ventions is significant, crossing risk still remains. Ashmead et al. (2005) posited that the probability of a dangerous crossing decision is given by 1–(1–pper crossing)n, where pper crossing is the observed intervention rate and n the number of crossing attempts. After 40 crossings (twice per day, 5 days a week, over 4 weeks), the probabilities of a risky decision at the rates of 1.4% and 2.9% are 43.1% and 69.2%, respectively. After 100 crossings, the post intervention rate of 2.9% at the SS-ONLY crosswalk results in a 94.7% likelihood of a risky decision. Exhibit 11 explores the hypothesis that the intervention rate was related to the time of day that the subjects participated in the study. Members of the research team anecdotally found the p.m. periods at the SS-ONLY corner very difficult to cross due to high afternoon turning volumes in the channelized turn lane. The exhibit further provides insight in the variability of interventions across participants. Note that participants 5 and 14 did not participate into the post study. Their post intervention rates are shown in the negative to distinguish them from participants with zero interventions. Given the inter- vention patterns, the hypothesis that the higher volume resulted in a greater degree of risk could not be supported. The figures further illustrate that the degree of risk varies across partici- pants, even after controlling for time of day. Channelized Turn Lane Results Summary The field evaluation at the channelized turn lane in Charlotte showed that participants experienced a lot of delay and risk at this site. Despite the fact that only a single lane needed to be crossed, the combination of background noise at the busy 108 Intervention Rate Pre Avg. Min. Max. Std. Dev. SS-ONLY (n = 16) 9.4% 0.0% 30.0% 9.5% SS+FB (n = 16) 5.6% 0.0% 15.0% 5.2% Post SS-ONLY (n = 14) 2.9% 0.0% 15.0% 4.7% SS+FB (n = 14) 1.4% 0.0% 10.0% 3.1% Exhibit 10. O&M intervention statistics for CTL crosswalks. -5.0% 6 12 16 1 3 5 8 11 15 2 4 7 9 10 13 14 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% 35.0% 1:30pm 10:30am 3:30pm 8:30am In te rv en tio n R a te Subject a) SS-ONLY PRE POST Exhibit 11. Intervention statistics by subject and time of day.

intersection and fast approach speeds in the turn lane caused higher delays and more frequent interventions than at the roundabout sites, including the two-lane roundabout in Golden. Exhibit 12 summarizes the crossing performance for the CTL location. The installation of the sound strip and lane delineator treatments at the SS-ONLY corner did not have a large impact on most of the availability and utilization performance meas- ures when aggregated for all participants. However, individ- ual participants seemed to benefit from the treatments, and the treatments further resulted in a significant reduction in interventions. 109 6 12 16 1 3 5 11 15 8 2 4 7 9 10 13 14 1:30pm 10:30am 3:30pm 8:30am -5.0% 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% 35.0% In te rv en tio n R a te Subject b) SS+FB PRE POST 6 12 16 1 3 5 8 11 15 2 4 7 9 10 13 14 Exhibit 11. (Continued). SS-ONLY Turn Lane SS+FB Turn Lane Performance Measure Pre Post Pre Post Yield Availability 18.4% 18.6% 15.2% 22.0% Gap Availability 34.9% 41.2% 44.7% 49.2% Yield Utilization 50.8% 40.5% 53.1% 64.6% Gap Utilization 60.3% 68.2% 63.2% 89.3% 85th percentile Delay (s) 40.9 32.7 38.6 17.9 O&M Interventions 9.4% 5.6% 2.9% 1.4% Exhibit 12. Crossing performance summary pre and post at channelized turn lane.

110 Introduction This section describes detailed analysis results of data col- lected at the single-lane roundabout in Golden, CO, at the intersection of Golden Road and Ulysses Road (Exhibit 13). The initial focus is on pedestrian-related measures, including the availability and utilization of yield and gaps, as well as pedestrian delay and O&M interventions. The single-lane roundabout was studied twice, coinciding with the pre and post studies at the nearby two-lane round- about. But contrary to that site, no pedestrian crossing treat- ments were installed at this single-lane roundabout. Without treatment installation, the pre–post comparison serves as a control for any learning effects or changes in driver behavior between the two studies. The analysis presents findings in the pre and post conditions for the studied crosswalk sequentially. Only the eastern crosswalk was included in the study. The results are also compared to those gathered at other single- lane roundabouts included in this study. The pre study was completed in July 2008 and had a total of 18 blind travelers. Thirteen of the original 18 participants returned for the post experiment in September 2008. Again, no treatments were installed at this roundabout, so the under- lying hypothesis is that overall performance in pre and post conditions are the same. Golden Single-Lane Analysis Results Site Description A picture of the studied crosswalk is shown in Exhibit 14. The roundabout has a central island diameter of 100 ft, includ- ing a 10-ft truck apron. The lanes at the studied crosswalk are 20 ft wide, partly to accommodate a nearby roadside bus stop. The crosswalk is located approximately 60 ft from the circu- lating lane measured at the exit side, and approximately 50 ft from the roundabout yield line at the entry. The two-stage crossing is divided by an 8-ft raised splitter island, but the cross- ing itself is at pavement elevation. No pedestrian-detectable warning surfaces were installed on the splitter island and so the study participants were instructed by the O&M specialist when they completed the first half of the crossing. Detectable warnings were installed on the outside curb ramps, and the crosswalk was outfitted with standard pedestrian signage. Crossing Statistics for Crosswalk The analysis of crossing performance focuses on aspects of pedestrian–vehicle interaction following the NCHRP Project 3-78A analysis framework. The first analysis compo- nent describes the availability and utilization of yields in both the pre and post treatment conditions. Two yield measures are used in the analysis: • P(Y_ENC): The probability of encountering a yield event, defined as the number of yields divided by the total of all events encountered by the pedestrian until he/she completes the crossing. • P(GO|Y): The probability of yield utilization, defined by the number of crossings in a yield divided by total number of yields encountered by the pedestrian. The P(Y_ENC) measure is somewhat different from the traditionally used probability of yielding, since it is calculated on the basis of all pedestrian–vehicle events and not just potential yielders. Chapter 3 provides additional discussion on these and other performance measures, including examples on the difference between the yielding measures. Exhibit 15 shows the statistics for the studied crosswalk. The figures shown represent the mean results by crossing leg considering all subjects. Each subject completed four crossing trials at the roundabout, with each trial consisting of four lane crossings (e.g., entry–exit–exit–entry). For example, a subject in the pre condition would have crossed the entry and exit portions of the crosswalk, respectively, eight times (twice in P A R T 2 Detailed Single-Lane Roundabout: Golden, CO

each of four trials). The average performance for the entry leg in the pre condition is then calculated from the mean of these eight crossings for all 18 subjects. The overall average is then calculated from 36 observations (18 entry and 18 exit) each representing eight individual crossing attempts. In the post conditions, 13 subjects participated in the experiment. In total, 31 subjects were included in the study (18 pre, 13 post) and each performed 16 lane crossings (four trials at four lanes each), resulting in a total of 496 crossing attempts at this location. Exhibit 15 shows that the probability of encountering a yield, P(Y_ENC), was higher in the entry lane than in the exit lane for both the pre (p = 0.0004) and post (p = 0.0624) conditions. The yield encounter probability did not change significantly in the pre and post studies, suggesting that driver behavior was comparable between the two studies. The exhibit further shows the rates of yield utilization, P(GO|Y), defined as the rate of yields that resulted in a pedes- trian crossing the roadway. The yield utilization rates are gen- erally on the order of 75% to 90%, and no significant differences were observed in either the pre–post or entry–exit comparisons due to large standard deviations in the mean estimates. A considerable fraction of yields further fell into the “forced yield” category, which is defined as the pedestrian stepping out into the roadway before the vehicle initiated the yielding process. The degree of risk associated with these events depends on the relative position and speed of the vehicle at the time of crossing initiation. Forced yield events should therefore not necessarily be interpreted as poor or risky decisions. In the pre condition, 32.5% and 40.6% of yields were forced at the entry and exit leg, respectively. In the post condition, the cor- responding forced yield percentages were 22.2% and 39.4%. The differences between pre and post percentages of forced yields are not statistically significant (p = 0.2294 and p = 0.8955 for entry and exit, respectively). The exit leg crossing did show a greater percentage of forced yields for both pre and post conditions, but these differences were also not statistically significant. The analysis next considered the availability and utilization of crossable gaps. For the purpose of this analysis, a crossable gap was defined as a gap greater than 8 s which was sufficient to cross the wide 21-ft crosswalk at a walking speed of 3.5 ft/s, while allowing for a 2-s safety buffer. This 2 s allows for some pedestrian reaction time before initiating the crossing, as well 111 Ph ot o by G oo gl e Exhibit 13. Aerial view of roundabout. Ph ot o by J an et B ar lo w Exhibit 14. The studied crosswalk. a) P(Y_ENC) Pre (n = 18) Avg. Min. Max. Std. Dev. Entry 51.1% 16.7% 100.0% 18.4% Exit 29.6% 7.1% 57.1% 13.7% Overall 40.4% 7.1% 100.0% 19.4% Post (n = 13) Entry 51.1% 18.8% 100.0% 21.1% Exit 36.5% 13.6% 62.5% 16.5% Overall 43.8% 13.6% 100.0% 20.0% b) P(GO|Y) Pre (n = 18) Avg. Min. Max. Std. Dev. Entry 82.8% 36.4% 100.0% 20.1% Exit 76.0% 25.0% 100.0% 26.1% Overall 79.4% 25.0% 100.0% 23.2% Post (n = 13) Entry 80.3% 21.3% 100.0% 23.3% Exit 89.2% 66.7% 100.0% 12.7% Overall 84.7% 23.1% 100.0% 18.9% Exhibit 15. Yield availability and utilization statistics for studied crosswalk.

as a safety buffer between a completed crossing and the next vehicle arrival. Similar to the yield statistics, two gap-related parameters are defined: • P(CG_ENC): The probability of encountering a CG event (gap greater than 8 s), defined as the number of crossable gaps divided by the total of all events encountered by the pedestrian. • P(GO|CG): The probability of crossable gap utilization, defined by the number of crossings in a CG divided by total number of CGs encountered by the pedestrian. Exhibit 16 shows the statistics for the studied crosswalk. The results in Exhibit 16 show that the P(CG_ENC) is slightly higher on the entry leg for the pre conditions, but this difference is not significant at p = 0.1125. There is no significant effect of P(CG_ENC) in a pre–post comparison, suggesting that traffic patterns with respect to gap availability remained largely unchanged between the two studies. Exhibit 16 further shows that the blind study participants generally had high crossable gap utilization rates, averaging in the 80% to 90% range. This may be the result of a very con- servative crossable gap definition that allows most pedestrians to cross. In fact, some assertive pedestrians crossed in gaps smaller than this threshold, resulting in P(GO|Gap>Min) values greater than 100% (capped in Exhibit 16). No significant difference in P(GO|CG) is detected in a pre–post or entry–exit comparison. The combined effect of gap and yield availability and utiliza- tion is reflected in the delay experienced by pedestrians. Delay statistics in Exhibit 17 are provided for two delay measures: • Observed Delay per Leg (s): The average pedestrian delay in seconds, defined as the time difference between when the trial started and when the pedestrian initiated the crossing. • Delay>Min (s): The delay beyond the first opportunity (Delay>Min), defined as the time difference between first yield or crossable gap encountered by the pedestrian and the actual crossing initiation. Statistics for all measures are for crossing one leg of the roundabout at either the exit or entry approach. The total average delay by crossing can be calculated by summing delay statistics for the entry and exit legs. Exhibit 17 shows that small differences in average pedestrian delay per leg were observed in an entry–exit leg comparison, although none of the small differences were statistically signifi- cant at the given sample size (all p > 0.40). Also, no significant differences in delay were observed in a pre–post comparison (all p > 0.50). In addition to the average delay for all participants, it is important to emphasize that some participants experienced much larger delays. The longest overall average delay by a participant was 51.4 s per leg. A 2-min time-out was used for all trials, but none of the participants ever reached that limit at this site. The results for Delay>Min also show no significant difference between for pre–post and entry–exit comparisons. Overall, the Delay>Min results suggest that the blind pedestrians did not 112 a) P(CG_ENC) Pre (n = 18) Avg. Min. Max. Std. Dev. Entry 26.3% 0.0% 44.4% 12.4% Exit 20.6% 4.8% 34.8% 8.4% Overall 23.5% 0.0% 44.4% 10.8% Post (n = 13) Entry 21.5% 0.0% 37.5% 9.4% Exit 21.1% 8.3% 50.0% 11.3% Overall 21.3% 0.0% 50.0% 10.2% b) P(GO|CG) Pre (n = 18) Avg. Min. Max. Std. Dev. Entry 83.2% 33.3% 100.0%* 23.7% Exit 86.8% 40.0% 100.0%* 23.4% Overall 85.1% 33.3% 100.0%* 23.2% Post (n = 13) Entry 80.9% 45.5% 100.0% 22.3% Exit 81.4% 40.0% 100.0% 22.5% Overall 81.2% 40.0% 100.0% 21.9% * These figures were capped at 100%, although the calculation resulted in estimates greater than 100%. This occurs if pedestrians utilize some non-crossable gaps and therefore have more utilized gaps than there are crossable gaps available. Exhibit 16. Crossable gap availability and utilization statistics for studied crosswalk. a) Observed Delay per Leg (s) Pre (n = 18) Avg. Min. Max. Std. Dev. Entry 10.9 4.0 31.3 7.3 Exit 13.0 3.5 29.4 7.9 Overall 11.9 3.5 31.3 7.6 Post (n = 13) Entry 13.3 2.7 51.4 13.6 Exit 11.0 3.4 27.8 7.3 Overall 12.1 2.7 51.4 10.7 b) Delay>Min (s) Pre (n = 18) Avg. Min. Max. Std. Dev. Entry 2.8 0.1 6.5 2.1 Exit 2.7 0.1 7.0 2.3 Overall 2.8 0.1 7.0 2.2 Post (n = 13) Entry 3.7 0.3 19.9 5.2 Exit 2.5 0.1 9.9 2.8 Overall 3.1 0.1 19.9 4.2 Exhibit 17. Average pedestrian delay statistics for studied crosswalk.

miss a lot of crossing opportunities. Despite these low averages, some pedestrians experienced Delay>Min up to 19.9 s. The null hypothesis that Delay>Min = 0 is rejected for both pre and post conditions (p < 0.0001 and p = 0.0030, respectively). Exhibit 18 shows the distribution of delay for all subjects in the pre and post conditions. The hypothesis that no significant changes took place between the studies is supported by the data. This implies that any effects observed at the neighboring two-lane roundabout are likely attributable to the installation of the treatments and not a learning effect by pedestrians. Exhibit 19 shows the 85th percentile delay estimate by subject. It appears that one participant (subject 7) experienced sig- nificantly greater delay in the post condition, while all other delay performances remained largely unchanged. The delay statistics are arranged by time of day during which subjects par- ticipated, but no trends can be identified. Note that subjects 1, 5, 10, 15, and 16 did not participate in the post study. The analysis further investigated two new parameters that were not previously used in Schroeder, Rouphail, and Hughes (2009). Both measures are intended to describe the efficiency 113 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 110 120 Pe rc e n til e Delay (sec.) Single-Lane Roundabout PRE POST 85%ILE DELAYPOST 21.7 sec. PRE 22.2 sec. Exhibit 18. Delay distribution all subjects – single-lane roundabout. 0 20 40 60 80 100 120 1 3 6 9 11 16 4 10 14 17 12 7 2 5 8 13 15 18 9:00am 11:30am 3:30pm 85 th Pe rc en til e De la y (se c. ) Subject Single-Lane Roundabout PRE POST Exhibit 19. 85th percentile delay by subject – single-lane roundabout.

with which a crossing opportunity is utilized for both gaps and yields: • Latency (s): Latency is defined as the time between when the last vehicle went through the crosswalk and the time the pedestrian initiated the crossing. • Yield Lost Time (s): The YLT is defined as the time between when a driver first yields and the time the crossing is ini- tiated. Note that in some cases, pedestrians may prefer to cross only after a car has come to a full stop (stopped yield), and so some inherent yield utilization time is expected. Exhibit 20 shows statistics for both measures. The latency results in Exhibit 20 suggest that on average pedestrians wait 4 to 6 s into a crossable gap before initiating the crossing, suggesting inefficiency in decision-making. No significant difference in latency was detected for a pre–post or entry–exit comparison. For the yield lost time measure, pedestrians on average wait 1.5 to 2 s before crossing in front of a yielding vehicle. The average maximum YLT was 11.1 s for the pre and 6.6 s for the post condition. Note that in many cases, drivers will not be willing to wait this long and a high YLT will therefore translate to an increased percentage of missed yields [lower P(GO|Y)]. Also note that some YLT values are negative, suggesting that some pedestrians forced vehicles to yield. Finally, the analysis includes the rate of O&M interventions that represent a measure of pedestrian risk during the crossings. The study participants were at all times accompanied by a certified O&M specialist who was directed to stop the partic- ipants if the crossing decision would have resulted in undue risk to pedestrian and/or driver. The resulting rate of O&M intervention is defined as follows: • Intervention Rate (%): The intervention rate is defined by the number of times the O&M specialist intervened for a particular subject divided by the total number of lanes crossed for a particular condition. For example, one inter- vention over a set of eight lane crossings at the roundabout entry corresponds to an intervention rate of 12.5%. The summary statistics for O&M Interventions are given in Exhibit 21. The results show that a total of four O&M interventions were observed in the pre case over a total of 72 lane crossings at the entry leg for a rate of 2.8%. In the post analysis the inter- vention rate at the entry was 1.0%; however, due to high stan- dard deviations that difference is not significant (p = 0.2616) at this sample size. No interventions were observed at the exit leg of this particular roundabout. The overall intervention rates for pre and post were 1.4% and 0.5%, respectively. Again, this difference is not statistically significant. The null hypothesis that the intervention rate is zero is rejected for the pre condition (p = 0.0344) but cannot be rejected at the given sample size for the post evaluation. Exhibit 22 explores the distribution of interventions by sub- ject and time of day. Given the rare occurrence of interventions, it is difficult to draw any conclusions about patterns at the given sample size. Participants who didn’t return for the post study are shown with negative intervention rates to visually distinguish them from zero-intervention subjects. The num- bers are shown as the percentage of interventions from 16 lane crossing per subject at this site. While the rates of interventions appear low, this does not mean that a crossing is safe. Ashmead et al. (2005) posited that the probability of a dangerous crossing decision is given by 1–(1–pper crossing)n, where pper crossing is the observed intervention rate and n the number of crossing attempts. After 40 crossings (twice per day, 5 days a week, over 4 weeks), the probabilities 114 a) Latency (s) Pre (n = 18) Avg. Min. Max. Std. Dev. Entry 5.7 1.8 16.0 4.0 Exit 4.9 2.4 10.5 1.9 Overall 5.3 1.8 16.0 3.0 Post (n = 13) Entry 5.0 1.3 12.6 3.1 Exit 4.7 2.7 7.3 1.6 Overall 4.8 1.3 12.6 2.4 b) Yield Lost Time (s) Pre (n = 18) Avg. Min. Max. Std. Dev. Entry 2.6 –2.1 11.1 3.8 Exit 0.5 –1.9 5.1 2.0 Overall 1.6 –2.1 11.1 3.2 Post (n = 13) Entry 2.8 0.1 6.6 2.3 Exit 1.1 –1.4 4.4 1.8 Overall 1.9 –1.4 6.6 2.2 Exhibit 20. Latency and yield lost time statistics for studied crosswalk. Intervention Rate Pre (n = 18) Avg. Min. Max. Std. Dev. Entry 2.8% 0.0% 12.5% 5.3% Exit 0.0% 0.0% 0.0% 0.0% Overall 1.4% 0.0% 6.3% 2.7% Post (n = 13) Entry 1.0% 0.0% 12.5% 3.5% Exit 0.0% 0.0% 0.0% 0.0% Overall 0.5% 0.0% 6.3% 1.7% Exhibit 21. O&M intervention statistics for single-lane roundabout crosswalk.

of a risky decision at the rates 0.5% and 1.4% are 18.2% and 43.1%, respectively. After 100 crossings, the entry intervention rate of 2.8% in the pre condition results in a 94.2% likelihood of a risky decision. Golden, CO, Single-Lane Roundabout Summary The field evaluation at the single-lane roundabout in Golden did not provide statistical evidence of a learning effect by the blind study participants or significant changes in driver behavior or traffic volumes. The crossing opportunity avail- ability measures, P(CG_ENC) and P(Y_ENC), remained un- changed, as did the rate of utilization of these opportunities. The analysis of the delay performance of study participants showed no conclusive pre and post difference. The number of experimenter interventions appeared to have dropped from four in the pre to only one intervention in the post; however, due to the low sample size and rare occurrence of this mea- sure, this effect is believed to be due to random variability. Exhibit 23 summarizes the crossing performance for the Golden single-lane roundabout. In a comparison to other single-lane roundabouts, the Golden roundabout showed a higher rate of driver yielding and a comparable rate of crossable gap occurrence. The rate of opportunity utilization was in the same general range as other sites. For all compared sites, the between-subject variability was very large. 115 -5.0% 0.0% 5.0% 10.0% 15.0% 20.0% 1 3 6 9 11 16 4 7 10 12 14 17 2 5 8 13 15 18 9:00am 11:30am 3:30pm O &M In te rv e n tio ns Subject Single-Lane Roundabout PRE POST Exhibit 22. O&M interventions by subject and by time of day. Performance Measure Pre Post Yield Availability 40.4% 43.8% Gap Availability 23.5% 21.3% Yield Utilization 79.4% 84.7% Gap Utilization 85.1% 81.2% 85th percentile Delay (s) 22.2 21.7 O&M interventions 1.4% 0.5% Exhibit 23. Crossing performance summary pre and post at single-lane roundabout.

116 Introduction This section describes detailed analysis results of data col- lected at two single-lane roundabouts in North Carolina. The analysis and comparison of these two sites were previously published in Schroeder, Rouphail, and Hughes (2009), although the present document shows some revised variable definitions compared to the published paper. The first roundabout is located at the intersection of 9th Street and Davidson Avenue in Charlotte. This roundabout was included in the original data collection scope of NCHRP Project 3-78A and was initially proposed to be evaluated in a pre and post experiment with treatment installation. However, the treatment installation and post study were aborted later in the project and funds reallocated to other purposes. One of the uses of these funds was the evaluation of the second single-lane roundabout described in this document. The roundabout at the intersection of Pullen Road and Stinson Drive in Raleigh was previously studied in a related project, using the same experimental protocol applied in NCHRP Project 3-78A. The analysis presented in this document was performed from video observations of that study. Another use of the funds was data collection at a single-lane roundabout in Golden, which is discussed in a separate document. The original data collection scope for NCHRP Project 3-78A included the evaluation of one single-lane roundabout at the intersection of 9th Street and Davidson Avenue in Charlotte (Site DAV-CLT). In the discussion at the interim NCHRP Project 3-78A panel meeting in January 2008 in Washington, D.C., concerns were raised that the low traffic volumes at this site were not representative of a typical single-lane U.S. round- about. As a result, the site was deemed to be accessible and would not substantively benefit from the installation of the proposed treatments. The NCHRP Project 3-78A team thus agreed to compare the crossing performance statistics to a higher-volume single-lane roundabout and contrast the acces- sibility criteria, which is the focus of this document. Several members of the research team are also involved in a separate research effort under sponsorship of the NIH. In this multi-year NIH project, research on the general crossing performance of blind pedestrians is performed, and a subset of studies focuses on roundabouts. In particular, a study in the fall of 2004 investigated the feasibility of an automated yield-detection system (AYDS) at a single lane roundabout at Pullen Road and Stinson Drive in Raleigh (Site PS-RAL). The data collection protocol at the PS-RAL study was comparable to DAV-CLT and included trials in both conditions with the AYDS treatment “on” and “off.” In this document, the PS-RAL trials conducted in the “off” condition are compared to the data in the “before” study at DAV-CLT. In the data collection at DAV-CLT, data from a total of 10 blind subjects were analyzed. The dataset for PS-RAL resulted in usable data from 12 blind participants as well as six sighted comparison subjects (not shown). At both sites, a full crossing consisted of four lane crossings (for example entry–exit–exit–entry) with the starting order of lanes random- ized for each subject. At DAV-CLT each subject completed three full crossings at the northern and three full crossings at the southern crosswalk, resulting in a total of 12 entry and 12 exit individual lane crossings. At the PS-RAL site each subject completed eight full trials at one crosswalk, resulting in 16 entry and 16 exit lane crossings. North Carolina Single-Lane Analysis Results Site Description The DAV-CLT roundabout has an inscribed diameter of approximately 140 ft and approach speed limits of 25 mph. The PS-RBT has a smaller inscribed diameter of 88 ft and also has approach speed limits of 25 mph. Exhibit 24 below shows aerial views of both sites. The tested crosswalks are highlighted. P A R T 3 Detailed Single-Lane Roundabout: North Carolina

When traffic volumes are compared, the major approaches on Davidson Avenue and Pullen Road have AADTs of about 9,900 and 15,000, respectively. The side streets on 9th Street and Stinson Drive respectively have much lower volumes. Exhibit 25 shows the peak hour entering volumes for both sites. The data suggest that the a.m. and p.m. peak hour vol- umes at the PS-RAL are about 50% and 90% higher than at the DAV-CLT site, respectively. More importantly, the lunch peak hour at PS-RAL has 240% more traffic, which is mostly a result of generally low daytime volumes at the DAV-CLT site. A similar trend was observed during the actual experi- mental trials. While the DAV-CLT has medium traffic vol- umes in the a.m. and p.m. peak hours, traffic during the actual experimental trials, which were carried out more often slightly outside the peak hours, was relatively low. The research team also gathered sample speed observations at both sites. The approach speeds on the entry approach lanes to the north and south crosswalk at the DAV-CLT site were 27.6 and 26.0 mph, respectively. Upon entry, the aver- age vehicle speed dropped to approximately 17.6 mph due to the roundabout geometry. The average approach speed at the southern crosswalk of the PS-RAL roundabout was lower than at DAV-CLT at 22.8 mph. The average entering speed to the PS-RAL roundabout was 15.6 mph. The average exiting speeds at DAV-CLT and PS-RAL were approximately 17.3 and 15.3 mph, respectively. The lower speeds at PS-RAL are likely 117 a) DAV-CLT b) PS-RAL Exhibit 24. Aerial views of DAV-CLT and PS-RAL roundabouts. a) PS-RAL b) DAV-CLT North East South West East South West TOTAL AM Peak (7:30-8:30AM) 779 3 461 36 1279 Lunch Peak (12:15-1:15PM) 583 38 560 113 1294 PM Peak (5:00-6:00PM) 454 20 887 123 1484 Peak Hour Volumes PS-RAL Total Entering Volumes, Sep-2007 North TOTAL AM Peak (7:30-8:30AM) 157 79 506 92 834 Lunch Peak (1:00-2:00PM) 198 26 272 39 535 PM Peak (5:00-6:00PM) 364 70 277 76 787 DAV-CLT Peak Hour Volumes Total Entering Volumes, Nov-2007 Exhibit 25. Peak hour entering volumes for sites.

attributable to the smaller inscribed diameter and associated lower design speed of the roundabout. Crossing Statistics The analysis of crossing performance focuses on aspects of pedestrian–vehicle interaction following the NCHRP Project 3-78A analysis framework. The first analysis compo- nent describes the availability and utilization of yields in both the pre and post treatment conditions. Two yield measures are used in the analysis: • P(Y_ENC): The probability of encountering a yield event, defined as the number of yields divided by the total of all events encountered by the pedestrian until he/she completes the crossing. • P(GO|Y): The probability of yield utilization, defined as the number of crossings in a yield divided by total number of yields encountered by the pedestrian. The P(Y_ENC) measure is somewhat different than the traditionally used probability of yielding, since it is calculated on the basis of all pedestrian–vehicle events and not just potential yielders. Chapter 3 provides additional discussion on these and other performance measures, including examples on the difference between the yielding measures. Exhibit 26a compares the yield encounters for the two sites. It shows generally higher probabilities of encountering a yield at the PS-RAL roundabout, which may be related to the prox- imity to a major college campus. The PS-RAL site further suggests lower yielding at the roundabout exit leg, which is not evident at DAV-CLT. Exhibit 26b shows the yield utilization rates at the two sites. A lower yield utilization rate is evident at DAV-CLT (67.4%) than at PS-RAL (85.4%). Both sites suggest a slightly higher yield utilization rate at the exit leg. By combining yielding and yield utilization rates, it can be stated that the PS-RAL site exhibits a higher likelihood of crossing in a yield than DAV-CLT. The range of observed yield utilization points to differences in crossing abilities among participants, with some utilizing 100% of yields while others don’t utilize any. The analysis next considered the availability and utilization of crossable gaps. For the purpose of this analysis, a crossable gap is defined as a gap that was sufficient to cross the width of the crosswalk at a walking speed of 3.5 ft/s, while allowing for a 2-s safety buffer. This 2 s allows for some pedestrian reaction time before initiating the crossing, as well as a safety buffer between a completed crossing and the next vehicle arrival. The resulting crossable gap thresholds for DAV-CLT and PS-RAL were 7 s and 6 s, respectively. Similar to the yield statistics, two gap-related parameters are defined: • P(CG_ENC): The probability of encountering a CG event (gap greater than CG threshold), defined as the number of crossable gaps divided by the total of all events encountered by the pedestrian. 118 a) P(Y_ENC) DAV-CLT Avg. Min. Max. Std. Dev. Entry 5.8% 0.0% 14.3% 4.8% Exit 6.7% 0.0% 20.0% 5.0% Overall 6.3% 0.0% 20.0% 4.9% PS-RAL Entry 37.9% 13.1% 66.7% 17.8% Exit 28.1% 8.1% 58.3% 14.4% Overall 33.0% 8.1% 66.7% 16.6% b) P(GO|Y) DAV-CLT Avg. Min. Max. Std. Dev. Entry 64.1% 0.0% 100.0% 41.2% Exit 70.4% 0.0% 100.0% 44.1% Overall 67.4% 0.0% 100.0% 42.3% PS-RAL Entry 83.0% 50.0% 100.0% 20.4% Exit 87.8% 60.0% 100.0% 14.1% Overall 85.4% 50.0% 100.0% 17.3% Exhibit 26. Yield encounters and utilization statistics for studied crosswalk.

• P(GO|CG): The probability of crossable gap utilization, defined as the number of crossings in a CG divided by total number of CGs encountered by the pedestrian. Exhibit 27 shows the statistics for the studied crosswalks. Exhibit 27a shows the encounters of crossable gaps at the two sites. Following the definition above, the minimum cross- able gaps for DAV-CLT and PS-RAL are approximately 7.0 and 6.0 s, respectively. The table shows that DAV-CLT has a slightly higher rate of gaps (28.8%) that are greater than the crossable gap than PS_RAL does (19.1%). For both sites, the gap occur- rence is comparable for entry and exit legs. Exhibit 27b shows gap utilization rates for DAV-CLT of approximately 60%. At PS-RAL the gap utilization rate is higher for the exit leg than the entry leg, with 63.6% and 52% utiliza- tion, respectively. Overall, the gap utilization rates across the two sites are comparable. Combining gap occurrence and utiliza- tion, there is a somewhat higher likelihood of crossing in a gap at DAV-CLT. The range of gap utilization again varies between 0% and 100%, emphasizing the need for a sufficient sample size given the variability of crossing behavior. In this context it is also important to point out that no utilized gaps below the defined crossable gap threshold were observed at either site, giving confidence to the assumed crossable gap thresholds. The combined effect of gap and yield availability and utiliza- tion is reflected in the delay experienced by pedestrians. Delay statistics in Exhibit 28 are provided for two delay measures: • Observed Delay per Leg (s): The average pedestrian delay in seconds, defined as the time difference between when the trial started and when the pedestrian initiated the crossing. • Delay>Min (s): The delay beyond the first opportunity (Delay>Min), defined as the time difference between first yield or crossable gap encountered by the pedestrian and the actual crossing initiation. Statistics for all measures are for crossing one leg of the roundabout at either the exit or entry approach. The total average delay by crossing can be calculated by summing delay statistics for the entry and exit legs. Exhibit 28a compares the observed delay experienced by the blind pedestrians at both sites and suggests significantly lower delays at PS-RAL. Interpreting this difference in light of the results in Exhibits 3 and 4, the lower delay is likely attributable to greater P(Y_ENC) and greater P(GO|Y) at this site. The delay at DAV-CLT correspondingly is higher because pedestrians wait for crossable gaps in the absence of yields. The delay is comparable for the entry and exit leg at both sites. The average total delay to get across both entry and exit lanes represents the sum of the two estimates. Exhibit 28b shows the delay beyond the first crossing opportunity for both sites. The findings are similar to those in Exhibit 28a, with pedestrians at PS-RAL experiencing less unnecessary delay compared to DAV-CLT. Again, the reason for the differences is likely related to P(Yield) and P(GO|Y). The difference in delay suggests that a crossing opportunity is utilized more quickly at PS-RAL. If these sites were analyzed using LOS definitions in the HCM, the average delay times at PS-RAL and DAV-CLT (approximately 11 and 25 s) would 119 a) P(CG_ENC) DAV-CLT Avg. Min. Max. Std. Dev. Entry 29.8% 18.5% 44.7% 6.9% Exit 27.8% 9.1% 40.0% 6.7% Overall 28.8% 9.1% 44.7% 6.8% PS-RAL Entry 17.7% 0.0% 30.0% 8.9% Exit 20.5% 0.0% 32.0% 9.7% Overall 19.1% 0.0% 32.0% 9.2% b) P(GO|Gap>Min) DAV-CLT Avg. Min. Max. Std. Dev. Entry 66.3% 25.0% 100.0% 20.6% Exit 60.3% 33.3% 100.0% 17.9% Overall 63.3% 25.0% 100.0% 19.3% PS-RAL Entry 52.0% 0.0% 100.0% 41.3% Exit 63.6% 18.8% 100.0% 26.6% Overall 57.8% 0.0% 100.0% 34.4% Exhibit 27. Crossable gap encounters and utilization statistics for studied crosswalk. a) Observed Delay per Leg (s) DAV-CLT Avg. Min. Max. Std. Dev. Entry 26.6 11.2 74.0 17.0 Exit 24.0 11.4 41.8 9.7 Overall 25.3 11.2 74.0 13.8 PS-RAL Entry 10.5 4.1 34.2 8.9 Exit 11.6 5.2 26.7 6.8 Overall 11.1 4.1 34.2 7.8 b) Delay>Min (s) DAV-CLT Avg. Min. Max. Std. Dev. Entry 18.8 4.8 59.4 15.5 Exit 17.2 5.2 35.1 9.6 Overall 18.0 4.8 59.4 12.8 PS-RAL (Min = 6 s) Entry 5.6 0.8 24.7 7.2 Exit 6.1 0.8 19.4 5.8 Overall 5.8 0.8 24.7 6.4 Exhibit 28. Average pedestrian delay statistics for studied crosswalk.

correspond to LOS scores C and D, respectively. To recall, the HCM defines levels of service on a scale from A (best) to F (worst) in terms of average delay per person. Finally, the analysis includes the rate of O&M interventions that represent a measure of pedestrian risk during the crossings. The study participants were at all times accompanied by a certified O&M specialist who was directed to stop the partic- ipants if the crossing decision would have resulted in undue risk to pedestrian and/or driver. The resulting rate of O&M intervention is defined as follows. • Intervention Rate (%): The intervention rate is defined by the number of times the O&M specialist intervened for a particular subject divided by the total number of lanes crossed for a particular condition. For example, one inter- vention over a set of eight lane crossings at the roundabout entry corresponds to an intervention rate of 12.5%. Exhibit 29 shows the rate of experimenter interventions. The intervention rates at PS-RAL are clearly higher than DAV-CLT, and the exit lane crossing is especially risky with an intervention rate of 5.8%. At the DAV-CLT site, one par- ticipant experienced a single intervention at the entry leg and another one a single intervention at the exit leg (1 intervention in 12 crossing results in a rate of 8.3%). Since no other subjects experienced any interventions, the resulting average inter- vention rate across 10 subjects was 0.8% for both the entry and exit leg. However, with repeated crossings even the 0.8% inter- vention rate at DAV-CLT could result in a high likelihood of a risky decision over time. Ashmead et al. (2005) discussed that the probability of a dangerous crossing decision is given by 1 – (1 – pper crossing)n, where pper crossing is the observed intervention rate and n the number of crossing attempts. Consequently, for a pedestrian who crosses this roundabout twice a day, the prob- ability of a dangerous decision after one month (10 crossings per week over 4 weeks) is 27.5%. At the 3.9% intervention rate for PS-RAL this likelihood increases to 79.6%. Discussion Based on all the criteria considered in the comparison of the two sites, the higher-volume PS-RAL site is in fact more acces- sible than the DAV-CLT site from a delay perspective. This is primarily due to the high frequency of yields at the PS-RAL site and a high yield utilization rate. As a result, blind pedestrians on average experience less than half the delay at PS-RAL com- pared to DAV-CLT. Similarly, the amount of unnecessary delay beyond the first crossing opportunity is about three times as high at the DAV-CLT site. These findings are some- what surprising, given that the availability of (long) crossing gaps at the PS-RAL is less than that at DAV-CLT. The cross- able gap utilization rates appear to be comparable for both sites. However, from a risk perspective, the PS-RAL clearly shows higher intervention rates and thus a more dangerous crossing situation. In light of these findings, it is evident that the question of roundabout accessibility is complex and cannot be reduced to a simple relationship to traffic volumes. While a low-volume site may appear to be easily accessible, a higher-volume site may result in higher accessibility if associated with a higher rate of yielding that is being utilized. The greater accessibility of the PS-RAL site is attributable to higher P(Yield) and P(GO|Y) probabilities. These two factors seemed to have a significant overall impact on reduced pedestrian delay, despite the fact that the site had higher volumes and consequently a lower availabil- ity of crossable gaps, P(Gap>Min). Given the higher propensity to yield at PS-RAL, the associated higher volumes resulted in more frequent crossing opportunities per unit of time. The analysis in this section shows that the studied higher- volume roundabout was in fact more accessible to blind pedes- trians based on the multi-criteria established in this study. The hypothesis that the DAV-CLT roundabout is “easily accessible because of low volumes” could not be supported by the compar- ison data from PS-RAL. Through the comparison it has become evident that a combination of crossable gaps, yields, and utiliza- tion rates all contribute to making a site more or less accessible. 120 P(Risky Crossing) DAV-CLT Avg. Min. Max. Std. Dev. Entry 0.8% 0.0% 8.3% 2.6% Exit 0.8% 0.0% 8.3% 2.6% Overall 0.8% 0.0% 8.3% 2.6% PS-RAL Entry 2.1% 0.0% 6.3% 3.1% Exit 5.8% 0.0% 25.0% 7.3% Overall 3.9% 0.0% 25.0% 5.8% Exhibit 29. O&M intervention statistics for single-lane roundabout crosswalk.

121 Introduction This section describes analysis results of data collected at the RCW of the two-lane roundabout in Golden at the inter- section of Golden Road and Johnson Road (Exhibit 30). The analysis focus will be on pedestrian-related measures, includ- ing the availability and utilization of yield and gaps, as well as pedestrian delay and O&M interventions. The measures are defined in the methodology chapter of this report. Because of the two-lane approaches at this site, the section distinguishes between near-lane and far-lane effects at the crosswalk. These describe the vehicle state in the near and far lane relative to the position of the waiting pedestrian. This section focuses on the south crosswalk with the treat- ment effect of the raised crosswalk. The analysis presents findings in the pre and post conditions for the studied cross- walk sequentially. The pre study was completed in July 2008, and a total of 18 blind travelers participated in the pre study. The treatment was installed following the pre study, and 12 of the original 18 participants returned for the post experiment in Septem- ber 2008. Raised Crosswalk Evaluation Raised Crosswalk Treatment Overview An RCW (Exhibit 31) was installed at the southern leg of the roundabout. The treatment had the objective of slowing driv- ers as they traversed the crosswalk, and the research team hypothesized that the speed impediment may also result in an increased likelihood of drivers yielding. Overall, the raised crosswalk was a lower-cost treatment than the PHB but also affected drivers in the absence of pedestrians (the PHB rested in “Dark” mode). The raised crosswalk was constructed from asphalt at a ver- tical elevation of 3 in. and a 1:15 slope transition from the existing pavement surface. This elevation and slope resulted in a fairly gentle transition for vehicular traffic, but was selected to mitigate concerns on the impacts on traffic flow. It is expected that a higher elevation and/or steeper transition slope would drastically alter driver behavior, and these results therefore cannot be transferred directly to raised crosswalks with different geometries. The raised crosswalk was further installed as a temporary installation, and therefore no reconstruction was done to the curb line. Due to drainage considerations, the raised cross- walk was at road surface level on the side of the street and then sloped upward from there. For pedestrians this resulted in a somewhat uncomfortable walking experience since the crosswalk curb ramp down slope was followed by an up slope onto the raised crosswalk. For a permanent installation, it is recommended that drainage considerations be incorporated into a raised crosswalk design that is flush with the sidewalk. Pretest Pedestrian Behavior at the RCW The NCHRP Project 3-78A analysis of single-lane cross- ings used a performance evaluation framework that described the availability of crossing opportunities, the rate of utiliza- tion of these opportunities, as well as the delay and risk asso- ciated with the crossings. For a single lane, the yield and gap events are uniquely defined by the vehicle state in the conflict lane. However, at a two-lane crossing the analysis needs to consider the vehicle state in both lanes. The following analy- sis distinguishes between driver behavior in the near lane (the lane closest relative to the position of the pedestrian) and the far lane. Depending on the crossing location (entry/exit and curb/island) the near lane can be the inside or outside lane of the two-lane approach. The analysis defines the vehicle state in the near lane in five event categories: 1. Rolling Yield (RY): Pedestrian encounters a driver who has slowed down for the pedestrian, but has not come to a full stop. P A R T 4 Detailed Two-Lane Roundabout: Golden, CO – RCW

2. Stopped Yield (STY): Pedestrian encounters a driver who has come to a stop, defined as moving at a speed less than 3 mph. 3. Forced Yield (FY): Pedestrian initiates crossing before the vehicle initiated the yield, but then forces the driver to slow down by entering the crosswalk. 4. Crossable Gap (CG): Pedestrian encounters a gap large enough to safely cross the street. A crossable gap is defined as the crossing width divided by 3.5 ft/s walking speed plus 2 s for start time and safety buffer. 5. Non-Crossable Gap (non-CG): Pedestrian encounters a gap between vehicles shorter than the crossable gap threshold. The vehicle state in the far lane will be defined relative to the near-lane condition in five principal categories: rolling yield, stopped yield, crossable gap, non-crossable gap, and multiple events. The last category indicates that more than one event took place in the far lane during one near-lane event. For example, several cars could have gone through the far lane during one large gap in the near lane. For multiple events, the last event in the sequence is considered for analysis. Exhibit 32 shows the near-lane and far-lane effects for the pre condition at the RCW. The event outcomes are broken down by whether the events were utilized by the pedestrians. The exhibit shows that 183 of 686 encountered events in the near lane (26.7%) were yields, but that only some of those events were associated with a yield or crossing opportunity in the far lane. Participants did not utilize 43 rolling yields and 40 stopped yields. The corresponding overall rate of yield uti- lization is 45.4%, but 16.4% of utilized yields in the near lane were forced by the pedestrian seizing the crosswalk. The analy- sis of the far-lane event shows that a majority of the non-utilized yields were attributable to either non-crossable gaps or multi- ple events in the far lane. Overall, 53.5% of yields were double yields with stopping or stopped cars in both lanes (including forced yields), of which 61.2% were utilized. 122 Ph ot o by G oo gl e Exhibit 30. Aerial view of roundabout. Ph ot o by B as tia n Sc hr oe de r Exhibit 31. Raised crosswalk. RY STY CG FY non-CG Rolling Yield Utilized 10 Non-Utlz. 3 43 Stopped Yield Utilized 7 60 Non-Utlz. 1 40 Forced Yield Utilized 2 30 Non-Utlz. 0 Crossable Gap Utilized 5 138 Non-Utlz. 1 37 Non-Cross. Gap Utilized 5 Non-Utlz. 8 3 0 1 8 22 14 6 12 2 7 2 0 0 0 9 0 3 31 0 0 1 0 0 0 17 1 12 74 3 15 2 0 2 28 1 18 2 1 21 266 0 0 2 1 2 0 1 0 0 2 1 0 0 1 0 3 0 1 0 0 0 0 0 1 2 1 3 12 1 3 1 0 0 0 0 0 1 2 0 0 0 0 0 1 0 2 0 0 9 22 0 0 0 323 Total 31 73 44 126 339 7 5 24 3 34 686 Multiple Events Total Far-Lane Event Near-Lane Event Lane Outcome Rolling Yield Stopped Yield Forced Yield X-Able Gap Non-X. Gap Exhibit 32. Near–far lane effects, pre condition, for RCW.

Of a total of 503 encountered gaps in the near lane, 175 were crossable (34.8%), and 78.9% of these crossable gaps were utilized by the pedestrian. The likelihood of encoun- tering a crossable gap from the 686 total events was 25.5%. Of the 37 non-utilized crossable gaps, 30 had non-crossable gaps in the far lane. The near–far lane evaluation makes it evident that both need to be considered in the evaluation of pedestrian behavior. Exhibit 33 shows a summary of the cross- ing opportunity availability and utilization statistics for the pre condition. The exhibit shows a relatively low rate of yield and crossable gap occurrence in both lanes, explaining the large portion of pedestrian–vehicle events that did not result in a crossing. Further, the rate of yield utilization is only 54.6% and 68.7% in the near and far lane, suggesting a lot of pedestrian uncer- tainty. Gap utilization is somewhat higher, at around 80%. The results in Exhibit 33 can further be interpreted as events that are potential crossing opportunities (in the form of yields and crossable gaps) and those that correspond to non-crossable gaps. Using this stratification, every cell in Exhibit 33 can be categorized as to whether the pedestrian correctly interpreted an event (for example utilized a crossable gap in both lanes) or not. Applying this framework to every cell, a total of five event outcome categories emerge: 1. Correctly Accepted Crossing Opportunity: Pedestrian “GO” in a crossable/safe situation. 2. Falsely Rejected Crossing Opportunity: Pedestrian “NoGO” in a crossable/safe situation. 3. Correctly Rejected Non-Crossable Event: Pedestrian “NoGO” in a non-crossable/unsafe situation. 4. Falsely Accepted Non-Crossable Event: Pedestrian “GO” in a non-crossable/unsafe situation. 5. Inconclusive Event: Pedestrian “GO” in a forced yield condition. The first four categories correspond to a classical 2x2 event matrix that relates the real-world condition to the pedestrian response. The fifth category was introduced, since it is “incon- clusive” whether a forced yield should be interpreted as an acceptable crossing strategy or not. Exhibit 34 summarizes the event classifications for the pre study at the RCW. Exhibit 34 suggests that for the total of 686 events, 23.8% of crossings were correct utilizations of crossing opportunities and 58.6% of events were correctly rejected events. Only 6.0% were classified as missed opportunities and inefficient behav- ior, and 1.3% fell into the “unsafe” category. Also, 10.3% of events were associated with a forced yield and were labeled as inconclusive. Note that any O&M interventions were removed from the dataset prior to analysis (discussed separately) and so none of these forced yields resulted in a truly dangerous situ- ation. However, in the absence of pedestrian and/or driver action, a forced yield can result in a collision. Posttest Pedestrian Behavior at the RCW The installation of the raised crosswalk was expected to assist pedestrians by encouraging more drivers to yield and by generally slowing down the conflicting traffic in the vicinity of the crosswalk. Since the treatment does not involve any form of signalization, the same near–far analysis framework was applied to the post dataset. Exhibit 35 plots the near- and far- lane events (same as pre analysis) for the RCW installation. Exhibit 35 shows that the probability of encountering a yield after the RCW installation increased from 26.7% to 51.3% in the near lane. Further, the rate of yield utilization increased from 54.6% to 92.0%. The presence of the RCW may have led to a modified driver behavior that made it easier for pedestri- ans to detect the yield. In the near lane, 13.8% of yields were forced by pedestrians, which is slightly less than the 16.4% in the pre condition. In the post data collection, pedestrians encountered 131 gaps, and 80 of those were crossable. The likelihood of encountering 123 Pre (n = 686) Near Lane Far Lane Availability Statistics P(Y_Enc) 26.7% 23.8% P(CG_Enc) 25.5% 21.9% Utilization Statistics P(GO|Y) 54.6% 68.7% P(GO|CG) 78.9% 83.3% Exhibit 33. Availability and utilization statistics for pre condition, RCW. Crosswalk Condition Pedestrian Decision Crossable/Safe Non-Cross./Unsafe Inconclusive GO 163 23.8% 9 1.3% 71 10.3% NoGO 41 6.0% 402 58.6% – Exhibit 34. Summary of pedestrian behavior, pre condition, RCW.

a crossable gap from all events was 29.7%, which is similar to the pre study (25.5%). However, the rate of crossable gap utilization increased from 78.9% to 93.8% in the near lane. Exhibit 36 shows the summary availability and utilization statistics. Exhibit 36 shows that the availability of yield crossing opportunities about doubled with the installation of the raised crosswalk, while the availability of gap crossing oppor- tunities remained largely unaffected. However, the rate of uti- lization of both types of opportunities increased drastically, with utilization rates well above 90%. Overall, far fewer events were (had to be) rejected by the pedestrian, as is evident in the summary statistics in Exhibit 37. Exhibit 37 shows that for the total of 269 events, 53.2% of crossings were correct utilizations of crossing opportunities and 21.6% of events were correctly rejected events. Only 1.5% were classified as missed opportunities and inefficient behav- ior, and 2.6% fell into the “unsafe” category. Also, 21.2% of events were associated with a forced yield and were labeled as inconclusive. A comparison of Exhibits 34 and 37 makes evident that the biggest difference between the pre and post data is a drastic reduction of rejected opportunities (reduced from 443 to 62 events). With the introduction of the raised crosswalk, drivers tended to yield more frequently, and many of these yields resulted in crossings. Furthermore, pedestrians seemed more comfortable accepting gaps, resulting in an overall drop of inefficient decisions from 6.0% to 1.5%. As a result of this more assertive behavior, the rate of potentially risky events about doubled, as did the rate of inconclusive events. Note that any O&M interventions were removed from the dataset prior to analysis (discussed separately) and so none of these forced yields resulted in a truly dangerous situation. Performance Statistics for RCW The installation of the RCW is expected to also affect the bottom-line delay and risk performance statistics for the pedes- trians. Delay statistics in Exhibit 38 are provided for pedestrian delay in seconds, defined as the time difference between when the trial started and when the pedestrian initiated the crossing. The exhibit further shows the delay beyond the first oppor- 124 Exhibit 35. Near–far lane effects, post condition, for RCW. RY STY CG FY non-CG Rolling Yield Utilized 27 Non-Utlz. 2 5 Stopped Yield Utilized 12 81 Non-Utlz. 0 6 Forced Yield Utilized 1 19 Non-Utlz. 0 Crossable Gap Utilized 5 75 Non-Utlz. 0 5 Non-Cross. Gap Utilized 5 Non-Utlz. 0 4 0 0 0 28 2 5 11 0 1 5 0 1 0 12 0 0 13 0 0 6 0 0 0 18 0 9 35 0 4 6 0 1 3 1 3 0 1 3 37 0 0 3 0 0 0 0 1 0 0 2 0 0 0 4 0 1 0 0 0 1 0 0 0 2 0 1 7 0 0 2 0 0 0 4 0 2 2 0 0 1 0 0 0 0 1 0 0 2 4 0 0 0 46 Total 24 53 31 73 51 3 6 12 9 7 269 Multiple Events Total Far-Lane Event Near-Lane Event Lane Outcome Rolling Yield Stopped Yield Forced Yield X-Able Gap Non-X. Gap Post (n = 269) Near Lane Far Lane Availability Statistics P(Y_Enc) 51.3% 46.8% P(CG_Enc) 29.7% 31.6% Utilization Statistics P(GO|Y ) 92.0% 96.0% P(GO|CG) 93.8% 95.3% Crosswalk Condition Pedestrian Decision Crossable/Safe Non-Cross./Unsafe Inconclusive GO 143 53.2% 7 2.6% 57 21.2% NoGO 4 1.5% 58 21.6% – Exhibit 36. Availability and utilization statistics for post condition, RCW. Exhibit 37. Summary of pedestrian behavior, post condition, RCW.

tunity (Delay>Min), which was defined as the time difference between the first yield or crossable gap encountered by the pedestrian and the actual crossing initiation. Statistics for all measures are for crossing one leg (two lanes) of the round- about. The statistics shown are calculated from the average crossing performance for each subject. The total sample size was 18 and 13 subjects in the pre and post studies, respectively. Each data point represents the average of 16 approach cross- ings, half at the entry and half at the exit of the roundabout. Exhibit 38 shows that the average pedestrian delay per leg decreased significantly between the pre and post conditions, from 17.0 s to 6.7 s (p = 0.0434). There was no significant dif- ference between the delay experienced at the entry and exit portions of the crossing in the pre study. In the post study, the delay difference of 2.7 s higher average delay at the exit is sig- nificant at p = 0.0440. In addition to the average delay for all participants, it is important to emphasize that some participants experienced much larger delays. The highest average delay was 84.9 s in the pre study, and the single highest delay experienced by a study participant was 115.8 s (not shown). These figures do not include trials that were beyond the 2-min time-out limit. The maximum average delay in the post was only 11.2 s, and the single highest delay in the post condition was 57.4 s. Overall, the 2-min time-out limit was reached 9 times for all subjects in the pre condition and never after installation of the raised crosswalk. With installation of the RCW, the observed range and standard deviation of average delay were reduced, sug- gesting more consistent behavior across subjects. The results for Delay>Min also show a reduction between the pre and post conditions from 3.4 to 2.3 s, but this differ- ence was not statistically significant (p = 0.2117). Overall, the Delay>Min results suggest that the blind pedestrians did not miss a lot of crossing opportunities, but rather were delayed due their infrequent occurrence. Despite these low averages, some pedestrians experienced Delay>Min up to 48.4 s in the pre condition and up to 36.9 s the post case (not shown). The max- imum average Delay>Min were 11.2 and 5.8 s, respectively. Exhibit 39 shows the cumulative distribution of pedestrian delay at the PHB. The 85th percentile delay is highlighted. The exhibit clearly shows a shift of the delay distribution, with pedestrians in the post condition overall experiencing lower delays. The 85th percentile delay across all participants was reduced from 31.0 to 13.4 s. The difference is also evi- dent when looking at the crossing performance of individual 125 a) Observed Delay per Leg (s) Pre Avg. Min. Max. Std. Dev. Entry (n = 18) 15.6 1.5 57.1 15.9 Exit (n = 18) 18.4 3.0 84.9 19.4 Overall (n = 36) 17.0 1.5 84.9 17.6 Post Entry (n = 13) 6.7 3.6 12.2 2.7 Exit (n = 13) 9.4 4.0 18.2 3.7 Overall (n = 26) 8.0 3.6 18.2 3.5 b) Delay>Min (s) Pre Avg. Min. Max. Std. Dev. Entry (n = 18) 3.1 0.0 11.1 3.6 Exit (n = 18) 3.8 0.2 11.2 2.8 Overall (n = 36) 3.4 0.0 11.2 3.2 Post Entry (n = 13) 1.7 0.1 4.3 1.3 Exit (n = 13) 2.8 0.2 5.8 1.7 Overall (n = 26) 2.3 0.1 5.8 1.5 Exhibit 38. Average pedestrian delay statistics for RCW. 0 10 20 30 40 50 60 70 80 90 100 Pe rc en til e Delay (sec.) RCW PRE POST 85%ILE DELAY POST 13.4 sec. PRE 31.0 sec. 0 10 20 30 40 50 60 70 80 90 100 110 120 Exhibit 39. Cumulative delay distribution of pedestrian delay at RCW.

participants. Exhibit 40 shows the 85th percentile delay for all participants in the pre and post condition. Note that partic- ipants 1, 5, 10, 15, and 16 did not participate in the post study. The figure shows that the 85th percentile delay was reduced for every participant in the post condition but also that the effect was greatest for those that experienced high delay in the pre. So, in addition to reducing the overall delay, the RCW also created a more uniform distribution of delay, even for participants with presumably worse travel skills. Exhibit 40 further explores the relationship between 85th percentile delay and the time of day of the participation. From a visual analysis, no trends are observed. The analysis further investigated two parameters intended to describe the efficiency with which a crossing opportunity is utilized. For utilized gaps, the latency is defined as the time between when the last vehicle went through the crosswalk and the time the pedestrian initiated the crossing. For utilized yields, the yield lost time is defined as the time between when a driver first slows down for a yield and the time the crossing is initiated. Note that in some cases, pedestrians may prefer to cross only after a car has come to a full stop (stopped yield) and so some inherent yield utilization time is expected. Exhibit 41 shows statistics for both measures. The latency results in Exhibit 41 suggest that on average pedestrians wait 7.3 s into a crossable gap before initiating the crossing, suggesting inefficiency in decision-making. With the installation of the RCW, the average latency decreases slightly to 5.4 s (p = 0.2767). For the YLT measure, pedestrians in the pre condition lose an average of 2.2 s before crossing in front of a yielding vehicle. However, the average maximum YLT was 9.0 s. In many cases, drivers will not be willing to wait this long, and a high YLT will therefore translate to an increased percentage of missed yields [lower P(GO|Y)]. Note that the minimum YLT is neg- ative, suggesting that some pedestrians forced vehicles to yield (the yield was “utilized” before it occurred). The instal- lation of the RCW had no significant effect on yield lost time. The above measures primarily focus on the efficiency of crossing, and largely ignore the risk experienced by pedestri- ans. While delay and other efficiency measures are used fre- quently by engineers, they fail to capture the human element of crossing risk. The selected surrogate risk measure for this study is the number of times the O&M specialist had to inter- vene in the crossing. Exhibit 42 shows the frequency and rate of O&M interventions for all trials. Exhibit 42 shows a drastic reduction in the occurrence of interventions. The percentage of trials that resulted in an O&M intervention is reduced from 2.8% to zero in the post condition. Following discussion in Ashmead et al. (2005), a 2.8% likelihood of a risky decision will result in a cumulative risk of 67.9% after 40 crossings (for example two crossings a day over 4 weeks with 5 working days per week). In the pre case, the exit lane had a slightly higher intervention rate than the entry, which is consistent with findings at other multi-lane roundabouts. However, given that interventions are very rare events, it is unlikely that the post intervention is an absolute zero, but rather small-enough to where it was not measura- ble at the given sample size. Exhibit 43 explores the distribu- tion of O&M intervention across subjects and by time of day. Subjects that didn’t participate in the post experiment are shown with a negative intervention rate to distinguish them from those with zero interventions that did participate. 126 0 20 40 60 80 100 120 1 3 6 9 11 16 4 7 10 12 14 17 2 5 8 13 15 18 9:00am 11:30am 3:30pm 85 th P e rc en til e De la y (se c. ) Subject RCW PRE POST Exhibit 40. 85th percentile delay by subject – RCW. a) Latency (s) Pre Avg. Min. Max. Std. Dev. Entry (n = 18) 6.5 1.3 15.3 4.0 Exit (n = 18) 8.1 1.9 41.4 8.7 Overall (n = 36) 7.3 1.3 41.4 6.7 Post Entry (n = 13) 4.5 2.3 6.8 1.5 Exit (n = 13) 6.1 2.2 11.2 2.8 Overall (n = 26) 5.4 2.2 11.2 2.4 b) Yield Lost Time (s) Pre Avg. Min. Max. Std. Dev. Entry (n = 18) 3.5 –0.8 9.0 2.5 Exit (n = 18) 0.8 –4.3 8.9 3.3 Overall (n = 36) 2.2 –4.3 9.0 3.2 Post Entry (n = 13) 3.4 1.2 6.2 1.3 Exit (n = 13) 2.4 0.5 3.6 0.9 Overall (n = 26) 2.9 0.5 6.2 1.2 Exhibit 41. Latency and yield lost time statistics for north crosswalk.

RCW Summary In summary, the installation of the raised crosswalk signal resulted in a large reduction of both delay and interventions for all study participants. The relative difference between pre and post studies was greatest for participants that experienced high delays in the pre condition, as the RCW created a more uniform distribution of delay across participants. The RCW further reduced the overall number of events, with much fewer rejected events. As more drivers yield, the relative frequency of gaps encountered decreases. The installation of the RCW further seemed to enhance the efficiency with which crossing opportunities were utilized. The higher utilization rates can potentially be explained by the hypothesis that pedestrians felt more confident with the RCW in place and its effect on drivers. Alternatively, the RCW may have altered driver behavior in a way that made it easier to detect crossing opportunities (e.g., yielding more abruptly and closer to the crosswalk). A learning effect by pedestrians is unlikely because the single-lane roundabout comparison site did not show improvement in behavior. Exhibit 44 presents a summary of the pre and post crossing performance at the raised crosswalk. In summary, the RCW seemed to have an impact on the availability of crossing opportunities in the form of yields and further increased the rate of opportunity utilization for both yields and gaps. The treatment therefore affects both driver behavior (yield availability) and pedestrian behavior (utiliza- tion). The effect on the behavioral parameters was large enough to significantly impact the bottom line in the form of greatly reduced delay and the reduction of O&M interventions to zero. While some pedestrians still encountered some high delay time in the post treatment installation, the overall effect was a reduction in the delay estimate itself as well as in the range and variability of the estimate. The biggest impact was notable for pedestrians who had high delays in the pre condition and whose post-treatment performance was within the range of that of more comfortable and experienced travelers. 127 Exhibit 43. O&M intervention rate by subject and time of day. O&M Interventions – RCW Crosswalk Pre Frequency # of Crossings Percent Entry 3 144 2.1% Exit 5 144 3.5% Overall 8 288 2.8% Post Entry 0 104 0.0% Exit 0 104 0.0% Overall 0 208 0.0% Exhibit 42. O&M interventions for RCW. Performance Measure Pre Post Yield Availability* 25.2% 49.1% Gap Availability* 23.7% 30.7% Yield Utilization* 61.7% 94.0% Gap Utilization* 81.1% 94.5% 85th percentile Delay (s) 31.0 13.4 O&M Interventions 2.8% 0.0% *Average of Near and Far Lane -5.0% 0.0% 5.0% 10.0% 15.0% 20.0% 1 3 6 9 11 16 4 7 10 12 14 17 2 5 8 13 15 18 9:00am 11:30am 3:30pm O &M In te rv en tio ns Subject RCW PRE POST Exhibit 44. Summary of crossing performance pre and post RCW installation.

128 Introduction This section describes analysis results of data collected at the northern crosswalk of the two-lane roundabout in Golden at the intersection of Golden Road and Johnson Road (Exhibit 45). The analysis focus will be on pedestrian-related measures, including the availability and utilization of yield and gaps as well as pedestrian delay and O&M interventions. The measures are defined in the methodology chapter of this report. Because of the two-lane approaches at this site, the docu- ment distinguishes between near-lane and far-lane effects at the crosswalk. These describe the vehicle state in the near and far lane relative to the position of the waiting pedestrian. This section focuses on the north crosswalk with the treat- ment effect of the PHB (also known as a HAWK signal). The analysis presents findings in the pre and post conditions for the studied crosswalk sequentially. The pre study was completed in July 2008, and a total of 18 blind travelers participated in the pre study. The treatment was installed following the pre study, and 13 of the original 18 participants returned for the post experiment in September 2008, allowing approximately 60 days for driver acclimation. Pedestrian Hybrid Beacon Evaluation Pedestrian Hybrid Beacon Treatment Overview A PHB was installed at the northern crosswalk at the two- lane roundabout, as shown in Exhibit 46. The treatment was installed at the existing crosswalk location and was outfitted with an APS device to provide additional assistance to blind study participants. The PHB is different from a conventional signal in that it remains dark for traffic unless a pedestrian presses the call but- ton. When the pedestrian presses the button, the approaching drivers are given a “Flashing Yellow” indication requiring them to reduce speed and be prepared to stop for a pedestrian in the crosswalk. The “Flashing Yellow” is followed by a “Solid Yellow” pro- viding additional emphasis on the need to reduce speed and be prepared to stop. “Solid Yellow” then changes to “Solid Red.” The law requires that drivers come to a complete stop when seeing a “Solid Red” signal indication. When approach- ing drivers see the “Solid Red,” sighted pedestrians see the customary “Walk” signal and may begin to cross. Visually impaired pedestrians hear a speech message saying, “Walk Signal On.” After a few seconds, the vehicle display will switch to a “Flashing Red” indication for the driver, as two red lights wig-wag back and forth. At this time drivers can proceed if the crosswalk to their immediate front is not occupied by any pedestrians. There may still be pedestrians completing their crossing when drivers see the signal turn from “Solid Red” to “Flashing Red.” If a pedestrian is still in the lane, a driver must remain stopped until the path is clear. Per proposed language in the MUTCD, a driver approaching the signal during “Flashing Red” must first come to a stop before proceeding through the crosswalk. The PHB is timed to allow pedestri- ans to cross one side of the street (two lanes) at a time (entry or exit). Once they reach the splitter island, pedestrians will place a second signal call to complete the crossing to the far- side curb. The phasing sequence is outlined in Exhibit 47. The PHB arrangement is intended to aid pedestrians who desire assistance in crossing from or to the median that sep- arates the two directions of traffic, especially when traffic is heavy. It provides visually impaired pedestrians audible infor- mation through an APS device about when the “Walk” sig- nal is on. Pretest Pedestrian Behavior at the PHB Crosswalk The NCHRP Project 3-78A analysis of single-lane cross- ings used a performance evaluation framework that described the availability of crossing opportunities, the rate of utiliza- tion of these opportunities, and the delay and risk associated with the crossings. For a single lane, the yield and gap events P A R T 5 Detailed Two-Lane Roundabout: Golden, CO – PHB

are uniquely defined by the vehicle state in the conflict lane. However, at a two-lane crossing the analysis needs to con- sider the vehicle state in both lanes. The following approach distinguishes between driver behavior in the near lane (the lane closest relative to the position of the pedestrian) and the far lane. Depending on the crossing location (entry/exit and curb/island), the near lane can be the inside or outside lane of the two-lane approach. The analysis defines the vehicle state in the near lane in five event categories: 1. Rolling Yield (RY): Pedestrian encounters a driver who has slowed down for the pedestrian, but has not come to a full stop. 2. Stopped Yield (STY): Pedestrian encounters a driver who has come to a stop, defined as moving at a speed less than 3 mph. 3. Forced Yield (FY): Pedestrian initiates crossing before the vehicle initiated the yield, but then forces the driver to slow down by entering the crosswalk. 4. Crossable Gap (CG): Pedestrian encounters a gap large enough to safely cross the street without the need for a driver yield. A crossable gap is defined as the crossing width divided by 3.5 ft/s walking speed plus 2 s for start time and safety buffer. 5. Non-Crossable Gap (non-CG): Pedestrian encounters a gap between vehicles shorter than the crossable gap threshold. The vehicle state in the far lane is defined relative to the near-lane condition in five principal categories: rolling yield, stopped yield, crossable gap, non-crossable gap, and multiple events. The last category indicates that more than one event took place in the far lane during one near-lane event. For example, several cars could have gone through the far lane during one large gap in the near lane. For multiple events, the last event in the sequence is considered for analysis. Exhibit 48 shows the near-lane and far-lane effects for the pre condition at the PHB crosswalk. The near lane event outcomes are clas- sified as to whether they are utilized by the pedestrians. Exhibit 48 shows that from a total of 603 events, partici- pants encountered 194 yield events in the near lane (32.2%) and did not utilize 33 rolling yields and 39 stopped yields. The corresponding overall rate of yield utilization in the near lane is 62.9%. A subset of these near-lane yields (15.5%) were 129 Ph ot o by G oo gl e Ph ot o by L ee R od eg er ts Exhibit 45. Aerial view of roundabout. Exhibit 46. Pedestrian hybrid beacon. Exhibit 47. Phasing sequence for PHB (source: MUTCD).

yields forced by the pedestrian. The analysis of the far-lane event shows that a majority of the non-utilized yields were attributable to either non-crossable gaps or multiple events in the far lane. Overall, only 22 events with yields in both lanes were non-utilized, including yields in the multiple events cat- egory, which indicates very good overall judgment of the per- ceived risk. Similarly, of the 22 non-utilized crossable gaps, 19 had non- crossable gaps in the far lane. In total, pedestrians encountered 28.7% crossable gaps in both the near and far lanes. In the near lane, 87.3% of crossable gaps were utilized, along with 89.0% in the far lane. The crossing opportunity availability and uti- lization statistics are summarized in Exhibit 49. The results in Exhibit 49 can be interpreted as events that are potential crossing opportunities (in the form of yields and crossable gaps) and as those that correspond to non-crossable gaps. Using this stratification, every cell in Exhibit 49 can be categorized as to whether the pedestrian correctly interpreted an event (for example, utilized a crossable gap in both lanes) or not. Applying this framework to every cell, a total of five event outcome categories emerge: 1. Correctly Accepted Crossing Opportunity: Pedestrian “GO” in a crossable/safe situation. 2. Falsely Rejected Crossing Opportunity: Pedestrian “NoGO” in a crossable/safe situation. 3. Correctly Rejected Non-Crossable Event: Pedestrian “NoGO” in a non-crossable/unsafe situation. 4. Falsely Accepted Non-Crossable Event: Pedestrian “GO” in a non-crossable/unsafe situation. 5. Inconclusive Event: Pedestrian “GO” in a forced yield condition. The first four categories correspond to a classical 2x2 event matrix that relates the real-world condition to the pedestrian response. The last category applies to events associated with forced yields. A forced yield may involve some risk if neither driver or pedestrian acts to avoid a collision. However, many participants appeared to be deliberately forcing yields, which makes it difficult to discern the level of true risk from the data. Exhibit 50 summarizes the event classification for the pre study at the RCW crosswalk. Exhibit 50 suggests that overall, 32.2% of crossings are clas- sified as correct utilizations of crossing opportunities and 49.4% of events were correctly rejected events. Only 4.5% were classified as missed opportunities and inefficient behav- ior, and 1.0% fell into the potentially risky category (after O&M interventions were removed from the data). In addi- tion, 12.9% fell into the inconclusive category and were asso- ciated with a forced yield in either the near or far lane. Posttest Blind Pedestrian Behavior at the PHB Crosswalk With the installation of the PHB, the analysis framework has to be modified from the pre condition. Pedestrians now encounter a signal indicating that the signal phase is either W, FDW, or DW. Blind pedestrians hear a locator tone during the DW and FDW phases and a speech message during the W phase. The appropriate crossing behavior is therefore 130 Exhibit 48. Near–far lane effects, pre condition, for PHB crosswalk. RY STY CG FY non-CG Rolling Yield Utilized 18 Non-Utlz. 3 33 Stopped Yield Utilized 2 74 Non-Utlz. 7 39 Forced Yield Utilized 2 30 Non-Utlz. 0 Crossable Gap Utilized 8 151 Non-Utlz. 0 22 Non-Cross. Gap Utilized 5 Non-Utlz. 6 1 0 0 4 24 6 4 7 1 10 4 0 1 0 10 0 8 22 0 0 3 0 1 1 21 1 11 82 1 13 6 0 2 24 0 19 0 0 17 185 1 0 0 1 2 1 1 2 0 2 0 0 0 0 4 0 0 2 0 0 1 0 0 0 8 0 3 19 1 2 1 0 1 0 3 0 1 9 0 0 1 0 0 0 0 5 0 0 2 13 0 0 0 231 Total 29 61 44 138 246 9 7 35 14 20 603 Multiple Events Total Far-Lane Event Near-Lane Event Lane Outcome Rolling Yield Stopped Yield Forced Yield X-Able Gap Non-X. Gap Exhibit 49. Summary of availability and utilization statistics, PHB crosswalk, pre. Pre (n = 603) Near Lane Far Lane Availability Statistics P(Y_Enc) 32.2% 27.2% P(CG_Enc) 28.7% 28.7% Utilization Statistics P(GO|Y) 62.9% 75.0% P(GO|CG) 87.3% 89.0%

linked to the signal indication, and the analysis of the concur- rent vehicle states becomes a secondary item of interest. Exhibit 51 shows the frequency of crossing initiation for the (blind) pedestrian relative to PHB signal phases. The results show that only 36.7% of pedestrians crossed in the intended “Walk” phase and that most (39.0%) actually ini- tiated the crossing just before the “Walk” phase (and the APS alert) in the vehicular solid yellow. In other words, they began to cross following their pressing the call button but prior to the APS message. Further, 11% crossed even earlier, in the vehi- cle “Flashing Yellow” phase, and 13.3% didn’t cross until the “Flashing Don’t Walk” phase. Overall, only three times (out of 208 lane crossings) did pedestrians not cross in the first cross- ing phase and have to reactivate the signal. These figures suggest that the study participants rely heavily on their own personal judgment, even with the signal beacon in place. Pedestrians tended not to cross in “Walk” if they were unsure about whether vehicles had in fact stopped. Even when the APS confirmed to the blind pedestrian that a “Red” signal indication was being presented to an approaching driver, some would still not cross until they were confident that it was safe to do so. Similarly, they would readily cross before the “Walk” phase if they perceived a crossing opportunity. To illustrate this point, Exhibit 52 shows the near and far-lane event outcomes (same as pre analysis) by signal phase. The results show that almost all of the early crossing events that started in “Flashing” or “Solid Yellow” phases were asso- ciated with crossing opportunities. The majority of these were yields in both lanes (68 out of 105 events, or 64.7%), although 52.9% of those were associated with a forced yield in at least one of the lanes. The rest were some combination of yields and crossable gap, with only one exception, where a non-crossable gap existed in the far lane. No rejected opportunities were observed in the two early phases, suggesting great efficiency for those pedestrians who chose to cross there. However, 57.1% of the crossings were associated with a forced yield, which may indicate some level of risk depending on pedestrian and driver awareness of the situation. Events in the “Walk” phase include a significant number of rejected events (28.0%), mostly in the form of non-crossable gaps. This suggests that a portion of drivers did not comply with the signal indication, a pattern that is explored in more detail later. Some events (4.7%) suggest inefficient behavior (i.e., failure to cross during “Walk” phase), pointing to uncer- tainty in crossing for some pedestrians. The 32.7% “inconclu- sive” events were all associated with forced yields, indicating that the pedestrian initiated the crossing before the driver ini- tiated the yield. Presumably, these events are acceptable at a signal; however, there is still some degree of risk if those drivers had been unaware of the pedestrian action. 131 Crosswalk Condition Pedestrian Decision Crossable/Safe Non-Cross./Unsafe Inconclusive GO 194 32.2% 6 1.0% 78 12.9% NoGO 27 4.5% 298 49.4% – Exhibit 50. Summary of pedestrian behavior, pre condition, PHB crosswalk. 11.0% 39.0% 36.7% 13.3% 0 10 20 30 40 50 60 70 80 90 Flashing Yellow Yellow Red Flashing Red / Don't Walk / Don't Walk / Walk / Flashing DW Fr e qu en cy Signal Phase n = 210 Exhibit 51. Blind pedestrian crossings at PHB by signal phase (% of all crossings).

132 d 4 4 7 2 0 0 . 0 . . . . . . . . . . . . . d 1 14 4 2 1 0 0 . . . . . . . . . . . d 6 5 8 5 0 0 . 0 . . . . . . . . . . . . . d 2 4 7 4 0 1 . 1 . . . . . . . . . . . . . d . . . . . . . . . . . . . . . . . . . . . d 4 5 5 2 0 0 1 0 0 . . 0 2 0 0 4 0 2 0 0 . d 4 10 2 3 0 0 1 0 0 . 0 1 0 0 1 0 0 0 0 . d 4 3 5 4 0 1 1 1 2 . . . . . . . . . . . . d 1 6 5 4 0 0 0 1 2 . . . . . . . . . . . . d . . . . . . . . . . 2 0 0 0 18 0 0 0 0 . d 0 3 0 0 . 1 0 1 1 . . . . . . . . . . . . d 0 4 0 0 0 0 0 1 . . . . . . . . . . . d 0 3 3 2 . 1 1 1 1 . . . . . . . . . . . . d 1 0 1 1 . 0 0 1 1 . . . . . . . . . . . . d . . . . . . . . . . . . . . . . . . . . . RY STY CG FY non-CG Rolling Yield Utilized 0 Non-Utlz. 0 Stopped Yield Utilized 5 Non-Utlz. 0 Forced Yield Utilized 10 Non-Utlz. 0 Crossable Gap Utilized 8 Non-Utlz. 0 Non-Cross. Gap Utilized 0 Non-Utlz. . . . . 1 2 . . 1 5 . . . 0 1 2 4 6 . . . . . . . . . . . . . 0 0 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 Total 0 2 8 12 1 0 0 0 0 0 0 23 17 0 22 0 24 0 19 0 0 0 13 27 26 13 1 1 0 1 0 0 82 17 8 20 2 21 0 19 0 0 20 15 27 17 13 23 1 5 2 4 0 107 6 0 5 0 12 0 5 0 0 0 1 10 4 3 0 2 1 3 4 0 28 Phase = Yellow Phase = Flashing Don't Walk/Flashing Red Phase = Walk/Red Rolling Yield Stopped Yield Forced Yield X-Able Gap Total Phase = Flashing Yellow Non-X. Gap Multiple EventsNear-Lane Event Near-Lane Outcome RY STY CG FY non-CG Rolling Yield Stopped Yield Forced Yield Crossable Gap Non-Cross. Gap Total 0 Rolling Yield Stopped Yield Forced Yield X-Able Gap Total Non-X. Gap Multiple EventsNear-Lane Event Near-Lane Outcome RY STY CG FY non-CG Rolling Yield Stopped Yield Forced Yield Crossable Gap Non-Cross. Gap Total 0 Rolling Yield Stopped Yield Forced Yield X-Able Gap Total Non-X. Gap Multiple EventsNear-Lane Event Near-Lane Outcome RY STY CG FY non-CG Rolling Yield Stopped Yield Forced Yield Crossable Gap Non-Cross. Gap Total 0 Rolling Yield Stopped Yield Forced Yield X-Able Gap Total Non-X. Gap Multiple EventsNear-Lane Event Near-Lane Outcome Utilize Non-Utlz Utilize Non-Utlz Utilize Non-Utlz Utilize Non-Utlz Utilize Non-Utlz Utilize Non-Utlz Utilize Non-Utlz Utilize Non-Utlz. Utilize Non-Utlz Utilize Non-Utlz Utilize Non-Utlz Utilize Non-Utlz Utilize Non-Utlz Utilize Non-Utlz Utilize Non-Utlz Exhibit 52. Near-far lane effects post condition for PHB crosswalk by signal phase.

Those pedestrians who rejected opportunities in the “Walk” phase ultimately crossed in the “Flashing Don’t Walk” phase. Again, the majority of events here are related to yield events with drivers stopped at the signal. No rejected opportunities or inefficient events were observed, but again many events fall into the inconclusive category. Exhibit 53 summarizes the avail- ability and utilization statistics for the post treatment installa- tion data. Since it appeared from the analysis above that most pedestrians crossed independently of the signal indication, the results are presented in light of the near–far lane framework dis- cussed above. This also ensures that the numbers are directly comparable to the pre condition results. The summary statistics suggest a large increase in the availability of both yields and gaps from the pre condition (Exhibit 49), as well as more efficient utilization of these crossing opportunities. Exhibit 54 shows a summary of all events for the post condition by signal phase. Exhibit 54 shows that crossing performance in the early phases (“Flashing Yellow” and “Solid Yellow”) was generally characterized by mostly correctly accepted crossing opportu- nities as well as a large portion of forced yields (inconclusive events). Virtually no risky or inefficient events were observed during these phases. For the intended crossing phase (Red/Walk), 39.3% of cross- ings were classified as correct utilizations of crossing oppor- tunities, and 23.4% of events were correctly rejected events. Further, 4.7% were classified as missed opportunities and inefficient behavior, and none were observed in the potentially risky category. Similar to the early phases, many events (32.7%) fell into the inconclusive category and were associated with a forced yield in either the near or far lane. Those pedestrians that waited to initiate crossing during the “Flashing Red” phase mostly made correct “GO” decisions, but more than half were once again associated with forced yields. A notable difference between Exhibits 49 and 53 for pre and post data is a drastic reduction in the rejected opportunities. With the introduction of the signal, drivers tended to yield much more frequently, and many of these yields resulted in crossings. The proportion of inefficient decisions was reduced slightly, as was the rate of potentially risky events. The rate of inconclusive events saw a large increase. As discussed above, these events are associated with forced yields, where none resulted in an O&M intervention. It is unclear whether a forced yield at a signal can truly be classified as a risky event since drivers are presumably prepared to stop given that the signal has been activated. However, in combination with red- light running events, some risk may remain. The discussion below examines in more detail the risk and delay performance measures. Performance Statistics at the PHB Crosswalk The changed pedestrian and driver behavior that took place with the introduction of the PHB affects the delay and risk per- formance measures. Delay statistics in Exhibit 55 are provided for pedestrian delay in seconds, defined as the time differ- ence between the time a trial started and when the pedestrian 133 Post (n = 242) Near Lane Far Lane Availability Statistics P(Y_Enc) 76.4% 68.6% P(CG_Enc) 70.2% 19.8% Utilization Statistics P(GO|Y) 92.4% 97.6% P(GO|CG) 100.0% 100.0% Exhibit 53. Summary of availability and utilization statistics, PHB crosswalk, post. Crosswalk Condition Pedestrian Decision Crossable/Safe Non-Cross./Unsafe Inconclusive Phase = Flashing Yellow/Don’t Walk (n = 23) GO 5 21.7% 0 0.0% 18 78.3% NoGO 0 0.0% 0 0.0% – 0.0% Phase = Yellow/Don’t Walk (n = 82) GO 39 47.6% 1 1 .2% 42 51.2% NoGO 0 0.0% 0 0.0% – 0.0% Phase = Red/Walk (n = 107) GO 42 39.3% 0 0.0% 35 32.7% NoGO 5 4.7% 25 23.4% – 0.0% Phase = Flashing Red/Flashing Don’t Walk (n = 28) GO 12 42.9% 0 0.0% 16 57.1% NoGO 0 0.0% 0 0.0% – 0.0% Exhibit 54. Summary of pedestrian behavior post condition, PHB crosswalk.

initiated the crossing. The exhibit further shows the delay beyond the first opportunity (Delay>Min), which was defined as the time difference between the first yield or crossable gap encountered by the pedestrian and the actual crossing initia- tion. The crossable gap definition assumed crossing of two lanes at one leg of the roundabout. All statistics shown are cal- culated from the average performance of each individual sub- ject. The sample sizes in the pre and post conditions are 18 and 13 participants, respectively. Exhibit 55 shows that the average pedestrian delay per leg in the post condition decreased significantly from that in the pre condition, from 16.0 s to 4.2 s (p = 0.0007). There was no sig- nificant difference between the delay experienced at the entry and exit portions of the crossing in either study. In addition to reporting the average delay for all participants, it is important to emphasize that some individual participants experienced much larger delays. The highest average delay for a participant was 46.5 s in the pre and 14.6 s in the post case. However, the single highest delay experienced by a study participant was 100.2 s (not shown in the exhibit). The single highest delay in the post condition was 56.3 s, indicating that some pedestrians did not cross during the first “Walk” phase. The reported delay figures further do not include trials that were terminated when the subject’s wait time exceeded the 2-min time-out limit. Overall, the 2-min time-out limit was reached in 3 of 288 lane crossings for all subjects in the pre condition and never with the PHB present (208 lane crossings). The results for Delay>Min also show a significant reduc- tion between the pre and post conditions from 4.5 s to 1.4 s (p = 0.0044). Overall, the Delay>Min results suggest that the blind pedestrians did not miss many crossing opportunities. Despite these low averages, some pedestrians experienced Delay>Min of up to 33.0 s in the pre condition and up to 11.4 s the post case (not shown). The highest average Delay>Min values were 11.9 s and 3.1 s, respectively. Exhibit 56 shows the cumulative distribution of pedestrian delay at the PHB. The 85th percentile delay is highlighted. The exhibit clearly shows a shift in the delay distribution, with pedestrians in the post condition experiencing much lower delays. The 85th percentile delay was reduced from 29.8 s to 134 a) Observed Delay per Leg (s) Pre Avg. Min. Max. Std. Dev. Entry (n = 18) 14.9 2.3 36.5 10.1 Exit (n = 18) 17.1 3.6 46.5 12.0 Overall (n = 36) 16.0 2.3 46.5 11.0 Post Entry (n = 13) 5.9 2.6 14.6 3.2 Exit (n = 13) 5.8 3.5 11.7 2.4 Overall (n = 26) 5.8 2.6 14.6 2.8 b) Delay>Min (s) Pre Avg. Min. Max. Std. Dev. Entry (n = 18) 4.5 0.2 11.9 3.5 Exit (n = 18) 3.9 0.0 11.8 3.9 Overall (n = 36) 4.2 0.2 11.9 3.7 Post Entry (n = 13) 1.4 0.2 3.1 1.0 Exit (n = 13) 1.4 0.2 3.0 0.9 Overall (n = 26) 1.4 0.2 3.0 0.9 Exhibit 55. Average pedestrian delay statistics for PHB crosswalk. 0 10 20 30 40 50 60 70 80 90 100 Pe rc en til e Delay (sec.) PHB PRE POST 85%ILE DELAY POST 8.7 sec. PRE 29.8 sec. 0 10 20 30 40 50 60 70 80 90 100 110 120 Exhibit 56. Cumulative distribution of pedestrian delay at PHB crosswalk.

8.7 s. The difference is also evident when examining the cross- ing performance of individual participants. Exhibit 57 shows the 85th percentile delay for all participants in the pre and post condition. Note that subjects labeled 1, 5, 10, 15, and 16 did not participate in the post study. Exhibit 57 shows that the 85th percentile delay was reduced for every participant in the posttest condition. Further, the effect appeared to be greatest for those subjects who experi- enced high delays in the pre study. So, in addition to reduc- ing the overall delay, the PHB also created a more uniform distribution of delay, even for participants with presumably modest travel skills. The data in Exhibit 57 are arranged by the time of day the subjects participated in the study. A visual comparison does not show a significant effect on performance by time of day. The team further investigated two parameters intended to describe the efficiency with which a crossing opportunity is utilized. For utilized gaps, the latency is defined as the time dif- ference between a vehicle entering the crosswalk and the time the pedestrian initiated the crossing. For utilized yields, the YLT is defined as the time difference between the driver first slowing down for a yield and the time the crossing is initiated. Note that in some cases, pedestrians may prefer to cross only after a car has come to a full stop (stopped yield) and so some inherent yield utilization time is expected. Exhibit 58 shows statistics for both measures in the pre and post cases. The latency results in Exhibit 58 suggest that on average pedestrians wait 5.9 s into a crossable gap before initiating the crossing, suggesting inefficiency in decision-making. With the installation of the PHB, the average latency decreases slightly to 4.8 s; however, that difference is not statistically sig- nificant (p = 0.2363). Both the range and standard deviation of the latency estimate are reduced in the post condition. For the YLT measure, pedestrians in the pre condition waited an average of 2.7 s before crossing in front of an already yielding vehicle. However, the maximum average YLT was 9.9 s, and individual YLT observations were even higher. In many cases, drivers may not be willing to wait this long and a high YLT will therefore translate to an increased percentage of missed yields [lower P(GO|Y)] or even an unsafe condition where both driver and pedestrian proceed simulta- neously. Note that the YLT can be negative, suggesting that some pedestrians forced vehicles to yield. After installation of the PHB, many pedestrians crossed with the signal and thus before the vehicles had yielded, resulting in an average YLT of –0.4 s. The maximum average YLT also decreased to 3.3 s, suggesting a quicker response to the yielding vehicle. The above measures primarily focus on the efficiency of crossing and largely ignore the explicit risk experienced by pedestrians. While delay and other efficiency measures are used frequently by engineers, they fail to capture the human element of crossing risk. The selected surrogate risk measure for this study is the number of times the O&M specialist had to intervene in the crossing. Exhibit 59 shows the frequency and rate of O&M interventions for all trials. Exhibit 59 shows a drastic reduction in the occurrence of interventions. The percentage of trials that resulted in an O&M intervention is reduced from 2.4% to zero in the post condition. In the pre case, the entry lane actually had a higher intervention rate than the exit, which is contrary to findings at other multi-lane roundabouts (Guth et al. 2005). Follow- ing discussion in Ashmead et al. (2005), a 2.4% likelihood of a risky decision will result in a cumulative risk of 62.2% after 135 0 20 40 60 80 100 120 1 11 16 4 7 10 12 14 17 2 5 8 13 15 18 9:00am 11:30am 3:30pm 85 th P e rc en til e D el ay (s ec .) Subject PHB PRE POST 3 6 9 Exhibit 57. 85th percentile delay by subject – PHB crosswalk. a) Latency (s) Pre Avg. Min. Max. Std. Dev. Entry (n = 18) 5.0 1.4 10.1 2.9 Exit (n = 18) 7.0 3.0 14.9 3.5 Overall (n = 36) 5.9 1.4 14.9 3.3 Post Entry (n = 13) 4.4 2.3 7.9 1.8 Exit (n = 13) 5.2 2.9 8.2 1.7 Overall (n = 26) 4.8 2.3 8.2 1.7 b) Yield Lost Time (s) Pre Avg. Min. Max. Std. Dev. Entry (n = 18) 2.7 -1.7 9.9 3.2 Exit (n = 18) 1.2 -4.9 8.7 3.3 Overall (n = 36) 1.9 -4.9 9.9 3.3 Post Entry (n = 13) -0.4 -3.2 3.3 1.9 Exit (n = 13) -0.3 -5.1 2.7 2.4 Overall (n = 26) -0.4 -5.1 3.3 2.1 Exhibit 58. Latency and yield lost time statistics for PHB crosswalk.

40 crossings (for example two crossings a day over 4 weeks with 5 working days per week). However, given that inter- ventions are very rare events, it is unlikely that the post inter- vention is an absolute zero, but rather is small enough to where it was not measurable at the given sample size. Exhibit 60 explores the distribution of interventions by sub- ject and by time of day. Subjects who didn’t return for the post experiment are shown with negative intervention rates to dis- tinguish them from participants with zero interventions. The intervention rates show no trend by time of day. The figure makes evident that several participants didn’t experience any interventions even in the before case at the given sample size of 16 lane crossings. Given the rare nature of the intervention measure, a zero rate should not be interpreted as a perfectly safe crossing. Driver Behavior at the PHB In the evaluation of the PHB, an important question of interest to traffic engineers is the effect of the signal on vehicle traffic flow. The driver behavior analysis described herein has two main components. First, the behavior of drivers relative to the signal phases is intended to capture driver understand- ing of and compliance with the signal indication. Second, the impact of the PHB installation on pedestrian-induced vehicle queues at the crosswalk is examined. Driver understanding of and compliance with the PHB can be evaluated by relating the driver stopping behavior to the indicated signal phase. Exhibit 61 shows a summary of 426 vehi- cle events that were observed during and just after the trial as a function of PHB signal phase. The exhibit shows the number of drivers who yielded (rolling or stopped yield) in each of five signal phases: “Blank,” “Flashing Yellow,” “Solid Yellow,” “Solid Red,” and “Flashing Red.” It then relates all vehicle events to the phase that was active when the vehicle crossed the plane of the crosswalk. The exhibit further contains a record of all vehicles that did not yield. The results in Exhibit 61 show that many drivers who encountered a pedestrian at a crosswalk yielded even before the signal was activated, while others didn’t stop at all, even when the signal was in the solid red phase. The events include all drivers who in some way interacted with the PHB signal or the pedestrian. The exhibit does not include any events that occurred before the signal was activated or after the trial was completed. Exhibit 62 plots two categories of driver behavior for each signal phase: (1) vehicles stopped or stopping, and (2) vehicles proceeding through the crosswalk. The exhibit shows that 34.1% proceeded through the cross- walk in “Flashing Yellow,” which is permitted behavior. As the signal changes to solid yellow, still 11.4% of drivers proceed through the crosswalk, which is allowable if the vehicles were too close to the crosswalk to come to a stop. However, even dur- ing the “Solid Red,” 12.6% of observed vehicles proceeded through the crosswalk. This figure is a concern, since drivers are legally required to stop for the red signal indication and because pedestrians expected a crossing opportunity. Driver behavior during “Flashing Red” shows that almost half of the drivers (48.2%) remained stopped, suggesting some inefficiency in driver behavior in response to the PHB. The second part of the analysis focuses on the impact of the PHB installation on vehicle queues. Exhibit 63 shows the sta- tistics for the maximum vehicle queue lengths in the pre and 136 O&M Interventions – PHB Crosswalk Pre Frequency # of Crossings Percent Entry 5 144 3.5% Exit 2 144 1.4% Overall 7 288 2.4% Post Entry 0 104 0.0% Exit 0 104 0.0% Overall 0 208 0.0% Exhibit 59. O&M interventions for PHB crosswalk. -5.0% 0.0% 5.0% 10.0% 15.0% 20.0% 1 3 6 9 11 16 4 7 10 12 14 17 2 5 8 13 15 18 9:00am 11:30am 3:30pm O &M In te rv en tio ns Subject PHB PRE POST Exhibit 60. O&M interventions by subject and by time of day.

post conditions. The maximum queue length was defined as the longest pedestrian-induced queue length that was observed during or just after a pedestrian crossing. Queues were mea- sured relative to the crosswalk and therefore do not include additional vehicles that were waiting to enter the roundabout downstream of the crosswalk (at the entry). Vehicle queues are combined for both lanes since no significant difference was observed between queues in the inside and outside lanes. Vehicle queue statistics are shown separately for entry and exit lanes. Exhibit 63 shows that the average maximum queue length increased from 2.3 to 5.0 vehicles at the entry and from 1.5 to 3.9 vehicles at the exit over both approach lanes. The increases in average maximum queues are significant at p < 0.0001. With available queue storage of two vehicles (one per lane) at the exit leg, it is evident that the maximum queue sometimes spilled back into the circulating lane, although the average queue is expected to be less than the reported max queue. The queue spillback effect is also evident in Exhibit 64, which shows the cumulative distributions of maximum queue lengths. The dashed line in Exhibit 64b represents the available queue stor- age at the exit leg. Exhibit 64 shows a shift in the cumulative queue distribu- tion toward higher queues associated with the installation of the PHB. The largest effect is a significant reduction in the occurrence of zero queues in the post condition, which results in the large discrepancy at queue length equal to zero. However, 137 Yielding Vehicles and Phase Yielding Is Initiated Signal Phase Blank Flash Y Solid Y Solid R Flash R TOTAL Non- Yielding Vehicles Flash Yellow 3 2 – – – – – 5 39 Solid Yellow 3 2 – 5 15 Solid Red 6 3 0 4 0 13 15 Flash Red 20 20 31 15 15 101 72 Cr os si ng P ha se Blank 13 18 41 36 53 161 n/a TOTAL 45 45 76 51 68 285 141 Total Vehicle Events 426 Exhibit 62. Evaluation of driver behavior at PHB. Exhibit 61. Vehicle events by yielding and stopping phases at PHB. 65.9% 88.6% 87.4% 48.2% 34.1% 11.4% 12.6% 51.8% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Flash Yellow Solid Yellow Solid Red Flash Red Signal Phase Proceeding through Crosswalk Stopped or Stopping Maximum Vehicle Queues for Both Lanes (in Vehicles) Pre Avg. Min. Max. Std. Dev. Entry (n = 104) 2.3 0.0 10.0 2.4 Exit (n = 104) 1.5 0.0 9.0 1.8 Total (n = 208) 1.9 0.0 10.0 2.1 Post Entry (n = 104) 5.0 0.0 19.0 3.9 Exit (n = 104) 3.9 0.0 18.0 3.0 Total (n = 208) 4.4 0.0 19.0 3.5 Exhibit 63. Maximum queue length statistics for PHB installation.

it is evident that very few long queues were observed in either the pre or post condition. With two lanes of storage, any total queue greater than two vehicles at the exit leg will cause some spillback into the circle, as shown by the dashed line. With the installation of the PHB, that proportion of maximum queues greater than two vehicles increased from 29.8% to 69.2%. However, the average queue is expected to be much lower, so that the overall effect of the PHB installation on vehicle queues is considered to be marginal. In fact, a determined yielder is likely to cause similar if not more delay to a driver waiting at the efficient PHB signalization scheme, as evident by some long queues observed in the pre study. PHB Crosswalk Summary In summary, the installation of the PHB or HAWK signal resulted in a large reduction in delay and elimination of O&M interventions for all study participants. The relative difference between pre and post studies was greatest for participants that experienced high delays in the pre condition since the PHB cre- ated a more uniform distribution of delay across participants. The PHB further reduced the overall number of pedestrian– vehicle interaction events, with far fewer rejected crossing opportunities. The reason for this was that drivers yielded (stopped at the light), thereby reducing the number of gaps encountered. Most drivers complied with the signal indication, although there was evidence for both misunderstanding (waiting until “Blank” to proceed) and non-compliance (proceeding through a red signal) on the part of drivers. It is expected that these num- bers may improve with additional public information material or enforcement. The installation of the PHB caused a marginal increase in vehicle queuing, although it is difficult to extra- polate that effect to higher-volume roundabouts. The analysis did confirm that queues caused by determined yielders can approach queues caused by the signal. Further, since many driv- ers did not proceed through the “Flashing Red,” the post queues are longer than expected with the PHB scheme. The impact on queues is therefore expected to be reduced with improved pub- lic education and driver understanding of the PHB. 138 Exhibit 64. Cumulative distribution. b) Exit Leg a) Entry Leg Queue Length in Both Lanes (in Vehicles) PRE POST 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Cu m u la tiv e Fr eq ue nc y Queue Length in Both Lanes (in Vehicles) 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Cu m u la tiv e Fr eq ue nc y PRE POST

Overall, the installation of the PHB greatly increased the availability and utilization of crossing opportunities, which is reflected in a reduction in pedestrian delay. The PHB further reduced O&M interventions to zero, suggesting enhanced safety performance. Exhibit 65 summarizes these key metrics for the PHB evaluation. But even given the improved pedestrian performance and the marginal vehicle impact, care needs to be taken extra- polating these results to higher-volume scenarios or round- abouts with different geometry. The PHB does appear to be a viable treatment for two-lane roundabouts, but it needs to be combined with pedestrian and driver education, as well as enforcement, to maximize its impact. 139 Performance Measure Pre Post Yield Availability* 29.7% 72.5% Gap Availability* 28.7% 45.0% Yield Utilization* 68.9% 95.0% Gap Utilization* 88.2% 100.0% 85th Percentile Delay (s) 29.8 8.7 O&M Interventions 2.4% 0.0% *Average of near and far lane Exhibit 65. Summary performance statistics pre and post PHB installation.

Next: Appendix N - IRB Approval and Consent Forms »
Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities Get This Book
×
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

TRB’s National Cooperative Highway Research Program (NCHRP) Report 674: Crossing Solutions at Roundabouts and Channelized Turn Lanes for Pedestrians with Vision Disabilities explores information related to establishing safe crossings at roundabouts and channelized turn lanes for pedestrians with vision disabilities.

Appendices B through N to NCHRP Report 674 were published as NCHRP Web-Only Document 160. The Appendices included in NCHRP Web-Only Document 160 are as follows:

• Appendix B: Long List of Treatments

• Appendix C: Team Treatment Survey

• Appendix D: Details on Site Selection

• Appendix E: Details on Treatment and Site Descriptions

• Appendix F: Details on PHB Installation

• Appendix G: Participant Survey Forms

• Appendix H: Details on Team Conflict Survey

• Appendix I: Details on Simulation Analysis Framework

• Appendix J: Details on Accessibility Measures

• Appendix K: Details on Delay Model Development

• Appendix L: Details on Roundabout Signalization Modeling

• Appendix M: Use of Visualization in NCHRP Project 3-78A

• Appendix N: IRB Approval and Consent Forms

On August 17, 2011, TRB co-sponsored a web briefing or "webinar" that presented information about the report. View the webinar page for more information and a link to the recorded webinar.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!