National Academies Press: OpenBook

Multimodal Level of Service Analysis for Urban Streets (2008)

Chapter: Chapter 3 - Literature Review

« Previous: Chapter 2 - State of the Practice
Page 17
Suggested Citation:"Chapter 3 - Literature Review." 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 17
Page 18
Suggested Citation:"Chapter 3 - Literature Review." 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 18
Page 19
Suggested Citation:"Chapter 3 - Literature Review." 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 19
Page 20
Suggested Citation:"Chapter 3 - Literature Review." 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 20
Page 21
Suggested Citation:"Chapter 3 - Literature Review." 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 21
Page 22
Suggested Citation:"Chapter 3 - Literature Review." 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 22
Page 23
Suggested Citation:"Chapter 3 - Literature Review." 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 23
Page 24
Suggested Citation:"Chapter 3 - Literature Review." 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 24
Page 25
Suggested Citation:"Chapter 3 - Literature Review." 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 25
Page 26
Suggested Citation:"Chapter 3 - Literature Review." 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 26
Page 27
Suggested Citation:"Chapter 3 - Literature Review." 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 27
Page 28
Suggested Citation:"Chapter 3 - Literature Review." 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 28
Page 29
Suggested Citation:"Chapter 3 - Literature Review." 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 29
Page 30
Suggested Citation:"Chapter 3 - Literature Review." 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 30
Page 31
Suggested Citation:"Chapter 3 - Literature Review." 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 31

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.

17 This chapter reviews the recent published research into multimodal level of service. The literature review is grouped by research into traveler perceptions of level of service for auto, transit, bicycle, and pedestrian, and research into mul- timodal level of service frameworks. 3.1 Auto Driver Perceptions of LOS Researchers have focused on auto driver perceptions of quality of service for urban streets, signalized intersections, and rural roads. Researchers have used field surveys (where subjects are sent into the field to drive a fixed course) and video laboratories and have laboratory interviews to identify key factors affecting perceived LOS and to obtain LOS ratings for different field conditions. Level of service has been defined by researchers in various ways. For example, LOS A may be defined as “excellent,” “best,” or “very satisfied” depending on the researcher. Others have defined LOS in terms of hazards and conflicts (e.g., num- ber of vehicle-to-vehicle and vehicle-to-pedestrian conflicts). Some have developed models that predict the average LOS rating, while others have developed models that predict the percentage of responses for each LOS grade. Several researchers have noted that drivers do not perceive six levels of service. Some researchers have proposed as few as three levels of service, while one researcher suggested a shift of the entire LOS spectrum by one level of service so as to combine LOS A and B and subdivide LOS F. Some of the latest research incorporates “fuzzy logic” in the translation of user perceptions into letter grade levels of service. Urban Street LOS While the HCM’s focus on measuring delay, percent of time spent following, and average travel speed (to name a few) offers a conceptual link to how the user perceives the transportation system’s level of service, a review of the exist- ing literature by Flannery et al. (2005) [10] found little re- search that empirically investigates these links. Flannery et al. conclude that a comprehensive research approach is needed to identify and prioritize the factors important to drivers fol- lowed by research that models and calibrates these factors. In a study comparing users’ perceptions of urban street service quality, Flannery et al. (2004) [11] found that HCM 2000 methods only predicted 35 percent of the variance in mean driver ratings, suggesting LOS does not completely rep- resent driver assessments of facility performance. Colman [12] sent 50 students to drive various arterial streets and compare the HCM level of service (based on speed) against their own perception of quality of service. The student’s perceived speed thresholds for urban street level of service tended to be 4% to 24% higher than the HCM speed thresholds. They expected better service for a given letter grade than the HCM. Seeking to identify the key factors that influence user per- ceptions of urban street LOS, Pecheux et al. (2004) [13] used an in-vehicle survey and interview approach to determine the factors that affect drivers’ perceptions of quality of service. They identified 40 factors that are relevant to these perceptions, including roadway design, urban street operations, intersec- tion operations, signs and markings, maintenance, aesthetics, and the behaviors of other road users. A study by Flannery et al. (2005) provides support for this collection of important fac- tors. Flannery et al. had drivers rate video segments of travel on urban streets and then select and rank from a list of 36 factors the 3 factors that they considered to be most important to LOS. Mean driver ratings had statistically significant correlations with operational and design characteristics, and aesthetics, in- cluding the following variables: travel time, average travel speed, number of stops, delay, number of signals, lane width, the presence of trees, and quality of landscaping. An FHWA-sponsored study of customer satisfaction (SAIC [14]) sought to determine what factors influence perceived C H A P T E R 3 Literature Review

18 driver satisfaction on urban streets. Drivers drove with two re- searchers in their vehicle and talked aloud about the factors that made them feel satisfied or dissatisfied with the drive they were experiencing in real time. The study was conducted in four locations and one pilot study location. The locations con- sisted of two small urban areas (Tallahassee, Florida, and Sacramento, California) and two large urban areas (Chicago, Illinois, and Atlanta, Georgia). In each location, routes re- quiring approximately 30 to 40 minutes of drive time were selected. Each of the routes incorporated characteristics in- cluded in Exhibit 26, taken from the HCM 2000. In small urban areas, the focus was on suburban and intermediate characteristics; in large urban areas, the focus was on inter- mediate and urban characteristics. Twenty-two participants were in the four study locations; their characteristics are described in Exhibit 27. The findings from this study resulted in 42 Quality of Ser- vice (QOS) factors for urban streets that can be categorized into several investment areas. Exhibit 28 contains the identi- fied factors according to driver transcripts and completed surveys. The researchers further refined the identified QOS factors into nine proposed measures of effectiveness (MOEs) shown in Exhibit 29. The proposed MOEs reflect the input provided by the participants in the study, but combine like QOS fac- tors into, for the most part, measurable performance meas- ures. For example, participants in the study often commented negatively when they were forced to slow down or stop be- cause of poor arterial design that did not provide for bus pull- outs, turning facilities, on-street parking maneuvers, and poor access management that created many merge/diverge situations. The authors of this study have concluded that the MOE number of stops best represents the views of the partic- ipants in this study. Intersection LOS Research Sutaria and Haynes [15] focused on determining the different levels of service at signalized intersections. The re- searchers investigated 30 signalized, isolated, fixed-time intersections in the Dallas-Fort Worth area and determined that only 1 intersection experienced the full range of LOS cat- egories (then based on Load Factor defined as the ratio of the total number of green signal intervals fully utilized by traffic during the peak hour to the total number of green intervals). The intersection of Lemmon and Oaklawn Avenues in Dallas Route (Design) Category Criterion Suburban Intermediate Urban Driveway/access density Low density Moderate density High density Arterial type Multilane divided; undivided or two-lane with shoulders Multilane divided or undivided; one-way two-lane Undivided one-way, two-way, two or more lanes Parking No Some Significant Separate left-turn lanes Yes Usually Some Signals/mile 1-5 4-10 6-12 Speed limit 40-45 mph 30-40 mph 25-35 mph Pedestrian activity Little Some Usually Roadside development Low to medium density Medium to moderate density High density Exhibit 26. Route Characteristics. Field Site Number of Participants Ages Se x Northern Virginia (Pilot location) 4 2 20 - 30 year olds 2 35 - 50 year olds 2 women 2 men Chicago 5 2 20 - 30 year olds 3 35 - 50 year olds 0 60 - 75 year olds 3 women 2 men Tallahassee 5 1 20 - 30 year old 2 35 - 50 year olds 2 60 - 75 year olds 3 women 2 men Atlanta 6 0 20 - 30 year olds 3 35 - 50 year olds 3 60 - 75 year olds 3 women 3 men Sacramento 6 1 20 - 30 year old 3 35 - 50 year olds 2 60 - 75 year olds 4 women 2 men Exhibit 27. Participant Characteristics.

19 Investment Area QOS Factor Cross-Section Roadway Design Lane width Pedestrian/bicyclist facilities # of lanes/roadway width Bus pull-outs Turning lanes/bays Parking Lane drop/add Access management Medians Two-way center left turn lane Arterial Operations Number of traffic signals Presence of large vehicles Volume/congestion Travel time Traffic flow Speed Intersection Operations Signal failure/inefficient signal timing Turning Timing of signals Traffic progression Signs and Markings Quality of pavement markings Advance signing Lane guidance—signs Too many signs Lane guidance—pavement markings Sign legibility/visibility Sign presence/usefulness Maintenance Pavement quality Overgrown foliage Aesthetics Presence of trees Medians with trees Visual clutter Cleanliness Roadside development Other Road Users Illegal maneuvers Careless/inattentive driving Driver courtesy Use of turn signals Aggressive drivers Pedestrian behavior Improper/careless lane use Blocking intersection Other Intelligent transportation systems Roadway lighting Planning Exhibit 28. Driver-Identified QOS Factors For Urban Streets. MOEs QOS Factors Number of stops Turning lanes/bays Bus pull-out areas On-street parking Two-way center left-turn lane Access management Lane drop/add Urban street capacity Heavy vehicles Lane width Number of lanes/roadway width Intersection efficiency Signal timing (cycle length/cycle split) Provision for turning vehicles Urban street efficiency Progression Number of traffic signals Travel time Travel speed Traffic volume Volume/congestion Traffic flow Speed Travel time Positive guidance Quality of pavement markings Sign legibility/visibility Sign presence/usefulness Lane guidance—signs Lane guidance—pavement markings Advance signing Too many signs (clutter/distracting) Visual clutter Pavement quality Pavement quality Perceived safety Presence of medians Lane width Pedestrian/bicycle facilities Access management Area type Roadside development Cleanliness Trees Visual clutter Exhibit 29. Proposed MOEs For Urban Streets.

20 Rating Description 5 = Excellent 4 = Very Good 3 = Good 2 = Fair 1 = Poor 0 = Very Poor Description Rating of Quality of Service I would describe the traffic situation presented in this film segment as a condition of: Free flow or as “free flowing” as can be expected if there is a traffic signal at the intersection under study. OR Tolerable delay, and nearly as good as could be expected at a signalized intersection. OR Considerable delay but typical of a lot of ordinary signalized intersections during busy times. OR Unacceptable delay and typical of only the busiest signalized intersections during the rush hour. OR Intolerable delay and typical only of the worst few signalized intersections I have seen. Exhibit 30. Sutaria and Haynes Scale A —Point Rating. Exhibit 31. Sutaria and Haynes Scale B – Descriptive Rating. was filmed using 16mm cameras for several hours to gather several film clips ranging from A to E Level of Service. For the study, 14 film clips, ranging from 42-193 seconds, were shown to the participants. The film clips were broken into two groups: microviews that showed the traffic situation from the view of an individual driver seated in an automobile and macroviews that showed the overall traffic situation on a given approach from high above. Seven clips, ranging from LOS A to LOS E, in each group were shown to participants. There were 310 participants in the study. The participants were given a questionnaire about their perceptions of signal- ized intersections before viewing the films collected in the field. The participants were asked to indicate, in order of impor- tance, the factors that affect their perceived views of quality of flow at signalized intersections. They were given five factors to rank: delay, number of stops, traffic congestion, number of trucks/buses, and difficulty in lane changing. It does not appear that definitions of the factors were provided to the participants. Before viewing the films, the participants ranked the fac- tors as follows: 1. Delay, 2. Number of stops, 3. Traffic congestion, 4. Difficulty in lane changing, and 5. Number of trucks/buses. After viewing the films, the rankings changed slightly as follows: 1. Delay, 2. Traffic congestion, 3. Number of stops, 4. Difficulty in changing lanes, and 5. Number of trucks/buses. After viewing each of the 14 film clips, the participants were also asked to score the service quality of the various film seg- ments on two different opinion scales: a 6-point scale (Scale A) and one of five descriptions (Scale B) (See Exhibits 30 and 31.) Based on input gathered from this study, the researchers developed a nomograph that depicted the relationship between Average Intersection Delay (AID), Load Factor (LF), and volume-to-capacity ratio (v/c) to perceived or rated level of service. The researchers went on to make three recom- mendations: • AID should be used to predict level of service. • Similar studies should be conducted on signalized inter- sections without full actuation. • Simultaneous filming and field studies should be con- ducted to allow for accurate measurement of traffic engi- neering measures captured on film. Based on the findings of this single research study, the Highway Capacity and Quality of Service Committee over- hauled the 1985 HCM to represent level of service at signal- ized intersections by AID versus LF. The authors state, “Field studies and the attitude survey provided data for the development of two psychophysical models. Statistical analysis indicated that average individual delay correlated better with level of service rating than with measured load factor and encompassed all levels of service. Of all parameters affecting levels of service, load factor was rated highest by road users.” Ha, Ha, and Berg [16] developed models for predicting the number of conflict opportunities (potential conflicts) at an intersection as a function of signal timing, intersection geom- etry, and turn volumes. Based on a review of previous inves- tigations, they limited their analysis to left-turn and rear-end accident analyses. The “total hazard” at an intersection is the sum of the likely number of rear-end and left-turn accidents multiplied by their severity. The total hazard is converted to a hazard index by dividing by the number of vehicles. The

21 letter-grade level of service is then determined from a hazard index look-up table. Zhang and Prevedouros [17] developed a model of vehicle-to-vehicle and vehicle-to-pedestrian conflicts and blended it with the existing HCM delay LOS criteria for sig- nalized intersections to obtain an LOS model that combines safety risk with traditional delay measures of LOS. Two delay and safety indices are computed—one for pedestrians, the other for vehicles. Each index is computed as a weighted sum of potential conflicts and delay. The weights are analyst spec- ified. The two indices are then weighted by pedestrian and ve- hicle volumes, respectively, to obtain a weighted average delay and safety index for the intersection. No surveys of traveler perception were performed. This paper was oriented toward methodological approaches rather than traveler perception. Several recent studies of intersection LOS have also cast some doubt on the HCM’s methods. Zhang and Prevedouros (2004) [18] investigated motorists’ perceptions of LOS at case study signalized intersections and found that, although the HCM 2000 predicts that permitted left-turn phases provide a higher level of service, users ranked protected left-turn phased intersections higher. This finding suggests that users may be including the perceived safety benefits of protected phasing at these case study locations in their assessments of LOS, in addition to delay. In a follow-up study, Zhang and Prevedouros (2005) [19] surveyed users’ perceptions of ser- vice quality at intersections and found that users consider multiple factors beyond delay (as calculated by the HCM 2000), including signal efficiency, left-turn treatment, and pavement conditions. Delay scored relatively low among im- portant factors. Drivers prefer to make left turns under pro- tected left-turn signals, especially at large intersections. Safety was stated to be 3 to 6 times more important than delay, de- pending on the type of conflict. The importance of safety in determining the level of service offered by an intersection is reflected in a study by Li et al. (2004) [20]. Li et al. used a “gray system” theory-based method to rank and evaluate the operational and safety performance of signalized intersections in mixed traffic con- ditions. The degree of saturation, average stopped delay, queue length, conflict ratio, and separation ratio are all used as parameters. Results of application in the urban area of Changsha, China, show that the method can be used to con- duct a comprehensive (safety and operations) performance under mixed traffic conditions. Pecheux, Pietrucha, and Jovanis [21] addressed users’ per- ception of level of service at signalized intersections. The re- search objectives were to examine delay distributions, assess the accuracy of delay estimates, determine if current levels of service are appropriate, and identify factors affecting per- ceptions. The research used a video laboratory to show 100 participants (in groups of 7 to 10) a tape of a series of signal- ized intersections. The intersections portrayed on the tape were chosen in cities outside the local area to eliminate fa- miliarity by the subjects, but in a location nearby so that local conditions were represented. The results of the study showed that, on average, subjects’ delay estimates were fairly accurate, but widely variable on an individual basis. The study also showed that subjects perceived three or four levels and were more tolerant of delays than suggested by the HCM. At least 15 factors emerged from the group discussions that subjects identified as influential in their LOS ratings. These included delay, traffic signal efficiency, arrows/lanes for turning vehi- cles, clear/legible signs and road markings, geometric design of intersection, leading left-turn phasing scheme, visual clutter/distractions, size of intersection, pavement quality, queue length, traffic mix, location, scenery/aesthetics, and presence of pedestrians. Use of Fuzzy Logic for LOS Modeling Recent research has begun to use “fuzzy logic” to identify delay thresholds for rating the level of service of signalized intersections. Fang and Pecheux [22] conducted a video laboratory of 98 subjects assessing the quality of service on 24 signalized in- tersection approaches. Cluster analysis (employing fuzzy thresholds) revealed that their subjects’ quality of service as- sessments did not distinguish between the delays at HCM LOS A or B. The LOS ratings of their subjects, however, did distinguish two classes of delay for delays at HCM LOS F. Zhang and Prevedouros [23] conducted a web-based stated preference survey of 1,300 volunteers. Their survey identified delay, pavement markings, presence of exclusive left-turn lanes, and protected left-turn phases, as factors sig- nificantly affecting the perceived level of service at a signal- ized intersection. Fuzzy inference was used to identify a distribution of LOS responses for a given physical condition. A percent confidence level was then reported for each LOS letter grade. Lee, Kim, and Pietrucha [24] exposed 27 subjects to video clips of 12 signalized intersections. Subjects were asked to (1) rate their intersection experience as “poor,” “acceptable,” or “good” and (2) describe the relative importance of six criteria to their rating of the intersections. The six criteria evaluated were delay, gaps in cross street traffic while waiting, efficiency of traffic signal operation, visibility of signal, signing/markings, and physical features of the intersection. Rural Road Research Nakamura, Suzuki, and Ryu [25] conducted a field driving survey on a rural motorway section under uncongested traffic

22 flow conditions and measured the driver’s satisfaction with the road. The test area was a 9.3-km, 4-lane, rural basic motorway section between an on-ramp and an off-ramp. Twenty-four participants drove subject vehicles in both directions in the study segment for a total of 105 test runs. Videocameras were mounted on the test vehicle to record travel time, number of lane changes, time of a car-following situation by lane, and elapsed travel time by lane. The factor that most influenced driver satisfaction was traffic flow rate. The number of lane changes, the elapsed time of a car- following situation, and the driver’s experience also affected the driver’s evaluation of traffic conditions. 3.2 Transit Passenger Perceptions of LOS Recent transit LOS research has focused on developing methods that incorporate more than just the characteristics of the available transit service, but measures of the environ- ment in which that service operates. Fu et al. [26] developed a Transit Service Indicator (TSI) that recognizes that quality of service results from the interaction of supply and demand. The proposed index uses multiple performance measures (e.g., service frequency, hours of service, route coverage, and various travel-time components as well as spatial and tempo- ral variations in travel demand). Tumlin et al. [27] developed a method that assesses transit performance in the context of different transportation environments. Quality of service cri- teria and scores reflect system performance in each area as well as provide for an aggregate measure of transit quality of service. Other transit LOS research efforts have focused on devel- oping or refining measures that can be easily calculated using existing transit agency data sources. Xin et al. [28] applied the recent edition of the TCQSM to evaluate the quality of transit service on several travel corridors in an urbanized area. Findings indicate that TCQSM measures (e.g., service frequency, hours of service, service coverage, and transit-auto travel time) are sensitive to planning/ design variables (e.g., service headway, route structure, and service span) and, therefore, can be easily calculated by tran- sit agencies using readily available data. Furth and Muller [29] noted that traditional transit service quality measures analyze waiting time and service reliability separately, under- estimating the total costs of service unreliability which cause patrons to budget extra time waiting for transit to account for unreliability. Using AVL data, actual plus bud- geted waiting time were measured and converted to costs. Findings indicate that service reliability improvements can reduce waiting cost as much as large reductions in service headways. A Handbook for Measuring Customer Satisfaction Morpace [30] presents a methodology for measuring cus- tomer satisfaction on an ongoing basis and the development of transit agency performance measures in response research findings. The authors point out that the results of a customer satis- faction measurement program cannot be expected to drive agency decisions. Agency personnel must choose between improvements to address customer expectations and better education of customers about service parameters. They state the premise that, “Customers must always be first, [but] cus- tomers may not always be right”. They identify 10 Determinants of Service Quality, which are applicable to most service industries. The contention is that consumers use basically similar criteria in evaluating service quality. The 10 criteria are as follows: 1. Reliability (consistent and dependable); 2. Responsiveness (timeliness of service, helpfulness of employees); 3. Competence (able to perform service); 4. Accessibility; 5. Courtesy; 6. Communication; 7. Credibility; 8. Security; 9. Understanding the Customer; and 10. Tangibles. They identify four transit market segments: 1. Secure customers very satisfied, definitely would repeat, definitely would recommend; 2. Favorable customers; 3. Vulnerable customers; and 4. At-risk customers. They recommend that telephone benchmark surveys be used to establish baseline customer satisfaction with the tran- sit service. These surveys are fairly expensive, so they also rec- ommend a simpler survey approach, based on “impact scores,” be used for tracking progress regularly. The “impact score survey” is administered on-board and distributed to transit riders annually or biennially. The goal is to identify those attributes that have the greatest negative effect on overall customer satisfaction and also affect the greatest number of customers. They suggest the use of an “Impact Score Technique” to identify the effect on customer satisfaction of “Things Gone Wrong” with the service. The score weights the effect of a

23 service problem on customer satisfaction by the percentage of customers experiencing the service problem. The resulting score gives the expected change in the customer satisfaction index for the operator. The steps to developing an impact score system are as follows: 1. Identify attributes with most impact on overall customer satisfaction. Compute gap scores. 2. Identify percent of customers who experienced the prob- lem. 3. Create a composite index by multiplying gap score by incidence rate. Result is attribute impact score. Example: Overall satisfaction rating for attribute 1 is 6.5 for those experiencing problem past 30 days 8.5 for those with no problem past 30 days The Gap score is 2.0 (8.5 − 6.5). If 50% of customers report having the problem, then the composite impact score is 2.0 * 50% or 1.00. A Guidebook for Developing Transit Performance Measurement System Although focused on implementing and applying a transit performance-measurement program, the Guidebook for Developing a Transit Performance Measurement System [31] provides useful information on more than 400 transit performance measures (including some for which levels of service have been developed) and on various means of measuring transit performance. The processes of developing customer satisfaction surveys and passenger environment surveys (a “secret shopper” approach to evaluating comfort- and-convenience factors) are summarized. Performance measures discussed in the guidebook cover the passenger, agency, community, and driver/vehicle points of view. Twelve case studies are presented in the guidebook on how agencies measure performance; 18 additional case studies are presented in a background document provided on an accompanying CD-ROM. Application of Transit QOS Measures in Florida Perk and Foreman (2001) [32] evaluated the process and results of the first year’s application of the quality of service measures contained in the TCQSM by 17 metropolitan plan- ning organizations (MPOs) in the state of Florida. Each MPO evaluated transit LOS in terms of service coverage, service fre- quency, hours of service, transit travel time versus auto travel time, passenger loading, and reliability. The evaluation procedure balanced comprehensiveness (covering as much of the area as possible) with cost. Service coverage and transit-auto travel time were evaluated for the system. For the remaining measures, 6 to 10 major activity centers within the region, resulted in 30 or 90 combinations of trips between activity centers. Service frequency and hours of service were evaluated for all origin-destination (O-D) combinations. Passenger load and on-time performance data were collected for the 15 O-D combinations that had the highest volumes (total of all modes), as determined from the local transportation planning model. Transit travel times, hours of service, and frequencies were obtained from local transit schedules. The travel demand between centers was obtained from the local travel model. Field measurements were required to obtain reliability data and passenger load- ing data. The authors point out that there were issues with the se- lection of major activity centers, including a general bias to- ward selecting for analysis those activity centers with the best existing transit service for analysis. The authors also found that the activity center selection method resulted in work ends of trips being over-represented and home ends of trips being under-represented There were also various issues with the difficulty and cost of data collection (e.g., the validity of mixing field data on passenger loads and transit travel times with model estimates of travel times and demand for the computation of some of the level of service measures). Training to improve consis- tency and reduce wasted efforts was also necessary. There was a strong concern about the costs of collecting and processing the data without receiving additional state funding to cover those costs. MPO-estimated costs ranged from “negligible” to $50,000, with most in the $4,000-$5,000 range. The $50,000 cost reflects an MPO that waited until the last minute to start the work and ended up contracting the work out. 3.3 Bicyclist Perceptions of LOS Researchers have used various methods to measure bicy- clist satisfaction with the street environment. Methods have included field surveys (e.g., having volunteers ride a desig- nated course), video laboratories, and web-based stated pref- erence surveys. One researcher intercepted bicycle riders in the middle of their trip in the field. Petritsch et al. [33] compared video lab ratings with field ratings of segment LOS and found they were similar. Some researchers have asked bicyclists which factors are most important to the perception of quality of service. Other

24 researchers have derived the factors by statistically fitting models of level of service to the bicyclist-reported level of service. Some researchers have used both methods. Many researchers have fitted models that predict the mean level of service that would be reported by bicyclists. Some have fitted ordered cumulative logit models that predict the percentage of bicyclists who will report a given LOS grade. The final LOS grade is then the one for which at least 50% of the responses were equal to or greater than that LOS grade. Some researchers (particularly the FDOT-sponsored research—See Landis for example) have defined LOS A as being the best and LOS F as being the worst. Others have de- fined LOS A as being “very satisfied” and LOS F as being “very unsatisfied” (see the Danish research reported by Jensen, below). Zolnick and Cromley [34] developed a bicycle LOS model based on the probability of bicycle/motor vehicle collision frequency and severity. One pair of researchers (see Stinson and Bhat below) sought to obtain measures of bicy- cle perceptions of quality of service by asking route choice questions. Their theory was that bicyclists will select the route that gives them the greatest satisfaction. Most of the research has focused on predicting bicycle level of service for street segments between signalized intersec- tions. A few research projects have focused on predicting the overall arterial street level of service. An Arterial LOS Model Based on Field Surveys and Video Lab Petritsch et al. [35] developed an arterial LOS model for bi- cyclists based on a mix of video laboratory and field surveys. LOS observations were obtained from 63 volunteers who rode the 20-mile course in Tampa, Florida, in November 2005. An LOS rating was obtained for each of the 12 sections of the course. A total of 700 LOS ratings were obtained. The average ratings for each section rated in the field ranged from LOS B to LOS E. The volunteers identified bike lanes, traffic volume, pave- ment condition, and available space for bicyclists as their most important factors for rating section LOS. The recom- mended arterial LOS model for bicyclists is as follows: BLOS Arterial = 0.797 (SegLOS) + 0.131 (unsig/mile) + 1.370 (Eq. 5) Where SegLOS = the segment level of service numerical rating (A ≤ 1.5, B ≤ 2.5, C ≤ 3.5, D ≤ 4.5, E ≤ 5.5) Unsig/mile = Number of two-way stop controlled intersec- tions per mile (arterial does not stop). SegLOS = 0.507 * ln(Vol15/lane) + 0.199 SPt (1 + 10.38 HV)2 + 7.066 (1/PC5)2 + −0.005 (We)2 + 0.760 (Eq. 6) Where Vol 15 = volume of directional traffic in 15-minute time period L = total number of through lanes SPt = effective speed limit (see below) = 1.12ln(SPP -20) + 0.81 And SPP = Posted speed limit (mi/h) HV = percentage of heavy vehicles PC5 = FHWA’s five point surface condition rating We = average effective width of outside through lane Petritsch [36] documented the video laboratory portion of the research. Seventy-five volunteers were shown video of eleven sections. The total viewing time for the video was 47 minutes. Comparison of the 615 LOS ratings by the video and the field participants found that the null hypothesis that there was no difference in the mean ratings between the field and video lab participants could not be rejected at the 5% probability of a Type I error (rejecting the null hypothesis when it is really true). Segment LOS Models Based on Field Surveys or Video Lab Jensen [37] showed 407 people video clips of 56 roadway segments (38 rural, 18 urban) in Denmark. A total of 7,724 LOS ratings were obtained for pedestrian LOS. Another 7,596 LOS ratings were obtained for bicycle LOS. A 6-point satis- faction scale was used (very satisfied, moderately satisfied, a little satisfied, a little dissatisfied, moderately dissatisfied, very dissatisfied). Jensen noted that walking against traffic, sounds other than traffic, weather, and pavement quality all affected perceptions of either bicycle or pedestrian LOS, but these variables were dropped from the model because they were not considered useful to the road administrators who would apply the models. Cumulative logit model forms were se- lected for both the bicycle and pedestrian LOS models. These models predicted the percentage of responses for each of the 6 levels of service. The single letter grade LOS for the facility was determined by the worst letter grade accounting for over 50% of the predicted responses for that letter grade and better (For example, if over 50% responded LOS B or better and less than 50% responded LOS A, then the segment LOS was B). Landis et al. [38] documented a field survey of 60 bicy- clist volunteers riding a 27-km (17-mi) course, in Orlando, Florida. The course included 21 intersections, of which 19 were signal controlled, 1 stop controlled, and 1 a roundabout. The volunteers ranged from 14 to 71 years of age (individuals 13 years and under were prohibited from participating because of safety concerns); 34 percent of the volunteers were female. Most of the volunteers were “experienced” bicycle

25 LOS Model Score A ≤ 1.5 B > 1.5 and ≤ 2.5 C > 2.5 and ≤ 3.5 D > 3.5 and ≤ 4.5 E > 4.5 and ≤ 5.5 F > 5.5 Exhibit 32. Correspondence Between LOS Grade and LOS Numerical Score (Landis). riders (i.e., those riding more than 200 miles per year). Riders with over 1,000 miles per year of riding experience represented a disproportionate share of the volunteers. The course consisted of roadways ranging from two to six lanes with average daily traffic (ADT) from 800 to 38,000 ve- hicles per day on the day of the survey. The percentage of trucks ranged from zero to 8.1. The posted speed limits ranged from 25 to 55 mph. Participants were given a score card to carry with them and instructed to “circle the number that best describes how com- fortable you feel traveling through the intersection” immedi- ately after crossing each subject intersection. The researchers defined Level A for the participants as “the most safe or com- fortable.” Level F was defined for the participants as “the most unsafe or uncomfortable (or most hazardous).” Videocameras were used to record (1) participant numbers and time at each intersection and (2) traffic conditions at the actual moment when the rider crossed the intersection. Ma- chine road tube counters were used to collect volumes at the time of the survey. Turn-move counts were also collected on the day of the survey. Participant starts were spaced so that bicycle-to-bicycle in- terference would not influence the LOS ratings. The letter grades were converted to numerical values (e.g., A = 1, F = 6) (see Exhibit 32) and a hypothesis test was per- formed to determine if sex had a significant effect on the mean LOS ratings. The mean rating for the 20 female partic- ipants was 2.86. For the 39 male participants, the mean rating was slightly lower—2.83 (The lower rating implies better perceived LOS). A t-test indicated that this difference was not significant at the 5% Type I error level. A second hypothesis test was made for delay. The 26 riders having to stop for the signal gave the intersections an average 2.93 rating, while the 33 not stopping rated the intersections 2.94 (the higher rating implied worse perceived LOS). This dif- ference was also insignificant at the 5% Type I error level. Those stopping at a signal were delayed an average of 40 seconds. A third test was for the effect of rider experience. The 55 experienced bicyclists reported an average LOS rating of 2.80. The four inexperienced cyclists reported an average LOS rating of 3.42 (the higher rating implied worse perceived LOS). This difference was found to be statistically significant. However, the four inexperienced cyclists’ results were in- cluded with the experienced cyclists’ results for the purpose of model development. The level of service model is as follows: LOS = −0.2144Wt + 0.0153CD + 0.0066(Vol15/L) + 4.1324 (Eq. 7) Where LOS = perceived hazard of shared-roadway environment for bicyclists moving through the intersection. Wt = total width of outside through lane and bike lane (if present). CD = crossing distance, the width of the side street (in- cluding auxiliary lanes and median) Vol15 = volume of directional traffic during a 15-minute time period. L = total number of through lanes on the approach to the intersection. The researchers reported a correlation coefficient (R-square) of 0.83 against the average repored LOS for each of 18 signal- ized intersections. The table below shows the author’s pro- posed correspondence between LOS letter grade and the scores reported by the volunteers. The authors selected the breakpoints. They are not based on an analysis of the reported scores. The lowest possible score that an individual could report was 1.00, so a preponderance of 1.00 responses was required for the average response to be less than 1.5. It was harder to get LOS A or LOS F than the other levels of service, because A and F require more agreement among the respondents than for the other levels of service. Harkey, Reinfurt, and Knuiman [39] developed a model for estimating bicycle level of service, based on users’ percep- tions. The model, known as the Bicycle Compatibility Index (BCI), was designed to evaluate the ability of urban and sub- urban roadways to accommodate both motor vehicles and bicyclists. The study included 202 participants, ranging from 19 to 74 years of age; approximately 60 percent were male. The expertise level of the participants ranged from daily com- muters to occasional recreational riders. The participants were surveyed in Olympia, Washington; Austin, Texas; and Chapel Hill, North Carolina. The study consisted of showing participants a series of stationary camera video clips taken from 67 sites in • Eugene and Corvallis, Oregon; • Cupertino, Palo Alto, Santa Clara, and San Jose, California; • Gainesville, Florida; • Madison, Wisconsin; and • Raleigh and Durham, North Carolina.

26 The video clips showed various characteristics, including a range of curb lane widths, motor vehicle speeds, traffic vol- umes, and bicycle/paved shoulder widths. Participants were asked to rate their comfort level based on a 6-point scale in the following categories: volume of traffic, speed of traffic, width or space available for bicyclists, and overall rating. In the end, eight variables were found to be sig- nificant in the BCI regression model: • Number of lanes and direction of travel; • Curb lane, bicycle lane, paved shoulder, parking lane, and gutter pan widths; • Traffic volume; • Speed limit and 85 percentile speed; • Median type (including two-way left turn lane); • Driveway density; • Presence of sidewalks; and • Type of roadside development. Given that this research was done in a laboratory setting, the subjects could not take into account the comfort effects of pavement condition, crosswinds, and suction effects caused by high-speed trucks and buses. These factors consequently either do not show up or show up to a lesser extent in the BCI model. Landis et al. [40] conducted a field survey of nearly 150 bi- cyclists who rode a 27-km (17-mile) course in Tampa, Florida. The subjects ranged in age between 13 and over 60 years of age, with 47 percent being female and 53 percent being male. The range of cycling experience was also broad—25 percent of the participants rode less than 322 km (200 miles) yearly to approximately 39 percent of the participants riding over 2,414 km (1,500 miles) yearly. In the study, participants were asked to evaluate the quality of the roadway links, not the in- tersections, on a 6-point scale (A to F) as to how well they were served as they traveled each segment. They were asked to only include conditions within or directly adjoining the right of way and to exclude aesthetics of the segments. Several significant factors were found to influence bicyclists’ perceived quality of service or perceived hazard rating: • Volume of directional traffic in 15-min period; • Total number of through lanes; • Posted speed limit; • Percentage of heavy vehicles in the traffic stream; • Trip generation intensity of the land adjoining the road segment; • Effective frequency per mile of non-controlled vehicular access (e.g., driveway and on-street parking spaces); • FHWA’s five-point pavement surface condition rating; and • Average effective width of the outside through lane. Between the two bicycle quality of service studies, the lab- oratory study conducted by FHWA found very similar factors that influenced quality of service ratings. However, the field studies revealed variables that would be difficult to simulate in a laboratory setting, such as percentage of heavy vehicles and pavement surface condition. The participants in the field study rode alongside traffic and rated the percentage of heavy vehicles as one of the top important factors followed by the condition of the pavement. This comparison of data collec- tion opportunities is the only one that can be made at this time for similar modes of travel, but may provide insight into the limitations of laboratory studies as compared with field studies. Measuring LOS Through Route Choice Stinson and Bhat [41] conducted a web-based stated- preference survey of 3,145 individuals. The individuals were recruited through announcements placed with 25 bicyclist- oriented listservers in the United States. Additional an- nouncements were made to a few non-bicyclist-oriented e-mail lists. The sample of respondents was heavily weighted toward members of bicycling groups. The authors identified 11 link and route attributes (each with multiple levels) for testing. To avoid participant over- load, no more than four attributes were considered in any given survey instrument; thus, nine different instruments were required so as to cover the full range of attributes (and levels) of interest. The respondent characteristics were as follows: • 91% were experienced bicycle commuters. • 22% were female. • About 9% lived in rural areas, 39% lived in urban areas, the rest of the respondents lived in suburbs. Stinson and Bhat identified travel time as the most impor- tant factor in choosing a route, followed by presence of a bicycle facility (striped lane or a separate path). Road class (arterial or local) was the third most important factor. Stinson and Bhat obtained 34,459 observations of route choice and found that the best model of route choice consid- ered the interactions between the bicyclist characteristics (e.g., age, residential location, and experience bicycling) and the route attributes. Stinson and Bhat noted however that the attributes of the route had a greater effect on route choice than the characteristics of the bicyclists themselves. Models of Rural Road Bicycle LOS Jones and Carlson [42] developed a rural bicycle compat- ibility index (RBCI) following a similar approach as that used

27 to generate the FHWA BCI (see Harkey). They employed a web-based survey consisting of questions and thirty-two 30-second video clips. The 30-second video clips were edited from 15-minute videos shot with image stabilization from a car moving 10 mph at a height 4.5 feet above the ground. Given that overtaking motor vehicle traffic tended to give wide clearance to the slow moving car on the shoulder, the video clips tended to show over-taking vehicles giving bicyclists more clearance than they would in reality. The clips were digitized in Windows Media Player compressed format for easy downloading by survey participants. Participants for the web-based survey were recruited through letters to various bicycle groups, flyers distributed at popular recreational bicycling facilities, and personal recruit- ing by the authors. A total of 101 participants (of which 56 were classified as ex- perienced) successfully completed the survey. The experience level of the respondents was determined by induction from the responses to a few key questions. Slightly fewer than 20% of the respondents were female. None were under 18 years of age. Three linear regression models (one for experienced riders, one for casual riders, and one for all riders) were fitted to the mean responses for each video clip. The best model included all bicyclists. The compatibility index in this model was a function of only two factors: shoulder width, and the volume of heavy vehicles traveling in the same direction as the bicy- clist. The model had an R-square value of 0.67. Jones and Carlson intentionally excluded pavement con- dition from the survey because of various data difficulties (in- cluding the difficulty of representing rough pavement in a video shot from a camera mounted on a car). All sites had rel- atively level grades, only two traffic lanes, and speed limits in excess of 50 mph. Noel, Leclerc, and Lee-Goslin [43] recruited bicyclists al- ready using various rural routes to participate in a survey of bicycle compatibility. A total of 200 participants were re- cruited at 24 sites. Bicyclists were stopped at the start of each test segment and asked to participate in the study. Those con- senting were then interviewed to determine their characteris- tics (e.g., age and city of residence). Participants were given segment and junction rating cards to evaluate six sites on each segment. The cards were collected at the end of the segment and the participants were then asked about various potential factors affecting safety at the junctions. The respondents were grouped into three experiential types: sport cyclists, moderate cyclists, and leisure cyclists. The survey found the following key factors affecting per- ceived comfort and safety (ranked by order of importance): riding space available to cyclist, traffic speed, presence of heavy vehicles, pavement conditions, presence of junctions, and finally, vertical profile of the route. The proposed CRC index includes the following variables: • Quality of Paved Shoulder; • Size of Cycling Space; • Auto Speed; • Auto Flow; • Truck Flow; • Roadside Conditions (e.g., sand, gravel, and vegetation); • Roadside Development; • Vertical Profile; • Longitudinal Visibility; and • Major Intersections. 3.4 Pedestrian Perceptions of LOS Researchers have used field intercept surveys and closed course surveys in the field to measure pedestrian perceptions of level of service. Some distributed questionnaires in the field to be returned later via the mail. Various definitions of level of service have been developed (e.g., LOS A is defined as “best,” “most safe,” “very satisfied,” or “excellent” depending on the researcher). Some researchers have asked pedestrians to directly rate the level of service of a sidewalk or intersection, while others have sought to derive the LOS rating indirectly from the pedestrian’s choice of which sidewalk and crosswalk to use. Several researchers have focused on the intersection cross- ing environment. Most have looked at the sidewalk environ- ment. A few have looked at mid-block crossings in between intersections. None of the researchers have incorporated Americans with Disabilities Act (ADA) considerations in their measurement or prediction of pedestrian LOS. None of the research is specifically applicable to individuals with disabilities. Intersection Crossing LOS Studies Several studies focused on specific pedestrian facility types to identify the key variables that determine level of service there. Some focused on methods of determining LOS for pedestrians at crossing locations. Hubbard, Awwad, and Bullock [44] developed a signal- ized intersection model for pedestrian LOS based on the percentage of pedestrian crossings affected by turning vehicles. Chilukuri and Virkler [45] sought refinements to the HCM 2000 equation for pedestrian delay at signalized intersections, which assumes pedestrians arrive at an intersection randomly. They performed a study of coordinated signal intersections and found that pedestrian delays were significantly different at these locations than expected if arrivals were random. The authors concluded that the HCM pedestrian delay equation

28 should be improved to incorporate the effects of signal coor- dination. Clark et al. [46] developed a pedestrian LOS method based on discrete pedestrian crossing outcomes: non-conflicting, compromised, and failed. Their case study results found that the greatest incidence of failed and compromised pedestrian crossings was observed was a moderately high number of vehicular right turns were served by an exclusive right-turn lane that subtended an obtuse angle with a large turning radius. Lee et al. [47] also looked at crossing LOS using a stated- preference survey. They found that the key determinants of LOS at signalized intersections were area occupancy, pedes- trian flow, and walking speed. Similarly, Muraleetharan et al. [48] identified the factors that describe pedestrian LOS at crosswalks and found that the most important factor was the presence of turning vehicles. While confirming these find- ings, Petritsch et al. [49] provided additional insights into the critical factors that determine pedestrians’ perceptions of LOS at signalized intersection crossings. They found that right-turn-on-red volumes for the street being crossed, per- missive left turns from the street parallel to the crosswalk, motor vehicle volumes on the street being crossed, midblock 85 percentile speed of the vehicles on the street being crossed, the number of lanes being crossed, the pedestrian’s delay, and the presence or absence of right-turn channeliza- tion islands were primary factors for pedestrians’ LOS at intersections. Sidewalk and Path LOS Studies Other studies focused on measuring pedestrian LOS on sidewalks or paths. Analysis of the results of these studies sug- gests that the most important variables that determine pedes- trian LOS—and therefore, the very definition of pedestrian LOS itself—change depending on the context. As described in more detail under the bicycle LOS model section, Jensen [50] used video lab observations to develop a pedestrian segment LOS model for Denmark. Bian et al. [51] conducted a sidewalk intercept survey to measure pedestrian perceptions of sidewalk LOS in Nanjing, China. A total of 501 people were interviewed on nine sidewalk segments. They identified lateral separation from traffic, motor vehicle volume and speed, bicycle volume and speed, pedes- trian volume, obstructions, and driveway frequency as the fac- tors influencing pedestrian LOS. They defined LOS 1 as “ex- cellent,” LOS 2 through 6 are “good,” “average,” “inferior,” “poor,” and “terrible,” respectively. A linear regression model was fitted to the data to predict the mean LOS rating. The nu- merical score predicted by the model was converted to a letter grade using the following limits: LOS A <= 1.5, LOS B <= 2.5, LOS C <= 3.5, LOS D <= 4.5, LOS E <= 5.5. Byrd and Sisiopiku [52] compared the more commonly accepted methods of determining pedestrian LOS for side- walks, including the HCM 2000, Landis, Australian, and Trip Quality methods. The comparison found that it is possible to receive multiple LOS ratings for the same facility under the same conditions from these methods and the paper concludes that a combined model could be developed that synthesizes the quantitative and qualitative factors that affect pedestrian operations. Muraleetharan and Hagiwara [53] used a stated prefer- ence survey to identify the variables most important to a pedestrian’s perception of the utility of the walking envi- ronment. A revealed preference survey with 346 respon- dents was used to develop a utility model that predicts which route a pedestrian will prefer to walk. LOS A was as- signed to the maximum computed utility among all of the sidewalks and crosswalks evaluated. LOS F was assigned to the lowest computed utility among all of the sidewalks and crosswalks evaluated. Muraleetharan et al. [48] found that the “flow rate” is the most important factor that determines pedestrian LOS on sidewalks. Hummer et al. [54] studied pedestrian path oper- ations and found that the path width, the number of meeting and passing events, and the presence of a centerline were the key variables that determined pedestrian path users’ percep- tions of quality of service. However, Patten et al. [55] noted that when paths are shared between pedestrians and bicy- clists, estimating LOS for each user group and designing a new facility to the appropriate width and whether to separate these different users on the right-of-way becomes difficult. Sponsored by FHWA, they developed a bicycle LOS estima- tion method for shared-use paths to overcome these limita- tions by integrating a path user perception model with path operational models developed in the project’s earlier phases. Petritsch et al. [56] found that traffic volumes, a sidewalk’s adjacent roadway width, and the density of conflict points along it (e.g., the number of driveways) are the most impor- tant factors determining pedestrian LOS along urban arteri- als with sidewalks. Taking a step back to revise the theoretical perspective on pedestrian LOS, Muraleetharan et al. [57] used conjoint analy- sis to develop a pedestrian LOS method based on total utility value. They found that total utility value can be used as an index of pedestrian LOS of sidewalks and crosswalks. Sisiopiku et al. [58] reviewed recent research on pedes- trian level of service. Their critique can be summed up as follows: 1. Non-HCM methods need to take into account the effect of platooning on pedestrian LOS. 2. All methods need to consider a variety of pedestrian groups. Different groups have different needs.

29 LOS Model Score A 1.5 B 1.5 and 2.5 C 2.5 and 3.5 D 3.5 and 4.5 E 4.5 and 5.5 F 5.5 Exhibit 33. LOS Categories. 3. All methods need to be applicable to a full range of pedes- trian facility types. Presence of sidewalk should not be a prerequisite. 4. The scale methodologies, although innovative, need fur- ther work to overcome problems with overlap of factors, small sample sizes, and nonlinear performance. 5. There is a need to consider a full and far broader range of factors for determining LOS. Landis et al. [59] developed a method to measure pedes- trian LOS, to aid in design of pedestrian accommodations on roadways, that is based on field measurements of pedestrian perceptions of quality of service. The survey included 75 volunteer participants walking a 5-mile (8-km) looped course consisting of 48 directional seg- ments. Traffic volumes ranged between 200 and 18,500 vehi- cles on the day of the survey. Heavy vehicles accounted for 3% or less of the traffic that day. Traffic running speeds ranged from 15 to 75 mph (25-125 km/h). The participants were asked to evaluate each segment ac- cording to a 6-point (A to F) scale (see Exhibit 33) how safe/comfortable they felt as they traveled each segment. Level A was considered the most safe/comfortable (or least haz- ardous). Level F was considered the least safe/comfortable (or most hazardous). Scoring fatigue was noticed as segment scores decreased as each participant walked the length of the course (Partici- pant’s expectations for the quality of the service drifted downward as they walked the course. Initial segments were rated more critically than later segments. It required about 2 hours to walk the length of the course). This problem was dealt with by walking people in opposite directions over the looped course and letting the fatigue effect cancel itself out through averaging of the responses. After eliminating outliers, a total of 1,250 observations were available for analysis. A stepwise linear regression was per- formed. The resulting equation had an R-square value of 85%, but later researchers have noted that this value was for the abil- ity of the model to predict the average LOS for a segment, not the actual LOS values reported by each individual participant. Human factors are completely absent from the pedestrian LOS model. Age, sex, physical condition, experience, and res- idential location (i.e., urban, suburban, or rural) have no effect on the perceived LOS in this model. Crowding and in- termodal conflicts with bicycles using the same facility are among the operational factors not included in the model. Grades, cross-slopes, and driveways are among the physical factors not included in the model. Ped LOS = −1.2021 ln (Wol + Wl + fp × %OSP + fb × Wb + fsw × Ws) + 0.253 ln (Vol15/L) + 0.0005 SPD2 + 5.3876 (Eq. 8) Where Wol = Width of outside lane (feet) Wl = Width of shoulder or bike lane (feet) fp = On-street parking effect coefficient (=0.20) %OSP = Percent of segment with on-street parking fb = Buffer area barrier coefficient (=5.37 for trees spaced 20 feet on center) Wb = Buffer width (distance between edge of pavement and sidewalk, feet) fsw = Sidewalk presence coefficient = 6 – 0.3Ws (3) Ws = Width of sidewalk (feet) Vol15 = Traffic count during a 15-minute period L = total number of (through) lanes (for road or street) SPD = Average running speed of motor vehicle traffic (mi/hr) Use of Visual Simulation Miller et al. [60] describes the use of computer-aided visualization methods for developing a scaling system for pedestrian level of service in suburban areas. A group of test subjects was presented with simulations (computer anima- tions and still shots) of scenarios of improvements to a sub- urban intersection at an arterial. The subjects were asked to rate each option from A (best) to E (worst) and also to give a numerical score from 1 to 75. These ratings were compared with a set of LOS ratings derived from a scale in which points were assigned based on various intersection characteristics: median type, traffic control, crosswalks, and speed limits. The results of the experiment led to a substantial revision of the scale ranges that correspond to specific levels of service. The authors concluded that, although visualization cannot replace real-world experience, it can be an appropriate tool for site-specific planning. The methods discussed are “. . . in- expensive, practical, and original ways of validating a scale that help ensure that the pedestrian environment is not unnecessarily compromised, especially on automobile- dominated arterials.” Midblock Crossing LOS Studies Chu and Baltes [61] developed a LOS methodology for pedestrians crossing streets at mid-block locations.

30 Thirty-three mid-block locations in Tampa and St. Peters- burg were identified to be included in the study. A total of 96 people were hired by a local temp-worker agency to test the mid-block crossings. They ranged in age from 18 to 77 years with a mean of 42.7 years of age. Sixty-eight percent of the participants were female and 32 percent were male. The participants were bused to each site and asked to ob- serve mid-block crossings for 3-minute periods and then rate the difficulty of crossing on a six-point scale (A to F). Cross- ing difficulty was defined as the risk of being hit by a vehicle, the amount of time to wait for a suitable gap in traffic, pres- ence of a median or other refuge, parked cars, lack of an ac- ceptable (wide enough) traffic gap, or anything else that might affect crossing safety in determining the crossing diffi- culty. It was stressed to the participants to only consider their crossing difficulty, not for others that might cross the road- way. A total of 767 observations were made. Results of the study showed that the level of crossing difficulty tended to in- crease with the width of painted medians, signal spacing, and turning movements, and that the presence of pedestrian sig- nals lowered the perception of crossing difficulty. The pres- ence of pedestrian signals and cycle length were also shown to be statistically significant. The final linear regression model had an R-square value of 0.34 and contained 15 variables re- lating to traffic volumes, turning volumes, age of pedestrian, average vehicle speed, crossing width, presence of pedestrian signal, cycle length, and signal spacing. Chu, Guttenplan, and Baltes [62] placed 86 people at 48 intersection and mid-block locations and asked them to identify one of six routes they might take to cross the street. They obtained a total of 1,028 observations of 4,334 cases. They fitted a 2-level nested logit model to the survey re- sponses. The first level predicted whether they would cross at an intersection or cross mid-block. The lower level then predicted which of various mid-block crossing routes they might pick. The significant explanatory factors were starting or ending point of trip, walking distance, crosswalk marking, and presence of traffic or pedestrian signal. Less significant factors included in the model were traffic volume and shoulder/ bike lane width. Delay at the signal was not an explicit factor. The presence of a signal positively encouraged crossings at the intersection. 3.5 Multimodal LOS Research Recent work on developing a method to estimate multi- modal LOS appears to wrestle with the issue of defining what LOS means in a multi-modal context. Several of these studies are working to establish a “common denominator” that can be used to compare the performance of different modes with- out unintentionally favoring one mode over the others. Winters and Tucker [63] reported on their work for the Florida Department of Transportation (FDOT) to develop a new approach to assess levels of service for automobile, bicy- cle, pedestrian, and transit modes of travel equally. To even the playing field among modes, they postulated a hierarchy of transportation user needs based on Abraham Maslow’s the- ory of personality and behavior. This transportation theory would consist of five levels: safety and security (the most basic need), time, social acceptance, cost, and comfort and con- venience (the least basic need). Perone et al. [64] provide an update to this FDOT work in their final project report. Their final multi-modal model is based on the work of Maslow as well as Alderfer and his Existence, Relatedness and Growth (ERG) Theory. The project provides evidence for the exis- tence of such a hierarchy in which most participants chose Existence over Relatedness over Growth needs and found that a lower motivator need not be substantially satisfied before one can more onto higher motivators. Dissatisfied with inadequacies of auto-based (i.e., HCM) and other multi-modal measures of LOS for project-level en- vironmental impact reviews, Hiatt [65] reports on the City of San Francisco’s efforts to develop an alternative method for use in that city’s urban, multi-modal context. The paper dis- cusses a proposed alternative to modal-based LOS measures that calculates automobile trips generated that would vary by land use typology and parking supply, and reflect expected mode shifts associated with projects such as bicycle or transit lanes. Winters et al. [66] looked at various methods for achiev- ing comparability of LOS significance across modes. They identified the issue of different letter grades implying “trav- eler satisfaction” for the various modes. LOS D for highway facilities is considered satisfactory by many public agencies for facility planning purposes. However, LOS D for bicycles may be a facility that only the hardiest bicyclists dare use. LOS D may not be a satisfactory level of service for planning bicycle facilities. The authors conducted a literature review and then devel- oped various options for reconciling the meaning of LOS across modes to an advisory panel of stakeholders consisting of potential technical users of the LOS methodology for state and local agency facility planning purposes. The authors looked at how the various modal measures of LOS addressed different degrees of travelers’ needs. Some, like FDOT’s bicycle and pedestrian LOS measures are based on travelers’ perception of safety, which is a higher priority need than “convenience” which is implicit in the auto LOS measure of speed. They suggested offsetting the scales against some standard of traveler satisfaction, i.e., using a sliding scale. LOS D is the threshold of acceptability for auto, but LOS C is the thresh- old of acceptability for bicycles. The advisory panel accepted (with reservations) this slide rule method, but recommended

31 that additional data be acquired for identifying common de- nominators across modes for level of service. Crider, Burden, and Han [67] developed a conceptual framework for the assessment of multimodal LOS at inter- sections and bus stops for the transit, bicycle, and pedestrian modes. For transit, a bus stop LOS measure based on fre- quency and pedestrian accessibility was recommended. For bicycles and pedestrians, intersection LOS measures based on conflicts, exposure, and delay to through movements were recommended. Various techniques for surveying traveler perceptions were considered and a selected set of techniques was recommended. Dowling [68] developed a methodology for assessing mul- timodal corridor level of service involving parallel facilities. The methodology generally relied on existing FDOT methods for estimating facility LOS and created new LOS measures to address aspects of corridor LOS not covered by current meth- ods. New LOS measures included difficulty of crossing of freeway LOS, freeway HOV lane LOS, rail LOS, off-street bike/pedestrian path LOS, and a congestion-based measure of auto LOS (i.e., the ratio of congested speed to free-flow speed). Phillips, Karachepone, and Landis [69] documented the results of a project to develop planning analysis tools for es- timating level of service for transit, pedestrian, and bicycle modes. This research built on prior research by Sprinkle Con- sulting and Kittelson & Associates and was adapted for use in the Florida Quality/Level of Service Handbook. The Phillips, Karachepone, and Landis report defined Qual- ity of Service as “The overall measure or perceived perform- ance of service from the passenger’s or user’s point of view.” The report defined Level of Service as “A range of six desig- nated ranges of values for a particular aspect of service, graded from “A” (best) to “F” (worst) based on a user’s perception.” It defined Performance Measures as “A quantitative or quali- tative factor used to evaluate a particular aspect of service.” The distinction between “service measures” and “performance measures” was that service measures represented only the pas- senger or user’s point of view, while performance measures could consider a broader range of perspectives, especially those of the public agency. Guttenplan et al. [70] discussed methods developed by FDOT to determine level of service to through vehicles, scheduled fixed-route bus users, pedestrians, and bicyclists on arterials. FDOT was concerned that the HCM assessment of arterial LOS focuses primarily on the automobile; LOS des- ignations for pedestrians and bicycles are based primarily on facility crowding. Recent research, however, has found that quality of service for pedestrians and bicyclists depends more on lateral separation of the mode, motorized vehicle volumes and speeds, and transit frequency of service. This paper presented the methods used by FDOT to calcu- late LOS for bicycles, pedestrians, and transit. For each mode, a score is computed using various characteristics of the road- way and traffic; LOS thresholds are used to transform the scores into LOS measures. Bicycle LOS depends primarily on effective width of the outside through lane (including bicycle lane width) and the volume of motorized vehicles. Pedestrian LOS depends on sidewalk presence, roadway widths, separa- tion from traffic, and vehicle speeds and volumes. Transit LOS depends on service frequency, adjusted for pedestrian LOS and hours of service per day. The methods described in the paper are primarily segment- based; additional research is under way to expand the appli- cability of the method (e.g., to area wide and point-level analyses). Separate LOS measures are provided for the differ- ent modes; but FDOT does not provide a single LOS measure that combines all modes because doing so could mask the ef- fect of less-used modes. A key feature of the method is that it captures interactions between modes, including the interac- tions of pedestrians and transit. Guttenplan et al. [71] describes the development of a multimodal areawide LOS methodology based on the FDOT Q/LOS Handbook procedures for individual facilities and modes. The steps of the methodology are 1. Define major modal facilities within study area. 2. Determine percentage of households and employment lo- cated within service areas of each major modal facility. The percentage of households and employment served by the major modal facilities sets the ceiling for the best pos- sible areawide LOS for the mode. 3. Determine modal LOS for each major modal facility. 4. Compute mean modal LOS across all major modal facili- ties in the study area. 5. Select the lower of mean modal facility LOS or the per- centage households and employment served LOS value.

Next: Chapter 4 - Data Collection »
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!