National Academies Press: OpenBook

Multimodal Level of Service Analysis for Urban Streets (2008)

Chapter: Chapter 4 - Data Collection

« Previous: Chapter 3 - Literature Review
Page 32
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 32
Page 33
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 33
Page 34
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 34
Page 35
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 35
Page 36
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 36
Page 37
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 37
Page 38
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 38
Page 39
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 39
Page 40
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 40
Page 41
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 41
Page 42
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 42
Page 43
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 43
Page 44
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 44
Page 45
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 45
Page 46
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 46
Page 47
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 47
Page 48
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 48
Page 49
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 49
Page 50
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 50
Page 51
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 51
Page 52
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 52
Page 53
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 53
Page 54
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 54
Page 55
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 55
Page 56
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 56
Page 57
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 57
Page 58
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 58
Page 59
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 59
Page 60
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 60
Page 61
Suggested Citation:"Chapter 4 - Data Collection." National Academies of Sciences, Engineering, and Medicine. 2008. Multimodal Level of Service Analysis for Urban Streets. Washington, DC: The National Academies Press. doi: 10.17226/14175.
×
Page 61

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.

32 The literature review found that various methods have been used to measure traveler perceptions of quality of ser- vice (i.e., field surveys, video laboratory surveys, simulator surveys, telephone surveys, and web surveys). The literature review revealed the wide range of customer satisfaction meas- urements used and the wide range of variables that re- searchers had determined to be critical for predicting or measuring traveler perceptions of quality of service. Some of the differences could be attributed to differences in survey methods. Other differences could be attributed to differences in the situations to which the survey participants were exposed. Still other differences could be attributed to differ- ences in how quality of service was defined (or left undefined) for the participants. In addition, all surveys were limited to a single metropolitan region, so it was not possible to rule out the potential effects of geographic location on the reported LOS models. The objective of the data collection task was, therefore, to develop and execute a set of quality of service surveys that could be uniformly and consistently implemented across all modes and in several different metropolitan areas of the United States. All prior quality of service surveys had been limited to a single site in a single urban area. One of the major purposes of the new data collection under NCHRP Project 3-70 was to gather data using a consistent method across multiple urban areas to determine if LOS perceptions vary significantly across urban areas of the United States. The data collection task of this project was conducted in two phases. Various data collection methods were pilot tested during Phase 1. The data collection effort for the project was completed in Phase 2. 4.1 Selection of QOS Survey Method Several different methods have been used in the literature to measure traveler perceptions of satisfaction. These meth- ods include traveler intercept surveys, field laboratory stud- ies, and video laboratory studies. Introductory material on customer satisfaction survey techniques can be found in Trochim [72]. • Traveler Intercept Surveys directly measure the LOS per- ceptions of actual travelers making real trips. These surveys intercept travelers mid-trip and either orally interview them on the spot or give them a postcard to report their LOS perceptions at a convenient time after they have com- pleted their trip. The Noel, Leclerc, and Gosselin [73] study of rural bicycle LOS used this method to measure bicycle LOS on rural roads. • Field Laboratory Studies recruit subjects (paid or unpaid volunteers) to travel over a fixed course in the field and report their LOS perceptions at strategic points along the course. The “Bike for Science” and “Walk for Science” stud- ies by Landis et al. [74] [75] are examples of this approach to measuring traveler perceptions of level of service. • Video Laboratory Studies show recruited subjects film clips of various street situations in a video laboratory setting. The Pecheux et al. [76] and Sutaria and Haynes studies [77] for intersection level of service are two examples of this ap- proach to measuring traveler perceptions of level of service. The level of service research to date is split fairly evenly be- tween the use of field laboratory settings and video laboratory settings for measuring traveler perceptions of level of service (see Exhibit 34). Traveler intercept surveys have been used by a few researchers to measure traveler LOS. Traveler Intercept Surveys, Field Laboratory Studies, and Video Laboratories Studies each have their relative strengths and weaknesses (see Exhibit 35). The traveler intercept surveys can gather responses from large numbers of individuals, but only for the particular trip that they made on the facility—the researcher obtains only one data point for each individual responding to the survey. C H A P T E R 4 Data Collection

33 Research Team Data Collection Method LOS Model Auto Hall, Wakefield, and Kaisy [78] Focus Group Discussions Freeway Pecheux et al.79] Video Laboratory (100 subjec ts) Signalized Intersection Sutaria and Haynes [80] Video Laboratory (310 subjec ts) Signalized Intersection Nakamura, Suzuki, and Ryu [81] Field Laboratory (24 subjects) Rural Road Colman [82] Field Laboratory (50 subjects) Urban Street Transit Morpace [83] Traveler Intercept Survey on-board vehicle Route Bicy cle Landis et al. [84] Field Laboratory (60 subjects) Intersection Harkey, Reinfurt, and Knuiman [85] Video Laboratory (202 subjects) Segment Jones and Carlson [86] Video Lab. Over Web (101 subjects) Rural Road Noel et al. [87] Traveler Intercept Survey (200 subjects) Rural Road Landis, Vattikuti, and Brannick [88] Field Laboratory (150 subjects) Segment Stinson and Bhat [89] Video Lab. Over Web (3,145 subjects) Segment Pedestrian Miller, Bigelow, and Garber [90] Simulated Video Lab Intersection Landis et al. [91] Field Laboratory (75 subjects) Segment Chu and Baltes [92] Field Laboratory (96 subjects) Mid-block Crossings Nadeir and Raman [93] 3-D Video Simulator Segment Exhibit 34. Traveler Perception Survey Methods in the Literature. Traveler intercept surveys may be the most realistic in that travelers are in an actual trip-making situation; however, the particular method used to intercept travelers may bias the results, especially, if the traveler is detained a long time or if certain “hard to stop” travelers are not interviewed or given a post card. Also, there may not be enough travelers on truly poor road sections to survey. Although the initial investment for traveler intercept surveys, and the cost per each subject are quite low, the cost per data point obtained (i.e., the product of the number of individuals surveyed and the number of situations they were exposed to) is higher than for the other data collec- tion methods (if one considers only the marginal costs and ignores the high initial investment costs of the video laboratory). The field laboratory studies also have low initial invest- ment costs, and they have the lowest cost per data point obtained. However, they are expensive to set up for a given site and have a high cost per subject. Each subject, however, is exposed to a wide variety of situations in the field, so this method generates numerous data points per individual. Field laboratory studies are realistic in that they expose the volunteer subjects to the full sensory experience (all five senses) of field conditions; however, because there is no penalty for arriving late at one’s appointment or job the re- alism of the trip experience is questionable. Video laboratory studies require an initial investment to create the video clips. If there is doubt about the ability of the video to capture all of the factors affecting a traveler’s perception of LOS, then there is also an added expense for calibration of the video lab LOS results to the field. Once the video clip has been assembled and calibrated, the cost per data point obtained is lower than that for traveler intercept surveys, but higher than for field laboratory studies. The cost per subject though is higher than for the traveler in- tercept surveys, because video labs typically test fewer subjects than would be found in an intercept survey. The video labs and field laboratories test fewer subjects than the traveler intercept surveys; however, both laboratory studies can expose single subjects to multiple conditions, thus enabling researchers to distinguish between a single subject’s reaction to a range of situations and differences in multiple subjects’ reactions to the same situation. This capability (highly valuable for model building) is not available in the traveler intercept surveys. The traveler intercept surveys are better for general model validation than for detailed model development. Video labo- ratory and field laboratory studies are better for model devel- opment because they give researchers more control over the variability of the results. Field surveys of traveler satisfaction, such as the FDOT/ Sprinkle “Walk For Science” surveys, come closest to the real world experience of travelers while controlling for the range of conditions they experience. However, this survey method is expensive and prone to agency liability problems (caused by exposing the participants to specified field conditions they might not otherwise attempt on their own). Conducting field surveys of traveler satisfaction would have cost $150,000 per mode per site to set up and conduct. For four modes and four cities, the data collection cost alone would have exceeded the entire research project grant. Thus field surveys were deemed infeasible.

34 Survey Type Strengths Weaknesses Cost Traveler Intercept Surveys: Surveyors stop people mid-trip to distribute post cards or conduct survey. 1. Most realistic of all methods. Only method that captures traveler’s response while making a real trip. 2. Can test for effects of travel time, wait time, and cost in combination with physical characteristics of facility. 1. No control over subject’s exposure to facility conditions. 2. Limited information on extent of subject’s exposure to facility. 3. Can’t test the same person’s response to conditions other than those of specific trip. 4. People don’t like to be interrupted while traveling, which may bias results. 5. Can’t sample extreme conditions. 6. Modal sample sizes depend on volumes. Bicycles are difficult to sample adequately. Initial Investment: $20,000 to pilot- test intercept methods. Data Collection: $15,000 per site for four modes. $60 per data point (not counting initial investment) Field Laboratory: Paid or unpaid volunteers travel specified course. 1. Second most realistic of survey types. It puts subjects in realistic physical situations, lacking only the realism of making the actual trip for an actual purpose (such as going to work). 2. Good control on subject exposure to facility. 1. Potential liability for accidents. 2. Can’t expose subjects to conditions not present in community or at time of test, particularly true for surveys using weekend volunteers. 3. Because subjects are not actually going anywhere the usual factors that influence trip- making behavior (travel time, wait time, and cost) cannot be reliably included or ruled out. 4. Unpaid volunteers are self- selected. Initial Investment: $-0- because method is well tested. Data Collection: $150,000 per site per mode. $10-$25 per data point. (not counting initial investment) Video Laboratory: Selected subjects shown video clips in laboratory setting. 1. Controlled exposure of subjects to audio- visual aspects of travel. 2. Little liability exposure. 3. Can expose subjects to wide range of conditions and time periods, thus enabling more in- depth analysis for each individual. 1. Not as realistic as simulator or field tests. Some important aspects of trip are excluded (e.g., pavement condition and rumble and back draft from trucks passing the subject). 2. Factors that influence trip- making behavior (e.g., travel time, wait time, and cost) cannot be reliably tested. 3. Needs calibration/validation against field conditions. 4. Not realistic for Transit. Initial Investment: $55,000 to develop videos for three modes. Another $125,000 to calibrate to field. Data Collection: $64,000 per lab site for three modes. $42 per data point (not counting initial investment) Exhibit 35. Validation Data Collection Options. Compared with the “Walk For Science” field surveys, trav- eler intercept surveys sacrifice the ability to “control” the range of physical conditions to which the participants are exposed. In addition, the travelers are self-selected (i.e., they would not be there to be intercepted, if it were not already their preferred mode and route). Nevertheless, among the remaining feasible survey methods, traveler intercept surveys were the best method for gathering transit rider quality of service perceptions. They were within the budget range of the research grant and travelers were exposed to the full physical experience of the transit experience. The traveler intercept survey method however was prob- lematic for auto and bicycle LOS because it is difficult to intercept auto drivers and bicyclists on the street without adversely affecting their perception of the quality of service. Consequently it was determined that this survey method could not be used for the auto or bicycle modes. This left video lab surveys as the best remaining method for surveying auto and bicycle level of service, because of its rel- atively low cost, the ability to control the environment to which each participant was exposed, the elimination of research agency liability exposure, and the ability to expose different people from different geographic areas to the same perceived street environment. Although it would have been feasible to use a traveler in- tercept survey method for pedestrians, the video lab survey method was considered superior because it would enable the team to expose survey participants to a controlled wider

35 range of physical conditions (including lack of sidewalk) that would not be easy to find in the field. The video lab approach also enabled testing of the signifi- cance of demographics and metropolitan area on the percep- tions of quality of service. 4.2 Phase I Data Collection (Pilot Studies) During Phase I, the video lab method for gathering traveler quality of service ratings was developed and tested. A video lab approach for measuring auto level of service was tested by George Mason University in Virginia. Sprinkle Consulting tested a similar video laboratory approach for pedestrian level of service in Florida. For the transit mode, a rider intercept approach for transit level of service was tested in three metropolitan areas of the United States (Ft. Lauderdale, Florida; Washington, DC; and Portland, Oregon). A total of 1,320 people were surveyed, and 2,535 observations of quality of service were gathered during Phase 1. Exhibit 36 provides key statistics on this Phase 1 data collection effort. The data gathered for each mode are summarized below. Auto: Fourteen video clips were developed and shown to 75 research subjects in the Washington D.C. metropolitan area. The results showed that a single factor, average travel speed, explained 64% of the variation in LOS ratings reported by the laboratory participants. Comparison of the video lab perceptions to field percep- tions of LOS identified the same key factor influencing LOS in the field (speed) as was found in the video lab. The corre- lation of the lesser factors to LOS varied between the field and the lab. The influence of other operational factors (signals and stops), design, maintenance, and aesthetics on LOS was less pronounced in the field than in the lab. The one signifi- cant exception was pavement condition, which had a stronger influence in the field than in the lab (as expected, given that the video gives only a visual input on pavement condition, while the field gives both visual and tactile inputs). The researchers noted that the limited number of video clips in the video library for Phase 1 resulted in some factors being spuriously correlated (for example: speed and the pres- ence of trees). This makes it difficult to build statistically robust models of LOS from the video laboratory data that accurately reflect the separate contribution of each correlated factor to a person’s perceived LOS. Thus for Phase 2 it was recommended that the video clip library be expanded to in- clude a wider range of cases. Transit: The Phase 1 data collection effort obtained a large amount of data (1,170 observations) for three urban areas (Miami; Portland; and Washington, DC). The re- search team noted that the specific routes surveyed in those metropolitan areas for Phase 1 did not exhibit significant crowding at the dates and times of the surveys. This gap in the transit data caused crowding to drop out as a significant explanatory factor of transit LOS. Therefore it was recom- mended that a few additional surveys be conducted in Phase 2 of more crowded bus routes with standees in one of the metropolitan areas. Pedestrian: Eight video clips were developed and shown to 45 participants in one metropolitan area (Sarasota, FL). These clips, however, did not cover a very wide range of LOS conditions (most being LOS C according to the FDOT method). Thus the research team recommended that addi- tional video clips of a wider range of conditions be obtained for Phase 2. Bicycle: No data collection was performed for bicycles in Phase 1, so an entire new video clip library was developed for Phase 2. 4.3 Development of Video Clips Auto Video Clips Based on findings from Phase I, the most influential factors to driver perceived level of service were selected by the re- search team. These included in no particular order • Presence of median (Yes/No); • Landscaping (Yes/No); • Progression (no progression is stopped at more than 50% of signals); • Posted speed (surrogate for arterial type); and • LOS depicted in clip using HCM methods. Contractor Mode Method Number of Metro. Areas Persons Data Points Cost GMU Auto Video Lab 1 75 975 $ 75,660 KAI Transit Field Intercept 3 1,170 1,170 $ 40,000 SCI Ped Video Lab 1 45 360 $ 30,500 Total Phase I 5 1,290 2,505 $ 146,160 GMU = George Mason University, KAI = Kittelson Associates, SCI = Sprinkle Consulting A data point is defined as one person providing an LOS rating for a single facility condition. Thus a person watching 10 video clips generates 10 data points. Exhibit 36. Phase 1 Data Collection Efforts.

36 These factors were chosen by the research team, with input from the project panel, as those factors that could most eas- ily be measured by engineers, those that were most important to drivers (as determined in previous studies and Phase I of the study), and those that could be captured in the field through videotaping. Arterials were selected in the Washington, DC, metropol- itan area that captured the required combination of condi- tions. As noted, some of the video clips were developed in Phase I of the study; an additional subset of video clips were developed by GMU in the summer/fall of 2005 in preparation of the data collection in the summer of 2006. As with the Phase I pilot test, videos were created for day- light conditions only. Taping was also limited to clear days without precipitation, and for the most part, snow is not a feature on the majority of tapes. In order to film the video clips, the following testing mate- rials were used: • Vehicle; • Two video cameras (one to capture the driver’s perspective and one to capture the speedometer); and • Two camera tripods. Standard vehicles (e.g., station wagons, sedans, and, in a few cases, small sports utility vehicles) were used for video- taping. Vehicles were rented from the GMU motor pool so as to standardize the vehicle set up and ride quality. Researchers set up two cameras and the GPS unit when they arrived at the vehicle rental location. A professional JVC digital videocamera, loaned to the project by the GMU Media Laboratory, was used to capture the roadway scene from the driver perspective (typically a full windshield view and pe- ripheral views of the roadside) and a palm-sized digital video- camera was used to capture the speedometer view. After the initial taping runs took place, the individual clips needed to be extracted. Based on the requirement of 1/2-mile on urban arterials (as determined through Phase I efforts), these clips were developed. The emphasis was on extracting segments from the videos that met several criteria including: After the videotaping took place, the researchers used the following to extract the videos: • Video editing decks available in the GMU Media Laboratory; • Adobe Premiere 9.0 video editing software; • Microsoft MapPoint; • Microsoft Excel; • Original mini-Digital Videos (DV) created in the field; and • Mini-DV player. In order to depict a consistent scene to study participants, it was necessary to identify video clips that had consistent cross section. For example, efforts were made to identify sections of video in which the roadway width did not change during the drive or that the sidewalk conditions were relatively consis- tent. Using a portable mini-DV player, students identified the portions of roadway to be made into a clip based on criteria such as arterial type, consistent cross section, lane position, and speed limit. After the general area of the clip was identi- fied, the researchers turned to Microsoft MapPoint. After each section of roadway was identified, individual clips needed to be made. The video feed needed to be syn- chronized with the speedometer feed. This was done using the mini-DV player and the time stamps on it. The field team had announced the run orally while the videocameras were filming the study arterials. The researcher’s voice was used to synchronize time stamps of the videocameras. Then, the researchers found the location of the beginning and end of the proposed clip and determined the tape length equivalen- cies for the two video feeds, for example 1 minute 6 seconds into the tape was when the voice was first heard on tape 1, 1 minute 20 seconds into the tape was when the voice was first heard on tape 2. After identifying the time stamps for both the road video and the speedometer, the team began editing using the video editing equipment available at GMU’s Media Laboratory to cut the clips and merge the speedometer video into the lower righthand corner of the video screen to simulate driving the vehicle. Adobe Premiere 9.0 was used to merge the two videos and create each clip. Once all the clips were made, transitions were put in between each clip on the final media to help proc- tors and participants identify each clip (for example, Clip #3) using the same software package. Then, the clips were merged and burned onto DVDs. Exhibit 37 summarizes the characteristics of the auto clips. Bicycle Video Clips Bicyclists are among the most vulnerable of travellers and are affected by a broader variety of traffic and roadway envi- ronmental factors (stimuli) than that of the motorized modes. Consequently, when collecting data and modeling perceptions, care must be taken to capture this sensitivity to the many environmental factors. Previous research, model development, and nationwide deployment of non-motorized LOS mode models have demonstrated that field-based studies are desirable to capture accurate perceptions of bicyclists. Such studies place the par- ticipants in typical real-life situations and capture the partic- ipants’ response to the host of stimuli present in roadway environments affecting bicyclists. However, field studies can be expensive and, depending on the range of conditions and variables being explored, represent the highest risk for par- ticipants of any method.

37 Cl ip # Cl ip D ist an ce (m ile s) St re et N am e H CM C la ss LO S as p er H CM N um be r o f T hr ou gh La ne s Pr es en ce o f M ed ia n To ta l T ra ve l T im e (se co nd s) Sp ac e M ea n Sp ee d PE D o n si de wa lk # St op s (be low 5 mp h) To ta l # o f S ig na ls Pr es . O f E x. L T La ne - Si gn al s Pr es . O f R t T ur n La ne - Si gn al s Tr ee P re se nc e Av er ag e La ne W id th (ft ) W id th o f M ed ia n (ft) R ig th S ho ul de r w id th (ft ) Le ft Sh ou ld er W id th (ft ) W id th o f p ar kin g la ne (ft ) W id th o f s id ew al k (ft) Se pa ra tio n fro m ri gh t-o f- w a y to s id ew al k (ft) W id th o f b ike la ne (ft ) 1 0.50 Rt 234 1 1 3 3 119 15.1 0 1 2 1 1 2 12 54 0 3 0 4 3 0 2 0.46 Gallows Road 3 6 2 3 48 34.5 0 0 3 1 1 2 13 4 0 0 0 4 3 0 5 0.50 Wilson Blvd 3 5 2 3 60 30.0 2 0 3 1 1 1 14 0 0 0 7 10 0 5 6 0.43 Clarendon 3 3 2 1 87 18.3 2 1 2 1 0 1 14 0 0 0 7 4 0 0 7 0.48 Wilson Blvd 3 4 2 1 86 20.1 2 0 3 1 0 1 14 0 0 0 7 10 0 5 8 0.49 Wilson Blvd 3 2 2 1 130 13.6 2 2 5 1 1 1 12 0 0 0 8 14 0 6 10 0.53 Washington Blvd 3 3 1 0 113 16.9 2 2 3 0 0 3 12 0 0 0 8 6 0 0 12 0.47 Wilson Blvd 3 3 2 0 118 14.3 0 2 2 0 0 1 11 0 0 0 8 11 5 0 13 0.50 Washington Blvd 3 5 1 0 71 25.4 1 0 1 0 0 3 12 0 0 0 8 6 0 0 14 0.50 Glebe Road 2 1 3 3 161 11.2 2 3 3 1 1 1 11 4 0 0 0 8 0 0 15 0.50 Glebe Road 2 1 3 3 229 7.9 2 3 3 1 1 1 11 4 0 0 0 8 0 0 16 0.55 Fairfax Drive 3 1 2 3 163 12.1 2 4 4 1 1 1 11 10 0 0 8 16 0 5 19 0.52 23rd St 4 4 2 0 116 16.1 2 3 8 0 0 2 10 0 0 0 7 6 5 0 20 0.55 Rt 50 1 2 2 3 122 16.2 2 1 2 1 0 1 11 17 8 2 0 0 0 0 21 0.50 Rt 50 1 2 2 3 89 20.2 2 2 3 1 1 2 11 17 8 2 0 0 0 0 23 0.54 M St 4 2 2 0 243 8.0 2 3 8 0 0 1 10 0 0 0 10 10 0 0 25 0.54 M St 4 3 2 0 179 10.9 2 2 8 0 0 1 10 0 0 0 10 10 0 0 29 0.50 Rt 234 2 4 3 3 79 22.8 0 1 3 1 1 2 12 54 0 3 0 0 0 0 30 0.55 M St 4 1 2 0 298 6.6 2 8 8 0 0 1 10 0 0 0 10 10 0 0 31 0.50 M St 4 1 2 0 471 3.8 2 9 8 0 0 1 10 0 0 0 10 10 0 0 51 0.44 M St 4 1 2 0 240 6.5 2 4 9 0 0 1 10 0 0 0 10 10 0 0 52 0.41 M St 4 2 2 0 186 7.9 2 3 7 0 0 1 10 0 0 0 10 10 0 0 53 0.60 Prosperit y 2 3 2 3 121 18.5 0 1 2 1 1 2 12 15 0 0 0 4 4 0 54 0.60 Lee Hw y 2 4 2 2 93 24.5 0 2 4 1 1 3 12 14 4 4 0 4 10 0 55 0.45 Braddock Rd 2 1 2 3 128 12.7 0 1 1 1 1 3 12 15 0 0 0 6 0 0 56 0.50 Sunset Hills Rd 2 4 2 3 77 23.1 0 1 1 1 0 3 12 8 0 0 0 0 0 0 57 0.61 Sunset Hills Rd 2 3 2 0 129 17.4 0 2 2 0 0 3 12 0 0 0 0 4 2 0 58 0.60 Sunrise Valley Rd 2 1 2 3 144 11.2 0 1 3 1 0 3 12 10 0 0 0 3 4 0 59 0.61 Sunset Hills Rd 2 1 2 0 182 12.1 0 3 2 0 0 3 12 0 0 0 0 4 4 0 60 0.50 Lee Hw y 2 2 2 2 120 15.0 0 1 3 1 0 1 12 14 0 0 0 4 4 0 61 0.70 Rt 50 1 4 3 0 91 27.7 0 1 3 1 0 3 12 0 0 0 0 0 0 0 62 0.50 Rt 50 1 5 3 0 49 36.7 0 0 2 1 0 3 12 0 0 0 0 0 0 0 63 0.50 Rt 50 1 6 2 3 53 41.9 0 0 2 1 1 3 12 6 4 4 0 0 0 0 64 0.50 Rt 50 1 2 2 3 92 19.6 0 1 3 1 0 3 12 6 0 0 0 0 0 0 65 0.50 Lee Hw y 2 6 2 2 50 36.0 0 0 3 1 0 2 12 14 0 0 0 0 0 0 Exhibit 37. Summary of Auto Clip Characteristics.

38 Video simulation, however, potentially provides some significant advantages to real-time field surveys, particu- larly if the “moving camera” approach is used. The moving camera perspective gives the video simulation a greater re- flection of reality as opposed to the stationary camera. Moving camera simulation also allows for a wider range of geographic participants and the testing of a greater range of variables, particularly the potentially hazardous higher truck volumes and the high frequencies of driveway/curb cut common in jurisdictions with minimal roadway access management practices. Finally, moving camera (video) simulation, if done based on lessons learned through pre- vious bicycle research, can approximate real-time condi- tions without the real-life hazards to participants in field studies. The research team chose to use a video simulation method- ology for this effort. The bicycle LOS research methodology used was designed to achieve the following objectives: • Obtain bicyclists’ perceptions of the level of accommoda- tion provided by arterial roadways using a real-time field- data collection event; • Coincident with the field data collection event, use video simulations to obtain bicyclists’ perceptions of the level of accommodation provided by arterial roadways; • Develop an equation to correlate the video simulation responses to the real-time event responses; and • Provide the information necessary to develop the research team’s initially proposed model form. For this NCHRP Project 3-70, a video simulation was used to collect data for the bicycle LOS model development. However, the research team took advantage of a coincident bicycle facility LOS project being conducted by FDOT’s Central Office and District 7 which combined approach of field based studies with video simulation. This timely FDOT study involved a real-time event in which bicyclists rode a study course and evaluated facilities along the course. As part of this project, we filmed moving camera videos of the event route under similar conditions expected for the actual event. The videos were edited into digital sequence videos for the creation of simulation videos for video-to- field calibration. Following the FDOT project, NCHRP Project 3-70 pro- duced additional video for testing in a separate video simula- tion laboratory effort to obtain responses from additional users. To ensure the consistency of the NCHRP research video survey results, the original Ride for Science (described below) video clips were re-edited to match the format of those produced specifically for NCHRP 3-70. The NCHRP 3-70 laboratory simulation clips were shown at four locations across the United States. Because the video simulation and its fidelity to a real-time event was an important consideration and because the NCHRP Project 3-70 team was able to take advantage of the FDOT study, the real-time event and coincident video simu- lation are described below. Staff from Dowling & Associates and Sprinkle Consulting, Inc., initially developed a matrix with 30 specific combina- tions (“runs”) of geometric and operational criteria. The matrix is provided as Exhibit 38. The research team used the matrix as a guide to identify filming candidate locations in Tampa. Dr. Huang and Mr. Petritsch field-checked the locations to verify their geo- metric and operational characteristics. Some runs identified when filling out the matrix involved unlikely combinations (for example, Run #11, which specified traffic volume in out- side lane > 800 vph and speed limit < 30 mph). Consequently, some of the combinations of variable ranges were not taped for the NCHRP project 3-70 study. After discussions with Mr. Reinke, the research team selected alternative locations so that there would be locations for each value of each criterion. For example, traffic volumes of < 400, 400-800, and 800+ vph were all represented. Theo Petritsch of Sprinkle Consulting, Inc. and Mr. Michael Munroe (a professional videographer) videotaped the bicycle locations during March and April 2006. The video platform used was a Viewpoint bicycle with Glidecam, as described above and shown in Exhibit 39. All traffic laws were obeyed during the filming of the bicy- cle clips. To ensure a consistent recording methodology, and one which reflects typical bicyclists’ scanning behavior, a pro- tocol was developed, tested, and used by the researchers and videographer for proper camera panning techniques and to keep the roadway ahead in the right-center of the frame to focus on the roadway and capture driveway conditions while not focusing on objects outside the right of way. One or two “takes” were filmed at each location. The researchers started about one city block upstream of the in- tersection, taped while riding at approximately 12 mph, and finished about one city block downstream of the intersection. The team also used several video clips from those filmed for the video simulation portion of the Ride for Science 2005. Those were filmed using the same procedures. With guidance from Mr. Petritsch, Dr. Huang selected 30 bicycle clips for inclusion in the bicycle DVD. The geomet- ric and operational characteristics of the locations depicted in these clips are shown in Exhibit 40. Pedestrian Video Clips Pedestrians are among the most vulnerable of travellers and are affected by a broader variety of traffic and roadway environmental factors (stimuli) than that of the motorized

39 Segment variables Intersection variables Run Width of outside lane (ft) Presence / width of bike lane (ft) Veh flow in outside lane (vph) Speed limit (mph) Crossing width (ft) Control delay (s) 1 < 12 No bike lane 400 - 800 30 - 40 36 - 60 No stop 2 < 12 No bike lane 800+ < 30 60+ No stop 3 < 12 No bike lane < 400 30 - 40 < 36 < 40 4 < 12 No bike lane 400 - 800 < 30 36 - 60 < 40 5 < 12 No bike lane 800+ 40+ 60+ < 40 6 < 12 ≤ 4 < 400 < 30 36 - 60 40+ 7 < 12 ≤ 4 400 - 800 30 - 40 60+ 40+ 8 < 12 ≤ 4 800+ 40+ < 36 40+ 9 < 12 ≤ 4 < 400 40+ 36 - 60 No stop 10 < 12 ≤ 4 400 - 800 30 - 40 60+ No stop 11 < 12 ≤ 4 800+ < 30 < 36 No stop 12 < 12 > 4 < 400 30 - 40 60+ < 40 13 < 12 > 4 400 - 800 < 30 < 36 < 40 14 < 12 > 4 800+ 40+ 36 - 60 < 40 15 < 12 > 4 400 - 800 30 - 40 < 36 40+ 16 < 12 > 4 800+ 40+ 36 - 60 40+ 17 12 + No bike lane 400 - 800 30 - 40 36 - 60 No stop 18 12 + No bike lane 800+ < 30 60+ No stop 19 12 + No bike lane < 400 30 - 40 < 36 < 40 20 12 + No bike lane 400 - 800 < 30 36 - 60 < 40 21 12 + No bike lane 800+ 40+ 60+ < 40 22 12 + ≤ 4 400 - 800 30 - 40 60+ 40+ 23 12 + ≤ 4 800+ 40+ < 36 40+ 24 12 + ≤ 4 < 400 40+ 36 - 60 No stop 25 12 + ≤ 4 400 - 800 30 - 40 60+ No stop 26 12 + ≤ 4 800+ < 30 < 36 No stop 27 12 + > 4 400 - 800 < 30 < 36 < 40 28 12 + > 4 800+ 40+ 36 - 60 < 40 29 12 + > 4 < 400 < 30 60+ 40+ 30 12 + > 4 800+ 40+ 36 - 60 40+ Exhibit 38. Bicycle Video Clip Sampling Plan. modes. Previous research, model development, and nation- wide deployment of non-motorized LOS mode models have demonstrated that field-based studies are the most desirable means to capture accurate perceptions of pedestrians. They place the participants in typical real-life situations and capture the participants’ response to the host of stimuli pres- ent in urbanized roadway environments affecting pedestri- ans. However, field studies can be expensive and, depending on the range of conditions and variables being explored, rep- resent the highest risk for participants of any method. Video simulation, however, potentially provides some sig- nificant advantages to real-time field surveys, particularly if the moving camera approach is used. The moving camera per- spective gives the video simulation a greater reflection of real- ity than the stationary camera. Moving camera simulation also allows for a wider range of geographic participants and the test- ing of a greater range of variables, particularly the potentially hazardous higher truck volumes and high driveway/curb cut frequencies common in jurisdictions with minimal roadway access management practices. Finally, moving camera (video) simulation, if done based on lessons learned through recent pedestrian research, can approximate real-time conditions without the real-life hazards to participants in field studies. Given these advantages of video simulation, the project team used video simulation to collect data for the pedestrian LOS model. Video clips were created and then shown to par- ticipants in video simulation laboratories. Exhibit 39. Bicycle Video Camera Mount.

40 Exhibit 40. Characteristics of Bicycle Video Clips. Staff from Dowling & Associates and Sprinkle Consulting, Inc., initially developed a matrix (see Exhibit 41) with 22 specific combinations (“runs”) of geometric and operational criteria to represent the typical ranges of urban arterials in metropolitan areas throughout the United States. The research team used the matrix as a guide to identify can- didate locations in Tampa and San Francisco. Herman Huang, Ph.D., of Sprinkle Consulting, Inc., field-checked the Tampa locations and Dowling & Associates staff field-checked the San Francisco locations to verify their geometric and operational characteristics. Some runs involved unlikely combinations (for example, Run #11, which specified sidewalk width < 4ft and high pedestrian volumes) and so were not included in the data collection video simulation video. After discussions with David Reinke of Dowling & Associates, the research team selected alternative locations so that there would be locations for each

41 Exhibit 40. (Continued). Segment variables Intersection variables Run Sidewalk width (ft) Separation of walkway from traffic Traffic speed (mph) Traffic volume outside lane (vph) Pedestrian volumes Number of lanes crossed Signal delay (sec) 1 < 4 No 30-40 400-800 Medium 2 < 30 2 4+ No 40+ 800+ High 2 < 30 3 No sidewalk No < 30 400-800 High 2 < 30 4 4+ No 40+ < 400 Medium 2 < 30 5 No sidewalk No < 30 800+ Medium 4+ < 30 6 < 4 No 30-40 < 400 High 4+ < 30 7 4+ No 40+ 400-800 Low 4+ < 30 8 < 4 No 40+ < 400 Medium 4+ < 30 9 4+ No <30 400-800 High 4+ < 30 10 No sidewalk No 30-40 800+ Medium 4+ < 30 11 < 4 No 40+ < 400 High 4+ < 30 12 4+ Yes <30 800+ Medium 2 > 30 13 < 4 Yes 40+ 800+ Low 2 > 30 14 4+ Yes <30 < 400 Medium 2 > 30 15 < 4 Yes <30 800+ Medium 2 > 30 16 4+ Yes 30-40 < 400 High 2 > 30 17 < 4 Yes <30 800+ High 4+ > 30 18 4+ Yes 30-40 < 400 Low 4+ > 30 19 < 4 Yes <30 < 400 Low 4+ > 30 20 4+ Yes 30-40 400-800 Medium 4+ > 30 21 < 4 Yes <30 400-800 High 4+ > 30 22 4+ Yes 30-40 800+ Low 4+ > 30 Exhibit 41. Pedestrian Sampling Plan.

42 value of each criterion. For example, traffic volumes of < 400, 400-800, and 800+ vph were all represented. The locations with high pedestrian volumes were mostly in San Francisco, as many parts of San Francisco are characterized by high levels of pedestrian activity. The locations with high traffic speeds and traffic volumes were mostly in Tampa, as many parts of Tampa are characterized by high speeds and volumes. Dr. Huang and a professional videographer videotaped the pedestrian locations in Tampa and San Francisco during March and April 2006. The filming protocol followed that pioneered and tested in 2004 by Sprinkle Consulting in their research Arterial Level of Service for Arterials project for FDOT. Videotaping was performed with a steady-cam unit. A stereo microphone mounted on the camera was used dur- ing videotaping. The videographer filmed the environment while walking the intersections and facilities, obeying all pedestrian signals in the process, while Dr. Huang provided recommendations concerning filming protocol and start and end points and served as a safety coordinator (see Exhibit 42). To ensure a consistent recording methodology that reflected typical pedestrians’ scanning behavior, the research team de- veloped, tested, and used a protocol for proper camera pan- ning techniques to keep the sidewalk on the right edge of the frame to focus as much as possible on the roadway, rather than objects outside the right-of-way. At each location, the researchers filmed multiple “takes,” each with a different length of signal delay. The researchers started about 100 yards upstream of the intersection, taped while walking at a normal (approximately 4 ft/sec) pedestrian speed, and finished about 100 yards downstream of the intersection. Dr. Huang selected 32 of the video pedestrian clips for inclusion in the DVD that would be used during the pedes- trian roadway LOS data collection events. Exhibit 43 lists the geometric and operational characteristics of the locations shown in these clips. Development of Master DVDs The research team members decided in the spring of 2006 that a maximum of 10 video clips for each mode were to be viewed in each study location and ratings gathered for each from participants. The decision to limit the videos to 10 clips per mode was partially based on the need to maintain the attention of study participants and also to maintain a total testing time of between 2 and 3 hours, including time for an informal focus group. The team also decided to select four specific clips for each mode to be shown in each of the four cities, so that one could later attempt to isolate the influence of variables such as population density, population, and expectation of travel conditions on traveler ratings of LOS across the four cities. Next, six additional clips per mode were selected to be shown in each of the four study locations. Finally, a pilot test clip for each mode was selected and shown in each of the fours cities to help orient participants to the mode they were to rate for that portion of the study. Exhibit 44, Exhibit 45, and Exhibit 46 show the specific sequence of video clips shown in each of the four study loca- tions. Clips shown in all four locations are highlighted—they show up at different points in the sequence. The specific sequence of clips shown in each city was intentionally ran- domized so as to minimize the likelihood of respondent fatigue biasing the results. Efforts were made to normalize the length of testing time in each of the four study locations while pro- viding a range of factors to participants in each study location. Using the GMU Media Laboratory facilities and staff, a set of master DVDs was created for each of the four testing locations. Efforts were made to maintain the highest possible quality of video to enhance the video presentation portion of the study; this requirement resulted in the creation of one DVD per mode per city, resulting in 12 master DVDs, which were later used in the data collection process. To maintain consistency among the clips, GMU Media Laboratory staff worked with the video production crew hired by Sprinkle Consulting to ensure that video clips had the same look and feel of those created by GMU. GMU Media Laboratory staff provided detailed editing instructions, which were followed perfectly by the production crew in Florida, while creating the pedestrian and bicycle clips. Each video clip was edited to include an opening title which read “Clip XXX” on a black background, next the title would fade out and the video clip showing a particular trip would begin. At Exhibit 42. Pedestrian Video Camera Mount.

Exhibit 43. Geometric & Operational Characteristics of Pedestrian Video Clip Locations. (continued on next pg)

Exhibit 43. (Continued).

Exhibit 43. (Continued).

46 Location of Video Laboratory – Pedestrian Clips Shown Presentation Order New Haven, CT Chicago, IL Oakland, CA College Station, TX Pilot Clip 212 212 212 212 1 223 201 215 208 2 208 226 220 217 3 226 225 206 215 4 204 208 201 214 5 205 219 227 201 6 203 228 226 230 7 201 211 209 218 8 231 215 216 232 9 215 229 224 226 10 210 222 208 221 Total Clip Time 16.2 min 18 min 16 min 19.4 min Note: Table shows the sequence of clips shown in each city. Entries are the clip identification numbers. Shaded clips were shown in all four cities. Sequence of clips shown was intentionally randomized in each city to counteract fatigue effects. Exhibit 44. Pedestrian Clip Sequence at Testing Locations. the conclusion of each video clip, GMU Media Laboratory staff looped the DVD back to a consistent title page which in- cluded a complete list of the video clips on each of the DVDs. This allowed the operator to then click the mouse on the next appropriate clip when participants were ready to begin rating the next video clip. Using the information provided in Tables 3-3 through 3-5, one DVD per mode per city was generated, resulting in 12 unique DVDs. These DVDs were then labeled by the GMU Media Laboratory staff to ensure ease of selec- tion by the facilitator of each laboratory session. 4.4 Video Lab Protocol Selection of Video Lab Cities Four metropolitan areas were selected for the Phase 2 auto, bike, and pedestrian video labs. They were Chicago, Illinois; San Francisco, California; New Haven, Connecticut; and College Station, Texas. They were selected to obtain a range of population and climates of the United States based on the hypothesis to be tested that the population of the urban area and the climatic area of the US might influence the degree of satisfaction reported by subjects in the video laboratories. There are 922 urban areas in the United States with a pop- ulation of at least 10,000 (U.S. Census [94]). These urban areas are ranked by population and then stratified into four groups, each group representing approximately one-quarter of the urban area population in the United States. The results are as shown in Exhibit 47. The eight largest metropolitan areas of the United States (i.e., New York, Los Angeles, Chicago, Philadelphia, Dallas, Miami, Washington DC, and Houston) hold one-quarter of the urban area population of the United States. Chicago was selected to represent these largest metropolitan areas of the United States. Location of Video Laboratory – Bicycle Video Clips Shown Presentation Order New Haven, CT Chicago, IL Oakland, CA College Station, TX Pilot Clip 326 326 326 326 1 301 319 302 311 2 323 308 310 328 3 321 306 305 324 4 320 309 324 315 5 317 320 327 309 6 312 318 321 313 7 309 304 309 303 8 307 324 322 319 9 314 321 330 320 10 324 329 320 321 Total Clip Time 13 min 13 min 13 min 13 min Note: Table shows the sequence of clips shown in each city. Entries are the clip identification numbers. Shaded clips were shown in all four cities. Sequence of clips shown was intentionally randomized in each city to counteract fatigue effects. Exhibit 45. Bicycle Clip Sequence at Testing Locations.

47 Location of Video Laboratory – Auto Clips Shown Presentation Order New Haven, CT Chicago, IL Oakland, CA College Station, TX Pilot Clip 25 25 25 25 1 21 20 12 15 2 55 56 56 7 3 52 10 8 52 4 60 51 65 13 5 53 14 59 58 6 56 2 29 56 7 54 62 6 2 8 2 63 15 1 9 15 52 2 61 10 57 15 52 64 Total Clip Time Note: Table shows the sequence of clips shown in each city. Entries are the clip identification numbers. Shaded clips were shown in all four cities. Sequence of clips shown was intentionally randomized in each city to counteract fatigue effects. Exhibit 46. Automobile Clip Sequencing at Testing Locations. The next-largest metropolitan areas hold another quarter of the urban area population in the United States: Detroit, Boston, Atlanta, San Francisco, Riverside, Phoenix, Seattle, Minneapo- lis, San Diego, St Louis, Baltimore, Pittsburgh, Tampa, Denver, Cleveland, Cincinnati, Portland, Kansas City, Sacramento, San Jose, San Antonio, Orlando, Columbus, Providence, Virginia Beach. San Francisco was selected to represent this group of large metropolitan areas. The other 800+ metropolitan areas constituting the re- maining 50% of the U.S. urban area population are too numerous to conveniently list here. The research team se- lected from the Census list of these cities the following two metropolitan areas to represent the lesser populated metro- politan areas of the United States: New Haven, Connecticut (population between 300,000 and 1.5 million), and College Station, Texas (population under 300,000). Thus, the four metropolitan areas for the Phase 2 auto, bike, and pedestrian video labs were Chicago, Illinois; San Francisco, California; New Haven, Connecticut; and College Station, Texas. IRB Review Most research institutions require, when working with human or animal subjects, that the study undergo a review by an independent review board to ensure that no undue harm will occur to study participants. George Mason University has its own Internal Review Board (IRB) to oversee research stud- ies within the University. The effort to obtain approval to proceed with the study included • Completing the IRB application for approval of study • Providing the IRB with an overview of the study protocol • Providing the IRB with sample survey instruments and testing material Researchers for this study received approval from the GMU IRB in June of 2006 to proceed with the study as de- scribed. Appendix B includes the materials submitted to the IRB including Study Protocol and the Application for Human Subjects Research Review. Recruitment Based on input from the team and ultimately the project’s Principal Investigator, a decision was made that the minimum number of participants in each location was 30 and a maxi- mum of 35 participants was budgeted for each study location. Phase I study results revealed that, although age influenced participant ratings—which is consistent with studies con- Group Population Range Number of SMSAs Total Population Percentage of U.S. Population 1 Pop.> 5M 8 65,154,790 24.9 2 1.5M < Pop < 5M 25 64,389,536 24.6 3 300K < Pop < 1.5M 104 66,586,646 25.5 4 13K < Pop < 300K 785 65,404,019 25.0 Total 922 261,534,991 100.0% MSA = Metropolitan Statistical Area, as defined by U.S. Census Exhibit 47. Stratification of MSAs Into Equal Population Groups.

48 New Haven, CT Chicago, IL San Francisco, CA College Station, TX Age Group (years of age) Male Female Male Female Male Female Male Female Total Young (18-35) 2 4 4 4 6 12 3 5 40 Middle (36-50) 9 8 9 6 9 8 8 6 63 Older (60+) 2 9 6 6 1 2 6 10 42 Total 13 21 19 16 16 22 17 21 145 Exhibit 48. Characteristics of Participants. 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% New Haven, CT Chicago, IL Oakland, CA College Station, TX Study Location Pe rc en ta ge o f P ar tic ip an ts Never At Least Once a Day >1 a Week but not Everyday About Once a Week Less than 1 a Month Exhibit 49. Non-Recreational Pedestrian Travel By Participants (More than Two Blocks). ducted by Sprinkle Consulting, gender was not found to be a statistically significant contributor to participant ratings. Based on these findings, the study team determined that recruiting of participants should be based on the following criteria in order of importance: • Age (seek equal distribution between young, middle, and older aged participants) • Gender (equal distribution between males and females) • Regular users of modes other than private vehicle, in par- ticular bicyclists Dr. Flannery of GMU recruited subjects in each location by establishing contact through the following: • Community senior citizen centers • Bicycle clubs • Community/neighborhood associations To assist in recruiting, posters and flyers, developed for each location, included all relevant information for the study (e.g., location, time, date, and participant requirements). These posters and flyers were sent to the contacts established through the various organizations. Posters also included tear- off contact information to register for the study. Appendix C contains an example flyer used in the Chicago location. Exhibit 48 breaks down the participants by age and gender in each of the four study locations. From the demographic survey, information was extracted on the regularity of participants to use modes other than pri- vate automobile in their travel. Researchers sought to include participants who regularly take non-recreational bike and pedestrian trips, as well as, regular transit users. Exhibits 49 through 51 show a breakdown of participant mode use by study location. Chicago had the highest per- centage of daily walkers among the cities surveyed. Oakland had the highest percentage of daily bicycle riders. College

49 Never At Least Once a Day >1 a Week but not Everyday About Once a Week Less than 1 a Month 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% New Haven, CT Chicago, IL Oakland, CA College Station, TX Study Location Pe rc en ta ge o f P ar tic ip an ts Exhibit 50. Non-Recreational Bicycle Usage By Participants. Never At Least Once a Day >1 a Week but not Everyday About Once a Week Less than 1 a Month 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% New Haven, CT Chicago, IL Oakland, CA College Station, TX Study Location Pe rc en ta ge o f P ar tic ip an ts Exhibit 51. Transit Usage By Study Participants.

50 Station had the highest percentage of participants who never walked, never biked, and never rode transit. Validity of Video Lab Respondent Sample The selected demographic characteristics of the video lab participants were compared with national averages (presented in Exhibit 52). With the exception of seniors (who were over- sampled) and single-family home residents (who were under- sampled), the video labs generally secured a representative national average of participants. Survey Instrument Survey instruments were developed to standardize data collection of input from the study participants. Study partic- ipants were asked to rate 10 video clips per mode on a six- point scale (A-F). Study participants were instructed that the A to F scale was similar to that used in grade school in which A was to represent the highest performance and F to repre- sent the worst performance. Pilot tests conducted by GMU in Phase I of NCHRP 3-70 revealed that trip purpose (i.e.,leisure versus time-constrained trips) influenced participant ratings of service quality; as a result, the team decided to focus study participants on time- constrained trips to better align with procedures in the HCM which are typically focused on peak-hour conditions. Study participants were instructed to choose the rating that best rep- resented their assessment of quality as a commuter after watching each video clip. A demographic survey instrument was also developed by team members for use in data analysis to better understand the motivation for responses by groups or individual partic- ipants. The survey contains questions about participant age group, gender, typical travel mode, and the use of all modes (including transit) during a typical day, week, and month. The survey instruments were developed to be easily under- stood and easily completed by the participants. Appendix A includes the survey instruments created for the study. Pilot Tests Pilot test sessions, using 14 GMU graduate students, were held to test the study methodology, to ensure that the surveys were easily understood by the participants, to refine our pres- entation of the materials, and to refine the study materials (such as how many clips to show and needing to increase font size for older drivers). Pilot test session data will not be included in the final database, but is available for review in hard copy format. One of the primary goals of the pilot test sessions was to de- termine if the order of video presentation by mode influenced ratings. For example, should the videos be shown in the order Auto Driver, Bicycle Rider, Pedestrian or in the order Pedes- trian, Bicycle Rider, Auto Driver? To control for the order, one order was presented to one study group and the other order was presented to the second study group. It was deter- mined that order of presentation had a slight influence on participant ratings, in particular for the auto video clips. Using this information, the team determined that the videos needed to be shown in a consistent order at each of the four locations to control for potential mode order bias in partici- pant ratings. The pilot sessions also revealed that the study materials were understandable to the participants, but that the font needed to be increased in some cases to account for vision loss in older participants. In addition, some of the questions in the demographic survey were rearranged to pro- vide better consistency in terms of those questions that required filling in a blank versus circling a response. Video Lab Sessions Laboratory sessions were held in four locations, as previ- ously noted. The study locations selected (large hotels in each of the four cities) had access to transit facilities, as well as, available parking; this provided the participants with easy to reach locations as well as a sense of security. Two study ses- sions were held in each location to enable older participants to attend daytime sessions (to address their desired times of arrival) and to enable working professionals to attend Group Nationa l Av erage Sample Bias in results? Male 49% 45% No, video lab participants mirror national average. Age over 60 16% 29% The video lab oversampled people over 60, which might possibly have a slight positive effect on LOS ratings for the bicycle video clips. Single-family detached dwelling unit 60% 36% The video lab undersampled people living in single-family homes, which might possibly have a slight positive effect on LOS ratings for the pedestrian video clips. Has vehicle available 90% 91% No, video lab participants mirror national average. Source for national averages: US Census, 2000, American Fact Finder, Tables P8, H32, and H44. Exhibit 52. Comparisons of Socioeconomic Characteristics of Sample with National Averages.

51 evening sessions. In each location, the daytime session was held from 10am-12:30pm and the evening session was held from 6:00-8:30pm. Hotel meeting rooms were set up classroom style with two participants seated at each roughly 10-ft-long table. There was an aisle between two rows of tables in which the video equipment was placed, and a large-screen projector screen was set up at the front of the room. Light refreshments were provided during each session. As participants arrived, they were given a unique identifier code which was written on their survey sheets for them and also corresponded to the receipt sheets generated for each location. The unique identifier scheme is required by GMU to keep the participants’ information confidential while allowing the researchers to later make correlations between responses and some other demographic (i.e., age or sex). The participants were then asked to complete a demographic questionnaire while waiting for the remaining participants to arrive. Once all participants had arrived, Dr. Flannery thanked them all for attending and gave a short introduction to help them understand the task at hand. The opening re- marks explained the study’s purpose, who the study sponsor was, general procedures, location of facilities, explanation of the forms (survey forms, Informed Consent Form), their rights as study participants, and the schedule of the study. The clips were shown in the order of pedestrian mode, bicy- cle mode, and finally auto mode. Participants were asked to keep their opinions to themselves during the study so as to not influence their neighbors and were informed that, at the end of the clips, a short focus group would be conducted in which they could provide more details on their opinions. A practice clip was shown to participants at the beginning of each mode to familiarize them with the task at hand. Ques- tions were clarified, if needed, once participants had com- pleted rating the practice clip. The participants were not informed what specifically to rate each clip on—only that they should rate the clip on how satisfied as a traveler. Upon completion of each mode, typically 20 to 25 minutes after the session had started, the participants had a 10- to 15-minute break before beginning the next mode video session. After all video clips had been rated by the participants, a short break was taken to set up for the focus group session. During the focus group session, participants were asked to discuss what factors greatly influenced their ratings in each of the mode video sessions. These comments were noted by Dr. Flannery on her laptop, and efforts were made to focus the participants on one mode at a time and to complete discussion of that mode before moving to the next. At the end of the session, the participants were allowed to ask questions and then were compensated for their time with a $75.00 honorarium paid in cash and required to sign a re- ceipt. Then forms were collected, and participants were thanked for their contribution to the study. 4.5 Effects of Demographics on LOS This section presents the results of an investigation into the effects of various socioeconomic and location factors on clip ratings for auto, bicycle, and pedestrian modes. The effects of metropolitan area location on transit LOS ratings could not be tested because of differences in the transit services pro- vided in each metropolitan area. For each of the auto, bicycle, and pedestrian modes, four common film clips were shown in each of the four metropol- itan regions. We used the ratings on these clips to test for effects of socioeconomic and location factors. For each factor, we divided the respondents into a test group and a control group. The test group contained those respondents for which the factor was present (e.g., persons having one or more cars available); the control group con- tained those respondents for which the factor was not pres- ent (e.g., persons not having a car available). Once the groups were defined, we used a nonparametric randomization (bootstrap) (see Davison and Hinkley [95]) test to determine whether the difference in mean ratings for an individual clip was significant. The sample size is denoted by N and the size of the control group is denoted by k. For hypothesis testing, the bootstrap method works as follows: 1. Compute the difference in means between the test group and the control group. 2. Generate a random permutation of cases. For each per- mutation, compute the mean for the first k cases and the last N − k cases. If the difference is greater than or equal to the difference computed in Step 1, add 1 to an indicator variable X. Repeat this step B times. 3. Divide X by B − 1. The result is the estimated probability that the difference computed in Step 1 results from chance alone. The bootstrap method has significant advantages over traditional hypothesis testing, mainly because it is nonpara- metric and, therefore, makes no assumptions about the shape of the distribution of responses. For this analysis, we defined significance to be at the 10% level (i.e., the probability that the difference in ratings could have arisen by chance alone is less than 10%.). We judged the bootstrap test to be superior to other tests that might be used for the following reasons: • Standard analysis of variance requires that cell sizes be equal, or nearly so (see Searle [96]). This assumption is violated for all the tests considered. • Classic hypothesis tests (e.g., the t-test) assume that responses are normally distributed.

52 Test group Control group Metro area is Chicago Metro area is San Francisco Bay Area Metro area is College Station Metro area is New Haven All other regions Metro area is San Francisco Bay Area Metro area is College Station Metro area is Chicago All other regions Metro area is College Station Metro area is San Francisco Bay Area All other regions Metro area is College Station All other regions Metro area population ≥ 1 million All other respondents Age is 18 - 35 All other respondents Age is 36 - 60 All other respondents Age is 60+ All other respondents Sex is male All other respondents Has a vehicle available All other respondents Has a bike available All other respondents Respondent is employed All other respondents Dwelling unit is single-family home All other respondents Respondent owns the home All other respondents Walks non-recreational > 2 blocks more than once a week All other respondents Cycles non-recreational > 2 blocks more than once a week All other respondents Uses transit more than once a week All other respondents Commutes by auto (drive alone or shared ride) All other respondents Commutes by transit All other respondents Exhibit 53. Test and Control Groups for Socioeconomic and Location Factors. Group Group Sample Size Test Control Test Control Mean Rating Differencea Highly Significant Differences None Significant Differences Metro area is New Haven Metro area is College Station 34 38 -1.02 Metro area is New Haven All other respondents 34 111 -0.85 Metro area is New Haven Metro area is Chicago 34 35 -0.84 Has a vehicle available All other respondents 132 13 0.67 Region is College Station All other respondents 38 107 0.50 a Mean of test group rating minus control group rating Exhibit 54. Significant Differences in Ratings—Auto. For each clip, we defined an indicator variable y as follows: • y = +1 if the mean rating for the test group is higher than the mean rating for the control group, and the difference is significant. • y = −1 if the mean rating for the test group is lower than the mean rating for the control group, and the difference is significant. • y = 0 if the mean rating for the test group is not signifi- cantly different from the mean rating for the control group. The scores for each of the four control clips for a given mode were added to form a cumulative score for the individ- ual factor for that mode. Given that there were four common clips for each mode, the score for each factor for a given mode could range from −4 to +4. A factor was deemed to be signif- icant if the score for that mode was −3 or +3; this meant that for three of the four common clips for that mode, the differ- ences in ratings between the test and control groups were sig- nificant and in the same direction. A factor was deemed to be highly significant if the score for that mode was −4 or +4; this meant that for all four common clips for that mode, the dif- ferences in ratings between the test and control groups were significant and in the same direction. The tested socioeconomic factors are listed in Exhibit 53. Effects of Demographics on Auto LOS Ratings Significant differences in auto clip ratings are shown in Exhibit 54. Although there were several significant differences (three of the four clips consistently rated higher or lower), none was highly significant.

53 Group Group Sample Size Test Control Test Control Mean Rating Differencea Highly significant differences Metro area is New Haven Metro area is San Francisco Bay Area 34 38 -0.87 Metro area is New Haven All other respondents 34 109 -0.75 Metro area is New Haven Metro area is College Station 34 36 -0.71 Metro area is New Haven Metro area is Chicago 34 35 -0.68 Male All other respondents 65 78 0.47 Metro area population ≥ 1 million All other respondents 73 70 0.40 Significant differences Age is 60+ All other respondents 42 101 0.61 a Mean of test group rating minus control group rating Exhibit 55. Significant Differences in Ratings—Bicycle. The following are the main findings for auto: • Respondents in the New Haven metro area consistently rated the clips lower than did respondents from other met- ropolitan areas. • Respondents with a vehicle available tended to rate the clips higher than did respondents from other metropolitan areas. • Respondents from the College Station metro area tended to rate the clips slightly higher than did other respondents. However, none of these differences were found to be “highly significant,” thus all data from all metropolitan areas and demographic groups were pooled for auto LOS model development. Effects of Demographics on Bicycle LOS Ratings Significant differences in bicycle clip ratings are shown in Exhibit 55. The following factors resulted in highly significant differences in the LOS ratings: • Respondents from the New Haven metro area consistently rated the clips lower than did respondents from the other metropolitan areas. • Male respondents tended to rate the clips slightly higher than did female respondents. • Respondents from metro areas with a population of over 1 million (Chicago and San Francisco Bay Area) tended to rate the clips slightly higher than did respondents from the other two metro areas. The following factor was found to be significant: • Respondents aged over 60 tended to rate the clips slightly higher than did other respondents. For these reasons “metropolitan area” was included as an explanatory variable for the bicycle LOS model development. However, analysts would not generally have information on the sexual split between bicyclists, so sex was excluded from the bicycle LOS model development. Effects of Demographics on Pedestrian LOS Ratings Significant differences in pedestrian clip ratings are shown in Exhibit 56. The following factors resulted in highly signif- icant differences in the pedestrian LOS ratings: • Respondents who walk more than two blocks for non- recreational purposes more than once a week tended to rate the clips lower than did other respondents. The following factors resulted in significant differences in the LOS ratings: • Respondents from the College Station metro area tended to rate the clips higher than did other respondents. • Respondents from the Chicago metro area tended to rate the clips lower than did respondents from the College Station metro area. • Respondents who have a bicycle available tended to rate the clips slightly lower than did other respondents. • Respondents who live in single-family detached dwelling units tended to rate the clips slightly higher than did other respondents. Only the extent of non-recreational walking was a highly significant factor affecting pedestrian LOS ratings. However, this a demographic variable unlikely to be known by analysts using the pedestrian LOS method. Consequently, this vari- able was excluded from the pedestrian LOS model.

54 Group Group sample size Test Control Test Control Mean rating differencea Highly significant differences Walks non-recreational > 2 blocks more than once a week All other respondents 97 48 -0.60 Significant differences Metro area is College Station All other respondents 38 107 0.78 Metro area is Chicago Metro area is College Station 35 38 -0.77 Has a bike available All other respondents 107 38 -0.76 Dwelling unit is single-family home All other respondents 52 93 0.36 a Mean of test group rating minus control group rating Exhibit 56. Significant Differences in Ratings – Pedestrian. The metropolitan area showed up as a significant factor affecting pedestrian LOS ratings for a couple of metropolitan areas, so this factor was included in the pedestrian LOS model development. 4.6 Transit On-Board Surveys The transit survey methodology used for this project was designed to achieve the following objectives: • Confirm the quality of service factors important to pas- sengers who have already decided to make a trip by transit; • Ask questions in a form relevant to passengers (relating to their trip), but provide results in a form relevant to the project (relating to a specific urban street facility); • Maximize the amount of useful information that could be gleaned from a limited number of survey locations; and • Provide the information necessary to develop the project team’s initially proposed transit model form, while also providing data that could be used to develop alternative model forms, if necessary. Agency Coordination Five transit agencies were contacted to obtain permission to conduct surveys: TriMet in Portland, Oregon; Washington Metropolitan Area Transit Authority (WMATA) for North- ern Virginia; Broward County Transit (BCT) for the Fort Lauderdale, Florida area; the San Francisco Municipal Railway (MUNI) (which operates bus and rail services within the City of San Francisco); and AC Transit (which operates express and local bus services in and between several cities in the San Francisco metropolitan area). These agencies were chosen for geographic variety, a range of service and demand conditions, and their proximity to research staff offices. All of the agencies were provided with an explanation of the purpose and expected outcomes of the NCHRP 3-70 project, a draft copy of the survey form, and the route(s) desired to be surveyed. All readily agreed to participate. Under TriMet’s union contract, drivers of buses on which surveys will occur must be notified in advance, which required that specific trips to be surveyed had to be identified well in advance. This requirement did not exist at the other agencies; however, the drivers there were also given advance notice, so that they would be aware that the surveys would be occurring. Surveyors at all sites carried a letter from the transit agency au- thorizing them to be on the bus, in case a driver had any ques- tions. WMATA also required that the names of the surveyors be provided in advance so that they could be listed on the letter. Field Data Collection The following roadway-related information was collected along the entire route: • Average stop spacing (bus stops/mile); • Stop-specific data: – Presence of shelter (yes/no); – Presence of bench (yes/no) [including a bench inside a shelter]; – Presence of sidewalk or path (yes/no); – Presence of ditch or other obstacle between sidewalk and street (yes/no); – Bus stop waiting area separation from auto traffic (curb-tight, sidewalk set back from street, on median or traffic island, off-street); – Street width (lanes); – Median type (raised/painted/none); – Traffic control at stop (signal/all-way stop/bus street stops/side street stops/roundabout/mid-block location/ off-street location); and – Crosswalk type at stop (marked/unmarked/no legal crosswalk). The following transit-related information was collected: • Stop location for each stop on the surveyed routes; • Survey route frequency—peak and midday (bus/h);

55 • Effective frequency on arterial—peak and midday (bus/h); • Survey route service span (h/day); • Effective service span on arterial (h/day); • Scheduled bus arrival/departure times; • Number of seats on the bus; and • Available standee area on bus. “Effective frequency” and “effective span” included all routes along a portion of the urban street serving the same destination as the surveyed route. For example, if the survey route operated two trips per hour during the peak hour, and another route on the same street serving the same destination also operated two trips per hour during the peak hour, the effective peak-hour frequency was four trips per hour. Survey Form Development An initial draft of the survey form was described in the Phase 1 “Transit Data Collection Plan” memo and subse- quently approved by the project panel. In working with TriMet to obtain permission to conduct surveys on their buses, TriMet’s Marketing Information Department offered to review and comment on the survey form, based on their experience conducting on-board surveys. Their review re- sulted in wording changes to some of the questions to shorten the descriptions, while keeping the original meaning. A Span- ish version of the pilot survey was also developed. The revised survey form and the survey procedures were pilot tested on April 22, 2004, on TriMet Line 15. Based on user feedback, the final question on the pilot survey, which asked persons to rank the quality of service factors most important to them, was substantially changed to reduce con- fusion. In addition, the number of factors presented on the final version of the survey was reduced from 29 to 17 by elim- inating factors that received few to no responses among users’ five most important factors. Spanish and large-print (22- point font) versions of the final survey were also developed. TriMet requested that the survey form used on their buses resemble an official TriMet survey, so the TriMet logo, a TriMet-tailored explanation of the survey purpose, and a TriMet information phone number were included on the TriMet survey form. The other agencies wanted the surveys dis- tributed on their buses to not resemble official agency surveys, so a generic survey purpose description resembling the one shown on the initial survey draft was used for those agencies. The final version of the Phase 1 survey forms are given in Exhibit 57, reduced in size from legal-size paper. Each survey was given a unique four-digit serial number, with the first digit indicating the route it was used on: • Portland: Line 15 (pilot test) • Portland: Line 14 • Portland: Line 44 • Northern Virginia: Line 38B • Northern Virginia: Line 2B • Broward County: Line 18 The questions asked on the Phase 1 survey were as follows: 1. The stop where the person would get off the bus. (Survey- ors recorded the stop where each passenger boarded and received a survey, using the survey’s serial number, elimi- nating the need to ask persons where they boarded.) 2. The number of times a week the person rode the bus. 3. The major reason why the person chose to ride the bus (had a car but preferred the bus, chose not to own a car because bus service was available, did not own a car, didn’t drive or know how to drive). 4. The person’s satisfaction with their trip today on the bus, using a 1 (very dissatisfied) to 6 (very satisfied) scale: a. Getting to the bus stop b. Waiting for the bus c. Riding on the bus d. The overall trip 5. The person’s satisfaction in general with the following aspects of the bus route, using the same 1-to-6 scale: a. Close to home b. Close to destination c. Sidewalk connects to stop d. Crossing street to stop is easy e. Shelter is provided f. Bench is provided g. Frequency of buses h. Times of day the route operates i. Reliability of service j. Seat available k. Wait time for bus l. Not overcrowded m. Friendly drivers n. Amount of time to reach destination o. Seat comfort p. Smooth ride q. Temperature inside bus is comfortable 6. Finally, persons were asked to rank up to five factors from the above list that were the most important to them. In Phase 2, the length of the survey was substantially reduced by eliminating questions 5 and 6. As described in Section 3 of this working paper, the Phase 1 responses to these questions confirmed the importance of the quality-of-service factors given in the TCQSM, and it was not thought necessary to con- tinue to ask these questions. In addition, one objective of the Phase 2 surveys was to sample more crowded routes than were sampled in Phase 1, and it was thought that a shorter survey

56 Exhibit 57. Phase 1 Transit Survey Form.

57 SERVICE QUALITY SURVEY 1. Where did you get on this bus? (Street & cross street) 2. Where will you get off this bus? (Street & cross street) 3. How often do you ride Muni? 5 or more days per week 1 – 2 days per week 3 – 4 days per week Less than 1 day per week 4. Could you have used a car for this trip? Yes No 5. How satisfied are you with your trip today on this bus? Please circle a number below for each answer, where 1 means very dissatisfied and 6 means very satisfied. Dissatisfied Satisfied Very dissatisfied Very satisfied a) Getting to the bus stop 1 2 3 4 5 6 b) Waiting for the bus 1 2 3 4 5 6 c) Riding on this bus 1 2 3 4 5 6 e) This bus trip overall 1 2 3 4 5 6 Please take a few minutes to fill out this survey. We want to know how satisfied you are with service on this bus. When finished, please hand this form back to the survey taker. Exhibit 58. Phase 2 Transit Survey Form. would be easier to administer under crowded conditions. Exhibit 58 shows an example of the Phase 2 survey form. In order to cover the different groups in the San Francisco Bay Area adequately, Spanish and Mandarin Chinese versions of the questionnaire were also produced. On the Phase 2 survey form, surveyors wrote a unique bus trip ID number in the upper right-hand corner for all ques- tionnaires collected from that one-way bus trip. This ID number was then used to tie in the individual questionnaire to a unique bus route, direction, and time period. Survey Distribution In Phase 1, surveyors worked in teams of two. Two teams were assigned to a given route in Portland and Virginia, while three teams were assigned in Broward County, due to the longer route length. The teams started during the a.m. peak hour (around 7 a.m.) and rode their assigned bus route back and forth, for approximately 4 hours. The surveyors rode the Portland and Virginia routes from end-to-end and rode the Broward County route for most of its length within the county (about three-quarters of its total length, which ex- tends just over the county line to the north and several miles beyond the county line to the south). On the bus, the first surveyor sat immediately behind the driver and was responsible for (1) handing out surveys and golf pencils as persons boarded the bus, (2) occasionally re- moving surveys and pencils from the collection envelopes, and (3) cleaning up any surveys or pencils that passengers might have dropped on the floor. Surveys were handed out in numerical order. The first surveyor also had self-addressed stamped envelopes to hand to passengers wishing to complete the survey later, as well as large-print versions of the survey for riders with visual impairments. TriMet also required that the first surveyor carry a TriMet-designed card that provided both printed and Braille information instructing riders with visual impairments to call TriMet’s information line to com- plete the survey. For this survey, TriMet’s operators were instructed to take the person’s name, phone number, and a call-back time and to pass the information to the researchers to complete the survey. Although cards were handed out, no one called TriMet to participate in the survey.

58 Location (Route) Distributed Returned Usable Virginia (2B) 186 182 172 Virginia (38B) 181 166 154 Portland (14) 218 204 198 Portland (44) 268 262 255 Florida (18) 306 276 165 San Francisco Muni (1) NA NA 201 San Francisco Muni (14) NA NA 366 San Francisco Muni (30) NA NA 112 San Francisco Muni (38) NA NA 339 San Francisco Muni (38L) NA NA 153 AC Transit (51) NA NA 199 AC Transit (72) NA NA 239 AC Transit (72R) NA NA 101 AC Transit (218) NA NA 24 Total NA NA 2,678 Exhibit 59. Transit Survey Distribution. The second surveyor sat in the first seat on the right side of the bus and was equipped with forms listing all of the bus stops served by the route in each direction. This surveyor was re- sponsible for recording the number of people getting on and off at each stop and for recording the last survey serial num- ber handed out at each stop. One team per route also carried a GPS unit that generated a log file recording the bus’ position and speed every second; the second surveyor was responsible for using it. A new log file was generated for each trip. The GPS unit used was tested in an automobile prior to use and worked as expected; however, when used on buses, the unit sometimes had problems receiving satellite signals. As a result, GPS data were not available for all bus trips. In Portland, the hand- held GPS data were supplemented with archived data from TriMet’s automatic vehicle location system. For the Phase 2 surveys, one surveyor was assigned per door to each bus that was surveyed; for example, two sur- veyors were assigned to regular buses, while three surveyors were assigned to articulated buses. Surveyors handed out a questionnaire to each person boarding the bus. Surveyors also attempted to interview standing passengers, asking them questions while the bus was in motion; consequently, sur- veyors with multi-lingual capability were assigned to specific routes that carry large numbers of non-English-speaking passengers. Exhibit 59 shows the number of surveys distributed and re- turned in each location. Not all returned surveys were filled out completely. For Phase 2, surveyors were unable to keep track of the number of surveys distributed due to the heavy workload on crowded buses. Route Characteristics Routes were selected to create variety in (1) particular route characteristics that previous research had determined to be important (e.g., the TCQSM’s LOS factors), and (2) specific pedestrian environment characteristics not previously re- searched. In addition, the routes selected in Northern Virginia included urban street segments also being used by GMU for the automobile LOS element of this project. In Northern Virginia, WMATA Route 2B starts in the dense urban portions of Arlington (Ballston-MU Metrorail) and travels past sprawling residential neighborhoods, golf courses, cemeteries, office parks, and strip commercial uses in Falls Church and Fairfax. Route 2B stops at several auto- oriented Metrorail stations (i.e., East Falls Church, Dunn Loring-Merrifield, and Vienna/Fairfax-GMU) before com- pleting its trip at the Fair Oaks Shopping Center in Fairfax. Some peak-period trips operate as Route 2G, deviating to serve an AT&T office building, but otherwise serving the same route. The eastern half of the route is duplicated by Route 2C, which effectively doubles the service frequency on that portion of the route. WMATA’s Route 38B also starts at the Ballston-MU Metro- rail station, but travels through the most dense, transit- oriented portions of Arlington, stopping near three Metrorail stations (i.e., Clarendon, Court House, and Rosslyn) sur- rounded by mixed uses and high-rise buildings, before cross- ing the Francis Scott Key Bridge into Washington’s dense Georgetown neighborhood and ending at Farragut Square, just two blocks from the White House. Route 38B is the only all-day WMATA route crossing the Key Bridge, although one of the Georgetown Metro Connection routes also crosses the bridge. In Portland, Route 14 is one of TriMet’s most frequent bus routes (eight buses per hour peak, five buses per hour mid- day) and has one of the longest service spans. Route 14 runs from the I-205/Foster Road interchange in southeast Portland (no park-and-ride provided) into downtown Portland via the Hawthorne Bridge (a drawbridge) and then runs along the downtown bus mall to Union Station. East of downtown, the street frontage is primarily commercial or mixed-use office and commercial in multiple-story buildings. Some low- to medium-density multi-family residential build- ings also front the streets served by the transit route. Past the immediate transit street frontage, land uses are primarily medium- to high-density single-family residential. TriMet Route 44 connects Portland Community College’s Sylvania campus in southwest Portland to downtown and Union Station via the bus mall, passing through the com- mercial districts of Multnomah Village and Hillsdale. Outside the commercial areas, land uses served by the route are a mix of medium-density single-family residential and low-density multi-family residential. A 1-mile section of the route south of Multnomah Village lacks sidewalks. The route is also one of many serving Portland State University at the south end of downtown. Service from Hillsdale into downtown is dupli- cated by several other routes, creating a better effective frequency.

59 TriMet Route 15, used for the pilot test, runs from the Parkrose/Sumner transit center, light rail station, and park- and-ride lot in northeast Portland south to Mall 205 (a small shopping center). The route continues west through down- town Portland (at right angles to the bus mall), past a hospital, and ends in northwest Portland. Alternate trips terminate in a residential area of northwest Portland or at Montgomery Park, a large office building (the surveyed trips went to Montgomery Park). Land uses vary along the route and include medium- to high-density single-family residential, low- to medium-density multi-family residential, and office and commercial uses. In Broward County, Route 18 runs north-south the length of the county along U.S. 441, from the south edge of Palm Beach County to the Golden Glades park-and-ride lot in northern Miami-Dade County, where transit connections can be made for trips continuing south. Service is provided at 15-minute headways during peak and midday periods. A weekday peak-period limited-stop version of the route (Route 18LS, not surveyed) duplicates all but the very northern end of the route and operates at 45-minute headways. Phase 2 sample selection was guided by the results of analy- sis of the Phase 1 data set. Based on that analysis, it was de- termined that the Phase 2 sample should be designed to round out the characteristics of the Phase 1 data set. In par- ticular, we sought a sample that included routes with one or more of the following characteristics: • Moderate to severe crowding, with load factors greater than 1; • High-demand density; • Frequent service; • Operation on reserved bus lanes; and/or • Low frequency. The San Francisco Bay Area has two transit agencies that operate routes with these characteristics. The San Francisco Municipal Railway (Muni) operates service within the City and County of San Francisco, with over 800,000 boardings on an average weekday. The Alameda-Contra Costa Transit District (AC Transit) operates service in the East Bay; cities within its service area include Oakland, Berkeley, Richmond, Hayward, and Fremont. For all routes surveyed on Muni and AC Transit, surveys were taken on bus trips in both directions during the AM peak and PM peak. The following routes were selected for the sample: • The Muni 1 California, a trolley bus, operates from the San Francisco Financial District westbound through Chinatown, Pacific Heights, and Laurel Heights to the Inner Richmond district. Surveyors rode the bus only from the Financial District to Laurel Heights. The bus runs on between 8 and 9-minute headways during the peak periods. • The Muni 14 Mission begins in the downtown Financial District and runs through the Mission District to Daly City. This is Muni’s busiest route, carrying over 60,000 passen- gers on an average weekday. Although articulated buses are used on this route, buses are frequently crowded, with most passengers standing. This bus runs on a priority bus lane (buses and right turns only) along Mission St. to about 10th St., but the bus lane is frequently violated by cars. • The Muni 30 Stockton, a trolley bus, runs from the Caltrain Depot at 4th and Townsend north through the Financial District, then through Chinatown and North Beach to the Marina District. Buses frequently bunch up; because it is a trolley bus, buses cannot pass each other. Buses move espe- cially slowly through Chinatown, where there are large numbers of boardings and alightings. The bus operates on 9-minute headways during the peak periods. • The Muni 38 Geary and 38L Geary Limited operate from the Transbay Terminal along Market St., then along Geary to western San Francisco. Past 33rd Avenue, the route splits into three branches. This is one of Muni’s busiest routes, carrying about 60,000 passengers per day. Buses are articu- lated, but are frequently crowded with most passengers standing. Effective headways east of 33rd Avenue are be- tween 4 and 8 minutes, depending on whether the local or limited service is used. The bus runs on dedicated priority bus lanes (buses and right turns only) from downtown to Van Ness Avenue, but these are frequently violated. • AC Route 51 Broadway, one of the busiest routes on the AC Transit system, runs from Alameda (an island off of Oakland) through the Posey Tube through downtown Oakland, then north along Broadway and College Avenue to Berkeley, and then west to the western part of Berkeley. This route serves the UC Berkeley campus, so many of the riders are UC students or faculty. The bus operates on about 8-minute headways during the peak periods. • AC Route 72 San Pablo runs along San Pablo Avenue from downtown Oakland along San Pablo Avenue through Berkeley to Hilltop Mall in Richmond. The 72 operates on 30-minute headways throughout the day, but service is paralleled by the 72M, which also operates on 30-minute headways on the part of the route that was surveyed. • AC Route 72R San Pablo Rapid is a new rapid bus service that runs from downtown Oakland along San Pablo Avenue through Berkeley to Contra Costa College. Stops are spaced about every half mile. The bus operates on 12-minute headways throughout the day. • AC Route 218 Thornton was chosen to provide a sample on a long-headway route (1 hour). The 218 runs from Ohlone College in Fremont to the Lido Faire shopping center in Newark. Exhibits 60, 61, and 62 summarize key characteristics of the selected routes.

60 Route Characteristic Virginia 2B Virginia 38B Portland 14 Portland 44 Florida 18 Peak frequency (bus/h) 2 4 8 4 4 Off-peak frequency (bus/h) 1 2 5 4 4 Maximum eff. frequency (bus/h) 4 7 38 38 5 Service span (h/day) 16.5 22 20.5 16 19.5 Stops with shelter (%) 13% 29% 34% 30% 23% Stops with bench (%) 15% 26% 47% 41% 75% Street width range (lanes) 2–9 1–7 2–6 1–6 5–9 Stops at traffic signals (%) 40% 68% 48% 40% 48% Stops with sidewalks (%) 89% 99% 99% 81% 88% Stops without legal crosswalks (%) 53% 19% 6% 9% 51% Average load (p/bus) 11 14 10 16 ** Average maximum load (p/bus) 18 28 27 25 ** Maximum load (p/bus) 34 44 37 42 ** **Due to a data collection problem, loads cannot be calculated for all trips. A few surveyed trips had standing loads. Exhibit 60. Route Characteristics—Phase 1. Route Characteristic 1 California 14 Mission 30 Stockton 38 Geary 38 Geary Limited Peak frequency (bus/h) 20 10 7 8 9 Off-peak frequency (bus/h) 10 7 7 8 9 Maximum eff. frequency (bus/h) 20 10 27 10 12 Service span (h/day) 20 24 20 24 14 Stops with shelter (%) 44 54% 44% 68% 84% Stops with bench (%) 44 56% 44% 69% 86% Street width range (lanes) 2 – 5 2 – 4 2 – 7 2 – 8 2 – 8 Stops at traffic signals (%) 58% 91% 54% 63% 75% Stops with sidewalks (%) NA NA NA NA NA Stops without legal crosswalks (%) 2% 1% 1% 3% 0% Average load (p/bus) NA NA NA NA NA Average maximum load (p/bus) NA NA NA NA NA Maximum load (p/bus) NA NA NA NA NA Exhibit 61. Route Characteristics—Phase 2, San Francisco. 4.7 Representation of Survey Results By A Single LOS Grade The automobile, transit, bicycle, and pedestrian surveys produced distributions of LOS ratings for any given condi- tion. However, a distribution of LOS results for any given mode on any given street is less convenient for decisionmak- ers than a single-letter LOS grade as is customary in the HCM). The HCM does not report the distribution of LOS grades for a given situation. The HCM reports a single LOS grade for a given situation. This research project was, there- fore, confronted with the issue of how to convert the distri- bution of LOS grades reported by the public into a single LOS grade for a given situation. In statistics, various single-value measures can be used to represent a distribution. The two most common single-value measures are the “mean” and the “mode” of the distribution. The mode is appealing, because it represents the most fre- quent LOS response of the public. However, the mode has one weakness, in that it is possible for a distribution with two “camel humps” to have two modes. It is possible for both humps to be an identical percentage of the total distribution, in which case, one cannot report a single LOS result. The mean is appealing because it always results in a single LOS grade, regardless of the distribution. However the mean has a major weakness in that it can reach LOS A or LOS F only in the rare cases when there is almost complete agreement by the members of the public that the LOS is A or F. Even if most respondents pick LOS A or F, it takes few dissenters to drag the mean LOS from A or F (see Exhibit 63). As shown for Distribution #1 in Exhibit 63, even when 50% of the people choose LOS A, the mean will still be 1.65, which is closer to 2.00 (LOS B) than to 1.00 (LOS A). The “mode” performs as desired for these example distributions, however; we have ruled it out because of its inability to re- solve a tie. The row labeled LOS 1 converts mean values to a letter grade using the same values of LOS A through LOS F that were used to compute the mean (see Exhibit 64). As can be seen, this approach cannot get LOS A for the mean of

61 Route Characteristic 51 Broadway 72 San Pablo 72R San Pablo Rapid 218 Thornton Peak frequency (bus/h) 8 4 5 1 Off-peak frequency (bus/h) 8 4 5 1 Maximum eff. frequency (bus/h) 8 4 5 1 Service span (h/day) 19 18 14 15 Stops with shelter (%) 28% 39% 74% 11% Stops with bench (%) 51% 46% 75% 15% Street width range (lanes) 2 – 7 2 – 5 2 – 5 2 – 6 Stops at traffic signals (%) 67% 60% 85% 60 Stops with sidewalks (%) NA NA NA NA Stops without legal crosswalks (%) 0% 8% 4% 18% Average load (p/bus) NA NA NA NA Average maximum load (p/bus) NA NA NA NA Maximum load (p/bus) NA NA NA NA Exhibit 62. Route Characteristics—Phase 2, AC Transit. LOS Dist 1 Dist 2 Dist 3 Dist 4 Dist 5 Dist 6 A 50% 25% B 35% 50% 25% C 15% 25% 50% 25% D 25% 50% 25% 15% E 25% 50% 35% F 25% 50% Mean 1.65 2.00 3.00 4.00 5.00 5.35 Mode 1 2 3 4 5 6 LOS 1 B B C D E F LOS 2 B B C D E E LOS 3 A B C D E F Exhibit 63. Example Distributions and Mean of Level of Service. distribution #1, but otherwise, performs reasonably for the other five example distributions. The row labeled LOS 2 converts the mean values to a letter grade using a shifted set of thresholds that divide at the mid- point between each LOS value. Unfortunately, this approach results in Distribution #6 being converted to LOS E, instead of the desired F and does not solve the problem of producing LOS B for Distribution #1, when LOS A is the desired result. The row labeled LOS 3 shows the results when a com- pressed range of thresholds is used to convert the mean values to letter grades. The compressed range squeezes together the thresholds for LOS B to LOS E, so that wider ranges are avail- able for LOS A and LOS F. Under this scheme, LOS A ranges from a mean of 1.0 to a mean of 2.0; LOS F ranges from a mean of 5.0 to 6.0. These larger ranges for the extreme LOS grades ensure that extreme LOS grades will be output for distributions where a large portion of the responses are at the extreme LOS grades. The LOS3 threshold scheme was tested on the auto video clip results and was found to produce a reasonable range of LOS A through F results for the mean LOS values for the video clips that were representative of the distribution of the reported LOS results. This threshold scheme was adopted for reporting the data collection results and for reporting single- letter grade results from the various LOS models developed under this research. LOS Numerical Value LOS 1 Straight Thresholds LOS 2 Thresholds Shifted to Midpoints LOS 3 Compressed Ranges A 1 Mean ≤ 1.00 Mean ≤ 1.50 Mean ≤ 2.00 B 2 > 1.00 to ≤ 2.00 > 1.50 to ≤ 2.50 > 2.00 to ≤ 2.75 C 3 > 2.00 to ≤ 3.00 > 2.50 to ≤ 3.50 > 2.75 to ≤ 3.50 D 4 > 3.00 to ≤ 4.00 > 3.50 to ≤ 4.50 > 3.50 to ≤ 4.25 E 5 > 4.00 to ≤ 5.00 > 4.50 to ≤ 5.50 > 4.25 to ≤ 5.00 F 6 Mean > 5.00 Mean > 5.50 Mean > 5.00 Exhibit 64. LOS Mean Value Threshold Schemes.

Next: Chapter 5 - Auto LOS Model »
Multimodal Level of Service Analysis for Urban Streets 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 616: Multimodal Level of Service Analysis for Urban Streets explores a method for assessing how well an urban street serves the needs of all of its users. The method for evaluating the multimodal level of service (MMLOS) estimates the auto, bus, bicycle, and pedestrian level of service on an urban street using a combination of readily available data and data normally gathered by an agency to assess auto and transit level of service. The MMLOS user’s guide was published as NCHRP Web-Only Document 128.

Errata

In the printed version of the report, equations 36 (pedestrian segment LOS) and 37 (pedestrian LOS for signalized intersections) on page 88 have been revised and are available online. The equations in the electronic (dpf) version of the report are correct.

In June 2010, TRB released NCHRP Web-Only Document 158: Field Test Results of the Multimodal Level of Service Analysis for Urban Streets (MMLOS) that explores the result of a field test of the MMLOS in 10 metropolitan areas in the United States.

  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!