Developing a Guide to Bus Transit Service Reliability (2020)

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.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-1 Appendix A – Literature Review

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-2 1.0 Introduction This report summarizes the results of Task 1 of the TCRP A-42 research, a literature review of documents addressing some aspect of bus service reliability. A total of 168 documents were identified as being related to bus service reliability and reviewed. Per the scope of the TCRP A- 42 research, the review focused on fixed-route bus service reliability. Topics of greatest interest in the review were:  reliability definitions,  metrics and measurement,  related factors,  improvement strategies and their impacts, and  the perception of reliability. Resources reviewed were grouped into categories based on their 1) source (educational, consultant, government, organization, transit agency, or some combination of these) and 2) country of origin, as shown in Figure 1.1 and Figure 1.2, respectively. Figure 1.1 – Count of Resources Reviewed by Source (Edu = Educational, Con = Consultant, Gov = Government, Org = Organization, Tra = Transit Agency) The vast majority of relevant resources (64.9 percent) were from educational institutions (Edu) only, plus several were produced by educational institutions in collaboration with consultants (9.5 percent), transit agencies (6.0 percent), government entities (1.2 percent), and transit agencies and government together (0.6 percent). Consultants (Con) produced 11.9 percent of the relevant resources independently, as well as another 1.8 percent in collaboration with government and 9.5 percent with educational institutions. Government entities (Gov) produced 1.8 percent of the relevant documents on their own, with collaborations as previously noted. Transit agencies (Tra) mostly collaborated with educational institutions in the research reviewed, but 1.2 percent of the documents were from transit agencies alone. Organizations (Org), such as the American Public Transportation Association (APTA), contributed another 1.2 percent of the documents reviewed. 1 2 2 2 3 3 10 16 20 109 0 20 40 60 80 100 120 Tra + Edu + Con Org Tra Edu + Gov Gov Gov + Con Tra + Edu Edu + Con Con Edu Source(s)

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-3 Figure 1.2 – Count of Resources Reviewed by Country of Origin The majority of documents reviewed (57.1 percent) came from researchers in the United States (U.S.), with several more coming from collaborations between the U.S. and other countries, including Canada and the Netherlands at 3.6 percent each, China at 1.8 percent, and Mexico and Singapore each with a single collaborative study (0.6 percent each). Researchers in Canada and the United Kingdom (UK) each produced 10 (6.0 percent) of the relevant documents included in the review. China was not far behind as the origin of 5.4 percent of reviewed relevant documents, as well as through a collaboration with Australia (0.6 percent) and those noted previously with the U.S. Researchers in Taiwan also produced one of the studies independently (0.6 percent), and one with researchers in Yugoslavia (0.6 percent). Researchers in Australia produced another six of the reviewed documents (3.6 percent), with two included from researchers in New Zealand (1.2 percent). Documents produced by researchers across Europe were also reviewed, including three (1.8 percent) from the Netherlands, plus one collaboration between the Netherlands and Sweden (0.6 percent), two (1.2 percent) each from Greece and Sweden alone, and one (0.6 percent) each from Finland, France, Italy, Spain, and Switzerland. Researchers from Chile and Israel also contributed one (0.6 percent) relevant document each to the list of documents reviewed. 6 1 10 1 9 1 1 2 1 1 2 1 2 1 1 1 1 3 10 96 6 3 1 1 6 0 20 40 60 80 100 120 Australia Australia + China Canada Chile China Finland France Greece Israel Italy New Zealand Spain Sweden Sweden + The Netherlands Switzerland Taiwan Taiwan + Yugoslavia The Netherlands UK US US + Canada US + China US + Mexico US + Singapore US + The Netherlands

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-4 The findings from relevant documents reviewed throughout this process are summarized in the sections that follow. A tabular summary of the findings from relevant documents is included in the Appendix A-1. 1.1 Early Bus Service Reliability Research Modeling and measuring bus service reliability has been a research area of interest for many decades now. Early work was often limited to high-level analysis based on small sample sets and aggregate measures. An example of this can be found in a 1978 report on transit service reliability commissioned by the U.S. Department of Transportation, which provided a comprehensive review of service reliability in research and in practice. The report established performance measures from the operator and customer perspectives, then explored major causes of unreliability, even going as far as to evaluate some potential solutions [2]. While this was a seminal work in the area of transit reliability at the time, it was based on a much more limited set of data than is commonly used for similar studies today. Most early studies on bus service reliability used measures such as schedule adherence or on-time performance for routes with schedules and headway adherence for more frequent routes. Levinson conducted a survey among 20 transit agencies in the United States and Canada, asking about their methods to monitor ridership and travel time [85]. Before automated vehicle location data was widely available, on-time performance was the most common reliability measure used by transit agencies, as stated in TCRP Synthesis of Transit Practice No. 10: Bus Route Evaluation Standards [12]. Models and simulations were also used to test hypotheses on the relationships between bus service reliability and a variety of factors. Turnquist and Bowman looked into the effects of network structure on transit service reliability, finding that transfer time variability and service frequency play important roles in determining overall system reliability. Grid networks were found to be less disrupted by transfers than radial networks. On-time arrival of vehicles at major transfer stations was found to be important, especially for radial networks [145]. A subsequent work by these authors suggested greater customer sensitivity to schedule deviation and less sensitivity to service frequency than found under the assumption of random passenger arrivals [16]. Another early work on analysis of reliability factors considered the relationship between employee absenteeism and overtime [128]. Beyond measuring and modeling bus service reliability, several researchers have investigated potential reliability improvement strategies and their impacts. An early work by Furth and Wilson aimed to optimize bus route frequencies with fixed subsidy, fleet size, and maximum headway, finding that simple rules of thumb are generally not the best way to set frequencies [59]. Their proposed method also assumed fixed demand and bounded travel times. Almost a decade later, researchers at Vanderbilt University studied headway-based control, including holding, and concluded that the method improved headway regularity and reduced running time [1]. However, this study was based on only two days of checkpoint data and six days of headway data, so lacks statistical significance. A very relevant piece by Strathman and Hopper considered the relationship between on-time performance and many factors. While they were able to recommend strategies to potentially improve bus service reliability, the impacts of proposed improvements were not fully assessed [134]. Early research into the use of adaptive control of transit operations indicated that this strategy may improve bus service reliability, effectively reducing user wait times, but often at the cost of increasing in-vehicle travel time. Other strategies, such as shortening routes, increasing stop spacing, transit signal priority, and providing exclusive right-of- way are also said to help improve transit speeds and reliability [90]. The development and deployment of automated data collection systems, such as automated vehicle location (AVL) systems and automated passenger counting (APC) systems, among transit

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-5 agencies and vehicles has led to a dramatic shift in the availability of data for measuring, monitoring, and improving bus service reliability. With these technologies, transit agencies are increasingly able to: 1) track service reliability and make adjustments as needed in real-time, 2) automatically calculate detailed reliability metrics, 3) model the impacts of various factors and improvement strategies on service reliability, and 4) conduct before and after studies to measure the real-world impacts of operational and environmental changes on the bus system at many different levels. Although several approaches to improving service reliability have been proposed, evidence of the actual impacts of these improvement strategies is still forthcoming in many cases. These and more recent findings related to reliability definitions, metrics and measurement, improvement strategies and their impacts, and the perception of reliability are summarized in the following sections and chapters of this report.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-6 2.0 Defining Bus Service Reliability Many different definitions of bus service reliability have emerged over the years. One of the early definitions of reliability that has been cited repeatedly is “the invariability of service attributes which influence decisions of travelers and transportation providers” [2]. Around the same time, Polus defined bus service reliability as “the amount of consistency associated with an operational performance measure from day to day” [116]. It is evident from most early definitions of bus service reliability that they are targeted toward transit agencies, rather than customers or the general public. Concern with providing definitions and measures of reliability that are more representative of the customer perspective increased in the early 2000’s with the availability of AVL data to measure performance more easily and accurately. In addition to reliability definitions and measures becoming more passenger-focused, two trends appear to be emerging: increasing use of broad, simple reliability definitions and breaking reliability into components. 2.1 Simple Definitions Several researchers have put forth simple definitions of reliability, but few of these have been found more than once or twice in the literature. Reliability may be defined as the level of consistency in transportation service for a mode, trip, route, or corridor for a time period [94]. An even simpler approach is to define reliability as a consistent result from a service over time [93]. Alternatively, a reliable service can be defined as having smaller deviations in arrival times [34]. Oort and Nes defined reliability as the match between planning and operations [151]. According to Tetreault and El-Geneidy, reliable transit service is that with short wait times and less variation [141]. Schil defines reliability as the quality of being consistently good in quality or performance [125]. Many different researchers have defined reliability as schedule and headway adherence or on-time performance [5,27,64,112,132,166]. A simple definition of unreliability is also provided in several cases, where unreliability is roughly equated to running or travel time variability or schedule deviation [20,23,57,108,121]. The role of running time variability in transit operations is shown in Figure 2.1. Figure 2.1 – The Role of Running Time Variability in Transit Operations [121]

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-7 2.2 Customer-Focused Definitions In TCRP Report 165: Transit Capacity and Quality of Service Manual (3rd Edition), reliability is defined from the passenger’s perspective as pertaining to arriving at destinations on-time and not having to wait too long at bus stops. This team of researchers also stated that from the operator perspective, reliability impacts schedule recovery time [82]. Similarly, in TCRP Report 95: Transit Scheduling and Frequency, reliability from the passenger perspective is equated to arriving at the intended time, or on-time performance from the operator perspective [53]. Another variation came from the authors of TCRP Report 113: Using Archived AVL-APC Data to Improve Transit Performance and Management, who defined reliability as on-time performance from the operator perspective, but equivalent waiting time from the user perspective [62]. Camus et al. present a novel methodology that considers delays as a distribution instead of a binary variable. They introduce the Weighted Delay Index, which is essentially the ratio of mean delay over headway [24]. Pangilinan defined reliable transit service as that which can be counted on over and over, with consistent wait times and travel times [111]. Another take on reliability from the passenger perspective is to define it as service that can be easily accessed by passengers, arrives predictably, has a short running time, and has low variance in running time [50]. This definition introduces the idea of accessibility as it relates to transit service reliability. In 2014, another group of researchers referred to service performance from the passenger perspective as including waiting times, travel times, headways, and running time performance [142]. Most recently, reliability from the passenger’s perspective has been defined as punctuality in arriving on time at the destination, short waiting times at the origin bus stop, and consistency of wait and travel times [64]. 2.3 Components of Reliability It is common for reliability definitions to incorporate several different, but often related, components. Eboli and Mazzulla define reliability as the ability of the transit system to adhere to a timetable, as well as the ability of vehicles to depart or arrive on-time [46]. Carrasco takes a slightly different approach in defining reliability as a state minus the probability of failure. The author claims that reliable travel times have consistency or dependability, measured from day to day for the same trip. Reliable transit service is considered to be on schedule, maintaining regular headways, and minimizing wait time variability for passengers [25]. Researchers in New Zealand break reliability into two components: reliability in arrival / departure time at the bus stop and reliability in the travel time spent on the bus [31]. Similarly, researchers at Ian Wallis Associates, Ltd. use reliability to describe two concepts, including 'reliability' (whether or not the service operates) and 'punctuality' (whether the service runs on time) [71]. Researchers in China proposed a similar breakdown of reliability, defining it as the probability of effectively transporting passengers and the smooth operation of the various bus lines in accordance with the planned schedule within the urban transit network [74].Along these same lines, researchers in Texas define reliability in terms of timeliness of vehicles and the difference between a passenger's scheduled and actual travel time [11]. Although many definitions of reliability exist, there is a great deal of overlap and similarities between them. As time passes, researchers seem to be less interested in explicitly defining reliability, and more focused on identifying the best metrics and measurement techniques for tracking and reporting on bus service reliability. A review of research on the various metrics and methods of measurement is provided in the next section.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-8 3.0 Reliability Metrics and Measurement Several decades of research have produced more than 150 different metrics for measuring reliability. While some of the metrics and measures from the research are synonymous or slight variations on each other (schedule adherence and on-time performance, for example), many are novel approaches to quantifying bus service reliability. Certain measures like schedule and headway adherence have been widely used for decades, but there is not one metric at this time that has emerged as clearly superior in all cases. Overall, it seems that one or more of a variety of reliability metrics could be suitable for transit agencies and other organizations to use, depending on the characteristics of the service(s) they offer and the audience(s) they are targeting, among other factors. There has been much debate regarding the merits of having a single measure of reliability as opposed to several measures that reflect different aspects of reliability; this matter is far from settled at this time. However, with widespread implementation of AVL and APC systems and other automated data collection technologies, transit agencies and researchers have gained access to an abundance of data, which is allowing them to use more fine-grained and data-intensive methods that were previously cost-prohibitive or impossible to use. While the reliability metrics included in this report do capture the vast majority of what was discovered during the literature review, they should not be considered as a completely exhaustive list. The authors of this report have summarized previously proposed metrics below. These are discussed more, and novel measures are proposed, in later portions of the study. 3.1 Early Metrics Even with the limited data that was available to transit providers in the late 1970’s and early 1980’s, researchers proposed some rather innovative reliability measures during this era. Polus proposed a single measure of reliability as one divided by the standard deviation of travel time for a route over a given period [116]. Turnquist and Bowman suggested four different metrics for measuring reliability, including 1) standard deviation of vehicle arrival times at stops, 2) coefficient of variation of arrival times, 3) coefficient of variation of transfer delays, and 4) twice the deviation of arrival times multiplied by the expected number of transfers, but then used the more common measure of schedule adherence in a similar work just a year later [16,145]. Around the same time, Furth and Wilson proposed using the weighted sum of total passenger waiting times as a reliability metric, but due to limited data this approach was based on the assumptions of fixed demand and bounded travel times [59]. As more fine-grained measures incorporating delays and wait times were proposed, researchers often found themselves limited to small datasets due mostly to cost, technology limitations, and time constraints, at least until automated data collection was widely used [1]. 3.2 Common Metrics Route-level service reliability metrics have been widely used, including measures of on-time performance, schedule adherence, and headway adherence, as well as the standard deviation and coefficient of variation of travel times [9,12,57,72,73,80,107,108,110,112]. On-time performance was cited in more than 30 studies, with schedule adherence cited about 20 times and headway adherence mentioned in at least eight of the reviewed documents. The most common statistical measures used were mean or average (cited in 21 studies combined), coefficient of variation (cited in 15 different studies), percent or ratio (cited in 15 studies combined), standard deviation (cited in 12 studies), and variance (cited in eight studies). Other common measures involve an index (cited in 10 studies), percentile (cited in five studies), or range

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-9 (cited in two studies). Figure 3.1 shows the top 15 reliability metrics identified from the literature review by count. Figure 3.1 – Top 15 Bus Reliability Metrics by Count of Reviewed Documents Using Each Passenger-focused metrics have also been very popular, as is evident from at least 26 different studies that mention the importance of measuring and reporting reliability from the perspective of customers [1,2,11,25,31,39,45,50,53,59]. Customer-oriented measures tend to focus on aspects of travel or journey time (mentioned in 39 studies combined), including wait time (mentioned in 28 different studies), and may include elements of customer feedback, such as survey ratings or the number of complaints. 3.3 Defining On-Time Performance Much of the controversy surrounding reliability metrics since the early 1990’s has revolved around the definition of “on-time”. While buses traveling between one minute early and five minutes late has become the standard for on-time performance across much of North America, many transit agencies use different definitions of punctuality, including between zero and two minutes early and between one and seven minutes late [8,25,75,78,114,117,118,134,139,160]. Determining which definition of on-time is best for a particular agency or service is largely based on service characteristics and the agency’s ability to achieve their stated on-time performance goal (percent of trips within the punctuality window). For example, the perception of reliability has been shown to depend on trip purpose. In a survey, Vincent and Hamilton found that on-time 38 20 18 15 15 13 12 11 10 10 9 9 8 8 7 0 5 10 15 20 25 30 35 40 On-time performance Schedule adherence Wait times Delay Travel times Travel time variance Schedule deviation Travel time coefficient of variation Excess wait time Headway regularity Buffer time Running time variance Headway adherence Headway variance Running times Me tri c

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-10 performance was very important on medical trips, somewhat important on commuting trips and not very important on social and shopping trips [154]. Trends in on-time performance can be easily tracked over time at a variety of levels, as shown in Figure 3.2, making it a useful and flexible metric for many audiences and applications. On-time performance tends to be used for infrequent services with a published or internal schedule for determining when a bus is “on-time”. Figure 3.2 – Example of On-time Performance Tracked at the Service Area Level over Several Years [117] 3.4 Novel Metrics As transit performance data became more widely available, a plethora of new reliability metrics were proposed and tested, with many researchers seeking a single measure or set of measures to provide a holistic understanding of reliability from the operator and customer perspectives. Similar to the widely used Traffic Level of Service (LOS), Perk, Thompson, and Foreman have proposed a Reliability Quality of Service (QOS) measure [114]. The following year, researchers in China proposed a Reliability Composite Index of Service (RCIS), which is comprised of a Punctuality Index based on Routes (PIR), Deviation Index Based on Stops (DIS), and Evenness Index based on Stops (EIS) [167]. Using data from London, England, graduate researchers at Massachusetts Institute of Technology (MIT) have also proposed new reliability metrics in the last few years. Uniman proposed the Excess Reliability Buffer Time (ERBT) measure, which is the amount of excess travel time that passengers need to budget to arrive on-time at their destinations 95 percent of the time [148]. A visual description of ERBT is shown in Figure 3.3 below. Figure 3.3 – Excess Reliability Buffer Time Example [148]

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-11 Schil introduced the Normalized Reliability Buffer Time (NRBT), which is defined as reliability buffer time divided by the median journey time for the O-D pair and period of interest [125]. Diurnal Mean Spread (DMS), a measure of overall random variability, was also proposed around this time by Sanchez-Martinez [121]. A group of North American researchers proposed three new measures to complement more common existing measures, such as on-time performance. The proposed new measures included an Earliness Index (EI), a Width Index (WI), and a Second Order Stochastic Dominance Index (SSDI) [120]. Figure 3.4 shows these three measures compared to headway adherence for an example bus route in Portland, Oregon. Figure 3.4 – Earliness Index, Width Index, and Second Order Stochastic Dominance Index vs. Headway Adherence [120]

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-12 Several new indicators, including many that are customer-focused, were proposed in 2015 alone, including the Journey Time Buffer Index (JTBI) [64], Total Bus Stop Time (TBST) [7], Public Transport Travel-time Variability (PTTV) [77], Individual Reliability Buffer Time (IRBT), and Platform to Platform Reliability Buffer Time (PPRBT) [159]. Figure 3.5 below shows measures of IRBT compared to reliability buffer time (RBT) aggregated by time of day and over six-week periods, for an example MTR rail line in Hong Kong. In a stated-preference survey, Hollander tested the cost of travel time variability to passengers compared to trip earliness and lateness. The web-based survey asked transit users to choose randomly generated combinations of trip departure times, arrival times, and fares. Respondents attributed greater cost to the necessity to start a trip early and much greater cost to the incidence of ending a trip late. The paper highlights to risk of considering summary performance metrics that can overlook valuable attributed of passenger reliability [69]. Figure 3.5 – Individual Reliability Buffer Time (IRBT) vs. Reliability Buffer Time (RBT) [159] 3.5 Metric Desired Qualities and Organization Given the variety of data sources and metrics available today, a common approach is for transit agencies and researchers to assemble reliability measures that fit with their particular services, needs, and goals. This has led to an explosion of possible metrics, with over 150 options identified in this literature review alone. As a result, many researchers have turned their attention toward organizing and analyzing these various measures. Strathman, et al. outlined four objectives for service reliability metrics. They stated that measures should:  be self-evident and easy to interpret,  permit direct comparison within and between routes,  be as comparable as possible across measures, and  retain as much information as possible [136]. A research team based in the United States suggested classifying reliability measures as statistical range measures, buffer time measures, and tardy trip indicators. They said that as measures are selected and calculated, several factors should be considered:  whether the measure is mode-specific or mode-neutral,  trip type and location,  controlling for length and time,

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-13  the target user of the measure, and  the area size. Of the measures considered, percent variation, misery index, and buffer time index were recommended [94]. Buffer time index is shown in Figure 3.6. Figure 3.6 – Buffer Time Index Example [94] In NCHRP Report 618: Cost-Effective Performance Measures for Travel Time Delay, Variation, and Reliability, a team of consultants proposed ten considerations for reliability measure selection:  relationship to goals and objectives  clarity of communication  inclusion of urban travel modes  consistency  accuracy  illustration of effects of improvements  application to existing and future conditions  application at several geographic levels  use of person- and goods-movement terms  use of cost-effective methods of data collection or estimation Definitions for a variety of reliability metrics were also provided in this report, as shown in Figure 3.7. In this case metrics were classified as either individual measures or area measures. Recommended reliability measures included buffer index (percent extra time to ensure on-time arrival), percent on-time arrival (percent of trips defined as on-time), planning time index (dimensionless factor indicating travel time for planning purposes), percent variation (percent of average travel time required for on-time arrival of given trip), and 95th percentile travel duration [23].

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-14 Figure 3.7 – Reliability Metric Definitions from NCHRP Report 618 [23]

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-15 Consideration of desirable qualities and useful classification systems for reliability metrics has extended well beyond the United States. In Switzerland, Carrasco organized his measures of bus service reliability as relating to:  Travel time (mean and 5th and 95th percentiles),  Speed (mean and 5th and 95th percentiles),  Punctuality (route-level schedule deviation frequency, mean schedule deviation at stop level, on-time performance, standard deviation from scheduled departures at stop level, coefficient of variation of schedule deviation at the stop level), or  Regularity (actual headway frequency distribution at route and stop levels, mean headway at stop level, coefficient of variation of actual headways at stop level) [25]. A team of researchers in New Zealand evaluated bus reliability measures using four criteria:  ease of understanding;  extent to which the measure has a customer focus;  accuracy, completeness, and objectivity of the measure; and  the relative cost / effort in collecting and analyzing the data. The results of this analysis, shown as a table in Figure 3.8, suggested that the two best measures are excess waiting time and customer delay, followed by customer complaints [31]. Figure 3.8 – Analysis of Bus Reliability Factors [31] # Description Ease of Understanding Customer Focus Fidelity & Objectivity Cost/Effort Efficiency Overall Rating Score Rating Rank 1 % Buses Cancelled High Low Low High Medium 4 4 2 % Departing On-Time High Low Medium Medium Medium 4 4 3 % Arriving On-Time High Low Medium Medium Medium 4 4 4 Excess Waiting Time High High Medium Medium High 6 1 5 Average Lateness High Low Medium Medium Medium 4 4 6 Variability Measures Low Low Low Medium Low 1 9 7 Reliability Buffer Low Medium Low Low Low 1 9 8 Passenger Ratings Medium High Medium Low Medium 4 4 9 Customer Complaints High High Low Medium Medium 5 3 10 Customer Delay High High High Low High 6 1 More recently, Canadian researchers have grouped reliability metrics as: travel time indicators, schedule adherence indicators, headway regularity indicators, wait time indicators, and composite indicators (64). Australian researchers went a step further in recommending that measures be developed for specific users, based on local and service-specific conditions, and outlining a process for developing such measures [100]. Schil discussed metrics as being:  operational (percentages of buses in service)  negative (number of delays)  retrospective (averages of past performance)  big facts (top causes of delay), and/or  future focused (bus frequencies).

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-16 Of these, big facts and future-focused metrics seemed to offer the most promise, though many passengers may not find reliability metrics to be useful and may instead choose to rely on their own experiences to judge a system's reliability. He also described desirable qualities for journey time reliability metrics as being customer-driven, simple, meaningful (for customers and operators), and standardizable [125]. Wood suggested the following design objectives for passenger-focused reliability metrics, in that they:  are inclusive of all sources of unreliability;  distinguish between service variability and schedule adherence, as well as early and late arrivals;  control for variation in passenger behavior and time of day;  exclude extreme delays; are calculated at the origin-destination pair level; and  are unbiased with regard to passenger demographics. He states that metrics should be meaningful for passengers and non-experts, such that they are understandable, objective, and useful for planning journeys. Measures should ideally be comparable across different services and times of day, independent of schedules, and absolute (as opposed to relative). Time period flexibility, meaning that a metric is calculable for short time intervals and allows for exclusion of weekends, holidays, and specific events, and service scope flexibility are also desirable traits [159]. Wood went on to recommend specific types of reliability buffer time measures, including IRBT and PPRBT, for certain applications, as shown in Figure 3.9. Figure 3.9 –Reliability Buffer Time Measure Recommendations by Use [159]

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-17 Major trends related to metrics that were noted during the literature review process include the move to more detailed and data-intensive measurement techniques, given the increasing volume and accuracy of data available, and the shift to making reliability information more accessible and customer-focused. Publicly available information about transit operations and reliability can help to not only improve public perception of a transit agency, but may also help customers make more informed decisions related to their bus use, increasing their actual and perceived reliability through reduced wait times and uncertainty. The trends and reliability metrics that emerged from the literature review, as well as some additional options, will be investigated in more detail in later stages of the study.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-18 4.0 Factors Affecting Bus Service Reliability A large number and wide variety of factors that might impact bus service reliability were found in the literature. Even after removing synonymous terms for the same factor, roughly 200 different factors were found. Most of these factors can be grouped using one or more of many possible classification systems. The most common factors and relevant documents are included in the following section. Priority was given to research involving analysis or organization with regard to factors impacting reliability. 4.1 Commonly-Cited Factors A total of 790 factors (with considerable overlap) were identified from the literature as potentially impacting bus service reliability, and then grouped at multiple levels. In Tier I, factors were declared to be either internal (366 factors) or external (424 factors), though some could involve a combination of internal and external influences. Internal factors were broken down as relating to 1) staff, especially drivers and supervisors, 2) planning, including scheduling and routing, 3) the bus fleet, including maintenance, and 4) service delivered, including elements of control and fare collection. The external factors were further broken down as being related to 1) infrastructure, which includes signals and stations, 2) traffic, as well as transit priority treatments, 3) temporal factors, such as weather and incidents, and 4) passengers, which include factors. The number of references reviewed that included each factor was tallied, as shown in Table 4.1. Table 4.1 – Prevalence of Factors by Tier I and Tier II Categories External Factors 424 Internal Factors 366 Infrastructure 66 Fleet 52 Passengers 116 Planning 83 Temporal 120 Service 184 Traffic 122 Staff 47

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-19 As shown in Figure 4.1, the most commonly-cited factors (by Tier III category) were related to service, passengers, and traffic impacts, followed by those related to incidents, routing, time, drivers, vehicles, weather, controls, signals, priority, scheduling, fare payment and stations. Somewhat less popular were factors relating to spatial, planning, and maintenance, as well as miscellaneous infrastructure- and staff-related factors. Figure 4.1 – Count of Factors from the Literature Review by Tier III Category 9 16 16 17 18 20 21 25 27 30 33 35 36 38 40 41 45 95 98 130 0 20 40 60 80 100 120 140 Staff Infrastructure Maintenance Planning Spatial Stations Fare Schedule Priority Signals Control Weather Fleet Driver Temporal Route Incident Traffic Passengers Service Ti er III F ac to r C at eg or y

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-20 With minimal grouping, the top 35 factors from the literature review were identified, as shown in Figure 4.2. From the chart below, it is clear that traffic congestion was the most-cited factor in the literature reviewed for this study. Traffic signals, weather conditions, and passenger activity, such as boarding and alighting, were not far behind in popularity, followed closely by traffic conditions, routing, passenger volumes, incidents, route length, passenger demand variability, and time of day. Whereas Figure 4.1 provided a comprehensive list of factor groups, Figure 4.2 shows only the top 35 of about 200 identified factors. The full list of factors is provided in the Appendix. Figure 4.2 – Count of Factors from the Literature Review with Minimal Grouping 7 7 7 7 7 7 7 8 8 9 9 10 10 10 11 11 12 12 13 13 13 14 14 15 17 18 18 19 19 19 20 24 24 25 37 0 5 10 15 20 25 30 35 40 Crowding Passenger load Planning Right-of-way Supervision Transit priority strategies Vehicle quality Departure delay at origin Stop spacing Driver experience Wheelchair lift or ramp usage Dwell times Road work Vehicle maintenance Direction of travel Fare payment methods Number of stops Scheduling Bus bunching Operations control strategies Peak period travel Driver behavior Traffic volumes Frequency of service Time of day Passenger demand variability Route length Incidents Passenger volumes Routing Traffic conditions Passenger activity Weather Signalization Traffic congestion Fa ct or

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-21 4.2 Internal Factors Internal factors were those identified as being entirely or primarily related to the organizational framework and services provided by transit agencies. . Fleet Various aspects of buses were found to be commonly-cited reliability factors, including vehicle characteristics, such as maintenance [11,21,71,93,111,115,146,146,147,159], quality [11,39,71,82,120,147], availability [11,53,82,93,94], breakdowns [9,48,82], floor height [42,141], bus type [5,6,37,38,99], bus control [116], and other fleet features [29,48,121]. Planning Commonly-cited bus reliability factors relating to planning [26,99,101,108,111,122,158] and operational outputs [26,99,101,108] included routing [5,7,8,18,25,42,48,54,56,64], route length [11,30,50,64,71,82,83,93,107,121], scheduling [8,11,16,25,29,48,54,60,70,71], bus priority strategies [5,6,38,48,39,41,167], stop spacing [5,39,50,64,72,73,134,139], stop location [7,29,93], and control strategies [11,39,41,71,82,101,122,148]. Network characteristics, especially as they relate to transfers and transfer times, have also been said to impact bus service reliability [5,25,54,86,90,119,140]. Service Service frequencies [5,6,15,16,34,59,70,83,86,99], bus bunching [2,9,25,56,66,68,71,83,90,161], operations control strategies [11,71,120,122,136,148], the number of stops [37,71,105,120,121,141,160], dwell times [5,7,18,50,60,107,102,117,120,121], departure delay [50,100,105,136], crowding [9,10,55,70,149,165], and supervision [71,82,100,108], are all said to impact bus service reliability. Fare collection and payment methods and technologies were also mentioned repeatedly in the literature as factors affecting bus service reliability [5,6,7,18,25,36,38,55,83,86]. Staff Drivers [74,121] or operators [48,108] and their behavior (6; 24; 65; 79; 83; 93; 101; 107; 135; 107; 145; 153) [6,28,65,78,83,93,101,107,134,144], experience [39,50,82,134,144,160], abilities [11,71,83,111], and availability [11,53,144] were also commonly-cited bus service reliability factors. 4.3 External Factors External factors are mostly out of control of transit agencies; they relate to environmental, mostly uncontrollable, concerns like infrastructure, passengers, time, and traffic. Infrastructure Signalization was the most popular infrastructure factor mentioned in the reviewed literature [2,3,18,23,25,29,32,38,39,41], followed by stop spacing [5,32,39,64,76,134,139], stop or station design [5,160], and stop locations [7,29,32,93,111]. Road characteristics, such as roadway geometry [39,54,86], road type [23,129] and intersections [90,93,150,160], and intelligent transportation systems [5,87] were also cited multiple times. Passengers Many bus service reliability factors referenced passengers, in terms of passenger loads [28,64,111,116,137,141], passenger volumes [48,88,93,100,162], passenger activity (boardings,

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-22 alightings, lift use, payment, luggage storage, etc.) [7,8,27,37,38,42,50,55,70,71], passenger demand [11,29,65,71,74,78,82,86,93,94], wheelchair lift or ramp usage [11,42,50,51,55,71,78,82,102], and direction of travel [3,38,50,51,66,93,95,99,138,141]. Time Temporal factors, such as time of day ) [7,23,27,37,37,41,42,50,52,64], peak periods [29,38,71,157,162], day of the week [100,121,141], and season [23,111,121,148], were also frequently cited in the literature. Figure 4.3 shows running time in minutes from 5:00am until 12:00am for a sample route, using various percentiles to indicate running time variation over the course of a typical day [121]. Figure 4.3 – Running Time Variation by Time of Day [121] Weather was listed as a bus service reliability factor in at least 22 cases [8,9,11,23,37,48,50,65,70,74], as well as specific types of weather, such as snow and ice [38,82,139], rain [111], fog [95], and heat [82]. Many studies also indicate that incidents [8,9,11,22,23,50,71,77,90,94], crashes [9,74,117,121], road maintenance or construction [23,74,82,99,111,117,121], and special events [25,94,111,117] contribute to bus service unreliability. Traffic Traffic congestion, which may be impacted by vehicle volumes, signals, priority treatments, incidents, time period, on-street parking, and many other conditions, was the most commonly- cited factor impacting bus service reliability in the literature review [2,4,5,7,8,9,11,15,18,23]. An example of the relationship between travel time variability and traffic signals is shown in Figure 4.4. Priority treatments are important factors affecting bus service reliability [72,73,116,122,132,139,155], with exclusive bus lanes [30,93,160,164] and running ways [5,87]thought to positively impact running times and reliability, though they do not necessarily eliminate all traffic impacts.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-23 Figure 4.4 – Coefficient of Variation of Travel Times vs. Traffic Signals per Kilometer [160] 4.4 Analysis of Factors Impacting Reliability Beyond simply identifying factors that may impact reliability, many researchers have sought to understand the extent and directionality of those impacts on bus service reliability. Fleet Factors related to the bus itself, such as its floor height [42], vehicle length [48], bus type [37,38], and distance along its route from the origin terminal [83,121,134], have been demonstrated to have a significant impact on dwell times and overall bus service reliability. As part of a study of STM in Montreal, Diab & El-Geneidy found that the introduction of articulated buses increased running times by 26.8s [37]. In a follow-up study of the same system, they found that while articulated buses increase running times, they can help to reduce running time variation [38]. This finding was largely attributed to the lower rates of acceleration and deceleration that are common for larger buses, but other factors could certainly be at play. Planning Some of the earliest research on reliability factors included in this review came from Turnquist and Bowman in the early 1980’s. In analyzing data from a simulation of theoretical transit networks in Chicago, they found that variability in time associated with making a transfer, which is directly influenced by frequency of service, appears to have a major impact on transit system reliability. Grid networks were found to be less disrupted by transfers than radial networks, but service reliability was found to be much more influenced by frequency of service than route density. On- time arrival of vehicles at major transfer stations was also found to be important, especially for radial networks [145]. More recently, Uniman used ordinary least squares regression analysis to estimate that when a trip involves transferring variability increases by about 56 seconds [148]. Many factors relating to planning and operations have been found to impact bus service reliability. Included among the statistically significant factors from the literature were:  route design [48,134,138],  route length, which has a negative correlation with reliability [30,48,50,64,100,134,151],  travel through congested areas and central business districts, which typically has a negative impact on reliability [48,64,100,121],  stop spacing or number of stops [18,50,64,100,121,134,139,141,151],  scheduling and bus frequencies [60,63,70,134,145,161],  automated vehicle location systems [48,54,78,112], y = 0.0374x + 0.0744 R² = 0.6729 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 2 4 6 8 10 12 14 CV o f T ra ve l T im e Traffic signals per kilometre CV Linear (CV)

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-24  dwell times [107],  control systems and strategies [60,68,75,90,101,133,153,161,163],  direction of travel [38,50,92,121,139], and  deviation from scheduled departure time at the origin terminal [28,50,107,133] with increasing frequencies improving reliability at least up to the point when bus bunching becomes a problem. The probability of later arrivals was found to increase as buses travel along their routes, which may be addressed by longer running times with many time points or shorter and less complex routes. Longer headways were associated with more late arrivals, with the highest chance of late arrival for services at headways of 70 minutes. Reliability appears to deteriorate with headways exceeding 15 minutes [134]. There also seems to be a tradeoff between running times and reliability when scheduling buses. Furth and Muller found that adding slack time into a schedule helps to improve reliability, but lowers operating speed, which can lead to other negative consequences [60]. In an extension paper, they found that reliability, as expressed in an amalgamated cost function to passengers, is a decreasing function of the number of time points with diminishing returns [58]. Graduate researchers at MIT have found compelling evidence for the propagation of schedule and headway deviations along a route being related to deviations in departure times at the route origin terminal. Through a simulation using data from the Chicago Transit Authority, Moses found that deviations in origin terminal departure have an impact on headway regularity [110]. In her master’s thesis, Cham considered numerous potential causes of bus unreliability, with the primary factor identified as deviations in departure time from terminals [28]. The effects of late departures from the origin terminal are illustrated in Figure 4.5. Figure 4.5 – Effects of Late Departures at the Origin Terminal on Bus Service Reliability [28] Several research teams have pursued this line of research more recently. El-Geneidy, Horning, & Krizek found a significant positive association between running time deviation and delay at the origin terminal, as well as a negative correlation between the coefficient of variation of running time and the coefficient of variation of delay at the origin terminal [50]. A few years later, researchers at the University of Illinois found that headway variability at any stop is an increasing function of the variability at departure [91]. Surprenant-Legault and El-Geneidy found that each second of delay at the beginning of a trip increases the odds of being late by 0.9 percent [139]. Seemingly contrary to this result, Diab and El-Geneidy found that buses getting a late start are generally found to have shorter running times, with running time decreasing by 0.22s for every second of delay at the beginning of the trip. This was attributed largely to drivers trying to catch up after a late start, and decreasing their running times overall [37]. In a follow-up study, the same researchers found that running time deviation decreased in their model when a bus trip is delayed at the origin [38]. Regression models were run for three measures of reliability used in a study of data from Metro Transit in Twin Cities, Minnesota. Significant variables in the model for running time deviation included scheduled stops and actual stops, which both had positive coefficients. Significant

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-25 variables in the model for headway deviation included scheduled stops, which had a negative coefficient, and actual stops, which had a positive coefficient. A significant variable in the model for coefficient of variation of running time was coefficient of variation of actual stops, which had a positive coefficient. In this study, it was found that running time deviation could be expected to increase by 0.9 percent for each scheduled stop and 1 percent for each actual stop. In this case only about 50 percent of stops were served, yet even unserved stops were found to increase running times and decrease reliability. Stop consolidation was recommended by the authors [50]. In another case, limited-stop service decreased the odds of being late by 66 percent [139]. Diab and El-Geneidy modeled running time deviation, coefficient of variation of running time, and coefficient of variation of running time deviation based on various factors, finding that the number of stops correlates positively with running time deviation and the coefficient of variation of actual stops correlates positively with the coefficient of variation of running time as well as the coefficient of variation of running time deviation. Limited-stop service was shown to improve running times, but worsen variation [38]. In one study northbound travel, which typically had more traffic, was found to increase the odds of being late by 75 percent [139]. Likewise, El-Geneidy, Horning, and Krizek found a significant positive correlation between running time deviation and westbound travel [50]. Diab and El- Geneidy found a significant negative correlation between northbound travel and running time deviation and a significant positive correlation between northbound travel and coefficient of variation of running time [38]. It should be noted that direction itself is not the actual factor of influence, but rather the direction of travel relative to the peak direction of travel, activity centers, and areas that may have relatively high and/or highly variable levels of traffic congestion. Service Fare collection and payment methods were found to have a statistically significant relationship with bus service reliability in several studies. A study of AC Transit, Route 51 found that fare payment was the factor contributing the most to dwell time variability [123]. Contactless fare cards [109], smartcards, flash passes [38], and off-board fare collection [18]can help to expedite the boarding process compared to magnetic strip cards or cash fare payment [55], thereby reducing dwell times, but this is not always the case [37,38,139]. Fare payment policies can also have a significant impact on bus service reliability, as they can be a major determinant of dwell times at stops [118]. Dorbritz et al. found that the most significant challenge in fare payment to reliability was the variability in time spent in the ticket sale process [13]. Faster, more efficient and consistent forms of fare payment are preferable from a service reliability perspective. For example, an agency may opt to eliminate cash fare payment from their system as a policy to improve reliability. Three studies focused on modeling bus service reliability for STM in Montreal, using the implementation of a new smartcard fare collection system as a parameter. Although each study reported findings that smartcard fare payment actually increased schedule deviation and running times, this was in comparison to the flash pass system that had been in place previously. In one study, the introduction of smartcards instead of flash passes for fare payment increased the odds of being late by 69 percent [139]. In another, the introduction of smartcard fare payment increased running times on the route in question by 5.83 seconds at the beginning of the implementation period and by 52.61 seconds by the end of the implementation period [37]. Most recently, smartcard fare payment was found to have a significant positive correlation with running time deviation, the coefficient of variation of running time, and the coefficient of variation of running time deviation, and deemed to be less ideal than flash passes with regard to running time and reliability [38]. Egge and Qian compared entry fare and exit fare policies on the Port Authority of

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-26 Allegheny County, Pennsylvania. They found that exit fares can reduce dwell time per passenger when boardings and alightings are concentrated at few major stops. They concluded that exit fare policies are most adequate for Bus Rapid Transit (BRT) or express service [47]. Fare collection and payment policies also seem to impact reliability. In a study of Translink in Queensland, Australia, the average boarding time per passenger was found to be 3 s for smartcard users, 30 s for onboard top-up, and 15 s for paper ticket users. Eliminating onboard top-up for smartcards resulted in a 15-18 percent improvement in reliability for the bus route in question. Smartcard users can still top-up their cards in a variety of places, so the inconvenience of this change was assumed to be minimal [118]. Staff Driver behavior, which is based on experience [50] and hours worked per week [128,134], among other factors [75,144], was shown to have a significant impact on bus service reliability in a few studies. Generally speaking, the more hours worked and years of experience associated with a driver, the more likely they are to have a positive impact on bus service reliability. Strathman and Hopper found that early arrivals are more likely during the first two weeks of a new sign-up (driver assignment). Their model results also indicate that part-time drivers are more likely to arrive late at stops and terminal points, which may be addressed through driver training and reduced use of part-time drivers [134]. In the mid-1990’s Shiftan and Wilson found that working overtime hours is not positively related to absenteeism. On the contrary, employees who worked overtime had a lower rate of absence [128]. In a regression model of coefficient of variation of running time, variation in driver experience (in years) had a significant negative relationship with the dependent variable. The researchers found that a 1 percent increase in coefficient of variation of driver experience led to a 5 percent decline in the estimated running time coefficient of variation from their model [50]. Strathman et al. analyzed the running times of bus operators on TriMet in Portland, OR. Their model found that operators’ running time decreased by 0.57 seconds for each month of additional experience. Although the marginal impact was small, its impact was important due to the wide variation in operator experience [135]. Bus drivers' lifestyles at home and at work play a critical role in their physical and psychological health, as well as their job performance. Stress factors such as traffic, violent passengers, and tight running schedules may have a negative impact on bus drivers, which can result in absenteeism, vehicle crashes, and other factors that may impact bus service reliability [144]. Now that automated vehicle location systems and real-time dispatching technologies are increasingly available, transit operators are able to make adjustments to reduce the impacts of driver disparities on bus service reliability. In a recent study by Ji, He, and Zhang, bus drivers' positive responses to real-time schedule information were shown to help improve bus service reliability [75]. Infrastructure In addition to transit priority, road type was found to significantly impact the service reliability of buses running in particular corridor types. A group of Australian researchers modeled average travel time, buffer time, and coefficient of variation of travel time against several factors, including whether buses run on arterial roads, busways, or motorways as opposed to other road types. Dummy variables for arterials, busways, and motorways versus other road types (local, district, and suburban roads) were used, with results indicating that the influence of these factors on average travel time, buffer time, and coefficient of variation of travel time varies depending on the operating environment. The average travel time model shows the greatest decrease in travel time associated with arterial roads, followed by motorways, and then busways, versus other road

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-27 types, though each of these three variables had negative coefficients. The buffer time model has the largest negative coefficient for motorways, followed by arterials, and then busways. The coefficient of variation of travel time model produced a similar result as the buffer time model [100]. Passengers Passenger activity, including boardings [42,50,100,121,134], alightings [42,50,100,134], loads [50,63,137], and passenger demand or ridership [48,93] have been shown to significantly influence bus service reliability. While there have been mixed results with regard to the relationship between these variables and bus service reliability, the vast majority of evidence indicates that increased passenger demand and activity often correspond to declining service reliability, due largely to increases and fluctuations in dwell times [42,55]. Figure 4.6 shows an example of historical passenger occupancy estimates by time of day (rows) and stop location (columns) [63]. Figure 4.6 – Example of Historical Passenger Occupancy Estimates by Time of Day and Stop Location [63] Passenger boardings were found to have a significant positive impact on the probability of arriving on-time, versus early, but no effect on the odds of buses arriving on-time versus late. More passenger alightings corresponded to increased likelihood of both early and late arrivals [134]. Furthermore, Strathman, Kimpel, & Callas reported a statistical correlation between passenger load and headway deviation at Portland State University [137]. As shown in Figure 4.7 this study demonstrated a correlation between headway variability and passenger loads, but the

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-28 researchers did not analyze the internal dynamics of the relationship. It is stated that "headway delays are a primary cause for passenger overloads", when the converse is also true. Adding to the confusion about the nature of this relationship, increased passenger demand appeared to decrease journey time variability in a study by Liu & Sinha [93]. Figure 4.7 – Passenger Load Versus Headway Delay for Morning Peak Inbound Trips on an Example Route [137] Another important factor influencing reliability is the use of wheelchair lifts or ramps [42,50,55,138] and bike racks [55], which relate to bus characteristics as well as the proportion of passengers requiring the use of these devices. Berkow et al. present a set of tools to visualize where lifts are used the most, and their impact on dwell time [14]. Using data collected from TriMet, Strathman, Kimpel, Dueker, Gerhart, and Callas developed a model to describe the relationship between several factors and bus running times. This model was used to estimate that stops involving a lift operation are estimated to require 68 seconds, while a stop involving a single boarding or alighting is estimated to require about 11 seconds [138]. A subset of this group of researchers conducted another study of TriMet in which they reported that the estimated impact of a lift operation on dwell time (for high-floor and low-floor buses combined) was an increase of 62.07 seconds. Low-floor buses correlated with a 0.11 second reduction in dwell time per dwell, which resulted in an average savings of 3.96 seconds in total running time per trip. Low-floor buses were also reported to help reduce dwell time for lift operations by 4.74 seconds, or 5.8 percent [42]. In a similar study of Metro Transit in Twin Cities, Minnesota, each lift activity increased running time variation by 24 percent and headway deviation by 3 percent [50]. Several models have been developed to better estimate the impacts of passenger activity on various measures of service reliability. Using data from Metro Transit in Twin Cities, Minnesota, researchers found positive correlations between:  running time deviation and boardings, alightings, and lift use;

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-29  headway deviation and boardings and lift use; and  coefficient of variation of running time and the coefficient of variation of average passenger load. Based on their models, each boarding is expected to increase running time deviation by 0.4 percent, while each alighting adds 0.2 percent. Each lift activity increased running time variation by 24 percent and headway deviation by 3 percent, in this case [50]. Similarly, Surprenant-Legault and El-Geneidy found that each passenger's activity (boarding, alighting, etc.) increases the odds of being late by 2 percent, and each rear boarding passenger increases the odds of being late by 6 percent [139]. In a recent study of the STM in Montreal, more passenger activity and rear-door activity corresponded to an increase in running time deviation. The coefficient of variation of passenger activity also showed a significant positive correlation with coefficient of variation of running time. Coefficient of variation of rear-door activity had a significant positive correlation with both coefficient of variation of running time and coefficient of variation of running time deviation [38]. Most recently, a group of Australian researchers found average travel time to increase with higher numbers of boardings and alightings, but to decrease with alightings squared, reflecting easier boardings and alightings in uncrowded conditions. Their buffer time model included significant positive coefficients for the standard deviation of boardings and the standard deviation of alightings. Coefficient of variation of travel time was also found to have a statistically significant positive relationship with these two parameters [100]. Models were also developed specifically for understanding the impacts of passengers on dwell times. Results from a regression model using dwell times as the dependent variable indicate that each boarding passenger adds 3.48 seconds to the base dwell time (plus 5.14 seconds for door operation), while each alighting passenger adds 1.70 seconds. The estimated impact of a lift operation on dwell time (for high-floor and low-floor buses combined) is an increase of 62.07 seconds [42]. Along these same lines, Fletcher and El-Geneidy developed a model to analyze factors impacting bus dwell times. Factors shown to significantly increase dwell times in this model included wheelchair ramp events, bike rack events, passengers with bulky items, front-door alighting, cash fare payment, and higher passenger volumes [55]. Time Time-related and seasonal factors have also proven to be significant influencers of bus service reliability, especially as it relates to peak travel periods [37,38,50,52,71,134], time of day [64,138], and seasonal travel patterns [121]. Peak period travel has been estimated to have a negative impact on bus service reliability [92,134], but not in every case [37,38,50,52]. Figure 4.8 shows a comparison of on-time performance statistics during the morning and afternoon peak periods [71].

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-30 Figure 4.8 – Morning and Afternoon Peak Period On-Time Performance Statistics for Example Bus Route [71] In an early work by Strathman and Hopper, peak period traffic, especially in the afternoon peak, was shown to have a negative impact on bus service reliability [134]. Almost a decade later, Strathman and his team estimated early morning and night trips to need about 250 (7.7 percent) fewer seconds than midday trips. Morning peak trips were estimated to need about 100 (3 percent) fewer seconds than midday trips, while afternoon peak trips may require 138 seconds (4 percent) less than midday trips [138]. A few years after that, El-Geneidy, Strathman, Kimpel, and Crout contracted past findings in reporting that running time reliability was estimated to increase during the morning peak period [52]. Distinguishing between the morning and afternoon peak periods seems to be critical in understanding their relationships to bus service reliability. To add to these findings, regression models were run for three measures of reliability used in a study of data from Metro Transit in Twin Cities, Minnesota. The afternoon peak period was found to have a positive impact in the running time deviation model, but a negative impact in the headway deviation model. Afternoon peak periods were found to increase running time deviation by 5 percent. Meanwhile, only the morning peak was found to have a significantly positive influence in the coefficient of variation of running time model [50]. Lin, Wang, & Barnum reported that buses traveling inbound to downtown Chicago tended to have worse schedule adherence in the morning peak [92]. This is not surprising, as typical commute patterns in major U.S. metropolitan areas still include more people traveling into major employment centers in the morning and away from them in the afternoon, and since urban core areas tend to be more popular and congested than lower-density areas. Diab and El-Geneidy conducted a fairly comprehensive analysis of factors impacting bus service reliability at STM in Montreal, in which they found that afternoon peak period trips are much longer than midday trips, while morning peak, nighttime, and midnight trips are faster by 47 seconds, 100 seconds, and 219 seconds, respectively [37]. In a follow-up study of the same system, these

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-31 researchers found that peak period travel and off-peak travel both significantly decreased running time deviation in their model, but that off-peak travel decreased running time deviation more than peak travel did. They also found that midnight and early morning travel decreased the model coefficient of variation of running time. Interestingly, morning peak travel and night travel both decreased the estimated coefficient of variation of running time deviation [38]. To investigate the relationship between seasons, travel patterns, and bus reliability, regression analysis was performed to compare peak to midday running times based on season, and the findings were as follows. The seasons (summer, fall, and spring) have different peaking characteristics. In the summer, morning peak running times were found to be lower than midday running times, possibly due to schools not being in session. During the spring and fall morning peak running times were estimated to be higher than in the midday period. Running times were also found to be highest in the fall. A linear model was developed to compare mean spread to season, route location and configuration, and bus line, but this was based on a small sample of data [121]. In his graduate thesis, Uniman used ordinary least squares regression analysis to estimate that when an incident occurs reliability buffer time increases by about 10 minutes above normal peak period performance. Overall findings indicated that incidents have a large impact on bus service reliability, which may be underestimated if only average metrics are considered [148]. Inclement weather conditions, such as precipitation, have been shown to have a significant negative impact on bus service reliability [37,38,48,100,139]. In a study of the STM in Montreal, researchers found that each centimeter of snow increases the odds of being late by 20 percent [139]. The following year, Diab and El-Geneidy produced a model for the same transit agency, in which running time is expected to increase by 0.79s per trip for each millimeter of rain or by 1.81s per trip for each centimeter of snow [37]. The same researchers found significant positive influences of precipitation and snow on running time deviation for the STM. The coefficient of variation of snow was also found to have a significant positive impact in their models of coefficient of variation of running time and coefficient of variation of running time deviation [38]. Most recently, statistically significant predictors of average travel time for Translink bus service in Brisbane, Australia were found to include rain versus good weather, which had a positive coefficient. The coefficient of variation in travel time model included light rain versus good weather with a positive coefficient and rain versus good weather, but with a negative coefficient [100]. These findings could be explained by drivers having more variable speeds under light rain conditions than under heavy rain conditions. Traffic Many researchers have analyzed bus reliability data and found traffic conditions to be a significant factor impacting bus service reliability and travel times [18,22,38,48,61,73,93,99,117,121]. Heavy traffic tends to have a negative impact on reliability and travel times in the absence of priority treatments, and in some cases where priority treatments are present but not able to eliminate all traffic impacts on bus service [73]. Of course period of travel, direction, location of travel, such as in the central business district, traffic signals, intersections, and transit priority treatments all impact traffic conditions. The Second Strategic Highway Research Program Report recommends the systematic consideration of external factors that may impact traffic condition [126]. These factors are discussed further in the following sections. Traffic conditions tend to have more of an impact on bus service reliability in highly congested urban areas. One example of this was provided in a report from the United Kingdom, which showed much higher rates of lost mileage due to traffic (scheduled bus mileage that is not

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-32 operated) in London as compared to the rest of England [117]. This relationship is clearly demonstrated in Figure 4.9. Figure 4.9 – Percent Lost Scheduled Bus Mileage by Cause and Geographic Area [117] Using simulated results, Liu and Sinha found that increased congestion increased the mean and variance of bus journey times [93]. In Ehrlich’s study on Transport for London bus service, the factors that were most highly correlated with the ratio of actual wait time (AWT) to scheduled wait time (SWT) included:  percent lost mileage due to traffic (+),  morning peak buses per hour (+),  ridership (+),  ridership per kilometer (+),  central London route length (+),  congestion charging zone route length (+),  vehicle length (+),  routes per workstation (a measure of service controller workload) (-), and  bus priority measures (+). A regression analysis of the highly-correlated variables showed greatest correlation between AWT/SWT and lost mileage due to traffic (+), precipitation (+), ridership (+), and route length (-) [48]. Figure 4.10 to the right provides further evidence of the relationship between service reliability and the percent of lost mileage due to traffic.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-33 Figure 4.10 – Percent Lost Mileage due to Traffic vs. AWT/SWT [48] Researchers in Australia modeled bus travel time reliability, in terms of average travel time, buffer time, and coefficient of variation in travel time, finding the recurrent congestion index, number of signals, central business district, and road type to be statistically significant for all three models. A Recurrent Congestion Index (RCI) was calculated as the mode speed divided by the free flow speed, and increases in RCI were found to significantly decrease average travel time, buffer time, and coefficient of variation of travel time in the models. Higher values of RCI correspond to less congestion, which is associated with shorter travel times and better reliability, as expected [100]. Transit priority treatments [124], such as exclusive bus lanes [18,30,37,38,139], busways [22,71,100], and transit signal priority [18,29,30,37,38,41,79,108], have been shown to have a significant impact on bus reliability and running times in several cases [37,38,48]. It should be noted that mixed results have emerged from the literature, with one study showing bus priority measures having a positive association with the ratio of actual to scheduled waiting times [48], rather than helping to improve reliability, as expected. In a case study from Montreal, a reserved bus lane was found to decrease the odds of being late by 65 percent [139]. Diab and El-Geneidy found that reserved bus lanes decreased route running time by 35.26 seconds on average, though this benefit could have been larger in magnitude if cars were not allowed to use the bus lanes to turn right (especially given that vehicles are not allowed to turn right on red) [37]. In a follow-up study, these researchers considered factors impacting running time deviation and coefficient of variation of running time. Reserved bus lanes may result in decreased running time, but increased variation, especially when general traffic is allowed to use bus lanes as right turn lanes but not allowed to turn right on red [38]. Busways were found to reduce average lateness of buses by 48 percent and standard deviation of lateness by 45 percent [71]. Transit signal priority was shown to decrease the standard deviation of arrival time deviation from schedule by 3.2 percent on average across 30 simulation cases in Arlington, Virginia [29]. Implementing transit signal priority (TSP) for articulated vehicles only decreased running times for all vehicles by 4.76s (0.3 percent) on average [37]. TSP has been shown to improve running times with no significant impact on variation in at least one case [38]. Signal priority was shown to have the biggest impact on speeds of any action in a case study of San Francisco, while New York City reported success with off-board fare collection and increased stop spacing, combined with transit signal priority, bus-only lanes, and all-door boarding. Far-side bus stops were also recommended to this end, especially as they facilitate transit signal priority [18].

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-34 4.5 Factor Organization With so many factors impacting or potentially impacting bus service reliability, grouping and organizing all the factors in a coherent and comprehensive way can be quite challenging. Many researchers have taken the approach of not grouping factors at all, which works quite well for a small to moderate number of factors. However, with nearly 200 factors identified through this literature review alone, the utility of organizing factors into coherent groups is apparent. These groups can be used to help identify factors that directly relate to transit operations, and can therefore be changed, versus those that are largely out of control of the transit agency. No classification system is perfect, but several researchers have proposed structures that may prove useful in dividing factors into more manageably sized groups. A few different examples of factor organization schemes are provided in this section. An early example of factor grouping came from Strathman and Hopper, who listed a variety of factors that might impact bus service reliability, as well as a few logical groups of factors. Their factors included:  driver experience and behavior,  sensitivity of the schedule to route conditions (headways, run times, layover times),  complexity of the route (length, stop spacing, passenger activity),  traffic congestion and incidents,  signal timing,  weather, and  disruptions from on-street parking [134]. Skabardonis divided reliability factors into three broad groups:  network configuration and characteristics (single arterial, grid network, signal spacing, number of lanes, pedestrian presence, type and operation of the traffic control system)  network traffic patterns (traffic volumes, turning movements, variability in traffic volumes, level of congestion, extent to which traffic congestion interferes with bus operations and the nature of the interference)  frequency and characteristics of transit service (bus volume, type(s) of bus operations, transit routes, bus stop locations and design, amount and variability of dwell times, and communication and monitoring equipment for transit vehicles) [129] The authors of the Transit Capacity and Quality of Service, 3rd Edition reported that factors influencing speed, capacity, and reliability of transit are similar. Internal factors identified were:  vehicle quality and age,  vehicle availability and breakdowns,  driver availability,  transit preferential treatments,  route length,  supervision,  control strategies,  driver experience, and  schedule achievability. External factors cited included:  snow and ice,

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-35  heat,  leaf fall,  traffic signals,  traffic congestion,  variability in traffic demand,  road construction,  passenger demand variability,  wheelchair lift or ramp usage, and  door holding [82]. In TCRP Report 95: Transit Scheduling and Frequency, factors identified were classified as either environmental factors or inherent factors. Environmental factors include traffic conditions, signals, variations in boardings and alightings, and availability of drivers and vehicles. Inherent factors include platooning, missed runs, and unplanned deviations [53]. Sterman and Schofer found a negative correlation between intersection density and reliability. They also found that intersection length and presence of a signal had a strong impact on reliability [131]. Liu and Sinha succeeded in creating a fairly comprehensive list of factors, sorted into one of four characteristic groups. Traffic characteristics included traffic composition, day-to-day and within- day variation in travel demand, and traffic congestion levels. Route characteristics included route length, number of lanes, bus stop location, provision of bus lanes, number of intersections, priorities at junctions for buses, on-street parking, passenger volumes, direction of travel, and driver behavior. Passenger characteristics referred to passenger volume at stops, variability in passenger demand, route choice, and arrival distributions. Bus operational characteristics included scheduling, staff shortages, fleet availability, vehicle maintenance, fare collection and ticketing system, and variability in driver behavior and experience [93]. In his master’s thesis, Uniman proposed organizing reliability factors as either recurrent or non-recurrent. Recurrent reliability factors were said to include characteristics of service, journey length, scheduled headways, and interchanges. Non-recurrent reliability factors included operations control interventions, incident-related disruptions, and seasonality. As shown in Figure 4.11, he also adapted a system from Abkowitz (1983) in which factors were classified as either controllable/intrinsic or exogenous/environmental and either long-term/planning or short- term/real-time [148].

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-36 Figure 4.11 – Factor Classification Scheme from Uniman, Adapted from Abkowitz (1983) [148] Mandelzys and Hellinga broke causes of unreliability into three groups: 1) travel time causes (traffic congestion, weather, signals, unscheduled stops), 2) dwell time causes (passenger activity and demand, traffic volumes, lift use), and 3) upstream causes (deviations at previous stops) [102]. Most recently, a group of Australian researchers proposed three different factor groups: 1) planning (link length, schedules, service frequency); 2) operational (departure delays, passenger activity, vehicle type, fare type, field supervision management); and 3) environmental (route characteristics, traffic conditions, weather, incidents, road work) [100].

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-37 5.0 Perception of Reliability Reliability and the perception of reliability are closely related, but not the same. While reliability can be measured rather precisely and accurately, the perception of reliability is difficult, if not impossible to quantify. Perceived reliability is also rather subjective, in that it varies from person to person, and perhaps even from day to day for an individual. While, from the operator perspective, objective measures of on-time performance and miles traveled between breakdowns may be their primary concerns with regard to reliability, perceived reliability is what matters most to the customer. As defined in the 3rd Edition of the Transit Capacity and Quality of Service Manual, reliability from the passenger perspective relates to arriving at destinations on-time and not having to wait too long at bus stops [82]. This section will focus on customer perceptions of bus service reliability, and how these perceptions may change based on changes in the actual reliability of a service, information available and other factors. An early mention of passenger perceptions was found in a U.S. Department of Transportation commissioned report on transit service reliability, wherein it was noted that reliability has a major impact on passengers’ perceived quality of service [2]. Decades later, researchers in the United Kingdom reported that punctuality is a highly valued characteristic of transportation systems, and that unreliability is especially costly when public transit is the chosen mode. Although these statements seem fairly intuitive, the researchers provided readers with little data to support their claims [10]. A more data-driven approach was taken by researchers in the Netherlands, who found evidence of a strong risk aversion when it comes to uncertainty in the reliability of public transport. Rietveld, Bruinsma, and vanVuuren used survey data to calculate the relationship between an "uncertainty minute" (one spent waiting, not knowing when a transit vehicle would arrive) and a “certainty minute” spent inside a transit vehicle that is proceeding along its route. In this study, the weight of an “uncertainty minute” was estimated to be 2.4 times the weight of a "certain" in-vehicle minute [119]. Shortly thereafter, Li found service reliability to be an important factor in transit users’ perceived travel time [89]. The following year, Evans et al. reported that on- time performance positively impacts riders and ridership due to less waiting, decreased travel times, fewer missed connections, more on-time arrivals, and reduced uncertainty [53]. The guidebook: Placing Value on Travel Time Reliability presents a methodology to quantify monetarily the different components of reliability. The guidebook provides a detailed methodology to assess the impact of unreliability and its treatments. The tradeoff between missed income for passengers due to excess waiting time and cost of treatments is explored in a cost/benefit analysis [81]. One difficulty with unreliability for many transit riders is the unknown wait time they will face. Riders stand at a corner scanning the horizon for an approaching bus, wondering when it will come, or if it will come. Another day one or more of these riders may time their arrival exactly to the scheduled minute, just in time to watch an early-running bus pass by their stop and realize that they have another 30 minutes (or longer) to wait for the next bus. To avoid this problem, many bus riders will time their arrival at the stop early enough to ensure that they do not miss the bus. However, this buffer time adds to the time and inconvenience associated with riding the bus. Less reliable services may increase uncertainty and require riders to budget even more buffer time, in the absence of real-time arrival information. By knowing when the bus is actually coming, the entire dynamic changes. Unreliability is less of an issue if the rider knows in advance when the bus is coming, even if it is a few minutes late. If transit agencies hope to retain choice riders and increase ridership, they need to allow their customers to maintain some control over their trips by providing them with real-time information. Giving passengers real-time information about the arrival of the next bus helps minimize waiting time, improves the perception of the wait, and

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-38 alleviates the stress of wondering when the bus is coming. Although bringing the perception of wait time in line with the actual wait time will not improve the reliability of transit, it can improve the perceptions that relate to reliability by reducing rider uncertainty. As of 2003, when TCRP Synthesis 48: Real-Time Bus Arrival Information Systems was written, only three U.S. and five international agencies had measured the rider reaction to real-time arrival information. According to the synthesis, London Transport’s Countdown program, which used at- stop real-time arrival signage, found that the perceived wait time dropped from 11.9 to 8.6 minutes. In addition, passengers felt less stress, and 64 percent of those surveyed thought that reliability had improved since implementation (although it had actually decreased) [127]. It has since been shown that real-time transit traveler information can result in a mode-shift to public transportation [109]. This stems from the riders’ ability to feel more in control of their trip, including their time spent waiting and their perception of safety. Furthermore, recent advances in mobile device technology are enhancing opportunities for more productive uses of travel time [96]. Now the introduction of more powerful personal mobile devices is also changing the wait time portion of the transit trip. In order to determine the possible benefit of real-time information, Mishalani, McCord, & Wirtz looked at the difference between perceived and actual waiting times at a bus stop. The study was conducted at campus bus service stops on the Ohio State University campus by waiting at bus stops and observing the arrival time, then asking riders how long they had been waiting when the bus approached. There was a statistically significant difference in the perceived versus actual waiting time, with perceived wait time exceeding actual wait time in most cases. Longer walking distance to a stop corresponded to greater perceived waiting times, whereas the presence of a time constraint reduced the degree of wait time over-estimation. However, this difference was small and no actual real-time arrival information was tested for comparison [106]. Katrin Dziekan has done significant work studying rider reactions to real-time arrival information via at-stop displays. In 2006, she and Vermeulen conducted a before and after study of the perception of wait time after the installation of real-time arrival information signage on a tramline in The Hague, Netherlands. The study was conducted via a survey mailed to the same respondents before the installation, three months after the installation, and again 16 months after the installation. The average perceived wait time decreased from 6.22 to 5.00, and then to 4.81 minutes, respectively, a difference of 20 percent over three months and 23 percent over 16 months [44]. In another paper, she stated that real-time arrival displays increase feelings of security, reduce uncertainty, increase ease-of-use, adjust travel behavior, and improve customer satisfaction. Most important to this investigation, the presence of permanent real-time arrival signage at transit stations and the ability to determine when the next vehicle is coming was shown to bring travelers’ perception of wait time in line with their true wait time [43]. In a before and after study of the ShuttleTrac system on University of Maryland College Park campus, seven models were estimated using panel data to determine behavioral and psychological responses [166]. The real-time information for ShuttleTrac is provided via terminals at selected stops, a large display at an activity center, telephone, and website. The results indicated that real-time information increased rider’s feelings of security after dark and boosted their overall level of satisfaction. However, real-time information was not found to significantly increase the number of trips taken, nor was it found to reduce waiting anxiety or the perception of on-time performance. Another study of perceived and actual wait times [157] found that for riders without real-time information, perceived wait time is greater than measured wait time, but having real-time information brings perceived wait time in line with actual wait time. In addition, mobile real-time information users in the study were observed to actually wait almost 2 minutes less than those

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-39 arriving using traditional schedule information (which, when aggregated over 100,000 riders per week, adds up to a considerable time savings). Studies in Chicago [140] and New York [19] found ridership increases of about 2 percent on routes with real-time information. In addition to the availability of real-time bus arrival information, other factors relating to transit service and operations have been shown to significantly impact the perception of waiting time. Daskalakis and Strathopoulos found a statistically significant relationship between perceived waiting time deviations and headway or frequency of service, though the relationship was not found to be linear. Larger headways correspond to greater perceived deviations, but at a diminishing rate. They suggested that transit operators wanting to improve passenger wait times and service quality should focus on headway reliability, and then try to shorten headways [34].

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-40 6.0 Bus Service Reliability Improvement Strategies More than 100 different strategies for improving bus service reliability were found in the literature. These are broken into three categories, which include operational, physical, and technological improvement strategies. The most commonly-cited types of improvement strategies are included in this section. Figure 6.1 shows the number of references citing the top 20 most popular improvement strategies from the literature review. Figure 6.1 – Top 20 Most Referenced Improvement Strategies from the Literature Review As shown above, transit signal priority and exclusive right-of-way were the two most-cited bus service reliability improvement strategies, followed by schedule adjustments, holding, control strategies, real-time bus arrival information, and route adjustments. Research on the effectiveness of many of these strategies is provided in the next section. 6.1 Operational Improvement Strategies Operational bus reliability improvement strategies include those relating to the routing, scheduling, and control of bus service. Furthermore, policies and practices related to drivers, stopping, maintenance, and boarding are included in this subsection. Route and Network Adjustments A variety of route and network design recommendations and adjustments [2,5,18,50,54,133,134,145] are proposed for improving reliability, including shorter routes 32 31 16 14 12 10 10 8 7 7 7 6 6 6 6 5 5 5 5 4 0 5 10 15 20 25 30 35 Transit signal priority Exclusive right-of-way Schedule adjustments Holding Control strategies Real-time bus arrival information Route adjustments Driver training Smartcard payment Stop consolidation Stop-skipping Automated vehicle location systems Limited-stop service Low-floor buses Queue jump lanes All-door boarding Fleet maintenance and periodic replacement Intelligent transportation systems Off-board fare collection Driver incentives Im pr ov em en t S tra te gy

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-41 [30,83,115,134], rerouting service around congested areas [54], and dynamic transit routing [27,98]. Schedule and Headway Optimization and Adjustments Schedule adjustments have also been proposed [2,27,48,50,53,70,71,78,82,88]. Recommended adjustments included pulse scheduling [145], increased transfer times [119], and improved transfer reliability [11]. Driver Training, Support, and Incentives Driver supervision and incentives [6,27,28,71,134], as well as training [6,27,50,71,144], are thought to have a positive impact on bus service reliability. Holding, Dispatching, and Control Many control strategies [6,35,39,50,68,70,71,82,101,111] are recommended, including those involving holding [1,27,35,48,54,60,70,77,83,101], dispatching [90,115,134,151], diverting [134], short-turning [48,101,133], and real-time monitoring and control [6,39,50,62,68,75,86,134]. An example of a short-turning strategy is shown in Figure 6.2. Figure 6.2 – Example Short-Turning Strategy [101] Limited-Stop Service, Stop skipping, and Stop Consolidation Express or limited-stop service [5,37,100,139,143], stop skipping [68,73,77,83,122,133,160], larger stop spacing [18], fewer stops [90,115,134], and stop consolidation [27,50,52,100,160] are also expected to improve bus service reliability. Fleet Maintenance Fleet maintenance and periodic replacement [27,67,147,150]was cited in several studies as critical to long-term bus service reliability.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-42 All-Door Boarding All-door boarding [6,18,82]and rear-door boarding [141] are thought to speed the boarding process, thereby reducing dwell times to improve running times and reliability. 6.2 Physical Improvement Strategies Physical bus service reliability improvements are those that relate to infrastructure, including right- of-way and stops, as well as bus characteristics. Bus Lanes, Shoulder Lanes, and Queue Jump Lanes Exclusive bus lanes [4,6,27,30,37,38,49,72,73,83], as well as busways [22,86,87,132], bus lanes with intermittent priority [49], and dynamic bus lanes [4], may help to improve bus speeds, travel times, and reliability. A few studies have suggested that shoulder lanes and bus on shoulder operations may improve bus service reliability [39,103,104]. Queue jump lanes and bus-only approaches at intersections may also reduce unreliability caused by delays at traffic signals [6,39,49,132,164]. An example of a bus-only approach is shown in Figure 6.3. Figure 6.3 – Bus-Only Approach in Portland, Oregon [32] Far-Side Stop Placement Far-side bus stops are recommended from a reliability standpoint, especially when transit signal priority is used [5,18,86]. Level Boarding, Low-Floor Buses, and Articulated Buses Several proposed reliability improvement strategies relate to buses themselves, and how well they complement bus stops and stations for ease of boarding. Buses and stops that allow for level boarding [6,82,87], including through the use of low-floor buses [39,42,55,67,87,141], and those with more and wider doors [82,87], such as articulated buses [37,39,100], are expected to improve reliability through faster boarding and shorter dwell times.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-43 6.3 Technological Improvement Strategies Technological bus service reliability improvement strategies are those that leverage various technical innovations, such as intelligent transportation systems and automated vehicle location, to improve bus service. New and more efficient methods of fare payment, including smartcards, mobile ticketing, and off-board fare payment, are also included in this section. Transit Signal Priority and Intelligent Transportation Systems Transit signal priority appears to be the most-cited bus service reliability improvement strategy from the literature [5,6,18,27,28,29,30,37,38,40]. Intelligent transportation systems, which include transit signal priority as well as better monitoring and control of traffic overall, have also been proposed to improve bus service reliability by several researchers [5,6,27,39,53]. Automated Vehicle Location Systems and Real-Time Bus Arrival Information Automated vehicle location (AVL) and monitoring systems are said to help with real-time tracking and control of bus service [39,48,50,62,75,86,110], which can lead to various reliability improvements when the collected information is properly utilized. Another key benefit of AVL systems is that they can be used to provide real-time bus arrival information to transit riders, which may help customers to better plan their transit trips with regard to reliability and minimizing wait times [19,27,39,49,43,44,50,106,115,127] Innovations in Fare Payment Fare payment methods have come a long way since the days of cash fares and tokens. Many fare collection and payment technologies are helping to reduce dwell times, thereby improving bus service reliability. Among the fare payment methods that may offer reliability benefits (although these are relative to existing methods and alternatives available) are contactless cards or smartcards [37,39,55,83,100,109,118,156], proof-of-payment or flash passes [82,156], and off- board fare payment [5,18,39,82].

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-44 7.0 Impacts of Bus Service Reliability Improvement Strategies While it is easy to propose bus service reliability improvement strategies, proving their effectiveness is quite another matter, especially given the myriad of factors that could potentially influence bus service reliability. As described in the following section, many researchers have used models and simulation to estimate the impacts of various bus service reliability improvement strategies, but the results may not accurately reflect the impacts that these strategies have when they’re implemented in the real world. Several researchers have also found empirical evidence regarding the effectiveness of certain improvement strategies, but this evidence is often limited to case studies that typically should not be generalized. Context and the specific implementation of a strategy will always play a role in determining the success of that strategy in a particular case, but many of the studies that evaluated the impacts of bus service reliability improvement strategies provide a basis for further investigation of the effectiveness of these techniques. 7.1 Model-Based Evidence Models and simulations have been used to demonstrate the effectiveness, or ineffectiveness, of several bus reliability improvement strategies. While much evidence has been gathered regarding the impacts of route adjustments, schedule optimization, control strategies, limited-stop service, exclusive right-of-way, bus type, transit signal priority, and fare payment methods, many of these studies provided mixed or inconclusive results. Furthermore, research on the impacts of far-side stop placement, vehicle maintenance, driver training, and automated vehicle location systems has been sparse. As such, these subsections are omitted below, due to lack of content. Route and Network Adjustments An early study by Turnquist and Bowman found through modeling and simulation that grid networks appear to be less disrupted by transfers than radial networks. On-time arrival of vehicles at major transfer stations was also found to be important, especially for radial networks [145]. Decades later, a network optimization model was proposed, which takes travel time reliability into account. The Chinese research team’s results indicate that this method may help to improve transit network reliability and reduce passenger travel times [162]. Speaking to route length, a group of researchers in Beijing found that shorter route lengths, specifically those less than 30 km, showed better reliability than longer routes for three measures of reliability [30]. Through simulation and a case study of the Land Transport Authority in Singapore, Lee, Sun, and Erath found that unreliability increases with bus line travel distance, in a positive linear fashion. Simulation was used to show that the level of service is improved with shorter operating distances, in terms of passengers' average waiting time and buses' occupancy. The researchers also found that shortening the distance between time points seems to be an effective method for improving bus service reliability [83]. Van Nes and Van Oort explore the tradeoff between long routes that avoid the need for connections and short routes that reduce the opportunity for delay propagation [152]. Their paper modeled the impact of splitting routes into shorter segments and evaluated the impact on waiting time. They found that the benefit of splitting routes is maximized at stops with few passengers riding through. Schedule and Headway Optimization and Adjustments Turnquist and Bowman found through modeling and simulation that variability in time associated with making a transfer, which is directly influenced by frequency of service, appears to have a major impact on transit system reliability. They also found service reliability to be much more influenced by frequency of service than route density. Transit service reliability may be enhanced by preventing vehicle bunching and/or by breaking up vehicle platoons when they form. They

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-45 state that, in general, operators should avoid providing excess slack time in their schedules, except where a large number of passenger transfers could benefit from having enough slack time to ensure successful connections [145]. More recently, Hong noted many tradeoffs in the scheduling process. He found that headway decisions impact the tradeoff between operational cost and service quality, both of which increase with smaller headways. Increasing schedule time and the number of time points can reduce deviation at time points, thereby improving schedule adherence. While this may help to reduce out-of-vehicle waiting times, it may increase in-vehicle waiting times. An analytical model was used to demonstrate the impacts of increasing schedule time, and the results indicated that this strategy does increase the mean travel time while decreasing the variability in travel time [70]. The effects of several improvement measures on three reliability measures, Punctuality Index based on Routes (PIR), Deviation Index based on Stops (DIS), and Evenness Index based on Stops (EIS), were also examined. Lengthening headways had a positive impact on the DIS reliability measure, but a negative impact on the EIS measure [30]. Peng et al. found that travel speed was an important factor leading to bus bunching [113]. Holding, Dispatching, and Control Many different control strategies have been tested over the last few decades, with various levels of success reported. Comparisons between these studies are difficult to make, since each study essentially tested a different approach to bus control. As such, these studies are presented individually below. An optimization algorithm was proposed by Eberlein, Bernstein, and Wilson to reduce passenger waiting time compared to a regular schedule by minimizing squared headways without using event-based simulation. They found that the more downstream the control point was the better. However, they also found that simply adding control points does not significantly improve performance [45]. A simulation conducted by Moses as part of his thesis indicated that adjusting driver and passenger behavior, as well as control strategies, for only bunched vehicles has a limited impact on service [107]. More generally, the TCRP Report 135, Controlling System Cost, highlights the conflicting objectives of maintaining short scheduled running times and reliability. The report recommends setting scheduled running time as the 93 percentile of observed values [17]. A study by Van Oort, Wilson, and Van Nes indicated that, assuming random arrival times, a schedule-based holding strategy is more effective in terms of minimizing total travel time than a headway-based holding strategy [153]. However, in reviewing this research, the authors of this report found that there was buffer time in the schedule, but none in the headways, which may have put the latter method at a disadvantage. The results of this experiment, shown in Figure 7.1 and expressed as average and 95th percentile irregularity, highlight the different impacts of four different holding strategies against a no holding scenario.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-46 Figure 7.1 – Impacts of 2 Scheduled-based and 2 Headway-based Holding Strategies [153] A study by researchers at the University of California, Berkeley indicated that dynamic holding strategies based on headways alone are not sufficient to help buses adhere to a schedule. These researchers proposed a dynamic holding strategy using bus arrival deviations from a virtual schedule. Using data on arrival times of the current and preceding bus, as well as a virtual schedule, the proposed "simple method" was used to remove about 40 percent of the slack in a conventional schedule-based system and produce a one-parameter indicator of schedule reliability that can be used for optimization purposes. The optimal control strategy proposed by the authors chooses control coefficients to minimize slack time required to avoid negative holding times while ensuring a maximum standard deviation from the schedule [161]. The partway deadheading strategy with real-time control proposed in a study by a Chinese research team was shown, through simulation, to improve bus service reliability for a lower operational cost than other deadheading strategies. This method is particularly effective for routes that experience heavy directional imbalances during peak periods [163]. A simulation-based control method consisting of holding buses, limiting boarding, and extending green lights was recently tested. The method was shown to reduce delays by 61 percent compared to no control, with slight increases in delays for car drivers [35]. Researchers considered control strategies and their impacts on bus bunching, passenger wait times, and wait time variability through simulations of a corridor with multiple bus services operating in it. The simulation results indicate that a central operator of public transport control systems in a particular corridor would lead to the greatest reduction in passenger wait times (55 percent compared to no control), as well as more balanced passenger loads, lower headway variability, and better reliability for all public transport users [68]. A graduate thesis by Maltzan supports the finding that real-time data can be used to inform operational improvements. Holding at terminals and strategies for reducing operator deviations from scheduled terminal departure times are shown to have a strong effect on operations. Holding

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-47 at midpoints and short-turning also seem to provide operational benefits, but more study is needed on the costs and benefits that these strategies offer to transit passengers [101]. Limited-Stop Service, Stop Skipping, and Stop Consolidation As mentioned previously, Diab and El-Geneidy found the actual stops a bus makes to be a significant factor in increasing running times. Further, the coefficient of variation of actual stops was estimated as significantly increasing the coefficient of variation of running times and the coefficient of variation of running time deviations. In their models, limited-stop service was shown to improve running times, but worsen variation [38]. All-Door Boarding Diab and El-Geneidy also found rear-door activity to be a significant factor in increasing running times. The coefficient of variation of rear-door activity was estimated as significantly increasing the coefficient of variation of running times and the coefficient of variation of running time deviations [38]. Bus Lanes, Shoulder Lanes, and Queue Jump Lanes Mixed results have been reported for the impacts of reserved bus lanes on bus service reliability, and one study sought to understand the impacts of bus lanes with intermittent priority (those that exclude general traffic only when buses are present or approaching) as compared to dedicated bus lanes (where general traffic is never allowed, expect to turn right in some cases). Chen, Yu, Zhang, and Guo found that bus services using exclusive bus lanes showed improved reliability compared to services without this amenity [30]. Diab and El-Geneidy found that reserved bus lanes decreased running time by 35.26 seconds on average for a frequent bus route in the STM system, though this benefit could have been larger in magnitude if cars were not allowed to use the bus lanes to turn right (especially given that vehicles are not allowed to turn right on red) [37]. In a similar study, these researchers found reserved bus lanes to significantly decrease running time deviation in their models, but significantly increase the coefficient of variation of running time. Overall they found that reserved bus lanes may result in decreased running time, but increased variation, especially when general traffic is allowed to use bus lanes as right turn lanes but not allowed to turn right on red [38]. According to the TRCP 118, Bus Rapid Transit Practitioner’s Guide, bus queue jumps can reduce travel time for buses at intersections by 5 percent to 15 percent [33]. Bus lanes with intermittent priority (BLIPs) were found to not reduce street capacity nearly as much as dedicated bus lanes, but they can reduce the impact of traffic congestion on bus operations. The amount of traffic disruption of a BLIP depends on the transit signal priority strategies used, among other factors. BLIPs can also help to reduce bus travel times, but this too is dependent on traffic saturation level, bus frequency, the improvement in bus travel time achieved by the special lane, and the ratio of bus and car occupant flows [49]. Level Boarding, Low-Floor Buses, and Articulated Buses Diab and El-Geneidy have been the primary researchers concerned with the impacts of level boarding, low-floor buses, and articulated buses on service reliability. In a study of the STM in Montreal, they found that the introduction of articulated buses increased running times by 26.8 seconds [37]. They also reported that the presence of articulated buses in a corridor corresponded to a significant increase in running time deviation, but significant decreases in the coefficient of variation of running time and the coefficient of variation of running time deviation in their models. They stated that, while articulated buses may increase running times, they can help to reduce

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-48 variation [38]. The differences noted between regular and articulated buses appear to be mostly due to the physical and performance characteristics of these vehicles, including speed of boarding and alighting (related to the number and width of doors, as well as floor height) and rates of acceleration and deceleration. Transit Signal Priority and Intelligent Transportation Systems Researchers have demonstrated the potential of transit signal priority (TSP) to improve transit running times and reliability in a few cases, but the estimated improvements are generally small, if not insignificant. Furth and Muller developed a model based on Eindhoven in the Netherlands, with which they were able to demonstrate significant improvement in schedule adherence after the implementation of conditional signal priority for buses at signalized intersections. Traffic impacts were also studied for three scenarios: no priority, conditional priority, and absolute priority. Absolute priority (when priority is always given for transit vehicles) significantly increased general traffic delays compared to the no priority scenario, while conditional priority had very little impact on traffic conditions overall [61]. Unfortunately the data used in this study was collected over only a few days, so the level of confidence in these findings and their relevance in the larger context is not as high as it could have been. The results of this study are shown in Figure 7.2. Figure 7.2 – Delay Time by Time of Day with No Priority, Absolute Priority, and Conditional Priority [61] Skabardonis from the University of California at Berkeley used simulations of an arterial corridor to test operational control strategies. Optimal timing plans favoring buses along the corridor were estimated to reduce delay by 14 percent, reduce stops by 1 percent, and improve average bus speeds by about 4 percent, resulting in a delay savings of about 2 seconds per bus per intersection and minimal increases in general traffic delays. These results, based on the baseline data, were found to be insensitive to bus volumes up to 30 buses per hour. Bus preemption at signals with offline (not connected to a traffic management system) fixed-time timing plans demonstrated a potential savings of 6 seconds per bus per intersection, or a 2-minute travel time savings, with more benefits for higher bus volumes. Even greater benefits were expected to result from active priority (signal reacts to bus transmitter) for buses experiencing delays, but the tests showed major negative impacts to general traffic. As such, this approach was deemed as unlikely to be implementable in any real-life system. System-wide transit priority with online (connected to a traffic management system) signal control and automatic transit vehicle location and monitoring showed more promise, with moderate improvements over preemption with fixed-time plans.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-49 Overall it was found that passive priority strategies (which improve signal timing for all corridor users) may be useful, easy, and affordable to implement for simple networks, high-frequency systems, and routes with predictable dwell times [129]. Transit signal priority was shown to decrease the standard deviation of arrival time deviation from schedule by 3.2 percent on average across 30 simulation cases in Arlington, Virginia [29]. Another simulation from Arlington showed that buses typically benefit from various types of transit signal priority, including fixed-time control, adaptive splits, and adaptive splits with offsets, while general traffic often experiences negative impacts. Adaptive signal control can help to improve transit operations, while minimizing the disbenefits to general traffic. Through a simulation on the Columbia Pike, adaptive transit priority was shown to have potential for successful deployment in arterial corridors with coordinated signals, without undue negative consequences for general traffic, under different types of signal control. Transit operations benefitted most when priority was given only to express buses. Adaptive signal control did not produce substantially better results for transit than fixed-time control, but it did mitigate many of the negative impacts on general traffic. Transit signal priority benefits depend greatly on traffic flow characteristics and the control strategies used in the corridor [41]. Bus services in Beijing using transit signal priority showed improved reliability compared to services without this amenity [30]. Diab and El-Geneidy have quantified the impacts of TSP on running times, using a model of a high-frequency bus route in Montreal. They estimated that implementing TSP for articulated vehicles only decreased running times for all vehicles by 4.76 seconds (0.3 percent) on average. For the vehicles equipped with TSP, running times decreased by 18.32 seconds (1.2 percent) [37]. The implementation of transit signal priority was also found to significantly increase the coefficient of variation of running time and the coefficient of variation of running time deviation in their models. They reported that transit signal priority appears to improve running times with no significant impact on variation [38]. Innovations in Fare Payment The Transit Capacity and Quality of Service Manual 3rd Edition offered some insights into estimating the differences between various forms of fare payment. As shown in Figure 7.3, visual fare inspection was estimated to be the fastest form of fare payment, followed by smart cards, then a single ticket or token, then exact change, and, lastly, magnetic stripe cards. So if the goal is to reduce dwell times to improve reliability, then flash passes or smart cards would be the preferred methods of payment [82].

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-50 Figure 7.3 – Comparison of Bus Facility Capacity Based on Fare Collection Method [82] Smartcard use has been tested against flash passes for fare payment, but little research could be found that compared smartcard use directly to cash or magnetic strip ticket payments. The introduction of smartcard fare payment increased running times on an STM bus route by 5.83 seconds total at the beginning of the implementation period and by 52.61 seconds by the end of the implementation period [37]. Smartcards were also shown to significantly increase running time deviation, coefficient of variation of running time, and coefficient of variation of running time deviation in their regression models. Smartcard fare payment was found to be less effective than flash passes with regard to running time and reliability [38]. It should be noted that a clear advantage of smartcards over flash passes is the ability to automatically collect origin and destination data, though this technology could potentially be applied to flash pass and mobile ticketing systems through the use of GPS technology. Bicycling to and from Transit While bicycling as an access and egress mode was not a commonly-cited improvement strategy, one study from the Netherlands found this to be among the most important of possible improvement strategies. Through a simulation of the effects of a variety of strategies to address unreliability in public transport chains in the Netherlands, the most promising approaches were identified as the use of a bicycle as the access and/or egress mode for those using public transportation because of their ability to facilitate travel for broader, more reliable access to the transit network [119]. Bicycles extend the typical range that people are willing to walk to transit, which can help transit users to avoid transfers and reduce wait times in many cases. While the interface of transit and cycling in the Netherlands may be a special case, this improvement strategy is certainly worth further investigation within the context of various cities around the world.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-51 7.2 Empirical Evidence and Case Studies At least one study was found to demonstrate the impact of each major bus service reliability improvement strategy identified in Section 6.0, except for fleet maintenance, which was found to be lacking in research on impacts. However, in many cases the evidence of specific impacts is based on case studies, anecdotes, or small, sometimes insignificant, data sets. Route and Network Adjustments A comparison of passenger productiveness revealed that long trunk segments of the Translink system in Brisbane, Australia are most productive, including a long high-speed trunk busway segment approaching the central business district, the bus-on-expressway river crossing, and two long trunk rail segments approaching the CBD. Segments found to be at highest risk for reliability incident events were long, highly productive bus segments, though busway corridors showed clear benefits over general traffic expressway and arterial road segments [22]. Another study of Translink in Brisbane showed several route-related factors as being significant in models of average travel time, buffer time, and the coefficient of variation in travel time. The length between two stops had a significantly positive coefficient for the travel time and buffer time models, but a negative coefficient for the coefficient of variation of travel time model. The number of signals had a positive coefficient for all three models, while the travel through the central business district had a negative coefficient for all three models [100]. Schedule and Headway Optimization and Adjustments Beduhn’s thesis presented a framework and model for network and route scheduling and planning, which takes travel time and transfer reliability into account. Through a case study demonstration in Austin, Texas, the Reliable Shortest Path (RSP) method was shown to improve transit service reliability for the transit network by reducing missed transfers by 18 percent when compared to the deterministic, schedule-based shortest path (DSP). While trips using the RSP method were expected to take about 11 percent longer than scheduled, trips using the DSP or FAST-TrIPs paths were estimated to take about 17 percent longer than scheduled [11]. This research also illustrated the tradeoff between scheduled transfer offset time and transfer reliability, as shown in Figure 7.4.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-52 Figure 7.4 – Scheduled Transfer Offset Time vs. Transfer Reliability [11] Driver Training, Support, and Incentives Researchers in China and the United States reported that bus drivers' positive responses to real- time schedule information were shown to help improve bus service reliability [75]. Holding, Dispatching, and Control Headway-based control improved headway regularity and reduced running time in a study of the Massachusetts Bay Transportation Authority bus service [1]. A decade later, researchers conducted an experiment to investigate the impacts of a combination of vehicle holds, swaps, and short turns on service reliability. The results indicated that, compared to the baseline measurements, headway ratio variances dropped 3.8 percent overall and 15.8 percent at the control point. Headway regularity was best at the location of the control action and in the early stages of trips. Although there were some mixed and insignificant results, the improvement strategies were found to produce a net benefit overall. Further automation and extension of vehicle location and monitoring technology into the field was recommended for potentially greater benefits to headway reliability [133]. Limited-Stop Service, Stop skipping, and Stop Consolidation Bus stop consolidation for TriMet in Portland, Oregon was shown to significantly reduce bus running times, while having no significant impact on passenger demand. However, bus stop consolidation was not shown to have a significant impact on running time variation or headway variation in this case [52]. In his graduate thesis, Boyle used several case studies as examples of how stop consolidation can improve transit speeds and reliability, but may be faced with community resistance [18]. The results of a study of STM in Montreal indicated that implementation of limited-stop service along an already heavily used bus route can improve running time for the existing bus service as well as the limited-stop service [141]. Another study of the same system revealed that limited-

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-53 stop service decreased the odds of being late by 66 percent [139]. Furthermore, the number of actual stops that buses make was estimated as significantly increasing average travel time, while decreasing buffer time and the coefficient of variation in travel time. Meanwhile, the standard deviation of actual stops seemed to have a positive influence on buffer time and coefficient of variation of travel time [100]. All-Door Boarding Mixed results have been shown related to all-door and rear-door boarding. While Surprenant- Legault and El-Geneidy found that each rear boarding passenger increases the odds of being late by 6 percent in their model for the STM in Montreal [139], the New York City MTA reported success with all-door boarding in combination with several other improvement strategies [18]. Bus Lanes, Shoulder Lanes, and Queue Jump Lanes Exclusive right-of-way for buses has generally proven to be beneficial for travel time and reliability, although the influence of using a busway versus a highway or arterial road is likely to be heavily context-dependent. Bus on shoulder operations also seems to offer service reliability benefits. In a study of the STM in Montreal, a reserved bus lane was found to decrease the odds of being late for a particular route by 65 percent [139]. Similarly, the New York City MTA reported success with bus-only lanes, in combination with several other reliability improvement strategies [18]. Busway corridors showed clear benefits over general traffic on expressways and arterial road segments in a study of the Translink system in Brisbane, Australia [22]. In a more recent study of the same system, the use of busways was found to decrease average travel time, buffer time, and coefficient of variation of travel time compared to local, district, and suburban roads, but not as much as arterial roads or motorways [100]. Martin recommended the use of highway shoulders to bypass congestion, as a cost-effective method for improving bus running times and reliability. This practice is said to be especially popular with bus passengers [103]. Furthermore, he found that customer perception of schedule adherence and trip reliability is higher when buses make use of freeway shoulder lanes during congested periods. Many of the case studies included in his guide for the Federal Transit Administration indicated that reliability is a primary reason behind implementing bus on shoulder (BOS) operations, as well as one of the key benefits of this practice. For example, Miami-Dade Transit reported that its three bus routes using bus on shoulder facilities improved their on-time performance by up to 19 percent. An inside-shoulder BOS system demonstrated in Chicago has also shown notable reliability benefits. Overall, BOS operations are recommended as a low investment means of improving transit service reliability [104]. Far-Side Stop Placement Far-side bus stops are recommended in Boyle’s graduate student thesis, especially as they facilitate transit signal priority [18]. Level Boarding, Low-Floor Buses, and Articulated Buses In a study of TriMet’s bus system, low-floor buses correlated with a 0.11 second reduction in dwell time per dwell, which resulted in an average savings of 3.96 seconds in total running time per trip. Low-floor buses also help to reduce dwell time for lift operations by 4.74 seconds, or 5.8 percent [42].

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-54 Transit Signal Priority and Intelligent Transportation Systems Transit signal priority (TSP) has had at least two cases of demonstrated success in the field, both in the U.S. and abroad. Lehtonen and Kulmala performed a case study in Helsinki, Finland, where TSP and real-time passenger information were implemented. Their results indicate a more than 40 percent reduction in delays at intersections, with noticeable improvements to service regularity and punctuality. Passenger volumes increased by 10-12 percent after implementation of the transit signal priority and real-time information improvements for the bus system [84]. In their guide on Transit Signal Priority, Smith, Hemily, and Ivanovic repeatedly stated that reliability improvements are a key benefit of TSP. Several examples of successful transit signal priority implementation were shared in this guide. For example, in Tacoma, Washington, TSP and signal optimization combined to reduce transit signal delay by 40 percent in two corridors. Similarly, TriMet was said to have demonstrated a 10 percent reduction in travel times and about 19 percent less travel time variability after implementing TSP, allowing them to reduce scheduled recovery time and avoid adding another bus to the route [130]. Impacts of transit signal priority were found to be mixed across routes and time periods in a study of TriMet in Portland, Oregon. The benefits of transit signal priority are said to be most fully realized when ongoing monitoring and adjustment programs are implemented. Over all analysis segments, on-time performance was found to decline after transit signal priority was implemented, mostly due to more early trips and roughly the same rate of late trips. The authors recommended adjusting bus schedules when implementing transit signal priority strategies. Mean and variance of headways was also found to increase after implementation of transit signal priority. The research team recommended using caution when considering whether or not to implement transit signal priority in a corridor. They proposed an action plan that includes selecting candidate bus routes based on identification of operational problems, performing a baseline analysis, undertaking regular performance monitoring following implementation of transit signal priority to identify any challenges, and being willing to adjust schedules, emitter activation thresholds, signal control logic, and other factors to improve system performance [79]. The results of a survey conducted for TCRP Synthesis 83: Bus and Rail Transit Preferential Treatments in Mixed Traffic revealed that transit signal priority (TSP) is the most popular preferential treatment on urban streets, as well as the lack of standard warrants for when to apply certain treatments. Twelve traffic engineering organizations also provided insights, which indicated a preference for transit signal priority, queue jump lanes, exclusive transit lanes, and greater stop spacing over median transitways, special signal phasing, and curb extensions. All in all, the analysis revealed that the greatest benefit is typically realized from systematic application of one or more preferential treatments along a corridor, with median transitways, exclusive lanes, and transit signal priority estimated to provide the most significant positive results. Examples of bus lanes with TSP were shown to represent reliability improvements from arterial bus lanes, including a 12-27 percent improvement in coefficient of variation on Wilshire Boulevard in Los Angeles and a 57 percent improvement on Madison Avenue in New York City. TSP was also shown to provide reliability benefits in several cases, such as reductions in travel times ranging from 0-38 percent, signal delay reductions between 20-57 percent, and up to 35 percent reductions in travel time variability. Significant increases in delays to general traffic were noted in a few cases. Bus queue jump lanes have been demonstrated to result in 5-15 percent reductions in travel time for buses through intersections. Curb extensions and stop consolidation were also shown to help reduce bus running times [32]. Most recently, Ma, Yang, and Liu developed a model to generate the optimal combination of priority strategies for intersection groups, to keep real delay in line with permitted delays. The resulting coordinated and conditional bus priority approach was demonstrated in the field through

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-55 a case study in China, which led to significant reductions in bus delays (35 percent per bus compared to no priority) and headway deviations (62 percent reduction compared to no priority, 51 percent reduction compared to unconditional priority) with minimal impacts on general traffic. The proposed strategy was deemed to be useful for improving bus service reliability. While a 9 percent increase in motor vehicle delay was noted, this was small in comparison to the delay increases measured for the unconditional priority approach [97]. Automated Vehicle Location Systems and Real-Time Bus Arrival Information In a study published for the Transportation Research Board, Parker included analysis of transit data from before and after implementation of AVL systems, which revealed improvements in schedule adherence, transfer coordination, dispatcher control, schedule adherence monitoring, incident response, bus tracking, and driver performance monitoring after implementation. Collected data can also be used to inform schedule adjustments and service planning [112]. In a case study of Transport for London, Ehrlich found that the automated vehicle location system, iBus, had a small effect on improving reliability in one model, but was not statistically significant in a second model with many more variables. The author indicated that automated vehicle location systems must be used effectively by transit agencies and operators for benefits to be realized [48]. Regarding real-time information, in a before and after case study of the Chicago Transit Authority Bus Tracker service, modest increases in ridership were noted after implementation of a real-time bus arrival information system, which could have an impact on reliability if not properly managed [140]. Innovations in Fare Payment The introduction of smartcards for fare payment increased the odds of being late by 69 percent when compared to the original flash pass system [139]. On the other hand, the New York City MTA reported success with off-board fare collection in combination with other strategies [18]. Most recently the results of a study of Translink in Brisbane, Australia indicated that average boarding time per passenger was 3 seconds for smartcard users, 30 seconds for onboard top-up, and 15 seconds for paper ticket users. Eliminating onboard top-up for smartcards resulted in a 15-18 percent improvement in reliability for the bus route in question [118].

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-56 8.0 Conclusion A great deal of research has been compiled in this literature review, covering many topics related to bus service reliability. The overall findings from the review are summarized as follows:  Bus service reliability research dates back to at least the 1950’s, but data availability was a limiting factor for much of this early research. As such, most research from before the mid-1990’s has limited relevance to modern transit agencies, which typically operate in a much more data-rich environment.  Bus service reliability has been defined in many different ways, with little or no consensus around a single “best” definition. Over the last few decades, definitions have become more customer-focused, though many operator-oriented definitions are still used. Many researchers have opted for simple definitions of bus service reliability, while others have proposed breaking reliability down into several components.  Measurement of bus service reliability has advanced significantly in recent years, thanks in large part to the greater availability of data from automated data collection systems. More than 150 different metrics were found in the literature review process, representing reliability in many ways, from various perspectives, and with differing degrees of detail. Despite the wide variety of metric options available, high-level metrics such as schedule adherence (or on-time performance) and headway adherence are still commonly used.  An even greater variety of factors (nearly 200) were identified as potentially impacting bus service reliability. Categories of commonly-cited factors included traffic conditions, right-of-way, road conditions, network structure, incidents, vehicle and service characteristics, planning and operational outputs, drivers, passengers, fare payment, temporal factors, and weather. There is a substantial amount of evidence demonstrating the relationship between commonly-cited factors and bus service reliability. Many strategies for organizing relevant factors have also been proposed.  While perceived reliability tends to be very important for transit customers, it is difficult to quantify, especially because it is subjective. Despite the difficulty, several researchers have successfully demonstrated the impacts of service reliability improvements on perceived reliability and service quality among transit riders. Greater availability and accuracy of real-time arrival information in particular has been shown to positively impact not only the perception of bus service reliability, but service reliability itself, as customers are able to make more informed decisions regarding their arrival times at stops and minimize their waiting times.  More than 100 strategies for improving bus service reliability have been proposed, though not all have been tested through simulation or before and after studies of real-life implementation. Among the proposed strategies are operational improvement strategies (those relating to the routing, scheduling, control, drivers, stopping, and boarding), physical improvements (those that relate to infrastructure, right-of-way, stops, and bus characteristics), and technological improvements (those that leverage various technical innovations, such as intelligent transportation systems, AVL, and new methods of fare payment).  Many studies have aimed to measure the impacts of improvement strategies on bus service reliability. While much model- and simulation-based evidence has been gathered on the impacts of route adjustments, schedule optimization, control strategies, limited-stop

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-57 service, exclusive right-of-way, bus type, transit signal priority, and fare payment methods, as well as bicycling to and from transit, many of these studies provided mixed or inconclusive results. Further, at least one case study was found to provide empirical evidence of the impact of each major bus service reliability improvement strategy identified. However, the results of case studies and anecdotal evidence should be taken in context, and may not be generalizable. Tasks 3 and 4 of the TCRP A-42 research draw upon the findings from this literature review, as well as responses to the transit agency survey from Task 2, diving even more deeply into each aspect of bus service reliability.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-58 Appendix A-1 - Text Citations

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-59 [1] Abkowitz, M., & Lepofsky, M. (1990). Implementing Headway-Based Reliability Control on Transit Routes. Journal of Transportation Engineering, 116(1), 49-63. Retrieved from http://ascelibrary.org/doi/abs/10.1061/%28ASCE%290733-947X% 281990%29116:1%2849%2. [2] Abkowitz, M., Slavin, H., Waksman, R., Englisher, L., & Wilson, N. H. (1978). Transit Service Reliability, Transportation Systems Center Report No. UMTA-MA- 6-0049-78-1. Urban Mass Transportation Administration, U.S. Department of Transportation. [3] Abkowitz, M., Tozzi, J. (1987). Research Contributions to Managing Transit Service Reliability. Journal of Advanced Transportation, Vol. 21. [4] Anderson, P., & Geroliminis, N. (2014). Dynamic Bus Lanes with Restricted Car Usage for Congested Arterial Routes. Presented at 93rd Annual Meeting of the Transportation Research Board. Washington, D.C. [5] APTA BRT Operations Working Group. (2010). APTA Standards Development Program Recommended Practice: Bus Rapid Transit Service Design. Washington, D.C.: American Public Transportation Association. [6] APTA BRT Operations Working Group. (2010). Operating a Bus Rapid Transit System. American Public Transportation Association. Retrieved from http://www.apta.com/resources/standards/Documents/APTA-BTS-BRT-RP-007- 10.pdf [7] Arhin, S., & Noel, E. (2015). Development of Bus-Stop Time Models in Dense Urban Areas: A Case Study of Washington DC. Mineta National Transit Research Consortium [8] Arhin, S., & Noel, E. C. (2013). Evaluation of Bus Transit Reliability in the District of Columbia. Washington, D.C.: Mineta National Transit Research Consortium. Retrieved from http://transweb.sjsu.edu/PDFs/research/1139-DC-bus-transit- reliability.pdf [9] Barker, D. P. (2002). Communication, information and responsibility distribution strategies for effective real-time transit service management. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.ne [10] Bates, J., Polak, J., Jones, P., & Cook, A. (2001). The Valuation of Reliability for Personal Travel. Transportation Research Part E, 37, pp. 191-229. Retrieved from http://www.sciencedirect.com/science/article/pii/S1366554500000119 [11] Beduhn, T. J. (2014). Reliable Routing in Schedule-Based Transit Networks. University of Texas. Retrieved from https://repositories.lib.utexas.edu/handle/2152/28096

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-60 [12] Benn, H. (1995). TCRP Synthesis 10: Bus Route Evaluation Standards. Washington, D.C.: TRB, National Research Council. Retrieved from http://onlinepubs.trb.org/onlinepubs/tcrp/tsyn10.pdf [13] Berkow, M., El-Geneidy, A., Bertini, R., & Crout, D. (2009). Beyond Generating Transit Performance Measures: Visualizations and Statistical Analysis with Historical Data. Transportation Research Record: Journal of the Transportation Research Board, No. 2111, Washington, D.C.: Transportation Research Board of the National Academies, pp. 158-168. [14] Berkow, M., El-Geneidy, A., Bertini, R., & Crout, D. (2009). Beyond Generating Transit Performance Measures: Visualizations and Statistical Analysis with Historical Data. Transportation Research Record: Journal of the Transportation Research Board, No. 2111. Washington, D.C.: Transportation Research Board of the National Academies, pp. 158-168. [15] Bertini, R., & El-Geneidy, A. (2003). Generating Transit Performance Measures with Archived Data. Transportation Research Record: Journal of the Transportation Research Board, No. 1841. Washington, D.C.: Transportation Research Board of the National Academies, pp. 109-119. Retrieved from https://journals.sagepub.com/doi/10.3141/1841-12. [16] Bowman, L. A., & Turnquist, M. A. (1981). Service Frequency, Schedule Reliability, and Passenger Wait Times at Transit Stops. Transportation Research Part A: General, 15(6), pp. 465-471. Retrieved from http://www.sciencedirect.com/science/article/pii/ [17] Boyle, D. K., J. Pappas, P. Boyle, B. Nelson, D. Sharfarz, and H. Benn. (2009). TCRP Report 135: Controlling System Costs: Basic and Advanced Scheduling Manuals and Contemporary Issues in Transit Scheduling. Washington, D.C.: Transportation Research Board of the National Academies. [18] Boyle, D. K. (2013). TCRP Synthesis 110: Commonsense Approaches for Improving Transit Bus Speeds. Washington, D.C.: Transportation Research Board of the National Academies. [19] Brakewood, C., Macfarlane, G., & Watkins, K. (2015). The Impact of Real-Time Information on Bus Ridership in New York City. Transportation Research Part C: Emerging Technologies, 53, pp. 59-75. Retrieved from http://www.sciencedirect.com/science/artic [20] Brownstone, D., & Small, K. (2005). Valuing Time and Reliability: Assessing the Evidence from Road Pricing Demonstrations. Transportation Research Part A, 39, pp. 279-293. Retrieved from http://www.sciencedirect.com/science/article/pii/S09658564040010 [21] Bubna, P., Brunner, D., Gangloff, J. J., Advani, S. G., & Prasad, A. K. (2010). Analysis, Operation and Maintenance of a Fuel Cell / Battery Series-Hybrid Bus for

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-61 Urban Transport Applications. Journal of Power Sources, 195(12), 3939-3949. Retrieved fr [22] Bunker, J. M. (2013). Planning for Transit System Reliability Using Productive Performance and Risk Assessment. TRB 92nd Annual Meeting Compendium of Papers. Washington, D.C.: Transportation Research Board. [23] Cambridge Systematics, Dowling Associates, System Metrics Group, and Texas Transportation Institute. (2008). NCHRP Report 618: Cost-Effective Performance Measures for Travel Time Delay, Variation, and Reliability. Washington, D.C.: Transportation Research Board of the National Academies. [24] Camus, R., Longo, G., & Macorini, C. (2005). Estimation of Transit Reliability Level- of-Service Based on Automatic Vehicle Location Data. Transportation Research Record: Journal of the Transportation Research Board, No. 1927. Washington, D.C.: Transportation Research Board of the National Academies, pp. 277-286. [25] Carrasco, N. (2012). Quantifying Reliability of Transit Service in Zurich, Switzerland: Case Study of Bus Line 31. Transportation Research Record: Journal of the Transportation Research Board, No. 2274. Washington, D.C.: Transportation Research Board of the National Academies, pp.114-125. [26] Carrel, A., Mishalani, R. G., Wilson, N. H., Attanucci, J., & Rahbee, A. B. (2010). Decision Factors in Service Control on a High-Frequency Metro Line: Importance in Service Delivery. Transportation Research Record: Journal of the Transportation Research Board, No. 2146. Washington, D.C.: Transportation Research Board of the National Academies, pp. 52-59. [27] Chakrabarti, S., & Giuliano, G. (2015). Does Service Reliability Determine Transit Patronage? Insights from the Los Angeles Metro Bus System. Transport Policy, 42, pp. 12-20. Retrieved from http://www.sciencedirect.com/science/article/pii/S0967070X153 [28] Cham, L. C. (2006). Understanding Bus Service Reliability: A Practical Framework Using AVL / APC Data. Massachusetts Institute of Technology. Retrieved from http://dspace.mit.edu/handle/1721.1/34381 [29] Chang, J. (2002). Evaluation of Service Reliability Impacts of Traffic Signal Priority Strategies for Bus Transit. Virginia Polytechnic Institute and State University. Retrieved from http://scholar.lib.vt.edu/theses/available/etd-10252002-181921/ [30] Chen, X., Yu, L., Zhang, Y., & Guo, J. (2009). Analyzing Urban Bus Service Reliability at the Stop, Route, and Network Levels. Transportation Research Part A, pp. 722-734. Retrieved from http://www.sciencedirect.com/science/article/pii/S09658564090006 [31] Currie, G., Douglas, N., & Kearns, I. (2012). An Assessment of Alternative Bus Reliability Indicators. Shaping the Future: Linking Research, Policy and Outcomes. Australasian Transport. Retrieved from http://atrf.info/papers/2012/index.aspx

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-62 [32] Danaher, A. (2010). TCRP Synthesis 83: Bus and Rail Transit Preferential Treatments in Mixed Traffic. Washington, D.C.: Transportation Research Board of the National Academies. [33] Danaher, A. R., Blume, K., Levinson, H. S., & Zimmerman, S. (2007). TCRP Report 118: Bus Rapid Transit Practitioner’s Guide. Washington, D.C.: Transportation Research Board of the National Academies. [34] Daskalakis, N., & Stathopoulos, A. (2008). Users' Perspective Evaluation of Bus Arrival Time Deviations in Stochastic Networks. Journal of Public Transportation, 11(4), 25-38. Retrieved from http://www.nctr.usf.edu/jpt/pdf/JPT11-4Daskalakis.pdf [35] Delgado, F., Munoz, J. C., Giesen, R., & Wilson, N. H. (2015). Integrated Real-Time Transit Signal Priority Control for High-Frequency Transit Service. Transportation Research Record: Journal of the Transportation Research Board, No. 2533. Washington, D.C.: Transportation Research Board, pp. 28-38. [36] Dell'Olio, L., Ibeas, A., & Cecin, P. (2010). Modelling User Perception of Bus Transit Quality. Transport Policy, 17(6), 388-397. Retrieved from http://www.sciencedirect.com/science/article/pii/S0967070X10000557 [37] Diab, E. I., & El-Geneidy, A. M. (2012). Understanding the Impacts of a Combination of Service Improvement Strategies on Bus Running Time and Passenger's Perception. Transportation Research Part A: Policy and Practice, 46(3), pp. 614-625. Retrieved fr [38] Diab, E. I., & El-Geneidy, A. M. (2013). Variation in Bus Transit Service: Understanding the Impacts of Various Improvement Strategies on Transit Service Reliability. Public Transport, 4(3), pp. 209-231. Retrieved from http://link.springer.com/article [39] Diab, E. I., Badami, M. G., & El-Geneidy, A. M. (2015). Bus Transit Service Reliability and Improvement Strategies: Integrating the Perspectives of Passengers and Transit Agencies in North America. Transport Reviews, 35, pp. 292-328. Retrieved from ht [40] Diakaki, C., Dinopoulou, V., Papamichail, I., & Garyfalia, M. (2015). State-of-the-Art and -Practice Review of Public Transport Priority Strategies. IET Intelligent Transport Systems. Retrieved from http://www.crossref.org/iPage?doi=10.1049%2Fiet-its [41] Dion, F., Rakha, H., & Zhang, Y. (2005). Integration of Transit Signal Priority within Adaptive Traffic Signal Control Systems. Presented at 84th Annual Meeting of the Transportation Research Board, Washington, D.C. [42] Dueker, K. J., Kimpel, T. J., Strathman, J. G., & Callas, S. (2004). Determinants of Bus Dwell Time. Journal of Public Transportation, 7(1), 2. Retrieved from http://scholarcommons.usf.edu/jpt/vol7/iss1/2/

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-63 [43] K., & Kottenhoff, K. (2007). Dynamic At-Stop Real-Time Information Displays for Public Transport: Effects on Customers. Transportation Research Part A, 41(6), pp. 489-501. Retrieved from http://www.sciencedirect.com/science/article/pii/S09658 Dziekan [44] Dziekan, K., & Vermeulen, A. (2006). Psychological Effects of and Design Preferences for Real-Time Information Displays. Journal of Public Transportation, 9, 71-89. Retrieved from http://scholarcommons.usf.edu/cgi/viewcontent.cgi?article=1251&context= [45] Eberlein, X. J., Bernstein, D. H., & Wilson, N. H. (2001). The Holding Problem with Real-Time Information Available. Transportation Science, 35(1), 1-18 [46] Eboli, L., & Mazzulla, G. (2011). A Methodology for Evaluating Transit Service Quality Based on Subjective and Objective Measures from the Passenger's Point of View. Transport Policy, 18(1), 172-181. Retrieved from http://www.sciencedirect.com/science [47] Egge, M., Qian, Z. & Silbermann, A. (2017). The Effect of Fare Policies on Dwell Time: A Case Study for the Pittsburgh Region. Transportation Research Record: Journal of the Transportation Research Board, No. 2649. Washington, D.C.: Transportation Research Board, pp. 20-27. [48] Ehrlich, J. E. (2010). Applications of Automatic Vehicle Location systems towards improving service reliability and operations planning in London. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handl [49] Eichler, M., & Daganzo, C. F. (2006). Bus Lanes with Intermittent Priority: Strategy Formulae and an Evaluation. Transportation Research Part B: Methodological, 40(9), pp. 731-744. Retrieved from http://www.sciencedirect.com/science/article/pii/S01912 [50] El-Geneidy, A., Horning, J., & Krizek, K. (2007). Using Archived ITS Data to Improve Transit Performance and Management. St. Paul, MN: St. Paul Minnesota Department of Transportation. Retrieved from http://conservancy.umn.edu/bitstream/handle/11299/55 [51] El-Geneidy, A., Horning, J., & Krizek, K. (2011). Analyzing Transit Service Reliability Using Detailed Data from Automatic Vehicle Locator Systems. Journal of Advanced Transportation, 45. [52] El-Geneidy, A., Strathman, J., Kimpel, T., & Crout, D. (2006). Effects of Bus Stop Consolidation on Passenger Activity and Transit Operations. Transportation Research Record: Journal of the Transportation Research Board, No. 1971. Washington, D.C.: Transportation Research Board of the National Academies, pp. 32-41.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-64 [53] Evans IV, J. E. (2004). TCRP Report 95: Traveler Response to Transportation System Changes Chapter 9—Transit Scheduling and Frequency. Washington, D.C.: Transportation Research Board of the National Academies. [54] Fijalkowski, J. A. (2010). Making Data Matter: The Role of Information Design and Process in Applying Automated Data to Improve Transit Service. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle [55] Fletcher, G., & El-Geneidy, A. (2013). Effects of Fare Payment Types and Crowding on Dwell Time: Fine-Grained Analysis. Transit 2013, 2351, pp. 124-132. Retrieved from http://trrjournalonline.trb.org/doi/10.3141/2351-14 [56] Frumin, M., Zhao, J., Wilson, N. H., & Zhao, Z. (2013). Automatic Data for Applied Railway Management: A Case Study on the London Overground. Transportation Research Record: Journal of the Transportation Research Board, No. 2353. Washington, D.C.: Transportation Research Board of the National Academies, pp. 47-56 [57] Furth, P. G. (2000). TCRP Synthesis 34: Data Analysis for Bus Planning and Monitoring. Washington, D.C.: TRB, National Research Council. [58] Furth, P. G., & Muller, T. H. (2009). Optimality conditions for public transport schedules with timepoint holding. Public Transport, 1(2), 87 [59] Furth, P. G., & Wilson, N. H. (1981). Setting Frequencies on Bus Routes: Theory and Practice. Transportation Research Record 818. Washington, D.C.: TRB, National Research Council, pp. 1-7. [60] Furth, P., & Muller, T. (2007). Service Reliability and Optimal Running Time Schedules. Transportation Research Record: Journal of the Transportation Research Board, No. 2034. Washington, D.C.: Transportation Research Board of the National Academies, pp. 55-61. [61] Furth, P., & Muller, T. H. (2000). Conditional Bus Priority at Signalized Intersections: Better Service with Less Traffic Disruption. Transportation Research Record: Journal of the Transportation Research Board, No.1731. Washington, D.C.: TRB, National Research Council, pp. 23-30. [62] Furth, P., Hemily, B., Muller, T., & Strathman, J. (2006). TCRP Report 113: Using Archived AVL-APC Data to Improve Transit Performance and Management. Washington, D.C.: Transportation Research Board of the National Academies. [63] Gayah, V. V., Yu, Z., & Wood, J. S. (2016). Estimating Uncertainty of Bus Arrival Times and Passenger Occupancies. San Jose, CA: Mineta National Transit Research Consortium [64] Gittens, A., & Shalaby, A. (2015). Evaluation of Bus Reliability Measures and Development of a New Composite Indicator. Transportation Research Record:

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-65 Journal of the Transportation Research Board, No. 2533. Washington, D.C.: Transportation Research Board. [65] Guo, G., Luo, H., Lin, X., & Feng, C. (2011). Headway-based Evaluation of Bus Service Reliability. 14th International IEEE Conference on ITS (pp. 1864-1868). IEEE. Retrieved from http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6082998 [66] Hammerle, M., Haynes, M., & McNeil, S. (2005). Use of Automated Vehicle Location and Passenger Count Data to Evaluate Bus Operations: Experience of the Chicago Transit Authority, Illinois. Transportation Research Record: Journal of the Transportation Research Board, No. 1903. Washington, D.C.: Transportation Research Board of the National Academies, pp. 27-34. https://journals.sagepub.com/doi/abs/10.1177/0361198105190300104 [67] Harrington-Hughes & Associates, Inc., (2010) TCRP Research Results Digest 96: Managing Increasing Ridership Demand. Washington, D.C.: Transportation Research Board of the National Academies. [68] Hernandez, D., Munoz, J. C., Giesen, R., & Delgado, F. (2015). Analysis of Real- Time Control Strategies in a Corridor with Multiple Bus Services. Transportation Research Part B: Methodological, 78, pp. 83-105. Retrieved from http://www.sciencedirect.c [69] Hollander, Y. (2006). Direct versus indirect models for the effects of unreliability. Transportation Research Part A: Policy and Practice, 40(9), 699-711 [70] Hong, Y. (2002). The development of more effective operating plans for bus service. Civil Engineering.Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/84840 [71] Ian Wallis Associates Ltd in association with The TAS Partnership. (2013). Research Report 527 - Improving Bus Service Reliability. New Zealand Transport Agency. Retrieved from http://www.nzta.govt.nz/resources/research/reports/527/ [72] Jacques, K. R., & Levinson, H. S. (2000). TCRP Research Results Digest 38: Operational Analysis of Bus Lanes on Arterials: Application and Refinement. Washington, D.C.: TRB National Research Council. [73] St. Jacques, K. S., & Levinson, H. (1997). TCRP Report 26: Operational Analysis of Bus Lanes on Arterials. Washington, D.C.: TRB, National Research Council. [74] Ji, J., Guo, Y., & Yan, Y. (2013). Optimization Model of the Public Transit Network Based on Operational Reliability. American Society of Civil Engineers. Retrieved from http://ascelibrary.org/doi/abs/10.1061/9780784413159.121 [75] Ji, Y., He, L., & Zhang, H. (2014). Bus Drivers' Responses to Real-Time Schedule Adherence and the Effects on Transit Reliability. Transportation Research Record: Journal of the Transportation Research Board, No. 2417. Washington, D.C.: Transportation Research Board of the National Academies, pp. 1-9.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-66 [76] Jordan, W. C., & Turnquist, M. A. (1979). Zone Scheduling of Bus Routes to Improve Service Reliability. Transportation Science, 13(3), 242-268. Retrieved from http://pubsonline.informs.org/doi/abs/10.1287/trsc.13.3.242 [77] Kieu, L.-M., Bhaskar, A., & Chung, E. (2015). Public Transport Travel-Time Variability Definitions and Monitoring. Journal of Transportation Engineering, 141(1). Retrieved from http://ascelibrary.org/doi/10.1061/%28ASCE%29TE.1943- 5436.0000724 [78] Kimpel, T. J., Strathman, J. G., & Callas, S. (2008). Improving Scheduling through Performance Monitoring. Computer-Aided Systems in Public Transport, 253-280. Retrieved from http://link.springer.com/chapter/10.1007/978-3-540-73312-6_13 [79] Kimpel, T., Strathman, J., Bertini, R., & Callas, S. (2005). Analysis of Transit Signal Priority Using Archived TriMet Bus Dispatch System Data. Transportation Research Record: Journal of the Transportation Research Board, No. 1925. Washington, D.C,: Transportation Research Board of the National Academies, pp. 156-166. [80] King, R. D. (2003). TCRP Synthesis 49: Yield to Bus—State of the Practice. Washington, D.C.: Transportation Research Board of the National Academies. [81] Kittelson & Associates, Inc. (2014). Gap-Filling Project 5: Guidebook: Placing a Value on Travel-Time Reliability. SHRP 2 Project L17. Washington, D.C.: Transportation Research Board. [82] Kittelson & Associates Inc., Parsons Brinckerhoff, Inc., KFH Group Inc., Texas A&M Transportation Institute, Houston, and Arup. (2013). TCRP Report 165: Transit Capacity and Quality of Service Manual, Third Edition. Washington, D.C.: Transportation Research Board of the National Academies. [83] Lee, D.-H., Sun, L., & Erath, A. (2012). Study of Bus Service Reliability in Singapore using Fare Card Data. 12th Asia-Pacific Intelligent Transportation Forum. [84] Lehtonen, M., & Kulmala, R. (2002). Benefits of Pilot Implementation of Public Transport Signal Priorities and Real-Time Passenger Information. Transportation Research Record: Journal of the Transportation Research Board, No. 1799. Washington, D.C.: Transportation Research Board of the National Academies, pp.18-25. [85] Levinson, H. S. (1991). Supervision strategies for improved reliability of bus routes (No. 15) [86] Levinson, H. S., Zimmerman, S., Clinger, J., & Rutherford, G. S. (2002). Bus Rapid Transit: An Overview. Journal of Public Transportation, 5(2). Retrieved from http://scholarcommons.usf.edu/jpt/vol5/iss2/1/ [87] Levinson, H., Zimmerman, S., Clinger, J., & Gast, J. (2003). Bus Rapid Transit. TRB 2003 Annual Meeting CD-ROM. Washington, D.C.: Transportation Research Board.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-67 Retrieved from http://nacto.org/docs/usdg/brt_synthesis_of_case_studies_levinson.pdf [88] Li, Y., & Lu, D. (2015). Bus Scheduling Model Based on Peak Hour Volume Clustering. American Society of Civil Engineers. Retrieved from http://ascelibrary.org/doi/10.1061/9780784479292.100 [89] Li, Y.-w. (2003). Evaluating the Urban Commute Experience: A Time Perception Approach. Journal of Public Transportation, 6(4), 3. Retrieved from http://scholarcommons.usf.edu/jpt/vol6/iss4/3/ [90] Lin, G.-S., Liang, P., Schonfeld, P., & Larson, R. (1995). Adaptive Control of Transit Operations. Federal Transit Administration. Retrieved from http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.41.537 [91] Lin, J., & Ruan, M. (2009). Probability-Based Bus Headway Regularity Measure. IET Intelligent Transport Systems, 3(4), 400-408. Retrieved from http://digital- library.theiet.org/content/journals/10.1049/iet-its.2008.0088 [92] Lin, J., Wang, P., & Barnum, D. T. (2008). A Quality Control Framework for Bus Schedule Reliability. Transportation Research Part E: Logistics and Transportation Review, 44(6), pp. 1086-1098. Retrieved from http://www.sciencedirect.com/science/article [93] Liu, R., & Sinha, S. (2007). Modelling Urban Bus Service and Passenger Reliability. Retrieved from http://eprints.whiterose.ac.uk/3686/ [94] Lomax, T., Schrank, D., Turner, S., & Margiotta, R. (2003). Selecting Travel Reliability Measures. Washington, D.C.: Federal Highway Administration. Retrieved from http://d2dtl5nnlpfr0r.cloudfront.net/tti.tamu.edu/documents/TTI-2003-3.pdf [95] Lyman, K., & Bertini, R. (2008). Using Travel Time Reliability Measures to Improve Regional Transportation Planning and Operations.Transportation Research Record: Journal of the Transportation Research Board, No. 2046. Washington, D.C.: Transportation Research Board of the National Academies, pp.1-10. [96] Lyons, G., & Urry, J. (2005). Travel Time Use in the Information Age. Transportation Research Part A, 39, pp. 257-276. Retrieved from http://www.sciencedirect.com/science/article/pii/S0965856404000977 [97] Ma, W., Yang, X., & Liu, Y. (2010). Development and Evaluation of a Coordinated and Conditional Bus Priority Approach. Transportation Research Record: Journal of the Transportation Research Board, No. 2145. Washington, D.C.: Transportation Research Board of the National Academies, pp. 49-58. [98] Ma, X., & Wang, Y. (2014). Development of a Data-Driven Platform for Transit Performance Measures Using Smart Card and GPS Data. Journal of Transportation Engineering, 140(12). Retrieved from http://ascelibrary.org/doi/10.1061/%28ASCE%29TE.1943-5436.0

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-68 [99] Ma, Z.-L., Ferreira, L., & Mesbah, M. (2013). A Framework for the Development of Bus Service Reliability Measures. Transport and the New World City: 36th Australasian Transport Research Forum. Retrieved from http://atrf.info/papers/2013/index.aspx [100] Ma, Z.-L., Ferreira, L., Mesbah, M., & Hojati, A. T. (2015). Modelling Bus Travel Time Reliability Using Supply and Demand Data from Automatic Vehicle Location and Smart Card Systems. TRB 94th Annual Meeting Compendium of Papers. Washington, D.C. [101] Maltzan, D. W. (2015). Using real-time data to improve reliability on high-frequency transit services. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/99541 [102] Mandelzys, M., & Hellinga, B. (2010). Identifying Causes of Performance Issues in Bus Schedule Adherence with Automatic Vehicle Location and Passenger Count Data. Transportation Research Record: Journal of the Transportation Research Board, No. 2143. Washington, D.C,: Transportation Research Board of the National Academies, pp. 9-15. [103] Martin, P. C. (2006). TCRP Synthesis 64: Bus Use of Shoulders. Washington, D.C.: Transportation Research Board of the National Academies. [104] Martin, P. C., Levinson, H. S., & Texas Transportation Institute. (2012). TCRP Report 151: A Guide for Implementing Bus on Shoulder (BOS) Systems. Washington, D.C.: Transportation Research Board of the National Academies. [105] Mazloumi, E., Currie, G., & Rose, G. (2010). Using GPS Data to Gain Insight into Public Transport Travel Time Variability. Journal of Transportation Engineering, 136, 623-631 [106] Mishalani, R., McCord, M. M., & Wirtz, J. (2006). Passenger Wait Time Perceptions at Bus Stops: Empirical Results and Impact on Evaluating Real-Time Bus Arrival Information. Journal of Public Transportation, 9(2), 89-106. [107] Moses, I. E. (2005). A transit route simulator for the evaluation of control strategies using automatically collected data. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/32413 [108] Muller, T. H., & Furth, P. G. (2000). Integrating Bus Service Planning with Analysis, Operational Control, and Performance Monitoring. Proceedings of the Intelligent Transportation Society of America Annual Meeting. Retrieved from http://www1.coe.neu [109] Multisystems, Inc., Mundle & Associates, Inc., & Simon & Simon Research and Associates, Inc., (2003). TCRP Report 94: Fare Policies, Structures, and Technologies Update. Washington, D.C.: Transportation Research Board of the National Academies.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-69 [110] Okunieff, P. E. (1997). TCRP Synthesis 24: AVL Systems for Bus Transit. Washington, D.C.: TRB, National Research Council. [111] Pangilinan, C. A. (2006). Bus supervision deployment strategies for improved bus service reliability. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/38238 [112] Parker, D. J. (2008). TCRP Synthesis 73: AVL Systems for Bus Transit: Update. Washington, D.C.: Transportation Research Board of the National Academies. [113] Peng, Z. R., Lynde, E. D., & Chen, W. Y. (2008). Improving service restoration using automatic vehicle location (No. MRUTC 08-09). Midwest Regional University Transportation Center, College of Engineering, Department of Civil and Environmental Engineering [114] Perk, V. & Foreman, C. (2001). Evaluation of First-Year Florida MPO Transit Capacity and Quality of Service Reports. Tampa: National Center for Transit Research, University of South Florida. Retrieved from http://intrans.iastate.edu/publications/_doc [115] Perk, V., Flynn, J., & Volinski, J. (2008). Transit Ridership, Reliability and Retention. Tampa: National Center for Transit Research, University of South Florida. Retrieved from http://www.nctr.usf.edu/pdf/77607.pdf [116] Polus, A. (1978). Modeling and Measurements of Bus Service Reliability. Transportation Research, 12(4), 253-256. Retrieved from http://www.sciencedirect.com/science/article/pii/0041164778900679 [117] pteg. (2014). Bus Punctuality: Towards a Structure that can Deliver. Leeds, United Kingdom. Retrieved from http://www.urbantransportgroup.org/system/files/general- docs/pteg%20bus%20punctuality%20web%20report%20June%202014_FINAL2.pdf [118] Qu, X., Oh, E., Weng, J., & Jin, S. (2014). Bus Travel Time Reliability Analysis: A Case Study. Proceedings of the Institution of Civil Engineers - Transport, 167(3), pp. 178-184. Retrieved from http://www.icevirtuallibrary.com/content/article/10.168 [119] Rietveld, P., Bruinsma, D., & vanVuuren, D. (2001). Coping with Unreliability in Public Transport Chains: A Case Study for Netherlands. Transportation Research Part A, 35(6), pp. 539-559. Retrieved from http://www.sciencedirect.com/science/article/pi [120] Saberi, M., Zockaie, A. K., Feng, W., & El-Geneidy, A. (2013). Definition and Properties of Alternative Bus Service Reliability Measures at the Stop Level. Journal of Public Transportation, 16. Retrieved from http://www.nctr.usf.edu/wp- content/upload [121] Sanchez-Martinez, G. E. (2013). Running Time Variability and Resource Allocation: A Data-Driven Analysis of High-Frequency Bus Operations. Civil Engineering.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-70 Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/172 [122] Sanchez-Martinez, G. E. (2015). Real-Time Operations Planning and Control of High-Frequency Transit. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/99550 [123] Sarna, P. and S.D. Lorgion. (2011) Route 51 Service and Reliability Report. AC Transit, Oakland, Calif [124] Schaller Consulting, "Unsustainable? The Growth of App Based Ride Services and Traffic, Travel, and the Future of New York City," 2017 [125] Schil, M. (2012). Measuring Journey Time Reliability in London Using Automated Data Collection Systems. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/74273 [126] List, G.F, Williams, B., & Rouphail, N. (2014). SHRP 2 Report S2-L02-RR-1: Establishing Monitoring Programs for Travel Time Reliability. Washington, D.C.: Transportation Research Board of the National Academies. [127] Schweiger, C. L. (2003). TCRP Synthesis 48: Real-Time Bus Arrival Information Systems. Washington, D.C.: Transportation Research Board of the National Academies. [128] Shiftan, Y. W. (1994). Absence, Overtime and Reliability Relations in Transit Workforce Planning. Transportation Research, Part A, 28(3), 245-258. [129] Skabardonis, A. (2000). Control Strategies for Transit Priority. Transportation Research Record: Journal of the Transportation Research Board, No. 1727. Washington, D.C.: TRB, National Research Council, pp. 20-26. [130] Smith, H. R., Hemily, B., & Ivanovic, M. (2005). Transit Signal Priority (TSP): A Planning and Implementation Handbook. Washington, D.C.: ITS. Retrieved from http://trid.trb.org/view.aspx?id=772546 [131] Sterman, B. P., & Schofer, J. L. (1976). Factors affecting reliability of urban bus services. Journal of Transportation Engineering, 102(ASCE# 11930) [132] Stewart, R. & Wong, R. (2013). Guidelines for Planning and Implementation of Transit Priority Measures in Urban Areas. 2013 Conference and Exhibition of the Transportation Association of Canada. Transportation Association of Canada. [133] Strathman, J. et al. (2000). Bus Transit Operations Control: Review and an Experiment Involving TriMet's Automated Bus Dispatching System. Transportation Northwest, Department of Civil Engineering, University of Washington.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-71 [134] Strathman, J. G., & Hopper, J. R. (1993). Empirical Analysis of Bus Transit On-Time Performance. Transportation Research Part A: Policy and Practice, 27(2), pp. 93- 100. Retrieved from http://www.sciencedirect.com/science/article/pii/096585649390065S [135] "Strathman, J. G., Kimpel, T. J., Dueker, K. J., Gerhart, R. L., & Callas, S. (2002). Evaluation of transit operations: Data applications of Trimet's automated bus dispatching system. Transportation, 29(3), 321-345 [136] Strathman, J., Dueker, K., Kimpel, T., Gerhart, R., Turner, K., Taylor, P., Callas, S,, Griffin, D., & Hopper, J. (1999). Automated Bus Dispatching, Operations Control, and Service Reliability: Baseline Analysis. Transportation Research Record: Journal of the Transportation Research Board, No. 1666. Washington, D.C.: TRB, National Research Council, pp. 28-36. [137] Strathman, J., Kimpel, T., & Callas, S. (2003). Headway Deviation Effects on Bus Passenger Loads: Analysis of TriMet's Archived AVL-APC Data. Portland, OR: Center for Urban Studies at Portland State University. [138] Strathman, J., Kimpel, T., Dueker, K., Gerhart, R., & Callas, S. (2002). Evaluation of Transit Operations: Data Applications of TriMet's Automated Bus Dispatching System. Transportation, 29(3), pp. 321-345. [139] Surprenant-Legault, J., & El-Geneidy, A. (2011). Introduction of a Reserved Bus Lane: Impact on Bus Running Time and On-Time Performance. Transportation Research Record: Journal of the Transportation Research Board, No. 2218. Washington, D.C.: Transportation Research Board of the National Academies, pp. 10-18. [140] Tang, L., & Thakuriah, P. (2012). Ridership Effects of Real-Time Bus Information System: A Case Study in the City of Chicago. Transportation Research Part C, 22, pp. 146-161. [141] Tetreault, P. R., & El-Geneidy, A. M. (2010). Estimating Bus Run Times for New Limited-Stop Service Using Archived AVL and APC Data. Transportation Research Part A: Policy and Practice, 44(6), pp. 390-402. [142] Tribone, D., Block-Schachter, D., Salvucci, F. P., Attanucci, J., & Wilson, N. H. (2014). An Automated Data Driven Performance Regime for Operations Management, Planning, and Control. Transportation Research Record: Journal of the Transportation Research Board, No. 2415. Washington, D.C.: Transportation Research Board of the National Academies, pp. 72-79. [143] Trompet, M., Liu, X., & Graham, D. (2011). Development of Key Performance Indicator to Compare Regularity of Service between Urban Bus Operators. Transportation Research Record: Journal of the Transportation Research Board, No. 2216. Washington, D.C.: Transportation Research Board of the National Academies, pp. 33-41.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-72 [144] Tse, J., Flin, R., & Mearns, K. (2006). Bus Driver Well-Being Review: 50 Years of Research. Transportation Research Part F: Traffic Psychology and Behavior, 9(2), pp. 89-114. Retrieved from http://www.sciencedirect.com/science/article/pii/S1369847805 [145] Turnquist, M. A., & Bowman, L. A. (1980). The Effects of Network Structure on Reliability of Transit Service. Transportation Research Part B: Methodological, 14(1), pp. 79-86. Retrieved from http://www.sciencedirect.com/science/article/pii/0191261580 [146] Tzeng, G.-H., Lin, C.-W., & Opricovic, S. (2005). Multi-Criteria Analysis of Alternative-Fuel Buses for Public Transportation. Energy Policy, 33(11), 1373-1383. Retrieved from http://www.sciencedirect.com/science/article/pii/S0301421503003811 [147] U.S. Department of Transportation, Federal Transit Administration. (2015, June 23). Bus Testing: Establishment of Performance Standards, a Bus Model Scoring System, a Pass/Fail Standard and Other Program Updates; Proposed Rule. [148] Uniman, D. L. (2009). Service reliability measurement framework using smart card data: application to the London Underground. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/52806 [149] Uniman, D. L., Attanucci, J., Mishalani, R. G., & Wilson, N. H. (2010). Service Reliability Measurement Using Automated Fare Card Data: Application to the London Underground. MIT [150] van Oort, N. (2011). Service Reliability and Urban Public Transport Design. Delft University of Technology, Netherlands. [151] van Oort, N., & van Nes, R. (2008). Improving Reliability in Urban Public Transport in Strategic and Tactical Design. Presented at 87th Annual Meeting of the Transportation Research Board, Washington, D.C. [152] van Oort, N., & van Nes, R. (2009). Line Length versus Operational Reliability: Network Design Dilemma in Urban Public Transportation. Transportation Research Record: Journal of the Transportation Research Board, No. 2112. Washington D.C.: Transportation Research Board of the National Academies, pp. 104-110. [153] van Oort, N., Wilson, N., & van Nes, R. (2010). Reliability Improvement in Short Headway Transit Services: Schedule- and Headway-based Holding Strategies. Transportation Research Record: Journal of the Transportation Research Board, No. 2143. Washington, D.C.: Transportation Research Board of the National Academies, pp. 67-76. [154] Vincent, M. P., & Hamilton, B. A. (2008). Measurement valuation of public transport reliability. Land Transport New Zealand.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-73 [155] Vuchic, V. R. (2002). Bus Semi rapid Transit Mode Development and Evaluation. Journal of Public Transportation, 5(2), 4. Retrieved from http://scholarcommons.usf.edu/jpt/vol5/iss2/4/ [156] Wallischeck, E. Y., Weisenberger, T., Berthaume, A., & Dinning, M. G. (2015). TCRP Report 177: Preliminary Strategic Analysis of Next Generation Fare Payment Systems for Public Transportation. Washington, D.C.: Transportation Research Board of the National Academies. https://www.nap.edu/catalog/22158/preliminary-strategic-analysis-of-next- generation-fare-payment-systems-for-public-transportation [157] Watkins, K., Ferris, B., Borning, A., Rutherford, S., & Layton, D. (2011). Where is My Bus? Impact of Mobile Real-Time Information on the Perceived and Actual Wait Time of Transit Riders. Transportation Research Part A, 45, pp. 839-848. [158] Wile, E. S. (2003). Use of automatically collected data to improve transit line performance. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/85755 [159] Wood, D. A. (2015). A Framework for Measuring Passenger-Experienced Transit Reliability Using Automated Data. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/99539 [160] WSP | Parsons Brinckerhoff. (2015). Bus Journey Time Variability in Urban Areas. Department for Transport. United Kingdom [161] Xuan, Y., Argote, J., & Daganzo, C. F. (2011). Dynamic Bus Holding Strategies for Schedule Reliability: Optimal Linear Control and Performance Analysis. Transportation Research Part B: Methodological, 45(10), pp. 1831-1845. [162] Yao, B., Hu, P., Lu, X., Gao, J., & Zhang, M. (2013). Transit Network Design Based on Travel Time Reliability. Transportation Research Part C: Emerging Technologies, 43, pp. 233-248. Retrieved from http://www.sciencedirect.com/science/article/pii/S09 [163] Yu, B., Yang, Z., & Li, S. (2012). Real-Time Partway Deadheading Strategy Based on Transit Service Reliability Assessment. Transportation Research Part A: Policy and Practice, 46(8), pp. 1265-1279. Retrieved from http://www.sciencedirect.com/science/ [164] Yu, S., Chen, J., Liu, Z., & Zhu, L. (2015). Study on a Model and Demonstration of Public Transit Priority Technology Strategy at an Urban Arterial Road. American Society of Civil Engineers. Retrieved from http://ascelibrary.org/doi/10.1061/978078447 [165] Zhang, C. & Teng, J. (2013). Bus Dwell Time Estimation and Prediction: A Study Case in Shanghai-China. Tongji University. Procedia - Social and Behavioral Sciences

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-74 [166] Zhang, F., Shen, Q., & Clifton, K. (2008). Examination of Traveler Responses to Real-Time Information about Bus Arrivals using Panel Data. Transportation Research Record: Journal of the Transportation Research Board, No. 2082. Washington, D.C.: Transportation Research Board of the National Academies, pp. 107-115. [167] Zhang, S.-y., Yang, X.-g., & Teng, J. (2011). Evaluation of Bus Service Reliability Based on AVL Information. 11th International Conference of Chinese Transportation Professionals. American Society of Civil Engineers.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-75 Appendix A-2 - Annotated Bibliography

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-76 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Abkowitz, Slavin, Waksman, Englisher, & Wilson, 1978 Abkowitz, M., Slavin, H., Waksman, R., Englisher, L., & Wilson, N. H. (1978). Transit Service Reliability, Transportation Systems Center Report No. UMTA- MA-06-0049-78-1. Urban Mass Transportation Administration, U.S. Department of Transportation. Transit service reliability U.S. DOT, Urban Mass Transportation Administration and Multisystems, Inc. in the United States The authors define service reliability as "the invariability of service attributes which influences the decisions of travelers and transportation providers". From the passenger's perspective, the main measures of reliability identified are distributions total travel time, wait time, in-vehicle time, and seat availability. The authors identify three statistics to evaluate reliability: mean time, coefficient of variation of time, and % of observations taking N minutes longer than the mean. From an operator's perspective, the authors consider reliability as the ability to fulfill scheduled operations on time. The measures of reliability from an operator's perspective are the average difference between mean and scheduled arrival time, standard deviation, and % of arrivals N minutes later than mean arrival time. The authors consider travel variability as the main cause of unreliability. They identify traffic congestion and signals as sources of unreliability. The report explains that unstable headway dynamics, where large headways tend to become larger, leading to bus bunching, is a major cause of unreliability. Since more passengers arrive during long headways and wait longer, unstable headways cause unreliability from passenger's perspective. The authors identify three categories of improvement strategies. The first is to prioritize transit vehicles through special treatment and dedicated right-of-way. In the second category, control methods make real-time adjustments to the original schedule. Lastly, operational measures adjust the schedule, route and resource allocation. This report presents an overview of service reliability both in research and in practice. The report establishes performance measures both from passenger and operator perspectives. The authors explore the main causes of unreliability and evaluate solutions. The reliability has a great impact in the quality of service as perceived by passengers and on the progress of operations as experienced by operators. Focusing on reliability, specifically when the system performs the worst, is the first step in securing fluid operations. Identifying and correcting unreliability can help increase the quality of service and reduce the operating cost of a transit system. Variability in travel time is often the main cause of unreliability. Operators can reduce this variability by applying treatments at every step on the planning and operations process. The analysis is based on much less data than contemporar y studies. Polus, 1978 Polus, A. (1978). Modeling and Measurements of Bus Service Reliability. Transportation Research, 12(4), 253-256. Retrieved from http://www.sciencedirect.co m/science/article/pii/004116 4778900679 Modeling and measuring bus service reliability Researcher based at Technion-Israel Institute of Technology in Haifa, Israel Bus service reliability is defined as the amount of consistency associated with an operational performance measure from day to day. A single measure of reliability is proposed as one divided by the standard deviation of travel time for a route over a period of time. Variability of travel time performance, Right-of-way exclusivity, Road and traffic conditions, Passenger loads, Weather, Physical route characteristics, Staff shortages, Bus control None This paper presents a strategy for predicting bus service reliability, using travel time as the key predictor. This research provides limited information and may be outdated. Jordan & Turnquist, 1979 Jordan, W. C., & Turnquist, M. A. (1979). Zone Scheduling of Bus Routes to Improve Service Reliability. Transportation Science, 13(3), 242-268. Retrieved from http://pubsonline.informs.org /doi/abs/10.1287/trsc.13.3.2 42 Zone scheduling of bus routes for reliability Researchers from Northwestern University in Evanston, Illinois, United States On-time performance, Variance of total trip time (wait time and travel time) for passengers between any pair of stops Route and bus characteristics (stop spacing, bus fleet size, bus capacity), Passenger activity, Peak period travel Zone scheduling A model was developed using reliability and average trip time as indicators of success. The impact of zone scheduling on reliability and trip time was estimated and the zone structures were optimized using a dynamic programming model. A case study was used to test the zone scheduling approach, and it was shown to improve reliability and reduce trip times on average, while lessening demands on bus fleet size when compared to all-local service. This study is from 1979, before many of the modern tracking technologies existed.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-77 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Turnquist & Bowman, 1980 Turnquist, M. A., & Bowman, L. A. (1980). The Effects of Network Structure on Reliability of Transit Service. Transportation Research Part B: Methodological, 14(1), pp. 79-86. Retrieved from http://www.sciencedirect.co m/science/article/pii/019126 158090034X The impacts of network structure on transit service reliability Researchers based at Northwestern University and Peat, Marwick, Mitchell and Co., Inc., Simulation using characteristics of services in Chicago, Illinois, United States Standard deviation of vehicle arrival times at stops, Coefficient of variation of arrival time, Coefficient of variation of transfer delays, Twice the deviation of arrival time plus the deviation in transfer time times the expected number of transfers Frequency of service, Coefficient of variation of link travel time, Demand to capacity ratio, Route density, Network orientation Vehicle passing policies (allowing "leap-frogging" when vehicles bunch), Pulse-scheduled operations (to enable immediate transfers), Improved signalization, Priority operation, Exclusive bus lanes, Active or real-time control over service operation Variability in time associated with making a transfer, which is directly influenced by frequency of service, appears to have a major impact on transit system reliability. Grid networks appear to be less disrupted by transfers than radial networks, but service reliability appears to be much more influenced by frequency of service than route density. Transit service reliability may be enhanced by preventing vehicle bunching and/or by breaking up vehicle platoons when they form. On-time arrival of vehicles at major transfer stations is important, especially for radial networks. In general operators should avoid providing excess slack time in their schedules, except where a large number of passenger transfers could benefit from having enough slack time to ensure successful connections. Results are based on simulations, rather than before and after studies. Bowman & Turnquist, 1981 Bowman, L. A., & Turnquist, M. A. (1981). Service Frequency, Schedule Reliability, and Passenger Wait Times at Transit Stops. Transportation Research Part A: General, 15(6), pp. 465-471. Retrieved from http://www.sciencedirect.co m/science/article/pii/019126 078190114X Vehicle frequency, schedule reliability, and passenger wait times at transit stops Researchers based at Cornell University and Peat, Marwick, Mitchell and Co. in the United States, Data from Chicago, Illinois, United States Reliability is defined as schedule adherence Frequency of service, Schedule reliability None The researchers' passenger-choice arrival model suggests greater sensitivity to schedule deviation and less sensitivity to service frequency than found under the assumption of random passenger arrivals. Focused on perception of wait time and schedule adherence, but offered no improvement strategies. Furth & Wilson, 1981 Furth, P. G., & Wilson, N. H. (1981). Setting Frequencies on Bus Routes: Theory and Practice. Transportation Research Record: Journal of the Transportation Research Board, No. 818, 1-7. Retrieved from http://www1.coe.neu.edu/~pf urth/Furth%20papers/1981% 20setting-frequencies.pdf This paper provides a method to optimize the frequency of bus routes with network- wide constrained resources. Researchers based at Massachusetts Institute of Technology, Focused on North America Weighted sum of total passenger waiting time Bus route frequencies Optimize frequencies on all bus routes with fixed subsidy, fleet size and maximum headway. Simple rules of thumb are generally not the best way to set frequencies. The method assumes fixed demand and bounded travel times. Abkowitz & Tozzi, 1987 Abkowitz, M., Tozzi, J. (1987). Research Contributions to Managing Transit Service Reliability. Journal of Advanced Transportation, Vol. 21. Managing transit service reliability Researchers based at Rensselaer Polytechnic Institute and R.W. Beck and Associates in the United States Vehicle run time, Run time variation (standard deviation), Headway variation, Passenger wait time On-street parking, Peak period travel, Direction of travel, Number of stops, Passenger activity, Number of signals Real-time reliability control The researchers affirm that a correlation exists between running time and signalized intersections, passenger activity, number of bus stops, direction of travel, and time of day. Headway variation was said to be highly correlated with traffic conditions and dwell time at stops. Holding strategies should be used selectively, when the benefits to downstream customers are estimated to outweigh the disbenefits to passengers waiting onboard. Review of other studies, with little or no new information. Abkowitz & Lepofsky, 1990 Abkowitz, M., & Lepofsky, M. (1990). Implementing Headway-Based Reliability Control on Transit Routes. Journal of Transportation Engineering, 116(1), 49-63. Retrieved from http://ascelibrary.org/doi/abs /10.1061/%28ASCE%29073 3- 947X%281990%29116:1%2 849%29 Implementi ng headway- based method on MBTA route Vanderbilt University and the MBTA in the United States Delay time at control point, at-stop waiting time, headway at control point, passenger load, and running time. Vehicle arrival time at control points Headway-based control: hold buses until their headway reaches a threshold. The method improved headway regularity and reduced running time. The authors could only record two days of point-check data and six days of headway data. As a result, the data has no statistical significance.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-78 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Dube, Schmitt, & Leclerc, 1991 Dube, L., Schmitt, B. H., & Leclerc, F. (1991). Consumers' Affective Response to Delays at Different Phases of a Service Delivery. Journal of Applied Social Psychology, Vol. 21, 10, 810-820. Consumer response to delays Researchers based at the Universite de Montreal in Canada and Columbia University and Massachusetts Institute of Technology in the United States Wait time None None Based on a field experiment in an educational setting, the researchers found that the perception of a wait over eight minutes differs during the wait than the perception of this wait either before or after it occurs. This study is not really focused on public transit service. Levinson, 1991 Levinson, H.S. NCTRP Synthesis of Transit Practice 15: Supervision Strategies for Improved Reliability of Bus Routes. Transportation Research Board, Washington, D.C., 1991. http://onlinepubs.trb.org/Onli nepubs/nctrp/syn15/syn15.p df Supervision strategies Transportation Research Board Supervision strategies None None Monitoring strategies were mainly based on having dispatchers on the ground Surveyed agencies did not have access to AVL systems Strathman & Hopper, 1993 Strathman, J. G., & Hopper, J. R. (1993). Empirical Analysis of Bus Transit On- Time Performance. Transportation Research Part A: Policy and Practice, 27(2), pp. 93-100. Retrieved from http://www.sciencedirect.co m/science/article/pii/096585 649390065S Analyzing bus service on-time performanc e Researchers based at Portland State University, Case study of TriMet in Portland, Oregon, United States On-time performance is defined as the probability of being between 1 minute early and 5 minutes late Driver experience and behavior, Sensitivity of the schedule to route conditions (headways, run times, layover times), Complexity of the route (length, stop spacing, passenger activity), Traffic congestion and incidents, Signal timing, Weather, Disruptions from on-street parking Holding, dispatching, and diverting strategies (waiting at time points, dispatching additional bus, turning before the end of a route), Passing policies, Driver behavior real-time monitoring and incentives, Schedule and route design modifications (longer running times, longer layovers, shorter routes, fewer stops) Passenger boardings were found to have a significant positive impact on the probability of arriving on time, versus early, but no effect on the on-time / late relative odds. More passenger alightings corresponded to increased likelihood of both early and late arrivals. The probability of later arrivals was found to increase as buses travel along their routes, which may be addressed by longer running times with many time points or shorter and less complex routes. Early arrivals are more likely during the first two weeks of a new sign-up (driver assignment). Longer headways were associated with more late arrivals, with the highest chance of late arrival for services at headways of one hour and ten minutes. Reliability appears to deteriorate with headways exceeding 15 minutes. Part-time drivers are more likely to arrive late at stops and terminal points, which may be addressed through driver training and reduced use of part-time drivers. Peak period traffic, especially in the afternoon peak, also has a negative impact on bus service reliability. Only considered one transit system, but otherwise a very relevant and thorough analysis. Shiftan & Wilson, 1994 Shiftan, Y. W. (1994). Absence, Overtime and Reliability Relations in Transit Workforce Planning. Transportation Research, Part A, 28(3), 245-258. Relationshi p between employee absenteeis m and overtime Researchers based at Cambridge Systematics and Massachusetts Institute of Technology in the United States Absences are divided in three categories, voluntary (unpaid), involuntary (paid), and sick (paid after first day). The model considers all categories of absences as a habit of approach-avoidance resulting from a decision process. The availability and usage of overtime hours (extraboard). None The authors test the hypothesis that availability of overtime may induce absence by providing a more profitable alternative to regular working hours, finding that working overtime hours is not positively related to absenteeism. On the contrary, employees who worked overtime had a lower rate of absence. The authors concluded that absenteeism is a habit that can be discourage by monitoring and incentivizing attendance. The authors did not find a significant explanation for absenteeism. Benn, 1995 Benn, H. (1995). TCRP Synthesis 10: Bus Route Evaluation Standards. Washington, D.C.: Transportation Research Board. Retrieved from http://onlinepubs.trb.org/onli nepubs/tcrp/tsyn10.pdf Bus route evaluation Researchers based in the United States Service reliability is measured in terms of on-time performance and headway adherence None None Most transit agencies surveyed reported using a definition of on-time performance as between 1 minute early and 5 minutes late. Limited discussion of reliability overall.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-79 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Lin, Liang, Schonfeld, & Larson, 1995 Lin, G.-S., Liang, P., Schonfeld, P., & Larson, R. (1995). Adaptive Control of Transit Operations. Federal Transit Administration. Retrieved from http://citeseerx.ist.psu.edu/vi ewdoc/summary?doi=10.1.1 .41.537 Adaptive control of transit operations Researchers based at the University of Maryland in College Park, Maryland, United States None Traffic congestion, Traffic signals, Incidents, Transfers, Bus bunching, Interrelated and congested intersections Reduction of the number of stops, Signal preemption, Exclusive right-of-way, Vehicle dispatching controls This report focuses on adaptive control of transit operations based on real-time information. The author indicates that control options may improve bus service reliability, effectively reducing user wait times, but often at the cost of increasing in-vehicle travel time. Short routes with larger stop spacing are said to improve reliability regardless of control strategies. Reserved right-of-way is expected to improve transit speeds and reliability. Bus service reliability is also said to decline as buses advance along their routes. Control strategies that might help to improve bus service reliability include holding and stop skipping. Transit signal priority may also improve reliability in some cases. Some relevance to, but little discussion of reliability overall. Jacques & Levinson, 1997 Jacques, K. S., & Levinson, H. (1997). TCRP Report 26: Operational Analysis of Bus Lanes on Arterials. Washington, D.C.: TRB, National Research Council. Retrieved from http://onlinepubs.trb.org/onli nepubs/tcrp/tcrp_rpt_26- a.pdf Analysis of bus lanes on arterials Researchers based in the United States Reliability seems to be represented in terms of standard deviation and coefficient of variation of travel times. Right-of-way, Frequency and duration of stops, Traffic congestion, Traffic signals Bus lanes, Stop skipping Even when busways are present, traffic signals and right turning traffic may still cause bus delays. Focuses on bus speeds, with little discussion of reliability. Okunieff, 1997 Okunieff, P. E. (1997). TCRP Synthesis 24: AVL Systems for Bus Transit. Washington, D.C.: TRB, National Research Council. Retrieved from http://www.tcrponline.org/PD FDocuments/TSYN24.pdf Automated vehicle location systems Researcher based in the United States Reliability is defined as schedule adherence None Automated vehicle location systems, Transit signal priority This was an early report on automated vehicle location (AVL) systems, which outlined opportunities for using AVL data to provide real-time schedule adherence information to transit operators. Results from the survey conducted for this report indicate that schedule adherence was the top objective for implementing AVL systems, with 56 percent of respondents claiming this as an objective of AVL. Early study, limited information regarding impacts on reliability. Strathman, et al., 1999 Strathman, J., Dueker, K., Kimpel, T., Gerhart, R., Turner, K., Taylor, P., Hopper, J. (1999). Automated Bus Dispatching, Operations Control, and Service Reliability: Baseline Analysis. Transportation Research Record: Journal of the Transportation Research Board, No. 1666, pp. 28-36. Retrieved from http://trrjournalonline.trb.org/ doi/abs/10.3141/1666-04 Bus operations control and service reliability Researchers based at Portland State University and TriMet in the United States, Case study of TriMet in Portland, Oregon, United States Average running times, Running times variation, Headways, On-time performance, Headway ratio, Run time ratio, Coefficient of variation of headways, Arrival delay, Excess wait time, User ratings of reliability Bus bunching, Dispatching, Operations control, Passenger activity, Time of day, Traffic congestion, Departure delays at origin Computer-aided bus dispatching, Automated vehicle location systems Four objectives are outlined for service reliability metrics. Measures should: be self-evident and easy to interpret; permit direct comparison within and between routes; be as comparable as possible across measures; and retain as much information as possible. The researchers found that service reliability challenges are most intense during the afternoon peak travel period. Based on passenger ratings of reliability, it was found that shorter headways are associated with higher perceptions of reliability. Delay was found to be significantly correlated with stops, but not passenger activity. Longer routes also were more prone to delays. This study focused on gathering and analyzing baseline data on eight bus routes over only two weeks. Furth, 2000 Furth, P. G. (2000). TCRP Synthesis 34: Data Analysis for Bus Planning and Monitoring. Washington, D.C.: TRB, National Research Council. Retrieved from http://www.tcrponline.org/PD FDocuments/TSYN34.pdf Data analysis for bus planning and monitoring Researcher based at Northeastern University in the United States Reliability is loosely defined as schedule adherence, with running time variation being an indicator of unreliability. None Automated data collection A primary conclusion of this study was that automated data collection is critical for statistical analyses of running time and route-level schedule adherence. Limited relevance to or discussion of reliability.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-80 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Furth & Muller, 2000 Furth, P., & Muller, T. H. (2000). Conditional Bus Priority at Signalized Intersections: Better Service with Less Traffic Disruption. Transportation Research Record: Journal of the Transportation Research Board, No. 1731, pp. 23-30. Retrieved from http://trrjournalonline.trb.org/ doi/abs/10.3141/1731-04 Conditional bus priority at traffic signals Researchers based at Northeastern University in the United States and Delft University of Technology in the Netherlands, Case study of Eindhoven, the Netherlands Delay Traffic volumes and congestion, Traffic signals Conditional priority for buses at signalized intersections (active or passive; full, partial, or relative; conditional or unconditional) The results of this study indicate significant improvement in schedule adherence after the implementation of conditional signal priority for buses at signalized intersections. Traffic impacts were also studied for three scenarios: no priority, conditional priority, and absolute priority. Absolute priority significantly increased general traffic delays compared to the no priority scenario, while conditional priority had very little impact on traffic conditions overall. Results are based on only a few days of data collection. Jacques & Levinson, 2000 Jacques, K. R., & Levinson, H. S. (2000). TCRP Research Results Digest 38: Operational Analysis of Bus Lanes on Arterials: Application and Refinement. Retrieved from http://www.tcrponline.org/PD FDocuments/TCRP_RRD_3 8.pdf Analysis of bus lanes on arterials Researchers and case studies based in the United States Reliability seems to be represented in terms of standard deviation and coefficient of variation of travel times. Right-of-way, Frequency and duration of stops, Traffic congestion, Traffic signals Bus lanes Bus lanes with no right turning vehicles are expected to have less delay when compared with bus lanes that have right turn delays and/or traffic, and especially better than buses operating in mixed traffic. Observed speeds were found to be lower than estimated speeds in most of the cases considered in this study. Focuses on bus speeds, with little discussion of reliability. Muller & Furth, 2000 Muller, T. H., & Furth, P. G. (2000). Integrating Bus Service Planning with Analysis, Operational Control, and Performance Monitoring. Proceedings of the Intelligent Transportation Society of America Annual Meeting. Retrieved from http://www1.coe.neu.edu/~pf urth/Furth%20papers/2000% 20integrating- bus%20service%20planning %20w%20analysis,%20cont rol,%20monitoring.pdf Integration of bus service planning with analysis, operational control, and performanc e monitoring Researchers based at Delft University of Technology in the Netherlands and Northeastern University the United States Unreliability is loosely defined as schedule deviation. Planning (including scheduling), Operators, Supervision, Operational control, Traffic volumes, Traffic control systems Operational control (holding and conditional priority), Schedule optimization and adjustments Two types of feedback loops are identified for delivering high- quality transit services: a real-time information loop to keep service stay on schedule and a planning loop that can be used to adjust service design in response to past performance. Real-time information can be used to help inform bus drivers, enabling them to adjust their behavior as needed to remain on schedule. Conditional signal priority is also said to help reduce schedule deviation. The authors report that it is common practice to use every bus stop as a time point in the Netherlands. They also say that the ideal scheduled running time balances service quality (high speeds and punctuality) and operating costs, while being especially careful to avoid early departures. Simulation is used to show potential impacts of improvement strategies, but no before and after data is given. Scottish Executive, 2000 Scottish Executive. (2000). Comparative Evaluation of Greenways and Conventional Bus Lanes - Research Findings. Edinburgh. Retrieved from http://www.gov.scot/Publicati ons/2000/05/fc190343-bf4a- 44fb-947f-f2282237e1dc Busways and bus lanes Researchers based in Scotland, United Kingdom None Traffic congestion Busways (greenways), Bus lanes The introduction of busways on a particular corridor is reported to have improved bus service reliability in the corridor. There is also evidence of the busway increasing delays for non-priority vehicles. This busway cost about five times as much as conventional bus lanes. Ridership also increased with the introduction of the busway. Case study in Edinburgh, Scotland. May not be generalizable.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-81 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Skabardonis 2000 Skabardonis, A. (2000). Control Strategies for Transit Priority. Transportation Research Record: Journal of the Transportation Research Board, No. 1727, pp. 20-26. Retrieved from http://trrjournalonline.trb.org/ doi/abs/10.3141/1727-03 Control strategies for transit priority Researcher based at the University of California at Berkeley in the United States Performance index (a weighted combination of delays and stops) Network configuration and characteristics (single arterial, grid network, signal spacing, number of lanes, pedestrian presence, type and operation of the traffic control system), Network traffic patterns (traffic volumes, turning movements, variability in traffic volumes, level of congestion, extent to which traffic congestion interferes with bus operations and the nature of the interference), Frequency and characteristics of transit service (bus volume, type(s) of bus operations, transit routes, bus stop locations and design, amount and variability of dwell times, and communication and monitoring equipment for transit vehicles) Transit priority (facility design or traffic control, passive or active) The author states that "the effectiveness of transit priority strategies depends on the amount and the source of delay to the transit vehicles", and that if signal delay is only a small portion of the overall route delay, then signal priority would only be able to have a small impact at best. Simulations of an arterial corridor were conducted to test operational control strategies. Optimal timing plans favoring buses along the corridor were estimated to reduce delay by 14%, reduce stops by 1%, and improve average bus speeds by about 4%, resulting in a delay savings of about 2 seconds per bus per intersection and minimal increases in general traffic delays. These results, based on the baseline data, were found to be insensitive to bus volumes up to 30 buses per hour. Bus preemption at signals with offline fixed-time timing plans demonstrated a potential savings of 6 seconds per bus per intersection, or a 2 minute travel time savings, with more benefits for higher bus volumes. Even greater benefits were expected to result from active priority for buses experiencing delays, but the tests showed major negative impacts to general traffic. As such, this approach was deemed as unlikely to be implementable in any real-life system. System-wide transit priority with online signal control and automatic transit vehicle location and monitoring showed more promise, with moderate improvements over preemption with fixed-time plans. Overall, it was found that passive priority strategies may be useful, easy, and affordable to implement for simple networks, high-frequency systems, and routes with predictable dwell times. The results are based on simulation, rather than a real-life before and after study. Strathman, et al., 2000 Strathman, J. et al. (2000). Bus Transit Operations Control: Review and an Experiment Involving Tri- Met's Automated Bus Dispatching System. Transportation Northwest, Department of Civil Engineering, University of Washington. Retrieved from http://www.researchgate.net/ profile/Kenneth_Dueker/publ ication/265308609_Bus_Tra nsit_Operations_Control_Re view_and_an_Experiment_I nvolving_Tri- Met's_Automated_Bus_Disp atching_System/links/54b68 1c50cf2bd04be32191d.pdf Bus transit operations control Researchers based at Portland State University and TriMet, Case study of TriMet in Portland, Oregon, United States Reliability is measured in terms of headway adherence (headway variance). Traffic congestion Operational methods (schedule modification, route restructuring, driver training), Priority methods (bus lanes, signal prioritization), Control methods (vehicle holding, short-turning, stop skipping, speed modifications) The researchers conducted an experiment to investigate the impacts of a combination of vehicle holds, swaps, and short turns on service reliability. The results indicate that, compared to the baseline measurements, headway ratio variances dropped 3.8% overall and 15.8% at the control point. Headway regularity was best at the location of the control action and in the early stages of trips. Although there were some mixed and insignificant results, the improvement strategies were found to produce a net benefit overall. Further automation and extension of vehicle location and monitoring technology into the field is recommended for potentially greater benefits to headway reliability. Mixed results of the experiment and a combination of strategies make the actual impacts of each strategy individually difficult if not impossible to assess. Bates, Polak, Jones, & Cook, 2001 Bates, J., Polak, J., Jones, P., & Cook, A. (2001). The Valuation of Reliability for Personal Travel. Transportation Research Part E, 37, pp. 191-229. Retrieved from http://www.sciencedirect.co m/science/article/pii/S13665 54500000119 Valuing reliability for personal travel Researchers based at Oxford, Imperial College, and University of Westminster in the United Kingdom Many definitions considered, authors loosely defined reliability as a combination of schedule adherence and variation in travel time, plus other applicable considerations Crowding None Punctuality is a highly valued characteristic of transportation systems. Unreliability is especially costly when public transport is the chosen mode. The authors also suggest that reliability is better measured through the median rather than the mean, and through the 90th percentile point of travel times rather than the standard deviation. Considered modes beyond public transport, and drew few conclusions relating to public transport specifically.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-82 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Eberlein, Bernstein, & Wilson, 2001 Eberlein, X. J., Bernstein, D. H., & Wilson, N. H. (2001). The Holding Problem with Real-Time Information Available. Transportation Science, 35(1), 1-18. This paper presents a holding method that aims to minimize the sum of squared headways on a rail line. Researchers based at the Massachusetts Institute of Technology and James Madison University in the United States Route performance is measured in terms of headway variability, which ultimately determines waiting time of randomly arriving passengers Vehicle Location Squared headway minimization mathematical program on a rolling horizon The optimization algorithm can reduce passenger waiting time compared to a regular schedule by minimizing squared headways without using event-based simulation. The more downstream the control point the better. Adding control points does not significantly improve performance. The model is fully deterministic and stationary. In addition, the optimization tool is presented as an algorithm, which could require intensive computation, if the model was extended to consider more complexity in the route. Fitzpatrick, Hall, Farnsworth & Finley, 2001 Fitzpatrick, K., Hall, K., Farnsworth, S., & Finley, M. (2001). TCRP Report 65: Evaluation of Bus Bulbs. Washington, D.C.: Transportation Research Board. Retrieved from https://nacto.org/docs/usdg/t crprpt65_fitzpatrick.pdf Bus bulbs Researchers based at the Texas Transportation Institute at Texas A&M University in the United States None None Bus bulbs Bus bulbs are commonly installed in cases of high transit ridership in a corridor and re-entry problems for buses during peak travel times, among other reasons. They may help to improve bus service reliability by facilitating passenger boarding and alighting, as well as bus movements. This guide provides detailed information on the proper use and expected impacts of bus bulbs. Limited relevance overall, and no discussion of impacts on reliability. Perk & Foreman, 2001 Perk, V. & Foreman, C. (2001). Evaluation of First- Year Florida MPO Transit Capacity and Quality of Service Reports. Tampa: National Center for Transit Research, University of South Florida. Retrieved from http://intrans.iastate.edu/pub lications/_documents/midco n- presentations/2003/PerkTra nsit.pdf Evaluation of service reports Researchers based at the University of South Florida in the United States Reliability is measured in terms of on-time performance (within 5 minutes of scheduled arrival time) and headway adherence. Reliability quality of service (QOS) is also used. Traffic LOS None This report detailed an evaluation of transit capacity and quality of service reports for a new MPO in Florida. Key recommendations related to reliability include: average service frequency and hours of service across the routes required for a trip involving one or more transfers, determine the maximum load point, and focus reliability data collection efforts on that point; reliability data should be collected over a period of 6-8 weeks; and reliability QOS measures should be analyzed to ensure that they are realistic for peak conditions. More research is recommended on the thresholds for transportation quality of service measures such as on-time performance. Focuses on a single MPO.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-83 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Rietveld, Bruinsma, & vanVuuren, 2001 Rietveld, P., Bruinsma, D., & vanVuuren, D. (2001). Coping with Unreliability in Public Transport Chains: A Case Study for Netherlands. Transportation Research Part A, 35(6), pp. 539-559. Retrieved from http://www.sciencedirect.co m/science/article/pii/S09658 56400000069 Unreliability in public transporta- tion Researchers based at Vrye, Universiteit, De Boelelaan in Amsterdam, The Netherlands Reliability may be measured as the probability that a vehicle arrives x minutes late, the probability of an early departure, the mean difference between the expected and scheduled arrival times, the mean delay of an arrival given that one arrives late, the mean delay of an arrival given that one arrives more than x minutes late, the standard deviation of arrival times, and the adjusted standard deviation of arrival times (ignoring early arrivals), among others. Bus frequency, Number of transfers, Transfer times, Access and egress mode, Reliability of arrival and departure times, Traffic congestion Bicycle used as access and/or egress mode, Increased transfer times, All frequencies at least twice per hour, All buses do not depart early Through a simulation of the effects of a variety of strategies to address unreliability in public transport chains in the Netherlands, the most promising approaches were identified as the use of a bicycle as the access and/or egress mode for those using public transportation. The authors also found evidence of a strong risk aversion when it comes to uncertainty in the reliability of public transport, as shown by the measured "uncertainty minute" being estimated at 2.4 the weight of a "certain" in-vehicle minute. The results of this study were largely drawn from a simulation, as opposed to a before and after study. Furthermore, the findings related to bicycling may not be easily transferable to places with less developed bicycle infrastructure. Strathman, Kimpel, Dueker, Gerhart, Callas, 2001 Strathman, J.G., T.J. Kimpel, K.J. Dueker, R.L. Gerhart, and S. Callas. Evaluation of Transit Operations: Data Applications of Tri-Met’s Automated Bus Dispatching System. Report TNW2001- 04. TransNow, Seattle, Wash., February 2001 Analyze reliability factors on TriMet bus routes Center for Urban Studies, Portland State University On-time performance recovery time and driver behavior None The study found that there was too much recovery time in the schedules of TriMet bus routes. The study also found a correlation between driver experience and travel time. None Barker, 2002 Barker, D. P. (2002). Communication, information and responsibility distribution strategies for effective real-time transit service management. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/ 8378 Real-time transit service manageme nt Researcher based at Massachusetts Institute of Technology in the United States, Case study of Chicago Transit Authority in Chicago, Illinois, United States Reliability is measured in terms of schedule adherence. Traffic congestion, Incidents, Mechanical breakdowns, Bus bunching, Crowding, Crashes, Weather, Route blockage Managing headways and schedules, Solving mechanical problems, Managing reliefs, Providing additional or altered service, Managing emergencies There seem to be inherent advantages in managing schedule adherence from the field, while managing incidents from a control center. Identified response techniques that might be used in real- time management include managing headways and schedules (hold bus, hold leader, drop off only, express to a later point, short turn, follower picks up passengers, spread the interval, spread the terminal / reschedule street), solving mechanical problems (supervisor repairs bus at terminal, supervisor repairs bus on site, truck repairs bus on site, maintenance brings bus change, maintenance tows bus, pulls in/out, jump buses), managing reliefs (wait for relief, relief operator relieves other than scheduled operator, relief operator relieves scheduled operator later, pull in, pull out instead of relieve, operator exchange), providing additional or altered service (use standby bus, fill from another street, fill with pull-in, put bus in place, emergency reroute), and managing emergencies (bus continues to point supervision, dispatch mobile supervision, dispatch police and/or medics). Major findings of this thesis include: disruptions have a predictable set of potential responses, but information is needed to choose among them; random disruptions occur with a predictable frequency; communication channel capacity may be a limiting factor on the effectiveness of real-time management strategies; and PDAs would allow for better service restoration decisions and schedule issue resolution. It is also concluded that adoption of new communication technologies should occur simultaneously with development of new procedures for maximum benefit. Little or no discussion of the impacts of various strategies on service reliability.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-84 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Chang, 2002 Chang, J. (2002). Evaluation of Service Reliability Impacts of Traffic Signal Priority Strategies for Bus Transit. Virginia Polytechnic Institute and State University. Retrieved from http://scholar.lib.vt.edu/these s/available/etd-10252002- 181921/ Impacts of bus transit signal priority Researcher based at Virginia Polytechnic Institute and State University, Case study in Arlington, Virginia, United States Schedule adherence (on-time performance, time reliability, perceived on-time performance, spacing, and arrival reliability) Peak period travel, Passenger demand and activity, Stop locations, Vehicle characteristics, Signal delay, Traffic congestion, Weather, Bus scheduling Transit signal priority Transit signal priority was shown to decrease the standard deviation of arrival time deviation from schedule by 3.2% on average across 30 simulation cases in Arlington, Virginia. Simulation- based results using data only from Arlington, Virginia. Hong, 2002 Hong, Y. (2002). The development of more effective operating plans for bus service. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/ 84840 Bus service operation planning Researcher based at the Massachusetts Institute of Technology in the United States, Case study on Chicago Transit Authority in Chicago, Illinois, United States Measures of service reliability include schedule adherence and passenger out-of-vehicle time. Passenger activity, Frequencies, Crowding, Operating plans, Schedule time, Recovery time, Distance from origin Holding, Control strategies, Increasing schedule time Quality of service is measured by in-vehicle time, crowding level, passenger wait time, and schedule adherence. Passengers are said to evaluate service reliability in terms of their waiting time, rather than in terms of schedule adherence as the operators do. Reliability is said to decline as a bus progresses along its route, though time points may help to correct for this trend. Vehicles are considered to be on-time if they are less than five minutes late. An example is used to demonstrate that recovery time is determined by the probability of an on-time departure at the terminal and schedule time. Many tradeoffs in the scheduling process were noted. Headway decisions impact the tradeoff between operational cost and service quality, both of which increase with smaller headways. Increasing schedule time and the number of time points can reduce deviation at time points, thereby improving schedule adherence. While this may help to reduce out-of-vehicle waiting times, it may increase in-vehicle waiting times. An analytical model is used to demonstrate the impacts of increasing schedule time, and the results indicate that this strategy does increase the mean travel time while decreasing the variability in travel time. Methods may be outdated or made obsolete when automated data collection and vehicle location technologies are widely used. Lehtonen & Kulmala, 2002 Lehtonen, M., & Kulmala, R. (2002). Benefits of Pilot Implementation of Public Transport Signal Priorities and Real-Time Passenger Information. Transportation Research Record: Journal of the Transportation Research Board, No. 1799, pp. 18-25. Retrieved from https://journals.sagepub.com /doi/10.3141/1799-03 Impacts of signal priority and real-time information Researchers based at VTT in Finland, Case study of Helsinki, Finland Reliability measures included travel time, regularity, and punctuality. Traffic signals, Traffic volumes, Passenger volumes Transit signal priority, Real-time passenger information The results of this research indicate a more than 40% reduction in delays at intersections, with noticeable improvements to service regularity and punctuality. Passenger volumes increased by 10- 12% after implementation of the transit signal priority and real-time information improvements for the bus system. Fuel consumption and emissions were also reduced by 1-5%. The return on capital investment for buses was increased by 6%. Only a few months’ worth of data collected. Levinson, Zimmerma n, Clinger, & Rutherford, 2002 Levinson, H. S., Zimmerman, S., Clinger, J., & Rutherford, G. S. (2002). Bus Rapid Transit: An Overview. Journal of Public Transportation, 5(2). Retrieved from http://scholarcommons.usf.e du/jpt/vol5/iss2/1/ Bus Rapid Transit implementa tion Researchers based at DMJM+HARRIS and the University of Washington in the United States Bus Rapid Transit is defined as a flexible, rubber-tired form of rapid transit that combines stations, vehicles, services, running ways, and ITS elements into an integrated system with a strong identity. Roadway geometry, Traffic signal controls, Curb parking and loading, Turn controls, Number of doors, Demand versus capacity, Fare collection methods, Frequencies, Transfers Running ways, Preferential treatments, Far-side bus stops, Guidance technologies, Real-time vehicle tracking Reliability is a key factor distinguishing Bus Rapid Transit from more general forms of bus transit. A combination of elements are grouped together to provide BRT service, including running ways, preferential treatments, queue bypasses, far-side bus stops, low- floor buses, transit signal priority, electronic fare collection, bus guidance technologies, and high-frequency service. Detailed recommendations for implementing BRT are included in this guide. Fairly comprehensive, but focuses on Bus Rapid Transit specifically.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-85 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Strathman, Kimpel, Dueker, Gerhart, & Callas, 2002 Strathman, J., Kimpel, T., Dueker, K., Gerhart, R., & Callas, S. (2002). Evaluation of Transit Operations: Data Applications of Tri-Met's Automated Bus Dispatching System. Transportation, 29(3), pp. 321-345. Retrieved from http://link.springer.com/articl e/10.1023%2FA%3A101563 3408953 Evaluating transit operations based on bus dispatching system data Researchers based at Portland State University and TriMet in the United States, Case study of TriMet in Portland, Oregon, United States Running time variation Bus operators, Route design, Time of day, Direction of service, Passenger activity, Season Using automatically collected data to improve scheduling, monitoring, and operation control, Shifting to more full-time and fewer part-time bus operators, Grouping the work assignments of operators with similar experience, Additional field supervision A model is developed to describe the relationship between several factors and bus running times. Stops involving a lift operation are estimated to require 68 seconds, while a stop involving a single boarding or alighting is estimated to require about 11 seconds. Early morning and night trips are estimated to need about 250 (7.7%) fewer seconds than midday trips. Morning peak trips are estimated to need about 100 (3%) fewer seconds than midday trips, while afternoon peak trips may require 138 seconds (4%) less than midday trips. Feeder routes are estimated to need about 418 seconds (12.5%) less running time than radial routes, with crosstown and peak express routes needing about 500 seconds (15%) and about 1000 seconds (32.5%) less running time, respectively. Changes in headway, operator differences, and variation between seasons were also noted as being significant factors in the model. The results show that part-time bus operators contribute to running time variability. No intercept in running time model, claims r- squared of 0.96. Turochy & Smith, 2002 Turochy, R. E., & Smith, B. L. (2002). Measuring Variability in Traffic Conditions by Using Archived Traffic Data. Transportation Research Record: Journal of the Transportation Research Board, No. 1804, pp. 168- 172. https://journals.sagepub.com /doi/abs/10.3141/1804-22 Measuring variability in traffic conditions Researchers based at Auburn University and the University of Virginia in the United States Variability index Traffic congestion None The variability index is demonstrated as a tool for identifying times of day that have the largest variation in traffic conditions, which are likely to correspond to unpredictable and unreliable bus service. It is unclear how this would be applied to make service more reliable. Vuchic, 2002 Vuchic, V. R. (2002). Bus Semi rapid Transit Mode Development and Evaluation. Journal of Public Transportation, 5(2), 4. Retrieved from http://scholarcommons.usf.e du/jpt/vol5/iss2/4/ Bus semi rapid transit Researcher based at the University of Pennsylvania in the United States None Right-of-way Exclusive right-of-way, Transit signal priority Fully-controlled exclusive right-of-way is said to offer the highest performance in terms of speed, reliability, capacity, and safety. Vuchic also recommends excluding all but transit vehicles from transit lanes for optimal performance. Focuses on bus semi rapid transit as a mode, not the reliability aspects of that mode compared to others. Bertini & El- Geneidy, 2003 Bertini, R., & El-Geneidy, A. (2003). Generating Transit Performance Measures with Archived Data. Transportation Research Record: Journal of the Transportation Research Board, No. 1841, pp. 109- 119. Retrieved from http://web.cecs.pdx.edu/~mo nserec/t.data/resources/met adata/TRB_Paper_Bertini_T ransit.pdf Transit performanc e measure ment Researchers based at Portland State University, Case study of TriMet in Portland, Oregon, United States Mobility measures include: Percentage of on-time performance, Percentage of scheduled departures that do not leave within a specified time limit, Travel time contour, Minute variation in trip time, Average transfer time and delay, Dwell time at intermodal facilities, Proportion of persons delayed, and Average wait time to board transit. Reliability is defined as schedule adherence, with consideration given to several measures, such as standard deviation. Frequency of service, Fluctuations in traffic volumes None This papers provides an overview of one approach to generating transit performance measures using archived bus dispatch system data. Conventional measures of reliability, including on- time performance, are used. King, 2003 King, R. D. (2003). TCRP Synthesis 49: Yield to Bus— State of the Practice. Washington, D.C.: Transportation Research Board. Retrieved from http://onlinepubs.trb.org/onli nepubs/tcrp/tcrp_syn_49.pdf Yield to bus policies Researcher based in Columbus, Ohio, United States Reliability measures include schedule adherence or on-time performance. Yield to bus policies Yield to bus policies In the survey conducted for this report, about one-third of transit agencies reported improvement in schedule adherence due to their yield to bus policies, although none had any data beyond anecdotal evidence to support their response. Some relevance to, but little discussion of reliability overall.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-86 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Levinson, Zimmerma n, Clinger, & Gast, 2003 Levinson, H., Zimmerman, S., Clinger, J., & Gast, J., (2003). Bus Rapid Transit. Washington, D.C.: Transportation Research Board. Retrieved from http://nacto.org/docs/usdg/br t_synthesis_of_case_studies _levinson.pdf Bus rapid transit Researchers based at DMJM+HARRIS and the University of Washington in the United States Bus Rapid Transit is defined as an integrated system of facilities, services, and amenities that collectively improve the speed, reliability, and identity of bus transit. Running ways, Stations, Vehicles, Route structure, Fare collection methods, Intelligent transportation systems Exclusive running ways, Larger stop spacing, Level boarding, Low- floor buses, More and wider doors, Automatic vehicle location systems, Passenger information systems, Traffic signal priority, Off-vehicle fare collection, Frequent service This report describes Bus Rapid Transit (BRT) systems from around the world. Key features of BRT include running ways, stations, vehicles, intelligent transportation systems, and service patterns. BRT is said to provide better speed, capacity, and reliability than typical bus service. Some relevance to reliability measures, but little or no discussion of improvement strategy effectiveness. Li, 2003 Li, Y.-w. (2003). Evaluating the Urban Commute Experience: A Time Perception Approach. Journal of Public Transportation, 6(4), 3. Retrieved from http://scholarcommons.usf.e du/jpt/vol6/iss4/3/ Time perception among public transit users Researcher based at Argosy University in the United States None None None Service reliability is said to be a factor in perceived travel time among transit users. Limited relevance to or discussion of reliability. Lomax, Schrank, Turner, & Margiotta, 2003 Lomax, T., Schrank, D., Turner, S., & Margiotta, R. (2003). Selecting Travel Reliability Measures. Washington, D.C.: Federal Highway Administration. Retrieved from http://d2dtl5nnlpfr0r.cloudfro nt.net/tti.tamu.edu/document s/TTI-2003-3.pdf Selecting travel reliability measures Researchers based at the Texas Transportation Institute and Cambridge Systematics in the United States Reliability is defined as the level of consistency in transportation service for a mode, trip, route, or corridor for a time period. Measures are classified as statistical range measures, buffer time measures, and tardy trip indicators. Traffic congestion, Bus availability, Incidents, Work zones, Weather, Fluctuations in demand, Special events, Traffic control devices, Inadequate base capacity None Statistical range reliability measures include travel time window (average travel time plus or minus one standard deviation), percent variation (standard deviation divided by average travel time times 100%), and variability index (difference in bounds of 95% confidence intervals during peak periods divided by the difference in bounds of 95% confidence intervals for non-peak periods). Buffer measures include buffer time (95th percentile travel time minus the average travel time), buffer time index (95th percentile travel rate minus average travel rate, divided by average travel rate, averaged across all sections), and planning time index (95th percentile travel time index of all peak period travel). Tardy trip indicators include the Florida reliability method (percent of unreliable trips as calculated by subtracting the percent of trips with travel times greater than expected from 100%), on-time arrival (100% minus the percent of travel rates greater than 110% of the average travel rate), and misery index (average of the travel rates for the longest 20% of trips minus average travel rates for all trips, divided by the average travel rate). As measures are selected and calculated, several factors should be considered: whether the measure is mode- specific or mode-neutral, trip type and location, controlling for length and time, the audience or target user of the measure, and the area size. Of the measures considered, percent variation, misery index, and buffer time index are recommended. No general answers are given, but a process is laid out for selecting and calculating reliability measures. Multisyste ms, Inc., Mundle & Associates, Inc., & Simon & Simon Research and Associates, Inc., 2003 Multisystems, Inc., Mundle & Associates, Inc., & Simon & Simon Research and Associates, Inc., (2003). TCRP Report 94: Fare Policies, Structures, and Technologies Update. Washington, D.C.: Transportation Research Board. Retrieved from http://www.tcrponline.org/PD FDocuments/TCRP_RPT_9 4.pdf Fare policies and technologie s Researchers based in the United States None Fare payment type and policies Contactless fare payment cards Contactless fare cards are said to allow faster boarding, which could help to improve bus service reliability. Smartcard systems also have lower maintenance requirements and higher reliability than other card-reader technologies that require swiping or inserting a ticket. Several examples of agencies that have reported reliability benefits of switching to contactless cards for fare payment are given in this report. Limited relevance to or discussion of reliability.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-87 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Schweiger, 2003 Schweiger, C. L. (2003). TCRP Synthesis 48: Real- Time Bus Arrival Information Systems. Washington, D.C.: Transportation Research Board of the National Academies. Retrieved from http://www.tcrponline.org/PD FDocuments/tsyn48.pdf Real-time bus arrival information systems Researcher based in Cambridge, Massachusetts, United States None None Real-time bus arrival information Real-time bus arrival information systems produce data that is sometimes used by agencies to improve bus reliability, system expansion, and real-time dispatching. London Transport passengers (64% of those surveyed) reported an improvement in bus service reliability after the implementation of real-time customer information, even though service reliability had actually declined. Little or no discussion of the impacts of real-time arrival information on service reliability. Strathman, Kimpel, & Callas, 2003 Strathman, J., Kimpel, T., & Callas, S. (2003). Headway Deviation Effects on Bus Passenger Loads: Analysis of Tri-Met's Archived AVL- APC Data. Portland, OR: Center for Urban Studies at Portland State University. Retrieved from http://ntl.bts.gov/lib/24000/24 100/24134/TNW2003-01.pdf Correlation of Passenger loads and headway variability Researchers based at Portland State University in the United States Headway Coefficient of Variation Passenger loads None Headway delay is correlated to passenger load. The study demonstrates the correlation between headway variability and passenger loads but it does not analyze the internal dynamics of the relationship. It states that "headway delays are a primary cause for passenger overloads" when the converse is also true. Wile, 2003 Wile, E. S. (2003). Use of automatically collected data to improve transit line performance. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/ 85755 Automated data collection and application Researcher based at the Massachusetts Institute of Technology in the United States, Case studies of the Chicago Transit Authority in Chicago, Illinois and the Massachusetts Bay Transportation Authority in Boston, Massachusetts, United States Reliability is discussed, as well as headway regularity and schedule adherence. Service management, Passenger information, Performance monitoring, Maintenance, Operations planning, Performance auditing, Scheduling, Route planning Automated vehicle location systems, Automated vehicle identification systems, Trip time analyzers, Automated passenger counters Data needed for segment-level run times and variances include: schedule-adherence figures (multiple from each time period), locations of time points, standard percentile of trips that should be on-time according to the schedule, standard percentile of buses that should be available to start the next trip on-time, and standards for data collection accuracy. Data required for tracking schedule adherence are: route-wide vehicle location and timetables. Some of the information is a bit outdated.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-88 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Dion, Rakha, & Zhang, 2004 Dion, F., Rakha, H., & Zhang, Y. (2004). Integration of Transit Signal Priority within Adaptive Traffic Signal Control Systems. Retrieved from http://www.researchgate.net/ profile/Hesham_Rakha/publi cation/228857827_Integratio n_of_Transit_Signal_Priority _Within_Adaptive_Traffic_Si gnal_Control_Systems/links/ 09e41508a2447e5f9300000 0.pdf Integration of transit signal priority with adaptive traffic signal control systems Researchers based at Virginia Polytechnic Institute and State University and Michigan State University, Simulation on Columbia Pike in Arlington, Virginia, United States Travel time, Delays, Stops Transit signal priority, Adaptive traffic signal control strategies, Time of day, Traffic congestion Transit signal priority, Adaptive traffic signal control strategies Buses typically benefit from transit signal priority, including fixed- time control, adaptive splits, and adaptive splits and offsets, while general traffic often experiences negative impacts. Adaptive signal control can help to improve transit operations, while minimizing the disbenefits to general traffic. Through a simulation on the Columbia Pike, adaptive transit priority is shown to have potential for successful deployment in arterial corridors with coordinated signals, without undue negative consequences for general traffic, under different types of signal control. Transit operations benefitted most when priority was given only to express buses. Adaptive signal control did not produce substantially better results for transit than fixed-time control, but it did mitigate many of the negative impacts on general traffic. Transit signal priority benefits depend greatly on traffic flow characteristics and the control strategy used in the corridor. Simulation- based results for a single corridor. Dueker, Kimpel, Strathman, & Callas, 2004 Dueker, K. J., Kimpel, T. J., Strathman, J. G., & Callas, S. (2004). Determinants of Bus Dwell Time. Journal of Public Transportation, 7(1), 2. Retrieved from http://scholarcommons.usf.e du/jpt/vol7/iss1/2/ Factors affecting bus dwell times Researchers based at Portland State University and TriMet, Case study of TriMet in Portland, Oregon, United States Dwell time is defined as the time in seconds that a transit vehicle is stopped for the purpose of serving passengers. It includes the total passenger service time plus the time needed to open and close the doors (HCM 1985). Passenger activity, Lift operations, Low-floor buses, Time of day, Route type Low-floor buses Results indicate that each boarding passenger adds 3.48 seconds to the base dwell time (5.14 seconds for door operation), while each alighting passenger adds 1.70 seconds. The estimated impact of a lift operation on dwell time (for high-floor and low-floor buses combined) is an increase of 62.07 seconds. Low-floor buses correlate with a 0.11 second reduction in dwell time per dwell, which resulted in an average savings of 3.96 seconds in total running time per trip. Low-floor buses also help to reduce dwell time for lift operations by 4.74 seconds, or 5.8%. Limited relevance overall, with some application to but very little mention of reliability. Evans IV, 2004 Evans IV, J. E. (2004). TCRP Report 95: Traveler Response to Transportation System Changes Chapter 9—Transit Scheduling and Frequency. Washington, D.C.: Transportation Research Board. Retrieved from https://www.nap.edu/catalog /23433/traveler-response-to- transportation-system- changes-handbook-third- edition-chapter-9-transit- scheduling-and-frequency Impacts of scheduling changes Researcher based in the United States Reliability from the passenger perspective is equated to arriving at the intended time, or on-time performance for the operator. Measures used include waiting time index and wait in excess of optimum as a percent. Environmental factors (traffic conditions, signals, variations in boardings and alightings, availability of drivers and vehicles), Inherent factors (platooning, missed runs, unplanned deviations) Frequency changes, Fare changes, Schedule changes, Reliability changes, Intelligent transportation systems On-time performance positively impacts riders and ridership due to less waiting, decreased travel times, fewer missed connections, more on-time arrivals, and reduced uncertainty. Limited relevance and new information overall. Shalaby & Farhan, 2004 Shalaby, A., & Farhan, A. (2004). Prediction Model of Bus Arrival and Departure Times Using AVL and APC Data. Journal of Public Transportation, Vol. 7, 1, 41- 61. Predicting bus arrival and departure times Researchers based at the University of Toronto and the City of Calgary in Canada Running times, Dwell times None User-interactive system to provide real-time information that can be used for proactive control The proposed system is said to allow users to predict downstream bus arrival and departure times, to help proactively control buses in a way that may avoid schedule deviations before they occur. This study focuses on a particular system for monitoring and controlling buses in real- time. Brownston e & Small, 2005 Brownstone, D., & Small, K. (2005). Valuing Time and Reliability: Assessing the Evidence from Road Pricing Demonstrations. Transportation Research Part A, 39, pp. 279-293. Retrieved from http://www.sciencedirect.co m/science/article/pii/S09658 5640400103X Valuing time and reliability in transportati on using road pricing demonstrati ons Researchers based at the University of California Irvine, Demonstrations in southern California, United States Unreliability is defined as the standard deviation of travel time across days. None None The results indicate that the value of time saved during the morning peak period is between $20-40 per hour, based on revealed behavior, but less than half of this amount when based on hypothetical behavior. Reliability is also highly valued, especially by women. Researchers also found that reliability can be effectively modeled as a property of the upper tail of travel time distribution across days. Limited relevance overall, with some application to but very little mention of reliability.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-89 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Hammerle, Haynes, & McNeil, 2005 Hammerle, M., Haynes, M., & McNeil, S. (2005). Use of Automated Vehicle Location and Passenger Count Data to Evaluate Bus Operations. Transportation Research Record: Journal of the Transportation Research Board, No. 1903, pp. 27-34. Retrieved from http://www.worldtransitresea rch.info/research/597/ Use of AVL and APC data to evaluated bus operations Researchers based at Creighton Manning Engineering, CTA, and the University of Illinois at Chicago in the United States, Case study of Chicago Transit Authority in Chicago, Illinois, United States Schedule adherence, Headway regularity Time of day, Traffic volumes, Peak travel, Passenger demand, Direction of travel, Bus bunching Use of AVL and APC data for bus operational improvements These researchers found that schedule adherence was reduced as buses traveled along their route towards downtown Chicago. Bunching was found to be attributable to buses running early in some cases. Results are based on only a few days of data collection. Kimpel, Strathman, Bertini, & Callas, 2005 Kimpel, T., Strathman, J., Bertini, R., & Callas, S. (2005). Analysis of Transit Signal Priority Using Archived TriMet Bus Dispatch System Data. Transportation Research Record: Journal of the Transportation Research Board, No. 1925, pp. 156- 166. Retrieved from https://journals.sagepub.com /doi/10.1177/036119810519 2500116 Analysis of transit signal priority Researchers based at Portland State University and TriMet, Case study of TriMet in Portland, Oregon, United States Measures considered include bus running times, on-time performance, and excess passenger wait times. Traffic signals Transit signal priority Impacts of transit signal priority were found to be mixed across routes and time periods. The benefits of transit signal priority are said to be most fully realized when ongoing monitoring and adjustment programs are implemented. Over all analysis segments, on-time performance was actually found to decline after transit signal priority was implemented, mostly due to more early trips and roughly the same rate of late trips. The authors recommend adjusting bus schedules when implementing transit signal priority strategies. Mean and variance of headways was also found to increase after implementation of transit signal priority. The research team recommends using caution when considering whether or not to implement transit signal priority in a corridor. They recommend an action plan that includes selecting candidate bus routes based on identification of operational problems, performing a baseline analysis, undertaking regular performance monitoring following implementation of transit signal priority to identify any challenges, and being willing to adjust schedules, emitter activation thresholds, signal control logic and other factors to improve system performance. Focuses on data from Portland, so results should not be generalized. Lyons & Urry, 2005 Lyons, G., & Urry, J. (2005). Travel Time Use in the Information Age. Transportation Research Part A, 39, pp. 257-276. Retrieved from http://www.sciencedirect.co m/science/article/pii/S09658 56404000977 Productive use of travel time and its implications on mobility and mode choice Researchers based at the University of the West of England and Lancaster University in the United Kingdom None None None This research focused on analyzing data from the United Kingdom related to productive use of travel time when using a car or public transport. Findings suggest that individuals are becoming increasingly productive while traveling, thanks in part to improved technology, and that this trend may have major implications for the utility of travel and resulting patterns of that travel. Limited relevance overall, with some application to but very little mention of reliability. Moses, 2005 Moses, I. E. (2005). A transit route simulator for the evaluation of control strategies using automatically collected data. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/ 32413 Transit control strategies and route simulation Researcher based at Massachusetts Institute of Technology in the United States, Case study on Chicago Transit Authority in Chicago, Illinois, United States Reliability is discussed in terms of schedule adherence and headway adherence. Dwell times, Vehicle bunching, Operator behavior, Passenger behavior, Controls, Route length, Time of day, Trip start punctuality Schedule-based holding, Headway-based holding A simulation indicates that adjusting driver and passenger behavior, as well as control strategies, for only bunched vehicles has a limited impact on service. Dwell times caused by passenger movements were found to greatly contribute to bus bunching and gapping in the simulation, and this trend is expected to hold true in real life. The simulation also indicated that deviations in origin terminal departure have an impact on headway regularity. This thesis focused on developing a model of transit service, but this model failed in the validation process.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-90 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Smith, Hemily, & Ivanovic, 2005 Smith, H. R., Hemily, B., & Ivanovic, M. (2005). Transit Signal Priority (TSP): A Planning and Implementation Handbook. Washington, D.C.: ITS America. Retrieved from http://trid.trb.org/view.aspx?i d=772546 Transit signal priority guide Researchers based in the United States None None Transit signal priority Reliability is repeated indicated as a key benefit of transit signal priority (TSP). Several examples of successful transit signal priority implementation were shared. In Tacoma, Washington, TSP and signal optimization combined to reduce transit signal delay by 40% in two corridors. TriMet demonstrated a 10% reduction in travel times and about 19% less travel time variability after implementing TSP, allowing them to reduce scheduled recovery time and avoid adding another bus to the route. This is a detailed guide focused on a single improvement strategy. Tzeng, Lin, & Opricovic, 2005 Tzeng, G.-H., Lin, C.-W., & Opricovic, S. (2005). Multi- Criteria Analysis of Alternative-Fuel Buses for Public Transportation. Energy Policy, 33(11), 1373- 1383. Retrieved from http://www.sciencedirect.co m/science/article/pii/S03014 21503003811 Analysis of alternative- fuel buses for public transportati on Researchers based at the National Chiao Tung University in Taiwan and the University of Belgrade in Yugoslavia, Analysis performed using Taiwan urban areas as the context None Reliability of energy supply and storage, Vehicle maintenance and capabilities None For Taiwan urban areas, hybrid-electric buses were found to be the most suitable replacements for standard diesel buses. If the range of pure electric buses was improved, these might become the better alternative. Limited relevance overall, with some application to but very little mention of reliability. Camus, Longo, Macorini, 2005 Camus, R., G. Longo, and C. Macorini. Estimation of Transit Reliability Level-of- Service Based on Automatic Vehicle Location Data. In Transportation Research Record: Journal of the Transportation Research Board, No. 1927, 2005, pp. 277–286. https://journals.sagepub.com /doi/abs/10.1177/036119810 5192700131 Novel methodol- ogy for on- time perfor- mance University of Trieste On-time performance Distribution of lateness None Considers delays as a distribution instead of a binary variable. They introduce the Weighted Delay Index, which is essentially the ratio of mean delay over headway. Lacks motivation. Cham, 2006 Cham, L. C. (2006). Understanding Bus Service Reliability: A Practical Framework Using AVL / APC Data. Massachusetts Institute of Technology. Retrieved from http://dspace.mit.edu/handle /1721.1/34381 Bus service reliability Researcher based at Massachusetts Institute of Technology in the United States Variability in running times and headways Deviations at terminals, Passenger loads, Running times, Environmental factors, Operator behavior Better supervision at terminal stops, Operator training, Corrective strategies, Transit signal priority This master's thesis considers numerous potential causes of bus unreliability, with the primary factor identified as deviations in departure time from terminals. To address this issue, better supervision at terminals is proposed. No before and after study of proposed improvement strategies. Dziekan & Vermeulen, 2006 Dziekan, K., & Vermeulen, A. (2006). Psychological Effects of and Design Preferences for Real-Time Information Displays. Journal of Public Transportation, 9, 71-89. Retrieved from http://scholarcommons.usf.e du/cgi/viewcontent.cgi?articl e=1251&context=jpt The customer effects of dynamic real-time information at transit stops Researchers based at the Royal Institute of Technology in Sweden and at HTM Rail in the Netherlands None None Real-time information at transit stops Several effects of real-time information at transit stops on customers were outlined, including reduced wait time and improved customer satisfaction. One of the studies presented in the paper indicated a 20% reduction in perceived wait times after the implementation of real-time information at tram stops in the Netherlands. Limited relevance overall, with some application to but very little mention of reliability.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-91 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Eichler & Daganzo, 2006 Eichler, M., & Daganzo, C. F. (2006). Bus Lanes with Intermittent Priority: Strategy Formulae and an Evaluation. Transportation Research Part B: Methodological, 40(9), pp. 731-744. Retrieved from http://www.sciencedirect.co m/science/article/pii/S01912 61505001104 Evaluation of bus lanes with intermittent priority Researchers based at the University of California Berkeley in the United States None None Bus lanes with intermittent priority, Transit signal priority, Exclusive bus lanes, Queue jump lanes Bus lanes with intermittent priority (BLIPs) do not reduce street capacity nearly as much as dedicated bus lanes, but they can reduce the impact of traffic congestion on bus operations. The amount of traffic disruption of a BLIP depends on the transit signal priority strategies used, among other factors. BLIPs can also help to reduce bus travel times, but this too is dependent on traffic saturation level, bus frequency, the improvement in bus travel time achieved by the special lane, and the ratio of bus and car occupant flows. Limited relevance overall, with some application to but very little mention of reliability. El- Geneidy, Strathman, Kimpel, & Crout, 2006 El-Geneidy, A., Strathman, J., Kimpel, T., & Crout, D. (2006). Effects of Bus Stop Consolidation on Passenger Activity and Transit Operations. Transportation Research Record: Journal of the Transportation Research Board, No. 1971, pp. 32-41. Retrieved from https://journals.sagepub.com /doi/10.1177/036119810619 7100104 Effects of bus stop consolidati on Researchers based at Portland State University, Case study of TriMet in Portland, Oregon, United States Reliability measures include running time coefficient of variation and headway coefficient of variation. Variation of passenger movements, Time of day Bus stop consolidation Bus stop consolidation was shown to significantly improve bus running times, while having no significant impact on passenger demand. Running time variability was estimated as increasing with increases in variation of passenger movements and decreasing during the morning peak. However, bus stop consolidation was not shown to have a significant impact on running time variation or headway variation in this case. This was a case study on two bus routes, so the results should not be generalized. Furth & Muller, 2006 Furth, P. G., & Muller, T. H. (2006). Service Reliability and Hidden Waiting Time: Insights from Automatic Vehicle Location Data. Transportation Research Record: Journal of the Transportation Research Board, No. 1995, pp.79-87. https://journals.sagepub.com /doi/10.1177/036119810619 5500110 Waiting time and reliability Researchers based at Northeastern University in the United States and Delft University of Technology in the Netherlands, Case study of Eindhoven, the Netherlands Waiting time None None Extreme values of waiting time, which relate to the budgeted and potential wait times, are more sensitive to service reliability than expected waiting time. This study focused on measuring wait time. Furth, Hemily, Muller, & Strathman, 2006 Furth, P., Hemily, B., Muller, T., & Strathman, J. (2006). TCRP Report 113: Using Archived AVL-APC Data to Improve Transit Performance and Management. Washington, D.C.: Transportation Research Board. Retrieved from https://www.nap.edu/catalog /13907/using-archived-avl- apc-data-to-improve-transit- performance-and- management Using archived ITS data for performanc e and reliability manageme nt Researchers based at Northeastern University and Portland State University in the United States, Hemily and Associates in Canada, and Delft University of Technology in the Netherlands Reliability is defined as on-time performance from the operator perspective, and equivalent waiting time from the user perspective. Vehicle reliability Real-time vehicle location monitoring and control The authors state that service reliability has been undervalued in the transit industry, in part due to the lack of reliability measures that reflect user experience. Equivalent waiting time is a measure proposed to account for increasing passenger wait times and travel time budgets. Some relevance to reliability, but mostly focused on methods of employing AVL and APC data. Hollander, 2006 Hollander, Y. (2006). The Cost of Bus Travel Time Variability. The University of Leeds Institute for Transport Studies. The cost of travel time variability Researcher based at the University of Leeds, United Kingdom Standard deviation of travel times None None In this study, Hollander improved the traffic micro-simulation model calibration such that it also produced bus travel times within the observed range of variability for buses and other vehicles for a particular study area, as opposed to merely reproducing the average travel time as the validation measure. No factors were used in the model other than travel times.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-92 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Hu & Jen, 2006 Hu, K.-C., & Jen, W. (2006). Passengers' Perceived Service Quality of City Buses in Taipei: Scale Development and Measurement. Transport Reviews, 26(5), 645-662. Retrieved from http://www.tandfonline.com/ doi/abs/10.1080/014416406 00679482 Measurem ent of perceived service quality of city buses Researchers based at Kainan University and National Chiao Tung University in Taiwan, Study based in Taipei, Taiwan None Traffic congestion None A problem related to reliability was identified as passengers lacking information about bus service reliability and having long wait times, which decreased passenger perception of service quality. Reliability is just a single component of a scale designed to measure overall service quality. Li, Rose, & Sarvi, 2006 Li, R., Rose, G., & Sarvi, M. (2006). Using Automatic Vehicle Identification Data to Gain Insight into Travel Time Variability and Its Causes. Transportation Research Record: Journal of the Transportation Research Board, No. 1945, pp. 24-32. https://journals.sagepub.com /doi/pdf/10.1177/036119810 6194500104 Measuring travel time variability Researchers based at Monash University in Australia Travel time variability Peak period travel, Capacity, Drivers None Travel times in morning and afternoon peak times show different characteristics and influencing factors. The morning peak tends to be more influenced by drivers, whereas the afternoon peak seems to be more influenced by limits on tollway capacity. This study focuses on a tollway, not public transit per se. Martin, 2006 Martin, P. C. (2006). TCRP Synthesis 64: Bus Use of Shoulders. Washington, D.C.: Transportation Research Board. https://www.nap.edu/catalog /13950/bus-use-of-shoulders Bus on shoulder operations Researcher based in San Francisco, California, United States, Case studies from the United States, Canada, and Ireland None Traffic congestion Bus on shoulder operations The use of highway shoulders to bypass congestion is recommended as a cost-effective method for improving bus running times and reliability. This practice is said to be especially popular with bus passengers. Focus is on bus on shoulder operations, not the impacts on reliability. Mishalani, McCord, & Wirtz, 2006 Mishalani, R., McCord, M. M., & Wirtz, J. (2006). Passenger Wait Time Perceptions at Bus Stops: Empirical Results and Impact on Evaluating Real- Time Bus Arrival Information. Journal of Public Transportation, 9(2), 89-106. Retrieved from http://nctr.usf.edu/jpt/pdf/JPT %209-2%20Mishalani.pdf Perceived wait times Researchers based at Ohio State University and Edwards and Kelcey, Inc. in the United States None None Real-time information at transit stops Results indicate that transit users perceive wait times to be greater than they actually are, in the absence of real-time arrivals information. Furthermore, longer walking distance to a stop corresponds to more exaggerated perceived waiting times, whereas the presence of a time constraint reduces the degree of this over-estimation. No discussion of reliability, only wait times.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-93 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Pangilinan, 2006 Pangilinan, C. A. (2006). Bus supervision deployment strategies for improved bus service reliability. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/ 38238 Bus supervision deployment strategies for service reliability Researcher based at Massachusetts Institute of Technology in the United States, Case studies of the Chicago Transit Authority in Chicago, Illinois and the Massachusetts Bay Transportation Authority in Boston, Massachusetts, United States Reliable transit service can be counted on over and over, with consistent wait times and travel times. Measures include the mean and coefficient of variation of travel time distribution, schedule adherence, and headway distribution. Weather, Traffic, Road construction, Planning, Maintenance, Driver abilities, Demand variability, Initial schedule deviations, Unbalanced passenger loads, Season, Time of day, Availability of personnel, Special events, Train crossings, Police activity, Equipment failures, Runs held-in at the garage, late garage pull-outs, Late or missing on-street reliefs, Early or late terminal departures Priority (exclusive lanes, signal priority), Control (service monitoring, holding, expressing), Operations (schedule improvements, fleet management, labor management) In this thesis, a bus supervision deployment strategy is developed for improving bus service reliability. The proposed strategy uses improved information and communication resources, while directing post supervisors to the busiest and most critical locations, using mobile supervisors for incident response, and redirecting many field-based supervisors to a central control center for headway and schedule management. An ideal service reliability hierarchy and feedback loop is defined, which incorporates long-term resource planning (sizing fleet, procurement, garage locations), short-term service planning (route structure, frequencies, service scheduling) day to day garage operations (fleet maintenance, operator training, staff scheduling, holding, late pull-outs), and real-time street operations (customer service, supervision). The MBTA uses service reliability standards for the beginning, middle, and end of a route. Routes with headways of 10 minutes or more should not depart the beginning or middle points early, but may arrive at the end point up to 3 minutes early. The standards for lateness on these less frequent routes is up to 3 minutes late from the start point, up to 7 minutes late at midpoints, and up to 5 minutes late to the end point. For routes with headways less than 10 minutes, the standards are for buses to depart the origin terminal within 25% of the scheduled headway and middle points within 50% of the scheduled headway. Running times should be within 20% of the scheduled running time. The overall test for a route is whether or not at least 75% of all trips comply with the applicable standards. Operations control strategies include spacing back (slowing down one or more runs ahead of a service gap), expressing (allowing passengers to alight only), moving up (departing a terminal early), short-turning (turning before reaching terminal), and filling-in (reallocating a bus from another route). Impacts of proposed supervision strategy are not fully known. Tse, Flin, & Mearns, 2006 Tse, J., Flin, R., & Mearns, K. (2006). Bus Driver Well- Being Review: 50 Years of Research. Transportation Research Part F: Traffic Psychology and Behavior, 9(2), pp. 89-114. Retrieved from http://www.sciencedirect.co m/science/article/pii/S13698 47805000872 Bus driver well-being Researchers based at the University of Aberdeen, King's College, in Scotland, United Kingdom None Bus driver experience, health, absenteeism, and behavior Driver training, Reduced stress and environmental demands Bus drivers' lifestyles at home and at work play a critical role in their physical and psychological health, as well as their job performance. Stressors such as traffic, violent passengers, and tight running schedules may have a negative impact on bus drivers, which can result in absenteeism, vehicle crashes, and other factors that may impact bus service reliability. Limited relevance overall, with some application to but very little mention of reliability.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-94 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Hollander, 2006 Hollander, Y. Direct versus indirect models for the effects of unreliability. In Transportation Research Part A, 40(9), 2006, pp. 699–711. cost of travel time variability to passengers compared to trip earliness and lateness University of Leeds Fare value Travel time reliability, early departure reliability, late arrival reliability None The model is based on a stated preference survey. Responses indicate that transit usage put little monetary value on travel time variability than early departure reliability and late arrival reliability. These results indicate that using travel time reliability as a summary performance metric may overlook passenger's reliability. None Dziekan & Kottenhoff, 2007 Dziekan, K., & Kottenhoff, K. (2007). Dynamic At-Stop Real-Time Information Displays for Public Transport: Effects on Customers. Transportation Research Part A, 41(6), pp. 489-501. Retrieved from http://www.sciencedirect.co m/science/article/pii/S09658 56406001431 Psychologi cal effects and design preference s for real- time information displays Researchers based at the Royal Institute of Technology in Sweden, Case study of the Hague, The Netherlands and Stockholm, Sweden None None Real-time information at transit stops A study of customer response to real-time information displays at transit stations was performed in The Hague, Netherlands. Before and after studies showed a 20% reduction in perceived customer wait time, subsequent to real-time transit information displays being implemented at transit stops. Limited relevance overall, with some application to but very little mention of reliability.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-95 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations El- Geneidy, Horning, & Krizek, 2007 El-Geneidy, A., Horning, J., & Krizek, K. (2007). Using Archived ITS Data to Improve Transit Performance and Management. St. Paul, MN: St. Paul Minnesota Department of Transportation. Retrieved from http://conservancy.umn.edu/ bitstream/handle/11299/559 7/200744.pdf?sequence=1 Using archived ITS data for performanc e and reliability manageme nt Researchers based at the University of Minnesota in the United States, Case study of Metro Transit in Twin Cities, Minnesota, United States Reliability from the passenger perspective is defined as service that can be easily accessed by passengers, arrives predictably, has a short running time, and has low variance in running time. Reliability measures include headway deviation, travel time deviation, and coefficient of variation of running time. Route length, Actual stops, Passenger activity, Traffic signals, Time of day, Driver experience, Driver period of service, Departure delay, Dwell time, Headways, Headway delay, Incidents, Travel direction, Weather, Lift usage Transit signal priority, Driver training, Schedule adjustments, Control strategies, Route modifications, Bus stop consolidation, Real-time operational controls, Real-time passenger information Regression models were run for each of the three measures of reliability used in this study. The significant variables in the model for running time deviation include: distance (-), scheduled stops (+), westbound travel (+), order of first stop (+), afternoon peak (+), actual stops (+), boardings (+), alightings (+), lift use (+), and delay at first stop (+). The significant variables in the model for headway deviation include: distance (+), scheduled stops (-), afternoon peak (-), actual stops (+), boardings (+), and lift use (+). The significant variables in the model for coefficient of variation of running time include: distance (-), westbound travel (+), order of first stop (+), morning peak (+), coefficient of variation of actual stops (+), coefficient of variation of average passenger load (+), coefficient of variation of delay at first stop (-), and coefficient of variation of driver experience (-). It was found that running time deviation is expected to increase by 0.9% for each scheduled stop and 1% for each actual stop. Segment length increases of 1 km are expected to decrease running time deviation by 6%. Afternoon peak periods are expected to increase running time deviation by 5%. Each boarding is expected to increase running time deviation by 0.4%, while each alighting adds 0.2%. Each lift activity increased running time variation by 24% and headway deviation by 3%, in this case. A 1% variation in driver experience was found to lead to a 5% decline in running time coefficient of variation. In this case only about 50% of stops were served, yet even unserved stops were found to increase running times and decrease reliability, so stop consolidation is recommended. This was a case study on one bus line, so the results should not be generalized. Furth & Muller, 2007 Furth, P., & Muller, T. (2007). Service Reliability and Optimal Running Time Schedules. Transportation Research Record: Journal of the Transportation Research Board, No.2034, pp. 55-61. Retrieved from http://www1.coe.neu.edu/~pf urth/Furth%20papers/2007% 20service%20reliability%20 &%20opt%20running%20tim es,%20Furth%20&%20Mulle r%20TRR.pdf Service reliability and scheduling Researchers based at Northeastern University in the United States and Delft University of Technology in the Netherlands Reliability is measured in terms of excess waiting time and potential travel time or buffer time. Scheduling, Variation in dispatching, Variation in running times, Variation in dwell times Holding at time points Researchers state that adding slack time into a schedule helps to improve reliability, but lowers operating speed, which can lead to other negative consequences. Holding at time points was found to greatly reduce excess waiting time, which is defined as the expected departure time minus the 2-percentile departure time at a stop. Travel time buffer, or potential travel time, is another user- focused reliability measure, defined as the 95th percentile arrival time minus the expected arrival time at a stop. Further analysis supports the policy of spreading holding patterns across many stops, rather than concentrating holding control at a few stops. The researchers also recommend setting route running times as the mean plus one standard deviation of uncontrolled running time, and cycle time as the mean plus two to three standard deviations of uncontrolled route running time. Focused on routes with longer headways.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-96 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Liu & Sinha, 2007 Liu, R., & Sinha, S. (2007). Modelling Urban Bus Service and Passenger Reliability. Retrieved from http://eprints.whiterose.ac.uk /3686/ Modelling urban bus service and passenger reliability Researchers based at the University of Leeds in the United Kingdom Reliability is defined as a consistent result from a service over time. Three measures used in this paper include travel time reliability, headway reliability, and passenger wait time reliability. Reduction in per passenger boarding time is proven to be effective for measuring improvement in reliability. Traffic characteristics (traffic composition, day to day and within-day variation in travel demand and traffic congestion levels), Route characteristics (route length, number of lanes, bus stop location, provision of bus lanes, number of intersections, priorities at junctions for buses, on-street parking, passenger volumes, direction of travel, driver behavior), Passenger characteristics (passenger volume at stops, variability in passenger demand, route choice, arrival distributions), Bus operational characteristics (scheduling, staff shortages, fleet availability, vehicle maintenance, fare collection and ticketing system, variability in driver behavior and experience) Reduced dwell time Using simulated results, increased congestion was found to increase the mean and variance of bus journey times. Increased passenger demand appears to improve journey time variability. Used simulation and a single case study, so results may not be accurate or generalizable. Cambridge Systematic s, Dowling Associates, System Metrics Group, & Texas Transportat ion Institute, 2008 Cambridge Systematics, Dowling Associates, System Metrics Group, and Texas Transportation Institute. (2008). NCHRP Report 618: Cost-Effective Performance Measures for Travel Time Delay, Variation, and Reliability. Washington, D.C.: Transportation Research Board. Retrieved from http://onlinepubs.trb.org/onli nepubs/nchrp/nchrp_rpt_618 .pdf Performanc e measures Researchers based in the United States Unreliability is defined as variability in travel time. Measures identified include delay per traveler, travel time, travel time index, buffer index, planning time index, total delay, congested travel, percent of congested travel, congested roadway, and accessibility. Metrics related to travel time reliability included percent variation, buffer index (3 times variance over the mean squared, minus 1), on-time arrival percentage, and misery index (inverse cumulative Gamma distribution for the 85th percentile divided by the mean, minus 1). Traffic volumes, Freight volumes, Incidents, Weather, Road work, Time of day, Season, Road type, Signalization Travel demand management, Increasing capacity at bottlenecks, Reduce probability of incidents, Reduce incident detection, response, and clearance times, Reduce impacts of incidents on capacity, Traveler information systems Considerations for reliability measure selection include: relationship to goals and objectives, clarity of communication, inclusion of urban travel modes, consistency, accuracy, illustration of effects of improvements, application to existing and future conditions, application at several geographic levels, use of person- and goods- movement terms, use of cost-effective methods of data collection or estimation. Recommended reliability measures include buffer index (% extra time to ensure on-time arrival), percent on-time arrival (% of trips defined as on-time), planning time index (dimensionless factor indicating travel time for planning purposes), percent variation (% of average travel time required for on-time arrival of given trip), and 95th percentile travel duration. Little or no discussion of the impacts of improvement strategies. Daskalakis & Stathopoul os, 2008 Daskalakis, N., & Stathopoulos, A. (2008). Users' Perspective Evaluation of Bus Arrival Time Deviations in Stochastic Networks. Journal of Public Transportation, 11(4), 25-38. Retrieved from http://www.nctr.usf.edu/jpt/p df/JPT11-4Daskalakis.pdf User perceptions of bus arrival time deviations in stochastic networks Researchers based at the National Technical University of Athens in Greece, Survey conducted in Athens, Greece A reliable service is defined as having smaller deviations in arrival times. Frequency of service None A significant factor impacting the perception of waiting time deviations is the headway or frequency of service, though this relationship is not linear. Larger headways correspond to greater perceived deviations, but at a diminishing rate. Transit operators wanting to improve passenger wait times and service quality should focus on headway reliability, and then try to shorten headways. This paper had a very narrow focus with little to no elaboration on improvement strategies.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-97 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Fellesson & Friman, 2008 Fellesson, M., & Friman, M. (2008). Perceived Satisfaction with Public Transport Service in Nine European Cities. Journal of the Transportation Research Forum, 47(3). Retrieved from http://journals.oregondigital. org/trforum/article/view/2126 Perceived satisfaction with public transportati on Researchers based at Karlstad University in Sweden, Data collected from respondents in Stockholm, Barcelona, Copenhagen, Geneva, Helsinki, Vienna, Berlin, Manchester, and Oslo None None None Reliability is shown to be a statistically significant factor in perceived customer satisfaction in 4 of the 9 European cities surveyed. Limited relevance overall. Gilliam, Chin, Black, & Fearon, 2008 Gilliam, C., Chin, T. K., Black, I., & Fearon, J. (2008). Forecasting and Appraising Travel Time Variability in Urban Areas. Association for European Transport. Travel time variability Case study of Department for Transport, United Kingdom Travel time variability (TTV), Standard deviation, Coefficient of variation Day-to-day variability, Incidents, Variability due to speed trends, Between-driver variability, Vehicles stopping for non-traffic reasons, data errors None A function was developed for the coefficient of variation of time in a given time slot, based on the congestion index and distance in that time slot. The focus of this paper is on forecasting and appraising variability in urban areas. Kimpel, Strathman, & Callas, 2008 Kimpel, T. J., Strathman, J. G., & Callas, S. (2008). Improving Scheduling through Performance Monitoring. Computer-Aided Systems in Public Transport, 253-280. Retrieved from http://link.springer.com/chapt er/10.1007/978-3-540- 73312-6_13 Improving scheduling through performanc e monitoring Researchers based at Portland State University and TriMet, Case study of TriMet in Portland, Oregon, United States Measures included on-time performance (between 1 minute early and 5 minutes late), headway adherence, heavily loaded trips, lightly loaded trips, and late trip departures. Operator behavior, Passenger demand, Traffic conditions, Lift operation Better management of bus operators, Efficient scheduling This report provides detailed information about the use of data from automated vehicle location and passenger counting systems to improve scheduling and on-time performance at TriMet. The authors recommend using information gathered in this process to help manage bus operator performance, and adjusting schedules as needed on a regular basis to maximize on-time performance. This research focuses on data from Portland, Oregon, which should not be generalized. Lin, Wang, & Barnum, 2008 Lin, J., Wang, P., & Barnum, D. T. (2008). A Quality Control Framework for Bus Schedule Reliability. Transportation Research Part E: Logistics and Transportation Review, 44(6), pp. 1086-1098. Retrieved from http://www.sciencedirect.co m/science/article/pii/S13665 54507001111 Developme nt and demonstrati on of a quality control framework for bus schedule reliability Researchers based at the University of Illinois at Chicago, Case study of the Chicago Transit Authority in Chicago, Illinois, United States Timepoint level running time adherence, headway regularity Bus overtaking policies and practices, Traffic congestion Better operations management, including the practice of using reliability thresholds in a quality control framework to identify routes in greatest need of improvement The running time adherence measures were defined as the average difference between the actual and scheduled running times divided by the scheduled running times. This measure was broken into two categories: Δ% shorter running time and Δ% longer running time. Headway regularity, defined as the average difference between the actual and scheduled headways relative to the scheduled headway, is similarly broken into Δ% shorter headway and Δ% longer headway. Through a case study, found that bus drivers have better control when running ahead of schedule than those running behind due to uncontrollable factors. Buses traveling inbound to downtown Chicago tended to have worse schedule adherence in the morning peak. On-time performance measures may not represent every aspect of reliability. Traffic conditions and environmental factors are not considered. Lyman & Bertini, 2008 Lyman, K., & Bertini, R. (2008). Using Travel Time Reliability Measures to Improve Regional Transportation Planning and Operations. Transportation Research Record: Journal of the Transportation Research Board, No. 2046, pp. 1-10. Retrieved from https://journals.sagepub.com /doi/abs/10.3141/2046-01 Travel time reliability measures Researchers based at Portland State University and URS Corporation in Portland, Oregon, Case study of TriMet in Portland, Oregon, United States Travel time reliability, Travel time, 95th percentile travel time, Travel time index, Buffer index, Planning time index, Congestion frequency Traffic congestion, Peak travel, Time of day, Precipitation, Ice, Fog, Work zones, Incidents, Direction of travel Better planning and operations The authors recommend using travel time reliability measures to set and track system-wide performance goals, evaluate road segments, and prioritize road segments for improvements. Case study data from Portland, Oregon, so may not be generalizable.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-98 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Pangilinan, Wilson, & Moore, 2008 Pangilinan, C., Wilson, N., & Moore, A. (2008). Bus Supervision Deployment Strategies and Use of Real- Time Automatic Vehicle Location for Improved Bus Service Reliability. Transportation Research Record: Journal of the Transportation Research Board, No. 2063, pp. 28-33. Retrieved from https://journals.sagepub.com /doi/10.3141/2063-04 Implementa tion of real- time holding method Researchers based at the Massachusetts Institute of Technology in the United States, Case study of the Chicago Transit Authority in the United States Headway ratio distribution and Headway Coefficient of Variation Vehicle location Holding method based on Turnquist 1982 Method improves headway stability compared to base schedule. Insufficient data for statistical significance Parker, 2008 Parker, D. J. (2008). TCRP Synthesis 73: AVL Systems for Bus Transit Update. Washington, D.C.: Transportation Research Board. Retrieved from https://www.nap.edu/catalog /22019/avl-systems-for-bus- transit-update Automated vehicle location systems Researcher based in Medford, Massachusetts, United States Reliability is defined as on-time performance. None Automated vehicle location systems Automated vehicle location (AVL) is described as a core system (central software used for tracking, communications, and operations management) with various optional features (schedule adherence monitoring, onboard mobile data terminals, managed voice communications, messaging, announcements, passenger counting, and real-time passenger information). An expected benefit of AVL systems is the sharing of schedule adherence information in real-time with dispatchers, operators, and supervisors to improve on-time performance. Systems that include automated passenger counting also offer an affordable way to collect data on boardings and alightings at stops, to help agencies plan for and manage demand. Systems that provide real-time information to transit users and the public may realize the additional benefits of reduced customer stress and enhanced perceived reliability. This study included analysis of transit data from before and after implementation of AVL systems, which revealed improvements in schedule adherence, transfer coordination, dispatcher control, schedule adherence monitoring, incident response, bus tracking, and driver performance monitoring. Collected data can also be used to inform schedule adjustments and service planning. Focused on automated vehicle location, rather than reliability. Perk, Flynn, & Volinski, 2008 Perk, V., Flynn, J., & Volinski, J. (2008). Transit Ridership, Reliability and Retention. Tampa: National Center for Transit Research, University of South Florida. Retrieved from http://www.nctr.usf.edu/pdf/7 7607.pdf Transit ridership, reliability, and retention Researchers based at the University of South Florida in the United States Reliability is considered in terms of travel time reliability, which is measured as on-time performance. Fleet maintenance, Route design, Scheduling Bus dispatching systems, Real-time customer information, Parking restrictions, Prohibiting left turns, Simplifying route structures, Avoiding long routes, Reducing the number of stops, Shortening dwell times, Transit signal priority, Bus lanes The relationship between travel time reliability and transit ridership seems apparent (in that there is one), but the exact nature of that relationship is not known. The authors recommend that transit agencies continue moving toward more customer-oriented reliability measures. They also note that values placed on reliability, travel time, and other characteristics of bus service may vary among individuals, as well as regions. Schedule adherence is said to be the most important reliability measure for occasional transit users, timed transfers, and infrequent service routes. For bus services with headways of 10 minutes or less, headway delay is a more appropriate measure. A couple of case studies of reliability improvements were noted, including a suburban bus service that adjusted headways on one of its routes to add two more trips per day while roughly maintaining service frequency. This Chicago-area agency, PACE, noted an almost 22% increase in ridership in one year on this route after addressing reliability concerns, as opposed to a system-wide average increase of 5%. On the other hand, Sacramento Regional Transit reported ridership decline even with an increase in on-time performance (roughly 65% to 85-90%). The researchers mostly use evidence from case studies and surveys, as well as a literature review, as the basis of their arguments.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-99 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations van Oort & van Nes, 2008 van Oort, N., & van Nes, R. (2008). Improving Reliability in Urban Public Transport in Strategic and Tactical Design. Presented at 87th Annual Meeting of the Transportation Research Board, Washington, D.C. Retrieved from http://www.goudappel.nl/me dia/files/uploads/14_Van_Oo rt_TRB_2008_Improving_reli ability_Revised.pdf Reliability improve ments through design Researchers based at HTM Urban Public Transport and Delft University of Technology in the Netherlands, Case study in The Hague, The Netherlands Reliability is defined as the match between planning and operations, with the primary measure given as schedule adherence. Traffic, Weather, Variable number of travelers, Driver behavior Service planning for reliability, Conditional priority, Central dispatching Variability in bus travel times mostly impacts waiting times, and can lead to lower customer comfort, certainty, appreciation, and ridership. In addition to operational strategies, planning plays an important role in schedule adherence. In timetable planning, one strategy is for driving time to be based on a high percentile value of the driving time distribution, such as the 85th percentile, to ensure good achievability, However, this must be balanced against the risk of buses running early and causing extra waiting time for passengers. Overall, using the average driving time is the recommended strategy for balancing these considerations, as it is expected to result in the least amount of extra waiting time. Line length should not be too long, as reliability is found to decrease as buses travel further from their origins, but this consideration should be balanced against the need for users to transfer when line lengths are shortened. Lines that share route regiments or may otherwise impact each other should be coordinated, when possible, to improve service reliability for all interacting lines. Stop spacing will also impact reliability, which must balance many considerations against each other. The authors recommend spacing stops far enough apart that passengers are expected to be at every stop, nearly all of the time. This strategy leads to less variability bus service, as the bus is expected to stop and pick up passengers at each stop for almost every trip. All data from The Hague in the Netherlands, so may not be indicative of other places and transit systems. Zhang, Shen, & Clifton, 2008 Zhang, F., Shen, Q., & Clifton, K. (2008). Examination of Traveler Responses to Real-Time Information about Bus Arrivals using Panel Data. Transportation Research Record: Journal of the Transportation Research Board, No. 2082, pp.107- 115. Retrieved from http://www.connexionz.us/w p-content/uploads/products- services/resources/08-2781- UMD-Connexionz.pdf Customer response to real-time bus arrival information Researchers and case study based at the University of Maryland in College Park, Maryland, United States Reliability is defined as on-time performance. None Real-time transit information This study showed positive psychological impacts of real-time transit information on transit riders, but no significant relationship between this dummy variable and reported perception of on-time performance. Limited relevance to or discussion of reliability. Vincent, 2008 Vincent. Measurement valuation of public transport reliability. In Land Transport New Zealand Research Report 339, 2008. Model reliability valuation Land Transport New Zealand Value of reliability Mode, trip purpose, geography None There is no appreciable difference in reliability value by geography and mode. The trip purpose, however, affects the stated importance of reliability. None Furth and Muller, 2009 Furth, P.G. and T.H. Muller. Optimality conditions for public transport schedules with timepoint holding. In Journal of Public Transport, Vol. 1, Issue 2, 2009, pp. 87–102. Optimal conditions for timepoint holding Northeastern University, and Delft University passenger travel time Number of timepoints, and total buffer time Optimal timepoint number and buffer time The benefits to reliability increase with the number of timepoints with diminishing returns None

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-100 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Dorobritz, Luthi, Weidmann and Nash, 2009 Dorbritz, R., M. Luthi, U. Weidmann, and A. Nash. Effects of Onboard Ticket Sales on Public Transport Reliability. In Transportation Research Record: Journal of the Transportation Research Board, No. 2110, 2009, pp. 112–119. https://journals.sagepub.com /doi/10.3141/2110-14 Measuring impact of fare payment on reliability Swiss Federal Institute of Technology, and ETH Zurich Dwell time fare payment none Although fare payment only takes up to 20% of total trip time, the variability in fare payment time causes substantial variability on the route. None Berkow, El- Geneidy, Bertini and Crout 2009 Berkow, M., A. El-Geneidy, R. Bertini, and D. Crout. Beyond Generating Transit Performance Measures: Visualizations and Statistical Analysis with Historical Data. Transportation Research Record: Journal of the Transportation Research Board, No. 2111, 2009, pp. 158–168. https://journals.sagepub.com /doi/abs/10.3141/2111-18 Visualizatio n of historical data Portland State University, McGill University, TriMet Travel time Passenger demand, fare payment, lift usage None The tools presented can help transit planning and operations. None Chen, Yu, Zhang, & Guo, 2009 Chen, X., Yu, L., Zhang, Y., & Guo, J. (2009). Analyzing Urban Bus Service Reliability at the Stop, Route, and Network Levels. Transportation Research Part A, pp. 722-734. Retrieved from http://www.sciencedirect.co m/science/article/pii/S09658 56409000688 Analyzing urban bus service reliability in Beijing at the stop, route, and network levels Researchers based at Beijing Jiaotong University, Texas Southern University, and Beijing Transportation Research Center in China, Case study of Beijing Public Transport Holdings, Ltd and Beijing Xianglong Bus CO. Ltd in Beijing, China Existing definition is on-time performance. New proposed parameters are Punctuality Index based on Routes (PIR), Deviation Index based on Stops (DIS), and Evenness Index based on Stops (EIS). Route length, Headway, Distance from the stop to the origin terminal, and Exclusive bus lanes Shorter route lengths, Exclusive bus lanes, Transit signal priority This paper proposed and tested three new parameters for measuring urban bus service reliability: Punctuality Index based on Routes (PIR), Deviation Index based on Stops (DIS), and Evenness Index based on Stops (EIS). "PIR is defined as the probability that a bus can arrive at the terminals in a given time period." "DIS is defined as the possibility that a bus will adhere [to] the headway between successive buses at each stop within a given time period." EIS is defined as one minus the fluctuation index based on stops, which is a measure of unreliability at each stop. The effects of several improvement measures on these three parameters are also examined. Shorter route lengths, specifically those less than 30 km, showed better reliability than longer routes for all three measures. Lengthening headways had a positive impact on the DIS reliability measure, but a negative impact on the EIS measure. Reliability declined overall with increased distance from the origin terminal. Bus services using exclusive bus lanes, as well as those with transit signal priority, showed improved reliability compared to services without these amenities. No before and after comparisons were performed. Factors analyzed were not comprehensive.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-101 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Lin & Ruan, 2009 Lin, J., & Ruan, M. (2009). Probability-Based Bus Headway Regularity Measure. IET Intelligent Transport Systems, 3(4), 400-408. Retrieved from http://digital- library.theiet.org/content/jour nals/10.1049/iet- its.2008.0088 This paper provides a new performanc e metric to model headway reliability Researchers based at the University of Illinois in the United States Cumulative distribution function of headway Dispatching headway and dwell times at intermediate stops None Headway variability at any stop is an increasing function of the variability at departure. The cumulative distribution of headways is expressed as a function of dwell time at intermediate stops. The dwell time of different buses at different stops are assumed to be independent and identically distributed. This determination of dwell time overlooks the unstable headway dynamics that lead to bus bunching. Uniman, 2009 Uniman, D. L. (2009). Service reliability measurement framework using smart card data: application to the London Underground. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/ 52806 Measuring service reliability using smartcard data Researcher based at Massachusetts Institute of Technology in the United States, Case study of the London Underground in London, England, United Kingdom Metrics defined for measuring reliability include reliability buffer time, excess reliability buffer time, and percentage of unreliable journeys. Recurrent reliability factors (characteristics of service, journey length, scheduled headways, interchanges), Non-recurrent reliability factors (operations control interventions, incident-related disruptions, seasonality) None The author of this thesis used ordinary least squares regression analysis to estimate that: for every additional minute that a journey typically takes the reliability buffer time increases by 7 seconds; when a trip involves transferring variability increases by about 56 seconds; and when an incident occurs reliability buffer time increases by about 10 minutes above normal peak period performance. Similar models were estimated using two feasible generalized least squares approaches. Overall findings were as follows. Incidents have a large impact on service reliability, which may be underestimated if only average metrics are considered. The occurrence and impacts of severe disruptions can be tracked through passenger travel times collected through smartcard systems. The proposed framework can be used to enhance reliability information provided to passengers, which can partially mitigate the impacts of uncertainty on service quality. Findings based on London Underground, not bus service. Boyle, Pappas, Boyle, Nelson, Sharfarz, and Benn, 2009 Boyle, D., J. Pappas, P. Boyle, B. Nelson, D. Sharfarz, and H. Benn. TCRP Report 135: Controlling System Costs: Basic and Advanced Scheduling Manuals and Contemporary Issues in Transit Scheduling. Transportation Research Board of the National Academies, Washington, D.C., 2009. Scheduling guide Dan Boyle and Associates, John Pappas Transit Consultant, Phillip Boayle and Associates, and Nelson/Nygaard On-time performance Running time Including buffer time the schedules None Only focuses on schedules

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-102 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations van Oort and van Ness, 2009 van Oort, R. and N. van Nes. Line Length Versus Operational Reliability: Network Design Dilemma in Urban Public Transportation. In Transportation Research Record: Journal of the Transportation Research Board, No. 2112, 2009, pp. 104–110. https://journals.sagepub.com /doi/abs/10.3141/2112-13 Tradeoff between route length and transfers HTM Urban Transport Company, and Delft University Travel time reliability Route length Split long route into smaller segment to avoid potential delays It is sometimes good to split a route in several segments. In those cases, the split should take place at stops where few passengers ride through. Based on limited definition of reliability. APTA BRT Operations Working Group, APTA Standards Developme nt Program Recommen ded Practice: Bus Rapid Transit Service Design, 2010 APTA BRT Operations Working Group. (2010). APTA Standards Development Program Recommended Practice: Bus Rapid Transit Service Design. Washington, D.C.: American Public Transportation Association. Retrieved from http://www.apta.com/resourc es/standards/Documents/AP TA-BTS-BRT-RP-004-10.pdf Bus rapid transit service design Researchers based in the United States and Canada Reliability is defined as on-time performance (schedule adherence and headway adherence) Service levels (frequencies, span of service), Stop spacing, Exclusivity of running way, Station design, Vehicle type, Fare collection methods, Intelligent transportation systems, Ridership, Transfer convenience, Traffic congestion, Route characteristics, Dwell times Effective routing and service structure, Peak travel times and directions, Express / limited-stop service, Prepaid off-board fare collection, Exclusive right-of-way, Intelligent transportation systems, Far-side stop placement, Transit signal priority This report serves as a guide for designing and implementing high- performance (in terms of speed and reliability) bus service, or bus rapid transit, including routing, scheduling, and operations strategies. Described many strategies, but did not discuss the effectiveness of any particular strategy for improving reliability. APTA BRT Operations Working Group, Operating a Bus Rapid Transit System, 2010 APTA BRT Operations Working Group. (2010). Operating a Bus Rapid Transit System. American Public Transportation Association. Retrieved from http://www.apta.com/resourc es/standards/Documents/AP TA-BTS-BRT-RP-007-10.pdf Bus rapid transit operations Researchers based in the United States None Transit priority strategies, Driver behavior, Operating policies and practices, Vehicle and stop configuration, Management, Frequency of service, Fare collection methods Driver training, flexibility, and incentives, Performance goals, Transit signal priority, Block signaling, Queue jump lanes, Intelligent transportation systems, Lane usage, Exclusive and limited access lanes, Level-boarding stations, Route monitoring and supervision, All-door boarding, Real-time operations control, Communication systems, Transit connection protection, Public education This report provides a thorough description of strategies to improve bus service as part of a bus rapid transit system, including: driver scheduling and incentives, driver training and development, service and performance goal setting, transit signal priority, block signaling, queue jump lanes, operating rules, exclusive lanes, shoulder lanes, the ability to pass, high-occupancy vehicle lanes, reversible lanes, contra-flow lanes, operator flexibility, level boarding, operating policies, proactive management, route monitoring, frequency of service, vehicle configuration, fare collection policies, communications systems, transit connection protection, and more. Described many strategies, but did not discuss the effectiveness of any particular strategy for improving reliability. Bubna, Brunner, Gangloff, Advani, & Prasad, 2010 Bubna, P., Brunner, D., Gangloff, J. J., Advani, S. G., & Prasad, A. K. (2010). Analysis, Operation and Maintenance of a Fuel Cell / Battery Series-Hybrid Bus for Urban Transport Applications. Journal of Power Sources, 195(12), 3939-3949. Retrieved from http://www.sciencedirect.co m/science/article/pii/S03787 75309023428 Fuel cell / battery series- hybrid buses for urban transit application s University of Delaware in Delaware, United States None Vehicle performance, durability, maintenance None Despite some troubleshooting of the fuel cell stack early on, these battery-heavy fuel cell buses showed similar maintenance requirements as their more conventional diesel counterparts. Limited relevance overall, with some application to but very little mention of reliability.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-103 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Carrel, Mishalani, Wilson, Attanucci, & Rahbee, 2010 Carrel, A., Mishalani, R. G., Wilson, N. H., Attanucci, J., & Rahbee, A. B. (2010). Decision Factors in Service Control on a High-frequency Metro Line and Their Importance in Service Delivery. Transportation Research Record: Journal of the Transportation Research Board, No. 2146, pp. 52-59. https://journals.sagepub.com /doi/abs/10.3141/2146-07 This paper evaluates the role of factors affecting reliability from an agency's perspective . Researchers based at the University of California, Berkeley and Massachusetts Institute of Technology in the United States, Case study of the London Underground Ltd. In the United Kingdom Institutional flow of operations in case of unpredicted event or disruption. Operations, planning, infrastructure, labor, and institutional framework. Extending traditional performance metrics to include organizational failures that lead to reductions in service quality. Many failures in operations can be traced back to organizational problems of institutional processes. Analysis is superficial, lacks data. Danaher, 2010 Danaher, A.R. (2010). TCRP Synthesis 83: Bus and Rail Transit Preferential Treatments in Mixed Traffic. Washington, D.C.: Transportation Research Board of the National Academies. Retrieved from http://www.tcrponline.org/PD FDocuments/tsyn83.pdf Transit preferential treatments Researcher based at Parsons Brinckerhoff in the United States, Four case studies in the United States On-time performance, Travel times Preferential treatments, Traffic, Signals, Ridership, Right turn volumes, On-street parking, Access density (driveways per mile), Right-of- way, Stop locations, Boardings, Stop spacing Median transitways, Exclusive transit lanes, Stop modifications, Transit signal priority, Special signal phasing, Queue jump lanes, Curb extensions The results of a survey conducted for this report revealed that transit signal priority (TSP) is the most popular preferential treatment on urban streets, as well as the lack of standard warrants for when to apply certain treatments. It was also found that most transit agencies at the time did not have comprehensive programs for preferential treatments. A majority of transit agencies did report intergovernmental agreements with traffic engineering organizations to work together to develop and operate preferential treatments in the agency's service area. Twelve traffic engineering organizations also provided insights, which indicated a preference for transit signal priority, queue jump lanes, exclusive transit lanes, and greater stop spacing over median transitways, special signal phasing, and curb extensions. All in all, the analysis revealed that the greatest benefit is typically realized from systematic application of one or more preferential treatments along a corridor, with median transitways, exclusive lanes, and transit signal priority estimated to provide the most significant positive results. Examples of bus lanes were shown to represent reliability improvements from arterial bus lanes, including a 12-27% improvement in coefficient of variation on Wilshire Boulevard in Los Angeles and a 57% improvement on Madison Avenue in New York City. TSP was also shown to provide reliability benefits in several cases, including reductions in travel times ranging from 0-38%, signal delay reductions between 20- 57%, and up to 35% reductions in travel time variability. Significant increases in delays to general traffic were noted in a few cases of TSP implementation. Bus queue jump lanes have been demonstrated to result in 5-15% reductions in travel time for buses through intersections. Curb extensions and stop consolidation were also shown to help reduce bus running times. This study focused on transit in the United States only. Dell'Olio, Ibeas, & Cecin, 2010 Dell'Olio, L., Ibeas, A., & Cecin, P. (2010). Modelling User Perception of Bus Transit Quality. Transport Policy, 17(6), 388-397. Retrieved from http://www.sciencedirect.co m/science/article/pii/S09670 70X10000557 Modelling bus transit quality from the perception of users Researchers based at the University of Cantabria in Spain, Case study of a public transport bus service in Santander, Spain None Factors impacting perceived quality of transit service were identified as waiting time, journey time, access time, safety in the vehicle, comfort when accelerating or decelerating, comfort during the journey, deviation from the optimal route, cleanliness of the vehicle, fare price, vehicle quality, vehicle reliability, and kindness of the bus driver. None Reliability of service and waiting time were the top two factors impacting perceived transit service quality with a combined weight of 51.4% of the overall service score, according to the researchers' survey and modeling results. Service reliability was the most important variable impacting perception of service quality for frequent users, women, those without a car, those without a driver’s license, people between the ages of 25 and 54, and those with incomes below 900 euros per month, as well as the population of interviewees overall. Survey results are from a single city in Spain, and may not be generalizable to the global population.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-104 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Ehrlich, 2010 Ehrlich, J. E. (2010). Applications of Automatic Vehicle Location systems towards improving service reliability and operations planning in London. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/ 60799 Automated vehicle location systems Researcher based at Massachusetts Institute of Technology in the United States, Case study of Transport for London in London, England, United Kingdom Reliability is defined as invariability of service attributes that influence travelers' and transportation providers' decisions (taken from Abkowitz et al., 1978). Measures include wait time, excess waiting time, in-vehicle travel time, journey time, excess journey time, and reliability buffer time. The dependent variable used to measure reliability in the case study was the actual wait time divided by the scheduled wait time. Driver controller ratio, Lost mileage, Frequency of breakdowns, Vehicle attributes, Location of route, Bus priority measures, Length of route, Scheduled peak headway, Variability in passenger volumes, Traffic congestion, Weather, Congestion charging, Recession, Period, Socio- economic factors, Automated vehicle location systems, Contract changes, New operators, Crossrail construction Service restoration actions include jumping, reassigning, shifting schedule timeframes, eliminating trips, adding trips, modifying headways, modifying scheduled running times, holding at bus stops, changing buses, passing, changing drivers, detouring, short- turning, extending the trip, and modifying the trip. From the case study, congestion charging and quality incentive contracts for private bus companies may have greatly improved bus service reliability. Guiding constraints for determining which restoration actions to take in the case of unreliability were identified as: demand, running times, topography, and route attributes. In London, congestion charging was found to have produced a 21% reduction in-vehicle traffic, while boosting bus network ridership by 6% during congestion charging hours. London bus reliability measures include excess waiting time for high-frequency routes (headways of 12 minutes or less), percent on time for low-frequency routes (headways of 15 minutes or more), and percent lost mileage. Excess waiting time is the actual waiting time (AWT) minus the scheduled waiting time (SWT), where waiting time is calculated as half of the expected headway times the sum of one and the headway coefficient of variation squared. On-time performance is defined as running between two minutes early and five minutes late. Percent lost mileage is defined as the percent of scheduled route miles that were not run over a given time period. Other reliability measures used informally include change of waiting longer than 10 minutes and the percentage of long gaps (headways greater than 4 times the scheduled wait time). Congestion charging and quality incentive contracts may have led to the most significant reliability improvements in London bus services during the study period. However, it is not clear whether or not the automated vehicle location system, iBus, had a positive impact on reliability. The factors that are most highly correlated with AWT/SWT include percent lost mileage due to traffic (0.64), morning peak buses per hour (0.56), ridership (0.56), ridership per kilometer (0.51), central London route length (0.43), congestion charging zone route length (0.42), vehicle length (0.28), routes per workstation (-0.28), and bus priority measures (0.28). A regression analysis of the highly correlated variables showed greatest correlation between AWT/SWT and lost mileage due to traffic (+), precipitation (+), ridership (+), and route length (-). iBus was also shown to have a small effect on improving reliability in one model, but was not statistically significant in a second model with many more variables. The author posits that automated vehicle location systems must be used effectively by transit agencies and operators for benefits to be realized. Three new reliability measures are proposed, including journey time, excess journey time (the difference between the median journey time and the scheduled journey time), and reliability buffer time (the difference between the 95th percentile journey time and the median journey time), which seem to be valuable reliability measures, especially from the passenger perspective. Details of the regression analysis are thin, and the results are from London only, so should not be generalized.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-105 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Fijalkowski, 2010 Fijalkowski, J. A. (2010). Making Data Matter: The Role of Information Design and Process in Applying Automated Data to Improve Transit Service. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/ 60800 Automated data collection and application Researcher based at Massachusetts Institute of Technology in the United States, Case studies of the Chicago Transit Authority in Chicago, Illinois and the Massachusetts Bay Transportation Authority in Boston, Massachusetts, United States Reliability performance metrics include percent of buses on-time, schedule run time minus actual run time, average and total excess wait time, excess journey time, reliability buffer time, percent of big gaps in service, and percent of bunched intervals. Scheduling, Routing, Span of Service, Transfers, Major trip attractors, Traffic conditions, Roadway geometry, Service changes Automated data collection, Service adjustments, Increasing frequencies, Increase half cycle time, Enforce on-time departures at terminals, Implement a mid-point holding policy, Implement prepaid boarding scheme, Reroute service around congested areas. Reliability metrics used by the Chicago Transit Authority at the time of this study were: adherence to scheduled headway for trips with headways less than 15 minutes, adherence to scheduled arrival times for trips with headways of 15 minutes or more, excess passenger wait times for trips with headways less than 15 minutes, and bus running times. Successful headways are defined as within three minutes for scheduled headways of 10 minutes or less, within five minutes for scheduled headways between 10-15 minutes, and on-time (between one minute early and two minutes late) for scheduled headways of 15 minutes or more. Reliability metrics used by the Massachusetts Bay Transportation Authority at the time of the study were schedule adherence for low-frequency service and headway adherence for high-frequency service. The author critiques these standards for having thresholds that may be improper and arbitrary, and for being operator-focused rather than passenger-focused. He recommends incorporating the degree of deviation (rather than just considering whether a bus is on time or not as a binary indicator) and passenger loads into reliability metrics. The major conclusions of this thesis are as follows. Automatically collected data can be utilized to track performance from the customer perspective, as well as the operator perspective, which can help to identify performance issues. Automatically collected data can be used to improve service planning processes by providing detailed information for every route and stop in a transit network, allowing for periodic adjustments to be made and their impacts to be monitored. Route information is critical to the service planning process. Automated data collection systems can help to lower the cost of data collection, but may require additional planning staff and resources to make use of the copious data. Little or no discussion of the impacts of reliability improvement strategies. Frumin, 2010 Frumin, M. S. (2010). Automatic data for applied railway management: passenger demand, service quality measurement, and tactical planning on the London Overground Network. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/ 61512 Automated data collection and application Researcher based at Massachusetts Institute of Technology in the United States Expected wait is defined as half the expected headway times the sum of one and the headway coefficient of variation squared. Service headway, Passenger incidence behavior None The author calls for unified estimation of excess journey time measures using automated passenger counting technologies. Focused on railway transit, not bus. Harrington- Hughes & Associates, Inc., 2010 Harrington-Hughes & Associates, Inc. (2010). TCRP Research Results Digest 96: Managing Increasing Ridership Demand. Washington, D.C.: Transportation Research Board. Retrieved from https://www.nap.edu/catalog /14418/managing- increasing-ridership-demand Managing increasing ridership demand Researchers based in the United States, Case studies from South America None Ridership demand Peak / off-peak fares, Electronic fare collection, Low-floor buses, Regular vehicle maintenance and replacement, Adding new service, Increasing fleet size, Surveillance systems, Control centers Several case studies demonstrate successful strategies for improving transit reliability across several modes and geopolitical boundaries. In at least one of these cases, improvements to service reliability have led to increasing ridership demand. The cyclical nature of the relationship between service reliability and ridership is important to note, as it gives rise to the need to periodically track reliability and adjust service to accommodate changes in transit user demand. Limited relevance to or discussion of reliability.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-106 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Ma, Yang, & Liu, 2010 Ma, W., Yang, X., & Liu, Y. (2010). Development and Evaluation of a Coordinated and Conditional Bus Priority Approach. Transportation Research Record: Journal of the Transportation Research Board, No. 2145, pp. 49-58. Retrieved from https://journals.sagepub.com /doi/abs/10.3141/2145-06 Evaluation of a proposed coordinated and conditional bus priority approach strategy Researcher based at Tongji University in China and at the University of Wisconsin Milwaukee in the United States, Case study in China Delay, Headway deviation Traffic congestion, Traffic signals Coordinated and conditional bus priority A model was developed to generate the optimal combination of priority strategies for intersection groups, to keep real delay in line with permitted delays. The resulting coordinated and conditional bus priority approach was demonstrated in the field, resulting in significant reductions in bus delays (35% per bus compared to no priority) and headway deviations (62% reduction compared to no priority, 51% reduction compared to unconditional priority) with minimal impacts on general traffic. The proposed strategy was deemed to be useful for improving bus service reliability. While a 9% increase in motor vehicle delay was noted, this was small in comparison to the unconditional priority approach. The field study is based on a since case. Mandelzys & Hellinga, 2010 Mandelzys, M., & Hellinga, B. (2010). Identifying Causes of Performance Issues in Bus Schedule Adherence with Automatic Vehicle Location and Passenger Count Data. Transportation Research Record: Journal of the Transportation Research Board, No. 2143, pp. 9-15. Retrieved from https://journals.sagepub.com /doi/abs/10.3141/2143-02 Identifying causes of unreliability Researchers based at the University of Waterloo in Canada, Case study in Waterloo, Ontario, Canada Schedule adherence Travel time causes (traffic congestion, weather, signals, unscheduled stops), Dwell time causes (passenger activity and demand, traffic volumes, lift use), Upstream causes (deviations at previous stops) Use of AVL and APC data for planning and operational improvements A process for using AVL and APC data to identify causes of schedule deviations is proposed and tested. Little or no discussion of impacts of potential improvements. Mazloumi, Currie, & Rose, 2010 Mazloumi, E., Currie, G., & Rose, G. (2010). Using GPS Data to Gain Insight into Public Transport Travel Time Variability. Journal of Transportation Engineering, 136, 623-631. Travel time variability Researchers based at Monash University in Australia Travel time variability (standard deviation of travel times and 90th minus 10th percentile travel times) Peak period travel, Departure delay at origin, Land use, Route length, Number of signals, Number of stops, Rain None The researchers described a model for predicting bus journey travel time variability in which length of route was an independent variable, as were number of stops and number of signal controlled junctions. They also identified signalized junctions as having a statistically significant impact on travel time variability. It is not clear from the work whether for buses the coefficient of variation will rise or fall with distance. Tetreault & El- Geneidy, 2010 Tetreault, P. R., & El- Geneidy, A. M. (2010). Estimating Bus Run Times for New Limited-Stop Service Using Archived AVL and APC Data. Transportation Research Part A: Policy and Practice, 44(6), pp. 390-402. Retrieved from http://www.sciencedirect.co m/science/article/pii/S09658 56410000480 Estimating bus running times for a new limited-stop service Researchers based at GENIVAR and McGill University in Canada, Case study of STM in Montreal, Quebec, Canada Reliable transit service is defined as that with short wait times and less variation. Passenger loads and activity, Low-floor buses, Day of the week, Time of day, Direction of travel, Number of stops, Weather, Delay at the beginning of a route Rear-door boarding, Low-floor buses, Limited- stop service The results of this research indicate that implementation of limited- stop service along an already heavily used bus route can improve running time for the existing bus service as well as the new limited- stop service. This was a case study on one bus line, so the results should not be generalized.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-107 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Uniman, Attanucci, Mishalani, & Wilson, 2010 Uniman, D. L., Attanucci, J., Mishalani, R. G., & Wilson, N. H. (2010). Service Reliability Measurement Using Automated Fare Card Data: Application to the London Underground. Transportation Research Record: Journal of the Transportation Research Board, 92-99. This paper defines new performanc e metrics using an automated fare collection system to analyze the impact of service disruptions Researchers based at the Massachusetts Institute of Technology and Ohio State University in the United States, and the Center for Sustainable Transport in Mexico Excess Reliability Buffer Time: Amount of excess travel time that passengers need to budget in order to arrive on time at their destination 95% of the time. Defective fleet in service, passenger crowding, track power failure & customer disruption None The current paradigm of averaging performance excludes key component of quality of service. Since passengers budget time to arrive at their destinations at a certain time, performance measures must be expressed in percentiles to reflect reliability. The metrics allowed to identify non-recurring disruptions as a major cause of unreliability in the route studied. The performance metric used is disconnected from the agency's operations. van Oort, Wilson, & van Nes, 2010 van Oort, N., Wilson, N., & van Nes, R. (2010). Reliability Improvement in Short Headway Transit Services: Schedule- and Headway-based Holding Strategies. Transportation Research Record: Journal of the Transportation Research Board, No. 2143, pp. 67-76. Retrieved from https://journals.sagepub.com /doi/10.3141/2143-09 The schedule- based and headway- based methods are compared on a light- rail route in The Hague, The Netherland s. HTM Transit Company, Massachusetts Institute of Technology, Delft University of Technology Both methods are examined in terms of additional travel time, and vehicle crowding Number of control points, maximum holding time, and holding factor Schedule-based and headway-based holding The results indicate that, assuming random arrival times, a schedule-based holding strategy is more effective in terms of minimizing total travel time. There was buffer time in the schedule, but none in the headways, which put the latter method at a disadvantage. Eboli & Mazzulla, 2011 Eboli, L., & Mazzulla, G. (2011). A Methodology for Evaluating Transit Service Quality Based on Subjective and Objective Measures from the Passenger's Point of View. Transport Policy, 18(1), 172-181. Retrieved from http://www.sciencedirect.co m/science/article/pii/S09670 70X10000958 Evaluating transit service quality from the passenger' s point of view Researchers based at the University of Calabria in Italy Reliability is defined as the ability of the transit system to adhere to a timetable, as well as the ability of transit vehicles to depart or arrive on time. None None The authors propose two measures of transit service reliability from the passenger perspective: reliability of runs that come on schedule and punctuality (runs that come on time). Runs that come on schedule was calculated as the ratio of runs executed versus scheduled in a period. Punctuality of runs that come on time was equivalent to on-time performance (runs up to 1 minute early or 5 minutes late). Limited relevance overall, with some application to but very little mention of reliability.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-108 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations El- Geneidy, Horning, & Krizek, 2011 El-Geneidy, A., *Horning, J., & Krizek, K., (2011). Analyzing transit service reliability using detailed data from automatic vehicular locator systems. Journal of Advanced Transportation, 45(1), 66-79. Retrieved from http://tram.mcgill.ca/Researc h/Publications/AVL_reliabilit y_MN.pdf Analyzing transit service reliability using AVL data Researchers based at McGill University in Canada and Cambridge Systematics and the University of Colorado Denver in the United States, Case study of Metro Transit in Twin Cities, Minnesota, United States Run time, run time deviation, headway deviation, coefficient of variation of run time, schedule adherence, and reliability are measured at the time point segment and at the route level. Number of scheduled stops, Direction of travel, Peak travel periods, Number of actual stops, Boardings, Alightings, Lift use, Passenger load, Delay at first stop, Headway delay at first stop, Driver experience Schedule revisions to improve run time and schedule adherence, Stop consolidation The results presented in this paper are very similar to those presented in the 2007 paper by the same authors. Their run time model was used to estimate that each scheduled stop adds 5 seconds to the run time regardless of whether or not the bus actually stops. The order of the starting time point for a segment is estimated to add 0.17 seconds to the run time, such that the run time between the first two time points should be 13 seconds faster than the run time along the last segment. Morning peak run times were found to be 17 seconds faster than off-peak trips, and 64 seconds faster than afternoon peak trips. Actual stops are estimated to add 11 seconds to bus trip run times, while each boarding is estimated to take 13 seconds and each alighting adds 6.5 seconds to the run time, but it should be noted that other aspects of dwell time may be incorporated in these variables. Regarding headway deviation, the model estimated lift activity to have the strongest effect, with an estimated 3% increase in headway deviation. Coefficient of variation of run time was found to decrease with longer distances between time points. This metric was also estimated to increase for buses traveling away from downtown. Morning peak buses were estimated to have higher variability in run time, compared to off-peak buses. Actual stops increasing in variation by 1% was shown to produce an estimated 5% increase run time variation between points. Variance in passenger load also added 11% to the variability of running times. Variance in delay at the beginning of a route, however, was found to reduce run time variation. Lastly, a 1% variation in driver experience was estimated to result in a 5% decline in run time coefficient of variation. This presents data analysis results from a single transit agency. Guo, Luo, Lin, & Feng, 2011 Guo, G., Luo, H., Lin, X., & Feng, C. (2011). Headway- based Evaluation of Bus Service Reliability. 14th International IEEE Conference on ITS (pp. 1864-1868). IEEE. Retrieved from http://ieeexplore.ieee.org/xpl /articleDetails.jsp?arnumber =6082998 Headway- based evaluation of bus service reliability Researchers based at Tongji University in China Headway average values, variances, and ranges Traffic congestion, Weather, Passenger demand, Driver behavior None The authors propose that transit service reliability can be measured in three ways using archived headway data. Average values can be used to describe the average headways, which can be compared with published or scheduled headways. Headway variance reflects the amount of fluctuation among headways. Headway deviation percentage can indicate the range of headway fluctuations. Limited detail and depth. Surprenant -Legault & El- Geneidy, 2011 Surprenant-Legault, J., & El- Geneidy, A. (2011). Introduction of a Reserved Bus Lane: Impact on Bus Running Time and On-Time Performance. Transportation Research Record: Journal of the Transportation Research Board, No. 2218, pp. 10-18. Retrieved from https://journals.sagepub.com /doi/10.3141/2218-02 Impacts of the introduction of a reserved bus lane on bus running time and on-time performanc e. Researchers based at McGill University in Canada, Case study of STM in Montreal, Quebec, Canada On-time performance (also considered running time) is defined as the probability of being between 1 minute early and 3 minutes late. Delay at the start of the trip, Snow precipitation, Passenger activity, Rear-door boardings, Smartcard use, Stop spacing, Right-of-way, Traffic conditions, Direction of travel Limited-stop service, Reserved bus lane (right turning vehicles allowed) The reserved lane was found to decrease the odds of being late by 65%. Limited-stop service decreased the odds of being late by 66%. Northbound travel, which typically had more traffic, was found to increase the odds of being late by 75%. The introduction of smartcards for fare payment increased the odds of being late by 69%. Each second of delay at the beginning of a trip increases the odds of being late by 0.9%. Each centimeter of snow increases the odds of being late by 20%. Each passenger's activity increases the odds of being late by 2%, and each rear boarding passenger increases the odds of being late by 6%. This was a case study on one bus line, so the results should not be generalized.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-109 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Trompet, Liu, & Graham, 2011 Trompet, M., Liu, X., & Graham, D. (2011). Development of Key Performance Indicator to Compare Regularity of Service between Urban Bus Operators. Transportation Research Record: Journal of the Transportation Research Board, No. 2216, pp. 33-41. https://journals.sagepub.com /doi/10.3141/2216-04 Bus regularity of service performanc e indicators Researchers based at Imperial College London in London, England, United Kingdom, Case studies from around the world Regularity is measured in terms of wait assessment, service regularity, excess wait time, and standard deviation of the difference between the actual and scheduled headways None None All four measures analyzed proved useful in some way, but excess wait time was the preferred method overall, due to its applicability to the customer. Fairly comprehensive international review, but provides little or no information on influencing factors or improvement strategies. Van Oort, 2011 Van Oort, N. (2011). Service Reliability and Urban Public Transport Design. Delft University of Technology, Netherlands. Public transit service reliability Researcher based at TRAIL Research School in the Netherlands Coefficient of variation of headways, Relative regularity, Average punctuality, Coefficient of variation of travel time, Difference between 90th and 50th percentile travel times, Difference between 80th and 50th percentile travel times, Passenger wait times, Average additional wait time, Reliability buffer time, Level of crowding, Perceived passenger travel time Terminal departure time, Trip time variability, Line length, Passenger activity, Number of signalized intersections Service monitoring (supply monitoring and customer satisfaction), Operational instruments (skipping stops, deadheading, headway control, speeding up, slowing down, detours, short turning, adding vehicles, vehicle holding), Preventive instruments (driver training, passenger education, spare drivers, maintenance, spare vehicles, trip time determination, vehicle design, priority at traffic lights, platform design, terminal capacity, stop capacity, exclusive lanes, line coordination, line length, stopping distance, line synchronization) Internal causes of travel time variability include driver behavior, other public transport, infrastructure configuration, service network configuration, and schedule quality. External causes of travel time variability include other traffic and weather conditions. Internal causes of variability in dwell times include driver behavior, vehicle design, and platform design. External causes of dwell time variability include passenger behavior and irregular loads. A great deal of discussion of metrics and improvement strategies are also included in this report. This thesis includes modes other than bus. Watkins, Ferris, Borning, Rutherford, & Layton, 2011 Watkins, K., Ferris, B., Borning, A., Rutherford, S., & Layton, D. (2011). Where is My Bus? Impact of Mobile Real-Time Information on the Perceived and Actual Wait Time of Transit Riders. Transportation Research Part A, 45, pp. 839-848. Retrieved from http://www.sciencedirect.co m/science/article/pii/S09658 56411001030 Effects of mobile real- time transit information on perceived and actual customer wait times Researchers based at Georgia Institute of Technology and the University of Washington, Case study of King County Metro in Seattle, Washington, United States On-time performance Real-time transit arrival information, Peak period travel, Bus frequency Provision of mobile real- time transit arrival information Transit customers without real-time information perceived their wait times to be greater than their actual wait times, while those using real-time information did not exhibit this discrepancy. The perceived wait times of riders using real-time information was about 30% less than for riders without this technology. The addition of real-time information was found to reduce perceived wait time by about 13%. Real-time information users were also found to wait almost two minutes less than those arriving without the use of real-time information. Application of a specific technology for reducing wait times and the perception of wait times.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-110 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Xuan, Argote, & Daganzo, 2011 Xuan, Y., Argote, J., & Daganzo, C. F. (2011). Dynamic Bus Holding Strategies for Schedule Reliability: Optimal Linear Control and Performance Analysis. Transportation Research Part B: Methodological, 45(10), pp. 1831-1845. Retrieved from http://www.sciencedirect.co m/science/article/pii/S01912 61511001093 Dynamic bus holding strategies Researchers based at the University of California at Berkeley in Berkeley, California, United States Schedule adherence Bus bunching Dynamic bus holding strategies This research indicates that dynamic holding strategies based on headways alone are not sufficient to help buses adhere to a schedule, so the researchers propose a dynamic holding strategy using bus arrival deviations from a virtual schedule. Using data on arrival times of the current and preceding bus, as well as the virtual schedule, the proposed "simple method" can be used to remove about 40% of the slack in a conventional schedule-based system and produce a one-parameter indicator of schedule reliability that can be used for optimization purposes. The optimal control strategy proposed by the authors chooses control coefficients to minimize slack time required to avoid negative holding times while ensuring a maximum standard deviation from the schedule. Results are based on simulations, rather than before and after studies. Zhang, Yang, & Teng, 2011 Zhang, S.-y., Yang, X.-g., & Teng, J. (2011). Evaluation of Bus Service Reliability Based on AVL Information. 11th International Conference of Chinese Transportation Professionals. American Society of Civil Engineers. Retrieved from http://ascelibrary.org/doi/abs /10.1061/41186%28421%29 287 Bus service reliability measures Researchers based at Tongji University in Shanghai, China Reliability Composite Index of Service (RCIS), comprised of Punctuality Index based on Routes (PIR), Deviation Index based on Stops (DIS), Evenness Index based on Stops (EIS) Traffic volumes, Peak travel periods None Researchers recommend using the proposed Reliability Composite Index of Service (RCIS) to evaluate bus service reliability for three target groups: government agencies, transit operators, and bus passengers. Variables used in the model may be chosen arbitrarily to weight the three components of the composite index. Sarna and Lorgion, 2011 Sarna, P. and S.D. Lorgion. Route 51 Service and Reliability Report. AC Transit, Oakland, Calif., May 13, 2011. Reliability on AC Route 51 AC Transit Dwell time variability Front-door boarding, back door boarding, fare payment, passenger demand None Stop removal and stop spacing Small scope of data and methodology Carrasco, 2012 Carrasco, N. (2012). Quantifying Reliability of Transit Service in Zurich, Switzerland: Case Study of Bus Line 31. Transportation Research Record: Journal of the Transportation Research Board, No. 2274, pp. 114- 125. Retrieved from https://journals.sagepub.com /doi/abs/10.3141/2274-13 Quantifying bus service reliability Researcher based at the Swiss Federal Institute of Technology, Case study based in Zurich, Switzerland Reliability is defined as a state minus the probability of failure, and usually measured in terms of its consequences. Travel time reliability has consistency or dependability in travel times, measured from day to day for the same trip. Reliable transit service is considered to be on schedule, maintaining regular headways, and minimizing wait time variability for passengers. Measures may include aspects of punctuality, as well as travel time variability. Often the result of the ever- changing nature of the transit operating environment. Service reliability may be influenced by external and internal factors including: Traffic signals, Traffic conditions, Unexpected events, Route structure, Stop patterns, Scheduling practices, Fare collection methods, Bus bunching, Connections at transfer points None Measures of bus service reliability used in this report include: Travel time (mean and 5th and 95th percentiles), Speed (mean and 5th and 95th percentiles), Punctuality (route-level schedule deviation frequency, mean schedule deviation at stop level, on-time performance, standard deviation from scheduled departures at stop level, coefficient of variation of schedule deviation at the stop level), and Regularity (actual headway frequency distribution at route and stop levels, mean headway at stop level, coefficient of variation of actual headways at stop level). In Zurich, on-time performance is defined as the percentage of trips departing a stop between 30s early and 60s late. Case study of a single bus line in Switzerland, where the service standards are much higher than in most of the world.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-111 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Currie, Douglas, & Kearns, 2012 Currie, G., Douglas, N., & Kearns, I. (2012). An Assessment of Alternative Bus Reliability Indicators. Shaping the Future: Linking Research, Policy and Outcomes. Australasian Transport. Retrieved from http://atrf.info/papers/2012/in dex.aspx Assessmen t of alternative bus reliability indicators Researchers based in New Zealand, Case study of Sydney, Australia Reliability is defined as having two components: reliability in arrival / departure time at the bus stop and reliability in the travel time spent on the bus. Measures assessed include Percentage of buses cancelled, Percentage of services departing on time, Percentage of services arriving on time, Excess wait time, Average lateness, Service variability indicators, Reliability buffer index (95th percentile on-board time / mean), Passenger ratings of reliability, Customer complaints, and Customer delay None None The identified measures of bus reliability were evaluated using a framework based on four criteria: ease of understanding; extent to which the measure has a customer focus; accuracy, completeness and objectivity of the measure; and the relative cost / effort in collecting and analyzing the data. The results of this analysis suggest that the two best measures are excess waiting time and customer delay, followed by customer complaints. Excess waiting time is described as a partial measure, only appropriate for high- frequency services, while customer journey time delay is a total measure applicable to frequent and infrequent services, incorporating at-stop and on-bus delay. Fairly comprehensive review, but focused on less conventional measures. May not include all the most common measures of bus reliability. Diab & El- Geneidy, 2012 Diab, E. I., & El-Geneidy, A. M. (2012). Understanding the Impacts of a Combination of Service Improvement Strategies on Bus Running Time and Passenger's Perception. Transportation Research Part A: Policy and Practice, 46(3), pp. 614-625. Retrieved from http://www.sciencedirect.co m/science/article/pii/S09658 56411001820 Impacts of combined service improveme nt strategies on bus running times and passenger perception Researchers based at McGill University in Canada, Case study of STM in Montreal, Quebec, Canada None Passenger activity, Passenger activity related to articulated buses, Number of stops made, Time of day, Delay at the start of a trip, Bus type, Weather conditions Smartcard payment, Limited-stop service, Reserved bus lanes, Articulated buses, Transit signal priority Based on the running time model, total passenger activity appears to increase running time by 1.39s per passenger. For every millimeter of rain, running time is expected to increase by 0.79s per trip, or by 1.81s per trip for each centimeter of snow. Buses that get a late start are generally found to have shorter running times, with running time decreasing by 0.22s for every second of delay at the beginning of the trip. Afternoon peak period trips are much longer than midday trips, while morning peak, nighttime, and midnight trips are faster by 47s, 100s, and 219s respectively. The introduction of smartcard fare payment increased running times on the route in question by 5.83s at the beginning of the implementation period and by 52.61s by the end of the implementation period. Reserved bus lanes decreased running time by 35.26s on average, though this benefit could have been larger in magnitude if cars were not allowed to use the bus lanes to turn right (especially given that vehicles are not allowed to turn right on red). The introduction of articulated buses increased running times by 26.8s. Implementing transit signal priority (TSP) for articulated vehicles only decreased running times for all vehicles by 4.76s (0.3%) on average. For the vehicles equipped with TSP, running times decreased by 18.32s (1.2%). The combined improvement strategies resulted in a 10.5% decline in running times along the new limited-stop service compared to the regular service, while the regular service running times increased by about 1% on average. Passengers were found to overestimate time savings associated with the combined strategies by 3.5-6 minutes for the regular route and 2.5-4.1 minutes for the limited-stop service. Focused on running times, not reliability.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-112 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Lee, Sun, & Erath, 2012 Lee, D.-H., Sun, L., & Erath, A. (2012). Study of Bus Service Reliability in Singapore using Fare Card Data. 12th Asia-Pacific Intelligent Transportation Forum. Retrieved from http://www.researchgate.net/ profile/Lijun_Sun3/publicatio n/249656894_Study_of_Bus _Service_Reliability_in_Sing apore_Using_Fare_Card_D ata/links/0c96051e689009ed 67000000.pdf Bus service reliability Researchers based at National University of Singapore and Singapore ETH Centre in Singapore and Massachusetts Institute of Technology in the United States, Case study of Land Transport Authority in Singapore Reliability is measured using average wait time (mean and standard deviation), as well as maximum standard deviation of headways. Traffic congestion, Signal control, Driver behavior and capabilities, Service frequency, Fare payment methods, Route length, Time point spacing, Bus bunching, Occupancy Holding strategies, Stop skipping, Smartcard fare payment, Shorter service distances, Exclusive bus lanes, Transit signal priority This paper outlines researchers' findings that unreliability increases with bus line travel distance, in a positive linear fashion. Simulation has been used to show that the level of service is improved with shorter operating distances, in terms of passengers' average waiting time and buses' occupancy. Shortening the distance between time points seems to be an effective method for improving bus service reliability. Case study of a single bus route. Martin, Levinson, & Texas Transportat ion Institute, 2012 Martin, P. C., Levinson, H. S., & Texas Transportation Institute. (2012). TCRP Report 151: A Guide for Implementing Bus on Shoulder (BOS) Systems. Federal Transit Administration. Washington, D.C.: Transportation Research Board. Retrieved from http://www.tcrponline.org/PD FDocuments/TCRP_RPT_1 51.pdf Bus on shoulder operations Researchers based at Wilbur Smith Associates and Texas Transportation Institute, Case studies based in the United States None Traffic congestion, Weather Bus on shoulder operations Customer perception of schedule adherence and trip reliability is higher when buses make use of freeway shoulder lanes during congested periods. Many of the case studies indicate that reliability is a primary reason behind implementing bus on shoulder (BOS) operations, as well as one of the key benefits of this practice. For example, Miami-Dade Transit reported that its three buses using bus on shoulder facilities improved their on-time performance by up to 19%. An inside-shoulder BOS system demonstrated in Chicago has also shown notable reliability benefits. Overall, bus on shoulder operations are recommended as a low investment means of improving transit service reliability. Focus is on implementing bus on shoulder operations, not the impacts on reliability.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-113 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Schil, 2012 Schil, M. (2012). Measuring Journey Time Reliability in London Using Automated Data Collection Systems. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/ 74273 Automated data collection and application Researcher based at Massachusetts Institute of Technology in the United States, Case study of Transport for London in London, England, United Kingdom Reliability is defined as the quality of being consistently good in quality or performance, and the ability to be trusted. Vehicle on-time arrivals and headways are said to be common measures of transit service reliability. Measures proposed include the median and 95th percentile journey time calculated for each O-D pair and time band, the reliability buffer time (the difference between the 95th percentile and the median journey times), the median and 95th percentile of the good journey time distribution, the percentage of passengers traveling in five minutes or more over the good journey time, and the Normalized Reliability Buffer Time (NRBT). None None Metrics are discussed as being operational (percentages of buses in service), negative (number of delays), retrospective (averages of past performance), big facts (top causes of delay), and/or future focused (bus frequencies). Of these big facts and future-focused metrics seem to offer the most promise, though many passengers may not find reliability metrics to be useful and may choose to rely on their own experiences to judge a system's reliability. Reliability of London's bus network high-frequency routes is measured in terms of excess wait time and percentage chance of waiting less than 10 minutes, 10-20 minutes, 20-30 minutes, and more than 30 minutes. For low-frequency routes, indicators such as percent on- time, percentage chance of a bus running early, percentage chance of a bus running late were used, and percentage chance of a bus not running. These metrics are deemed by the author to be lacking in that they don't take a user's entire journey into account, as well as the variability of travel times. Desirable qualities for journey time reliability metrics include being customer-driven, simple, meaningful for customers, meaningful for operators, and standardizable. A reliability measurement framework is developed that incorporates analysis at the origin-destination pair level, aggregation at the route level, and the definition of reliability standards and metrics. In the O-D analysis, journey time distribution and reliability buffer time are considered. When data is aggregated at the route level, average median and average 95th percentile journey times are calculated, along with the weighted average median and 95th percentile journey times. To measure reliability for comparison across modes and lines, the normalized reliability buffer time (reliability buffer time divided by the median journey time for the O-D pair and time band of interest) is proposed. When setting journey time reliability standards, reliability buffer time distributions can be helpful. Waiting time distribution is of more interest for buses, as this data can be analyzed to find the corresponding probability density function and waiting time reliability (95th percentile waiting time minus the 50% percentile waiting time). The journey time distribution and reliability is also considered an important indicator of overall service reliability. Headway distributions, waiting time reliability, and journey time reliability are also calculated at the route level. Major findings and recommendations were as follows. Automated data collection allows transit agencies to take a more customer-focused approach to measuring travel time variability. Metrics developed using AVL and other data can capture extreme values. Proposed waiting time and journey time metrics can be used to complement existing metrics, and their calculation is relatively straightforward. The distributions and metrics should be used to find a representative "good" journey time for the purpose of setting reliability standards and to provide information to customers. Case study focused on Transport for London, but some results may be generalizable.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-114 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Tang & Thakuriah, 2012 Tang, L., & Thakuriah, P. (2012). Ridership Effects of Real-Time Bus Information System: A Case Study in the City of Chicago. Transportation Research Part C, 22, pp. 146-161. Retrieved from http://www.sciencedirect.co m/science/article/pii/S09680 90X12000022 Ridership impacts of real-time bus arrival information Researchers based at the University of Chicago Illinois and MacroSys LLC in the United States, Case study of the Chicago Transit Authority in Chicago, Illinois, United States None Ridership Real-time bus arrival information In the before and after case study of the Chicago Transit Authority Bus Tracker service, modest increases in ridership were noted, which could have an impact on reliability if not properly managed. Limited relevance overall, with some application to but very little mention of reliability. Yu, Yang, & Li, 2012 Yu, B., Yang, Z., & Li, S. (2012). Real-Time Partway Deadheading Strategy Based on Transit Service Reliability Assessment. Transportation Research Part A: Policy and Practice, 46(8), pp. 1265-1279. Retrieved from http://www.sciencedirect.co m/science/article/pii/S09658 56412000845 Evaluation of real-time partway deadheadin g strategy Researchers base at Dalian Maritime University in China Passenger waiting times Headways between buses Real-time partway deadheading strategy (bus runs empty to the origin station or another stop for the peak direction when unreliability is occurring) The partway deadheading strategy with real-time control proposed in this paper was shown, through simulation, to improve bus service reliability for a lower operational cost than other deadheading strategies. This method is particularly effective for routes that experience heavy directional imbalances during peak periods. Simulation- based results for a single route. Arhin & Noel, 2013 Arhin, S., & Noel, E. C. (2013). Evaluation of Bus Transit Reliability in the District of Columbia. Washington, D.C.: Mineta National Transit Research Consortium. Retrieved from http://transweb.sjsu.edu/PD Fs/research/1139-DC-bus- transit-reliability.pdf Evaluating bus transit reliability Researchers based at San Jose State University and Howard University in the United States, Case study of WMATA in Washington, D.C., United States Reliability is measured in terms of on-time performance (up to 2 minutes early or 7 minutes late), travel time adherence, run time adherence, and customer satisfaction. Traffic congestion, Type of route, Scheduling practices, Location on the route, Passenger activity, Weather, Routing, Incidents Transit reliability management strategies About 50% of WMATA survey respondents selected improvement in on-time arrival / reliability as their most desired improvement for bus service, with young riders rating this as even more important than the overall population. On-time performance on the 15 bus lines studied did not meet WMATA's target threshold in 71% of cases, and 82% did not meet the industry standard threshold of between 1 minute early and 5 minutes late. Only considered on-time performance. Boyle, 2013 Boyle, D. K. (2013). TCRP Synthesis 110: Commonsense Approaches for Improving Transit Bus Speeds. Washington, D.C.: Transportation Research Board. Retrieved from http://onlinepubs.trb.org/onli nepubs/tcrp/tcrp_syn_110.p df Improving transit bus speeds Researcher and case studies based in the United States None Traffic conditions, Delays due to traffic signals, Entering and exiting bus stops, Routing, Dwell time, Fare collection policies Streamlined routes, Increased stop spacing, All-door boarding and alighting, Improved or judicious stop placement, Dedicated bus lanes, Signal priority, Yield-to-bus laws, Traffic engineering to improve traffic flow overall, Off- board fare collection, Far-side stop placement Improving bus speeds is possible through a variety of strategies. Working with traffic engineers to expedite the flow of transit vehicles has proven effective. Signal priority was shown to have the biggest impact on speeds of any action in the case of San Francisco, while New York City reported success with off-board fare collection and increased stop spacing, combined with transit signal priority, bus-only lanes, and all-door boarding. Stop consolidation can also improve transit speeds and reliability, but may be faced with community resistance. Changes in fare policies, vehicles, and schedules are also discussed as being effective strategies to improve bus speeds. Far-side bus stops are also recommended to this end, especially as they facilitate transit signal priority. Focused on transit speeds, but some discussion of reliability.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-115 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Bunker, 2013 Bunker, J. M. (2013). Planning for Transit System Reliability Using Productive Performance and Risk Assessment. TRB 92nd Annual Meeting Compendium of Papers. Washington, D.C.: Transportation Research Board. Retrieved from http://docs.trb.org/prp/13- 0036.pdf Planning for transit system reliability using productive performanc e and risk assessmen t Researcher based at Queensland University of Technology in Australia Transit productiveness is defined as transit work delivered over time by one or more services traversing a line. Reliability incident events are defined as those that cause delay to the transit passenger through later than expected punctuality and/or longer than schedule travel times. Congestion, Incident events Busways A comparison of passenger productiveness revealed that long trunk segments of the transit system are most productive, including a long high-speed trunk busway segment approaching the central business district, the bus-on-expressway river crossing, and two long trunk rail segments approaching the CBD. Segments found to be at highest risk for reliability incident events were long, highly productive bus segments, though busway corridors showed clear benefits over general traffic expressway and arterial road segments. Included modes other than bus. Chen, Wang, Li, & Deng, 2013 Chen, Q., Wang, X., Li, W., & Deng, W. (2013). Bus Arrival Headway Reliability: A Case Study in Hefei, China. TRB 92nd Annual Meeting Compendium of Papers. Washington, D.C.: Transportation Research Board. Retrieved from http://trid.trb.org/view/2013/ C/1240619 Introducing a new performanc e metric for on-time performanc e Researchers based at Southeast University, Old Dominion University, and University of South Florida in the United States Discrete function of arrival time in acceptable punctuality intervals Bus arrival times None Performance metric Unlike traditional performance metric, the proposed metric does not describe a physical phenomenon. Its calibration is completely arbitrary. As a result, it is not interpretable in operational terms. Diab & El- Geneidy, 2013 Diab, E. I., & El-Geneidy, A. M. (2013). Variation in Bus Transit Service: Understanding the Impacts of Various Improvement Strategies on Transit Service Reliability. Public Transport, 4(3), pp. 209-231. Retrieved from http://link.springer.com/articl e/10.1007%2Fs12469-013- 0061-0 Impacts of transit service reliability improveme nt strategies Researchers based at McGill University in Canada, Case study of STM in Montreal, Quebec, Canada Running time deviation, Coefficient of variation of running time, Coefficient of variation of running time deviation Direction of travel, Passenger activity, Rear-door activity, Actual stops, Precipitation, Snow, Delay at the start of a trip, Peak periods, Fare payment method, Reserved lane, Articulated buses, Transit signal priority Reserved bus lane, Transit signal priority Factors that appear to significantly increase running time deviation include: actual stops, passenger activity, rear-door activity, precipitation, snow, smartcard use, and presence of articulated buses. Factors that significantly decrease running time deviation in the model include: delay at the start of the trip, northbound travel, peak period travel, off-peak travel (more so than peak period), and reserved bus lanes. Significant factors in the coefficient of variation (CV) of running time model include (sign is + unless otherwise indicated): CV of actual stops, CV of passenger activity, CV of rear- door activity, CV of snow, northbound travel, midnight and early morning travel (-), smartcard use, reserved bus lane, articulated buses (-), and transit signal priority implementation. Significant factors in the CV of running time deviation model include (sign is + unless otherwise indicated): CV of actual stops, CV of rear-door activity, CV of snow, morning peak travel (-), night travel (-), smartcard use, articulated buses (-), and transit signal priority implementation. Smartcard fare payment was found to be less effective than flash passes with regard to running time and reliability. Reserved bus lanes may result in decreased running time, but increased variation, especially when general traffic is allowed to use bus lanes as right turn lanes but not allowed to turn right on red. While articulated buses increase running times, they can help to reduce variation. Limited-stop service was shown to improve running times, but worsen variation. Transit signal priority appears to improve running times with no significant impact on variation. Models based on data for one bus route in Montreal, so results may not be generalizable.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-116 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Fletcher & El- Geneidy, 2013 Fletcher, G., & El-Geneidy, A. (2013). Effects of Fare Payment Types and Crowding on Dwell Time: Fine-Grained Analysis. Transit 2013, 2351, pp. 124- 132. Retrieved from http://trrjournalonline.trb.org/ doi/10.3141/2351-14 Impacts of fare payment types and crowding on bus dwell times Researchers based at McGill University in Canada, Case study of Translink in Vancouver, Canada Dwell time is defined as the amount of time a bus spends while stopped to serve passengers. Fare payment type, Passenger activity, Crowding, Wheelchair, bike, and stroller users Smartcard payment, Low-floor buses, Fast ramp actuation, Efficient wheelchair tie-down systems The researchers developed a model to analyze factors impacting bus dwell times. Factors shown to significantly increase dwell times in this model included wheelchair ramp events, bike rack events, passengers with bulky items, front-door alighting, cash fare payment, and higher passenger volumes. The three case studies were all located in Vancouver, so may not be applicable to other locations and services. Frumin, Zhao, Wilson, & Zhao, 2013 Frumin, M., Zhao, J., Wilson, N. H., & Zhao, Z. (2013). Automatic Data for Applied Railway Management: A Case Study on the London Overground. Transportation Research Record: Journal of the Transportation Research Board, No. 2354, pp. 47-56. https://journals.sagepub.com /doi/abs/10.3141/2353-05 Assessing performanc e of new tactical plan on North London Line Researchers based at the University of British Columbia in Canada and the Massachusetts Institute of Technology in the United States, Case study of the Metropolitan Transportation Authority Bus Customer Information Systems Quality of service as experienced by passengers, measured in terms of Excess Journey Time and Observed Journey Time (Zhao, Frumin, Wilson & Zhao 2013) Bus bunching, Overlapping bus routes Coordinating the common trunk of two overlapping bus routes to avoid bus bunching with even headways. The plan reduced (improved) Excess Journey Time, but it made the journey times longer by increasing running time to accommodate the two routes on their overlapping portion. No data on customer satisfaction with performance metric.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-117 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Ian Wallis Associates, Ltd., 2013 Ian Wallis Associates Ltd in association with The TAS Partnership. (2013). Research Report 527 - Improving Bus Service Reliability. New Zealand Transport Agency. Retrieved from http://www.nzta.govt.nz/reso urces/research/reports/527/ Improving bus service reliability Researchers based in New Zealand, Case study for London, England, United Kingdom Reliability is used to describe two concepts, including 'reliability' (whether or not the service operates) and 'punctuality' (whether the service runs on time). Traffic congestion, Staff shortages, Bus boarding times, Inadequate route supervision, Variable speeds, Bus bunching, Passenger activity, Traffic signals, Scheduling, Parking, Incidents, Road construction, Vehicle and maintenance quality, Transit preferential treatments, Schedule achievability, Evenness of passenger demand, Operator driving skills, Wheelchair lift and ramp usage, Route length, Number of stops, Operations control strategies Ideal scheduling, Coordinating schedules across services, Control strategies, Schedule adjustments, Driver training and incentives, Development of route contingency plans, Intervention and improvement actions, Supervision and communication procedures, Options to respond to bus bunching and gaps, Deployment of spare resources Reliability measures should reflect service reliability (whether the service operates), punctuality and running times, and passenger perceptions. Efforts to improve reliability should begin with a 'root cause' analysis, and may take place at the service planning, operational planning, or on-the-day level. Recommendations from this report include: transitioning to automated vehicle location data collection, using collected data to optimize running times, including incentives and penalty structures for new drivers in their contracts, and fostering partnerships between regional authorities to improve reliability through implementation of best practices. Aspects of performance recommended for monitoring reliability include: cancellations, service arrival and departure times, service regularity, running times, and passenger perceptions. Related specific measures include: percent of scheduled trips cancelled, percent of trips arriving / departing 'on-time' (up to 1 minute early or 5 minutes late), average minutes early / late arriving / departing, average passenger waiting time, standard deviation of headways, coefficient of variation of headways, excess waiting time, wait assessment (percent of headways within X minutes of the scheduled interval), service regularity (percent of headways within Y% of scheduled interval), bunching index, mean running time, standard deviation of running time, coefficient of variation of running time, travel time buffer index (95th percentile travel time divided by the average travel time), Florida reliability index (100% minus percent of trips with travel time over the mean plus a specified margin), and the misery index (average travel time of the slowest 20% of trips minus the average travel time of all trips divided by the average travel time of all trips). Other measures mentioned include average lateness, reliability variability, reliability buffer, passenger ratings, customer complaints, and customer delay. Busways were found to reduce average lateness of buses by 48% and standard deviation of lateness by 45%. Although many strategies for improving reliability are noted, there is little evidence provided of the actual impacts of these improvements. Ji, Guo, & Yan, 2013 Ji, J., Guo, Y., & Yan, Y. (2013). Optimization Model of the Public Transit Network Based on Operational Reliability. American Society of Civil Engineers. Retrieved from http://ascelibrary.org/doi/abs /10.1061/9780784413159.12 1 Public transit network optimizatio n based on operational reliability Researchers based at Jiangsu Provincial Communication Planning and Design Institute and Transit College in China, Case study of Swiss Road Network Public transportation operational reliability is defined as the probability of effectively transporting passengers and the smooth operation of the various bus lines and the site in accordance with the planned schedule within the urban transit network. Demand fluctuation, Schedule, Drivers, Time deviation, Crashes, Road work, Weather, Traffic conditions, Passenger behavior None The authors propose an optimization model for public transit networks, which may support network planning and decision- making, based on operational reliability metrics. Limited detail is provided on the model, which is the basis of the research.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-118 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Kittelson & Associates, Inc., Parsons Brinckerhof f, KFH Group Inc., Texas A&M Transporta tion Institute, and Arup, 2013 Kittelson & Associates, Inc., Parsons Brinckerhoff, KFH Group Inc., Texas A&M Transportation Institute, and Arup. (2013). TCRP Report 165: Transit Capacity and Quality of Service Manual Third Edition. Washington, D.C.: Transportation Research Board. Retrieved from https://www.nap.edu/catalog /24766/transit-capacity-and- quality-of-service-manual- third-edition Transit capacity and quality of service Researchers based in the United States Reliability includes on-time performance and regularity of headways. Reliability is defined from the passenger's perspective as pertaining to arriving at destinations on time and not having to wait too long at bus stops. From the operator perspective, reliability impacts schedule recovery time. Internal factors (vehicle quality and age, vehicle availability and breakdowns, driver availability, transit preferential treatments, route length, supervision, control strategies, driver experience, schedule achievability), External factors (snow and ice, heat, leaf fall, traffic signals, traffic congestion, variability in traffic demand, road construction, passenger demand variability, wheelchair lift or ramp usage, door holding) Grade-separated and exclusive right-of-ways, Level boarding, Off- board fare collection, Proof-of-payment fare collection, All-door boarding, Increased stop spacing, More doors per bus, Schedule adjustments, Control strategies This manual reports that factors influencing speed, capacity, and reliability of transit service are similar. Internal factors relating to reliability include those related to finance, purchasing, maintenance, human resources, capital projects, service planning, operations, and scheduling. External factors include those relating to the environment, roadway conditions, and passengers. Fairly comprehensive, but limited information on impacts of improvement strategies. Ma, Ferreira, & Mesbah, 2013 Ma, Z.-L., Ferreira, L., & Mesbah, M. (2013). A Framework for the Development of Bus Service Reliability Measures. Transport and the New World City: 36th Australasian Transport Research Forum. Retrieved from http://atrf.info/papers/2013/in dex.aspx Modelling bus travel time reliability Researchers based at the University of Queensland, Case study of Translink in Brisbane, Australia Reliability tactical indicator (expected reliability buffer time divided by median travel time), Journey planning time (departure time to ensure an on-time arrival at destination), Value of reliability Traffic demand, Passenger volumes, Time of day, Day of week, Operation direction None In addition to reviewing commonly used reliability indicators, such as on-time performance, headway regularity, travel time, waiting time, transfer time, and buffer time, the authors of this paper propose three new ways to measure reliability from different perspectives. It is recommended that measures be developed for specific users, based on local and service-specific conditions, and a process for developing such measures is outlined. The proposed framework uses demand- and supply-side data to identify latent service reliability factors through factor analysis, as well as disaggregated performance categories through cluster analysis. Then mixture distribution models are developed for different scenarios. Transit service variability and reliability indicators are then developed based on the data analysis, with at least one indicator each proposed for the transit operator, passengers, and government support agencies. Proposed framework requires a local, context- sensitive approach, which cannot be generalized across different locations or services. New York Metropolita n Transportat ion Authority, 2013 New York Metropolitan Transportation Authority. (2013). Mission Statement, Measurements, and Performance Indicator Report Covering Fiscal Year 2012. New York. Retrieved from http://web.mta.info/mta/com pliance/pdf/2012_annual/Mis sionPerfMeasurementFY201 2.pdf Performanc e measures Metropolitan Transportation Authority in New York City, New York, United States On-time and reliable service is measured in terms of on-time performance, bus trips completed, and mean distance between failures. None None This report provides an example of how the largest transit agency in the United States reports on its performance internally. Reliability measures are included and discussed, but little or no discussion of impacting factors or improvement strategies. Saberi, Zockaie, Feng, & El- Geneidy, 2013 Saberi, M., Zockaie, A. K., Feng, W., & El-Geneidy, A. (2013). Definition and Properties of Alternative Bus Service Reliability Measures at the Stop Level. Journal of Public Transportation, 16. Retrieved from http://www.nctr.usf.edu/wp- content/uploads/2013/03/16. 1_saberi.pdf Assessmen t of existing bus reliability measures from the TCQSM, with three new complemen tary measures proposed. Researchers based at Northwestern University and Portland State University in the United States and McGill University in Canada, Case study of TriMet in Portland, Oregon, United States TCQSM definition of reliability (on- time performance and evenness of headways). New measures include: Earliness Index (EI), Width Index (WI), and Second Order Stochastic Dominance Index (SSDI). Considers on-time performance for infrequent service. (From TCQSM 2004) Traffic conditions, Road construction, Vehicle and maintenance quality, Vehicle and staff availability, Transit preferential treatments, Schedule achievability, Evenness of passenger demand, Driver differences, Dwell time, Route length and number of stops, Operations control strategies, Weather, Incidents None The three new proposed measures would complement standard bus reliability measures defined in the TCQSM, providing a more comprehensive characterization of bus reliability at the stop, route, and network levels. The Earliness Index (EI) is defined as the probability that a bus leaves a stop early (delay < 0). The Width Index (WI) is meant to capture extreme cases of bus unreliability, and also accounts for differences in the cost of early versus late bus departures. The Second Order Stochastic Dominance Index (SSDI) helps to show additional characteristics of the distribution of delays, which would not be portrayed through conventional reliability measures. Detailed information on bus location and schedule adherence (such as AVL data) is needed to calculate the proposed measures.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-119 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Sanchez- Martinez, 2013 Sanchez-Martinez, G. E. (2013). Running Time Variability and Resource Allocation: A Data-Driven Analysis of High-Frequency Bus Operations. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/ 79498 Running time variability and resource allocation Researcher based at Massachusetts Institute of Technology in the United States, Case study of Transport for London in London, England, United Kingdom Unreliability is discussed in terms of running time variability. Variability measures for homogeneous time periods are described as mean- based or percentile-based and absolute or normalized, and include standard deviation, spread, coefficient of variation, and normalized spread. Diurnal mean spread (DMS) was also introduced. Route length, Number of stops, Ridership, Resource level, Dwell time variability, Probability of lateness propagation, Traffic, Traffic signals, Vehicle characteristics, Driver characteristics, Crashes, Incidents, Roadwork, Diversions, Season, Day of the week, Time of day, Fleet size, Operating strategies Measurement and regular communications regarding running time variability and resource allocation, Modifications in stop location, Bus lanes, Signal priority A regression analysis was performed to compare peak to midday running times based on season, and the findings were as follows. The seasons (summer, fall, spring) have different peaking characteristics. In the summer, morning peak running times were found to be lower than midday running times, possibly due to schools not being in session. During the spring and fall morning peak running times were estimated to be higher than midday running. Running times were also found to be highest in the fall. A linear model was developed to compare mean spread to season, route location and configuration, and bus line, but this was based on a small sample of data. Another linear model compared period median running times to direction of travel relative to the central business district, distance along the route, number of stops, and boardings, finding all variables to be significant and positive. Travel from the CBD was estimated to have the least positive correlation with running time, while travel through the CBD had the largest positive correlation. A linear model of spread, using travel pattern with relation to the CBD, distance, number of stops, and boardings as predictors, and travel pattern was found to be the most significant variable, again with travel through the CBD having the largest positive influence on spread, and travel from the CBD having the least. Another linear model of spread with a better fit was developed using predictors of travel pattern relative to the CBD, boardings, running time median, and distance divided by the running time median. In this model boardings and distance divided by running time median had negative coefficients. Overall the regression analysis indicated that different routes have varying levels of running time variability, and that routes entering the CBD or with higher running times often had more variable running times. Distance, number of stops, and ridership were also found to contribute to unreliability, while higher operational speeds correlated with slightly lower running time variability. A combination of measures are proposed for understanding reliability, including diurnal mean spread (measure of overall random variability) and more detailed measures at the direction, time period, and segment levels. Visual analysis tools are also introduced, which show running times throughout the day and 10th, 50th, and 90th percentile lines. A simulation was developed and used to demonstrate that bus performance can be improved through optimal resource allocation. A method of optimizing resource allocation is proposed, based on processes of an optimizer, a simulator, and a vehicle profiler. Focuses on high-frequency bus service. The linear spread model is based on a small sample. Stewart & Wong, 2013 Stewart, R. & Wong, R. (2013). Guidelines for Planning and Implementation of Transit Priority Measures in Urban Areas. 2013 Conference and Exhibition of the Transportation Association of Canada. Transportation Association of Canada. Retrieved from http://conf.tac- atc.ca/english/annualconfere nce/tac2013/session10/stew art.pdf Planning and implementa tion of transit priority measures in urban areas Researchers based at IBI Group in Winnipeg, Manitoba, Canada Transit service reliability is defined as how well a transit vehicle complies with a schedule or predetermined headway. Traffic congestion, Right-of- way Transit priority measures, including Regulatory (lane use restrictions, time of day or part-time reserved transit lanes, on-street parking restrictions), transit signal priority (passive, active), and Physical Measures (exclusive running way, geometry changes, bus bulbs and bays, queue jump lanes) Guiding principles for transit priority measures include: safety, transit delay reduction, minimal disruption to road users, consistency, and pragmatism. The proposed assessment process includes consideration of local and corridor challenges, physical impacts of transit priority measures, operational impacts, public acceptance, safety, enforcement, organizational challenges, and financial implications. Assessment framework is context- sensitive, so there are no one-size-fits-all improvement strategies.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-120 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Yao, Hu, Lu, Gao, & Zhang, 2013 Yao, B., Hu, P., Lu, X., Gao, J., & Zhang, M. (2013). Transit Network Design Based on Travel Time Reliability. Transportation Research Part C: Emerging Technologies, 43, pp. 233- 248. Retrieved from http://www.sciencedirect.co m/science/article/pii/S09680 90X13002647 Designing transit networks for improved travel time reliability Researchers based at Dalian University of Technology in China Travel time deviation, On-time performance Passenger volumes, Peak period travel, Traffic congestion, Schedule buffer time Exclusive bus lane A network optimization model is proposed, which takes travel time reliability into account. The results indicate that this method may help to improve transit network reliability and reduce passenger travel times. Simulation- based results for a single corridor and a medium-sized network. Zhang & Teng, 2013 Zhang, C. & Teng, J. (2013). Bus Dwell Time Estimation and Prediction: A Study Case in Shanghai-China. Tongji University. Procedia - Social and Behavioral Sciences Estimating bus dwell times Researchers based at Tongji University in Shanghai, China Dwell times Passenger boardings, Passenger alightings, Crowding, Fare type Use of AVL and APC data for dwell time predictions The model developed by these researchers was shown to be useful for high demand bus lines in urban areas, especially when crowding is common. Usefulness of model is determined based on a single case study. Zhao, Frumin, Wilson, & Zhao, 2013 Zhao, J., Frumin, M., Wilson, N., & Zhao, Z. (2013). Unified estimator for excess journey time under heterogeneous passenger incidence behavior using smartcard data. Transportation Research Part C: Emerging Technologies, 34, 70-88 New performanc e measure for excess journey time Researchers based at the University of British Columbia in Canada and the Massachusetts Institute of Technology in the United States, Case study of the Metropolitan Transportation Authority Bus Customer Information Systems Quality of service as experienced by passengers, measured as excess journey time compared to perfectly on-time base case measured using smart card data. Excess journey is computed with respect to passenger incidence (intent to board specific bus or route within predefined trip) caused by delays. None None The performance metric allows to measure quality of service from passengers' perspective. No data on customer satisfaction with performance metric. Anderson, & Geroliminis 2014 Anderson, P., & Geroliminis, N. (2014). Dynamic Bus Lanes with Restricted Car Usage for Congested Arterial Routes. TRB 93rd Annual Meeting Compendium of Papers. Washington, D.C.: Transportation Research Board. Retrieved from http://trid.trb.org/view/2014/ C/1287502 Dynamic bus lanes with restricted car usage in congested arterial corridors Researchers based at Ecole Polytechnique Federale de Lausanne in France, Information and case studies taken from around the globe None Traffic congestion Reserved bus lane, Dynamic bus lane Reserved bus lanes are not recommended unless buses have an average occupancy of 25 or more passengers. Dynamic bus lanes can be implemented in congested corridors, but doing so requires a control strategy to continuously balance the needs of transit vehicles with the needs of other users, such as personal automobile drivers and passengers. Different user groups could be allowed to use the bus lanes as appropriate, with variable message signs communicating permitted user groups. Limited discussion of impacts on reliability.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-121 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Beduhn, 2014 Beduhn, T. J. (2014). Reliable Routing in Schedule-Based Transit Networks. University of Texas. Retrieved from https://repositories.lib.utexas .edu/handle/2152/28096 Reliable routing in schedule- based transit networks Researcher based at the University of Texas in Austin, Case study of Austin, Texas, United States Timeliness of vehicles and the difference between a passenger's scheduled and actual travel time. Traffic condition, Vehicle and maintenance quality, Vehicle and staff availability, Transit preferential treatments, Schedule achievability, Evenness of passenger demand, Operator driving skills, Wheelchair lift usage, Route length and number of stops, Weather, Incidents and construction, Operations control strategies Improved transfer reliability This thesis presents a framework and model for network and route planning and scheduling, which takes travel time and transfer reliability into account. Through the case study demonstration in Austin, Texas, this method is shown to improve transit service reliability for the transit network overall. Simulation- based results using data only from Austin, Texas. Ji, He, & Zhang, 2014 Ji, Y., He, L., & Zhang, H. (2014). Bus Drivers' Responses to Real-Time Schedule Adherence and the Effects on Transit Reliability. Transportation Research Record: Journal of the Transportation Research Board, No. 2417, pp. 1-9. Retrieved from http://trrjournalonline.trb.org/ doi/10.3141/2417-01 Bus driver response to real-time schedule adherence and impacts on transit reliability Researchers based at Tongji University in China and the University of California at Davis in the United States Unreliability is defined as schedule deviation of more than one minute early or two minutes late. Variance of schedule deviation is the measure used in this study. Traffic volumes, Real-time information Provision of real-time schedule adherence information for drivers and transit agencies Bus drivers' positive responses to real-time schedule information are shown to help improve bus service reliability. Focuses on one possible improvement strategy. Ma & Wang, 2014 Ma, X., & Wang, Y. (2014). Development of a Data- Driven Platform for Transit Performance Measures Using Smart Card and GPS Data. Journal of Transportation Engineering, 140(12). Retrieved from http://ascelibrary.org/doi/10. 1061/%28ASCE%29TE.194 3-5436.0000714 Transit performanc e measurem ent Researchers at Beihang University in China and the University of Washington in the United States, Data from Beijing, China Network-level speed, Route-level travel time reliability, and Headway variance, 95% travel time, Buffer index (extra time needed to ensure on-time arrival for 95% of trips), Segment-level travel time reliability indicator Traffic signals, Variable passenger demand Dynamic transit routing In this study, automated fare collection and automated vehicle location data are used to develop a data-driven platform for monitoring transit performance online. Some relevance to reliability measures, but little or no discussion of improvement strategies. pteg, 2014 pteg. (2014). Bus Punctuality: Towards a Structure that can Deliver. Leeds, United Kingdom. Retrieved from http://www.urbantransportgr oup.org/system/files/general - docs/pteg%20bus%20punct uality%20web%20report%20 June%202014_FINAL2.pdf Bus punctuality Researchers based at pteg in Leeds, England, United Kingdom A bus is considered to be on time if it is between one minute early and five minutes late. Roadwork and crashes, Highway design, Traffic signals, Timetable allowances, Parking and enforcement, Dwell time, Driving styles, Major events, Traffic volumes, Vehicle reliability, Weather conditions Context-specific punctuality policies and better collaboration between government and transport agencies. This report makes a case for local performance regimes, which reflect local circumstances, rather than one-size-fits-all national standards. The authors also advocate for more available and usable AVL data, which should be made public for use in open source platforms. Focused on punctuality regimes in the UK, and the particular government policies and practices that may hinder effective management of bus punctuality.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-122 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Qu, Oh, Weng, & Jin, 2014 Qu, X., Oh, E., Weng, J., & Jin, S. (2014). Bus Travel Time Reliability Analysis: A Case Study. Proceedings of the Institution of Civil Engineers - Transport, 167(3), pp. 178-184. Retrieved from http://www.icevirtuallibrary.c om/content/article/10.1680/tr an.13.00009 Case study on the impact of fare payment policies and technology on bus travel time reliability. Researchers based at Griffith University in Australia and Beijing Jiaotong University and Zhejiang University in China, Case study of Translink in Queensland, Australia Stop-level sum of boardings and alightings times punctuality (defined as the probability of being less than 3 minutes late). Passenger type, Fare payment method, Smartcard top-up policies Eliminate onboard top-up for smartcards Average boarding time per passenger was found to be 45 s for disabled persons, 3 s for smartcard users, 30 s for onboard top-up, and 15 s for paper ticket users. Eliminating onboard top-up for smartcards resulted in a 15-18% improvement in reliability for the bus route in question. Smartcard users can still top-up their cards in a variety of places, so the inconvenience of this change was deemed to be minimal. This was a case study on one bus line, so the results should not be generalized. Timmel, 2014 Timmel, C. (2014). Bus Transit Reliability in Metropolitan Boston: A Study of MBTA's Dropped Trips. Tufts University. Retrieved from http://gradworks.umi.com/15 /58/1558587.html Bus transit reliability in terms of dropped trips Tufts University, Case study of MBTA in Boston, Massachusetts, United States Dropped trips Community engagement None Service improvement recommendations include: establishing reliable funding mechanisms, being more transparent and engaged with the community, instituting and enforcing staff accountability, and investing in operations rather than infrastructure expansion. Limited relevance overall, with some application to but very little mention of reliability. Tribone, Block- Schachter, Salvucci, Attanucci, & Wilson, 2014 Tribone, D., Block- Schachter, D., Salvucci, F. P., Attanucci, J., & Wilson, N. H. (2014). An Automated Data Driven Performance Regime for Operations Management, Planning, and Control. Transportation Research Record: Journal of the Transportation Research Board, No. 2415, pp. 72-79. https://journals.sagepub.com /doi/abs/10.3141/2415-08 The authors developed a data display to be used by service controllers at the MBTA. Researchers based at the Massachusetts Institute of Technology in the United States, Case study of Massachusetts Bay Transportation Authority (MBTA) in Boston, Massachusetts, United States Service performance from passengers' perspective, including waiting times, travel times, headways and running time performance Passenger arrival rates, O-D pairs, AVL A collaborative process with Operations Control Center to find useful display of large quantities of data. The authors found that useful displays of big data should be developed in collaboration with the agency's operations staff, that the performance measures should use the passengers' perspective, and that the design affects utility. Little empirical data. Kittelson & Associates, 2014 Kittelson & Associates, Inc. (2014). Gap-Filling Project 5: Guidebook: Placing a Value on Travel-Time Reliability. SHRP 2 Project L17. Board, Washington, D.C.: Transportation Research Board. Modeling monetary value of reliability Kittleson and Associates The monetary value of reliability Travel time reliability None The guidebook explains how reliability can be quantified in monetary terms of missed income. The guidebook says how transit planners can include the quantification in the cost/benefit analysis. Stated preference survey.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-123 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations List, Williams, & Rouphail, 2014 List, G.F., Williams, B., & Rouphail, N. SHRP 2 Report S2-L02-RR-1: Establishing Monitoring Programs for Travel Time Reliability. Transportation Research Board, Washington, D.C., 2014 (San Diego case study). https://www.nap.edu/catalog /22612/establishing- monitoring-programs-for- travel-time-reliability Highway reliability Institute for Transportation Research and Education; Iteris/Berkeley Transportation Systems, Inc.; Kittelson & Associates, Inc.; National Institute of Statistical Sciences; University of Utah; Rensselaer Polytechnic Institute, Northwestern University, and Planitek Highway travel time traffic and special events Monitor traffic and special events None This report is mainly focused on vehicular traffic Arhin & Noel, 2015 Arhin, S., & Noel, E. (2015). Development of Bus-Stop Time Models in Dense Urban Areas: A Case Study of Washington DC. Mineta National Transit Research Consortium. Retrieved from http://transweb.sjsu.edu/PD Fs/research/1239-bus-stop- time-models-in-urban- areas.pdf Modelling bus stop time in dense urban areas Researchers based at San Jose State University and Howard University in the United States, Case study of WMATA in Washington, D.C., United States Total Bus Stop Time (TBST) Route, Traffic conditions, Time of day, Bus stop conditions, Passenger activity, Fare payment methods, Dwell time, Bus stop location Minimizing total bus stop time The results indicate that to maintain or improve bus service reliability, buses stopping at intersections should spend no more than 43, 47, and 67 seconds during the morning, midday, and evening peak periods, respectively. For midblock bus stops, no more than 36, 34, and 32 seconds should be spent by buses stopping during these peak periods. Models based on data for Washington, D.C. only, so results should not be generalized. Brakewood , Macfarlane , & Watkins, 2015 Brakewood, C., Macfarlane, G., & Watkins, K. (2015). The Impact of Real-Time Information on Bus Ridership in New York City. Transportation Research Part C: Emerging Technologies, 53, pp. 59-75. Retrieved from http://www.sciencedirect.co m/science/article/pii/S09680 90X15000297 Impacts of real-time mobile information on bus ridership Researchers based at City College of New York, Parsons Brinckerhoff, and Georgia Institute of Technology in the United States, Case study of New York City Transit in the United States None None Real-time mobile transit information A fixed effects model of average weekday unlinked bus trips per month indicates an increase of about 118 trips (1.7% of weekday route ridership) per route per weekday that could be attributed to the recent provision of real-time mobile transit information. For the largest quartile of routes (by revenue miles of service), about 340 additional trips per route per weekday (2.3% increase) appear to have resulted from the addition of mobile real-time information. Limited relevance overall, with some application to but very little mention of reliability.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-124 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Chakrabarti & Giuliano, 2015 Chakrabarti, S., & Giuliano, G. (2015). Does Service Reliability Determine Transit Patronage? Insights from the Los Angeles Metro Bus System. Transport Policy, 42, pp. 12-20. Retrieved from http://www.sciencedirect.co m/science/article/pii/S09670 70X15300068 Relationshi p between service reliability and transit patronage Researchers based at the University of Southern California, Case study of Los Angeles Metro bus system in Los Angeles, California, United States A reliable service is defined as one which consistently operates according to its schedule or plan. On-time performance is used as the primary measure of reliability. Factors found to affect bus line patronage include: Unreliability, Accessibility, Connectivity, Service quality, Time of day, Passenger activity, Transit stops, Transit alternatives, Route length, Rapid service Bus lanes, Transit signal preemption, Efficient scheduling, Real-time rerouting of vehicles, Holding strategies, Stop consolidation, Driver training and incentive programs, Periodic system maintenance, Real-time information sharing via mobile devices, Advances in intelligent transportation infrastructure and information technology Overall, the study shows a statistically significant positive association between bus line service reliability and patronage, and this relationship appears to be stronger during afternoon peak travel periods. Factors found to significantly impact bus line patronage include in terms of average per-hour bus line boardings are: Unreliability (-), Population density and employment access (+), Stops per mile (-), Scheduled headway (-), Total stops (+), Transit alternatives (-), Afternoon peak travel (+), and Night travel (- ). Factors found to significantly influence scheduled headway include: Average per-hour bus line boardings (-), Total stops (-), and Rapid service (-). Focused on data specific to the Los Angeles Metro system, which may not have broader applicability. Delgado, Munoz, Giesen, & Wilson, 2015 Delgado, F., Munoz, J. C., Giesen, R., & Wilson, N. H. (2015). Integrated Real- Time Transit Signal Priority Control for High-Frequency Transit Service. Transportation Research Record: Journal of the Transportation Research Board, No. 2533, Transportation Research Board, Washington, D.C., pp. 28-38. https://journals.sagepub.com /doi/abs/10.3141/2533-04 This paper presents the extension of a simulation- based bus control method developed by the same authors, where traffic signals can extend green light to avoid bus bunching. Pontifical Catholic University of Chile and Massachusetts Institute of Technology in the United States The method seeks to minimize the average waiting time of all transit riders and car drivers on the corridor. Delay Simulation-based control method consisting in holding buses, limiting boarding and extending green lights. The method can reduce delays by 61% compared to no control, with slight increases in delays for car drivers. The model assumes known and deterministic travel time and passenger arrival rates.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-125 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Diab, Badami, & El- Geneidy, 2015 Diab, E. I., Badami, M. G., & El-Geneidy, A. M. (2015). Bus Transit Service Reliability and Improvement Strategies: Integrating the Perspectives of Passengers and Transit Agencies in North America. Transport Reviews, 35, pp. 292-328. Retrieved from http://www.tandfonline.com/ doi/abs/10.1080/01441647.2 015.1005034?journalCode=t trv20 Bus service reliability and improveme nt strategies Researchers based at McGill University in Canada Many definitions and measures considered from the perspective of passengers and transit agencies. Reliability performance measures included on-time performance, delivered trips as a percent of scheduled trips, cancelled trips as a percent of scheduled trips, average transit vehicle speed, mean distance between failures, big gap interval, percentage of bunched intervals, bus wait assessment, headway adherence, running time variation, running time, dwell time, bus arrivals, and bus departures. From the passenger perspective it is noted that perception of wait time, perception of reliability, and value of time are all important considerations. To capture passenger views, overall satisfaction, customer complaints, and reliability ratings are often used. Factors impacting reliability overall include driver experience, control strategies, stop spacing, vehicle quality, road geometry, transit priority strategies, traffic signals, and more. Transit signal priority, Bus lanes, Queue jumpers, Express service, Bus bay improvements, Articulated buses, Bus rapid transit, Rapid transit network, Bypass lanes, Shoulder lanes, Intelligent transportation systems, Road geometry changes, New buses, Off-board fare collection, Managing fleet defects, Improved schedules, Real-time bus information, All-door boarding, Stop consolidation, Signal timing, Bus arrival information, Automated vehicle location system and centralized control center, Smartcard fare collection, Low-floor buses, Route adjustments Key findings were summarized for several studies on bus reliability improvement strategies, as follows. Strathman et al. (2000) found that implementing bus dispatching systems in Portland, Oregon decreased running time by 1.45 minutes (3%). Strathman, Kimpel, Dueker, Gerhart, and Callas (2002) found that bus drivers play an important role in running time variation in Portland, after controlling for factors like route design, time of day, direction of service, and passenger activity. In this study operators running time relative to other operators decreased by 0.57 seconds for each month of additional experience. Dueker, Kimpel, Strathman, and Callas (2004) went on to study the impact of low-floor buses on dwell time, finding that low-floor buses reduced dwell time by an estimated 0.11 seconds (0.93%) per dwell. Transit signal priority impacts on running time, running time variation, headway, and on-time performance were found to vary across routes and time periods by Kimpel, Strathman, Bertini, Bender, and Callas (2005a). El- Geneidy, Strathman, Kimpel, and Crout (2006) found that bus stop consolidation helped to decrease running times by roughly 6%, with no significant impact on passenger demand, running time variation, or headway variation in Portland. Milkovits (2008) studied the impact of smartcards and bus type on dwell times in Chicago, estimating smartcards to be 1.5s faster than magnetic strip tickets when crowding is not present, when this difference was found to become insignificant. El-Geneidy et al. (2011) estimated a 0.34 decrease in running time for each additional year of driver experience in Minneapolis. Based on studies in Montreal, Surprenant-Legault and El-Geneidy (2011) reported that reserved bus lanes saved 1.3-2.2% of total running time, especially in the direction of peak travel and congestion. Furthermore, El-Geneidy and Vijayakumar (2011) found that articulated buses reduced dwell times, especially during high passenger activity periods, but did not reduce running times overall. Yetiskul and Senbil (2012) conducted a study in Ankara, Turkey, finding that travel time variation is impacted by temporal, spatial, and service factors. Lastly, Diab and El-Geneidy (2013) assessed the impacts of a set of strategies on bus service in Montreal, Canada, finding that smartcard fare payment actually increased running times and variation, with articulated buses, limited-stop service, and bus lanes having mixed impacts, and transit signal priority having no apparent impact. Review of other studies, with little or no new information. Diakaki, Dinopoulou , Papa- michail, & Garyfalia, 2015 Diakaki, C., Dinopoulou, V., Papamichail, I., & Garyfalia, M. (2015). State-of-the-Art and -Practice Review of Public Transport Priority Strategies. IET Intelligent Transport Systems. Retrieved from http://www.crossref.org/iPag e?doi=10.1049%2Fiet- its.2014.0112 Public transit priority strategies Researchers based at the Technical University of Crete and the Technological Education Institute of West Macedonia in Greece None None Facility-design-based measures, Signal- control-based measures Identified facility-design-based transit priority strategies include: exclusive bus lanes (with-flow or contra-flow), shared bus lanes (HOV, reversible), intermittent bus lanes (dynamic fairways, bus lanes with intermittent priority, bus-only roads, busways, bus gates and rising bollards, queue jumper or bypass lanes, bus bypass lanes, and bus advance areas. Signal-control-based public transit priority strategies include systems granting priority based on schedule or headway adherence, route progression, and downstream congestion. Inhibition and compensation strategies are considered. Priority methods outlined include green extension, stage recall (early green or red truncation), stage skipping, stage reordering or rotation, special stages, offset modification, cycle extension, retaken start, special plan (green wave), queue dissipation or clearance, and priority stage truncation. Fixed-time and real-time priority strategies are described in detail. Lots of detail and depth on possible improvement strategies, but no information on their effectiveness.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-126 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Gittens & Shalaby, 2015 Gittens, A., & Shalaby, A. (2015). Evaluation of Bus Reliability Measures and Development of a New Composite Indicator. TRB 94th Annual Meeting Compendium of Papers. Washington, D.C.: Transportation Research Board. Retrieved from http://trid.trb.org/view/2015/ C/1337354 Bus reliability measures Researchers based at the University of Toronto in Canada, Case study in London, England, United Kingdom Authors adopt a broad definition of reliable service as adhering to its schedule and being consistent. Reliability from the passenger perspective is defined as punctuality in arriving on time at the destination, short waiting times at the origin bus stop, and consistency of wait and travel times. A new measure of reliability is proposed, called Journey Time Buffer Index (JTBI). Statistically significant factors included route length, stop spacing, time of day, route orientation relative to the city center, and passenger load. None Two versions of the Journey Time Buffer Index (JTBI) were developed, one for service with short headways and one for long headways. JTBI accounts for travel time variability, combining an arrival penalty and a wait penalty. As demonstrated for bus service in London, the JTBI measure appears to be better suited to identifying reliability factors than measures that focus on a single component of reliability. The authors identify the following characteristics of good reliability indicators: easy to interpret, allows comparison between routes, allows comparison across indicators, is continuous, take the user's perspective into account, are objective and independent of arbitrary thresholds, meet data condition requirements, are reasonably affordable in terms of data collection, depict what they perceive to be occurring, and are consistent with travelers' perspectives. Indicators identified, but found to be lacking in some way, include travel time indicators (Reliability Buffer Time Metric, Buffer Index, Travel Time Window, Variability Index, Coefficient of Variation of Travel Time, Reliability Factor, and Punctuality Index on Routes), schedule adherence indicators (On-Time Performance, Weighted Delay Index, and Earliness Index), headway regularity indicators (Coefficient of Variation of Headway, Headway Regularity, Wait Assessment, Deviation Index based on Stops, Width Index, and Second Order Stochastic Dominance Index), wait time indicators (Excess Wait Time, 95th Percentile Wait Time, Potential Wait Time, and Excess Equivalent Wait Time), and composite indicators (Customer Journey Time Delay). Customer Journey Time Delay was found to be reflective of user experience, but difficult to apply to a bus network. Comprehensive review with wide applicability, but models and vetting of this new measure may be lacking. Hernandez, Munoz, Giesen, & Delgado, 2015 Hernandez, D., Munoz, J. C., Giesen, R., & Delgado, F. (2015). Analysis of Real- Time Control Strategies in a Corridor with Multiple Bus Services. Transportation Research Part B: Methodological, 78, pp. 83- 105. Retrieved from http://www.sciencedirect.co m/science/article/pii/S01912 61515000934 Analyzing real-time control strategies in a corridor with multiple bus services Researchers based at Pontificia Universidad Catolica de Chile in Chile Waiting time, Extra waiting time of passengers prevented from boarding, In-vehicle waiting time aboard a bus being held at a stop, Extra waiting time of passengers who must wait for another bus due to stop skipping, Waiting time due to traffic signals, Variance in bus headways Bus bunching Real-time control strategies (threshold based control strategies and mathematical programming), Boarding limits, Stop skipping, Signal priority Researchers considered control strategies and their impacts on bus bunching, passenger wait times, and wait time variability through simulations of a corridor with multiple bus services operating in it. The simulation results indicate that a central operator of public transport control systems in a particular corridor would lead to the greatest reduction in passenger wait times (55% compared to no control), as well as more balanced passenger loads, lower headway variability, and better reliability for all public transport users. Simulation- based results for a single corridor. Kieu, Bhaskar, & Chung, 2015 Kieu, L.-M., Bhaskar, A., & Chung, E. (2015). Public Transport Travel-Time Variability Definitions and Monitoring. Journal of Transportation Engineering, 141(1). Retrieved from http://ascelibrary.org/doi/10. 1061/%28ASCE%29TE.194 3-5436.0000724 Defining and measuring public transit travel-time variability Researchers based at Queensland University of Technology in Brisbane, Queensland, Australia Public transport travel-time variability (PTTV) is measured using the coefficient of variation of travel time at the corridor and service levels. Traffic congestion, Incidents, Length Transit signal priority, Bus holding strategies, Stop skipping The lognormal distribution is recommended as the descriptor of public transportation travel-time variation. Paper focuses on defining and measuring reliability, but does not test any of the mentioned improvement strategies. Li & Lu, 2015 Li, Y., & Lu, D. (2015). Bus Scheduling Model Based on Peak Hour Volume Clustering. American Society of Civil Engineers. Retrieved from http://ascelibrary.org/doi/10. 1061/9780784479292.100 Bus scheduling model based on passenger peak hour volumes Researchers based at Tongji University in China, Case study based in Shanghai, China None Traffic volumes, Peak travel periods, Passenger volumes Schedule optimization The researchers establish a model of passenger flow to aid in bus scheduling, based on passenger peak hour volume clustering from Smartcard data. This model is successfully demonstrated using Shanghai as a case study. Limited relevance overall, bus some discussion of bus scheduling to meet passenger flows.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-127 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Ma, Ferreira, Mesbah, & Hojati, 2015 Ma, Z.-L., Ferreira, L., Mesbah, M., & Hojati, A. T. (2015). Modelling Bus Travel Time Reliability Using Supply and Demand Data from Automatic Vehicle Location and Smart Card Systems. TRB 94th Annual Meeting Compendium of Papers. Washington, D.C.: Transportation Research Board https://journals.sagepub.com /doi/10.3141/2533-03 Bus service reliability measures Researchers based at Queensland University of Technology in Australia, Case study of Translink in Brisbane, Queensland, Australia Average travel time, Buffer time, Coefficient of variation of travel time Planning (link length, schedules, service frequency), Operational (departure delays, passenger activity, vehicle type, fare type, field supervision management), Environmental (route characteristics, traffic conditions, weather, incidents, road work) Smartcard fare collection, Reserved bus lanes, Limited-stop service, Stop consolidation, Articulated buses, Transit signal priority Statistically significant predictors of average travel time were found to be: Length between two stops (+), Delay at first stop of link (-), Number of actual stops (+), Number of boardings (+), Number of alightings (+), Alightings squared (-), Recurrent congestion index (- ), Number of signals (+), Rain versus good weather (+), Central business district (-), Arterial road (-), Busway (-), and Motorway (-). The buffer time model included the following predictors: Length between two stops (+), Number of actual stops (-), Standard deviation of delay at first stop of link (+), Standard deviation of actual stops (+), Standard deviation of boardings (+), Standard deviation of alightings (+), Recurrent congestion index (-), Number of signals (+), Central business district (-), Arterial road (-), Busway (-), and Motorway (-). The coefficient of variation in travel time model included: Length between two stops (-), Number of actual stops (-), Standard deviation of delay at first stop of link (+), Standard deviation of actual stops (+), Standard deviation of boardings (+), Standard deviation of alightings (+), Recurrent congestion index (-), Number of signals (+), Signals squared (-), Light rain versus good weather (+), Rain versus good weather (-), Central business district (-), Arterial road (-), Busway (-), and Motorway (-). Data is specific to Brisbane, Australia and may not be generalizable. Maltzan, 2015 Maltzan, D. W. (2015). Using real-time data to improve reliability on high- frequency transit services. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/ 99541 Real-time data use for improving reliability Researcher based at Massachusetts Institute of Technology in the United States, Case study of the Massachusetts Bay Transportation Authority in Boston, Massachusetts, United States None Operations planning, Control strategies, Staff behavior and attitudes, Unexplained operator deviations from scheduled departure times Operational control (terminal-based holding, holding at midpoints, short-turning) This thesis supports the finding that real-time data can be used to inform operational improvements. Holding at terminals and strategies for reducing operator deviations from scheduled terminal departure times are shown to have a strong effect on operations. Holding at midpoints and short-turning also seem to provide operational benefits, but more study is needed on the costs and benefits these strategies offer to transit passengers. Recommendations for operational improvements made to the Massachusetts Bay Transportation Authority include better driver and supervisor training about on-time terminal departures, management intervention with drivers who are habitually not on time, use of departure time displays at bus berths, allowing boarding during layovers at terminal stations, rear-door boarding, and instruction on when and how inspectors should use express, deadheading, and short-turning strategies. This thesis relied largely on simulation results. New York Metropolita n Transportat ion Authority, 2015 New York Metropolitan Transportation Authority. (2015). MTA Performance Indicators. New York. Retrieved from http://web.mta.info/persdash board/performance14.html# Service indicators New York City Transit in New York City, New York, United States Reliability service indicators for MTA bus include percent of completed trips and mean distance between failures None None Reliability-related indicators also include bus passenger wheelchair lift usage, total ridership, collisions with injury rate, and employee lost time rate. This is not a study, but rather an online performance dashboard.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-128 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Sanchez- Martinez, 2015 Sanchez-Martinez, G. E. (2015). Real-Time Operations Planning and Control of High-Frequency Transit. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/ 99550 Real-time operations planning and control Researcher based at Massachusetts Institute of Technology in the United States None Signal failures, Disabled vehicles, Demand surges, Traffic, Right-of-way, Weather, Infrastructure, Resource allocation, Planning, Operations control, Lane blockages Control strategies (holding, optimization, stop skipping) Key contributions from this dissertation include: development of a holding control model using dynamic running time and demand data to, making it possible to adopt holding policies in real-time that reflect predicted changes in running times and demand; demonstration that dynamics information can be used in a holding strategy to improve the performance of crowded high-frequency transit, though this strategy may not have the same effect on less crowded system; a control framework for high-frequency transit using data about expected and unexpected events; evidence that accurate event information can be used to improve performance; development of a schedule-free paradigm high-frequency transit operations planning; a proposed real-time schedule-free planning methodology with holding and stop skipping; and simulation results indicating that the proposed schedule-free paradigm is not only feasible, but also leads to reduced average passenger waiting times. Focused on high-frequency transit. Sanchez- Martinez, Koutsopoul os, & Wilson, 2015 Sanchez-Martinez, G. E., Koutsopoulos, H. N., & Wilson, N. H. (2015). Event- Driven Holding Control for High-Frequency Transit. Transportation Research Record: Journal of the Transportation Research Board, No. 2535, pp. 65-72. https://journals.sagepub.com /doi/abs/10.3141/2535-07 This paper applies a holding method developed by the same authors on a simulated route with unexpected disruptions. Researchers based at Massachusetts Institute of Technology in the United States Cost function of weighted waiting time components Surges in demand with uncertain volumes and times. Simulation-based holding method The authors find that simulation-based optimization can help contain disruptions with accurate and real-time information, but that inaccurate information increases the cost relative to the naive strategy. The method only reduces the cost relative to naïve method if real-time predictions of passenger arrivals are accurate. U.S. Depart- ment of Transporta- tion, Federal Transit Administra- tion, 2015 U.S. Department of Transportation, Federal Transit Administration. (2015, June 23). Bus Testing: Establishment of Performance Standards, a Bus Model Scoring System, a Pass/Fail Standard and Other Program Updates; Proposed Rule. Federal Register, 80(III). Retrieved from http://www.apta.com/gap/fed reg/Documents/FTA%E2%8 0%932015%E2%80%93001 9-NPRM- Bus%20Testing.pdf Bus testing performanc e standards Federal Transit Administration in the United States None Vehicle quality and maintenance Fleet maintenance and periodic replacement The Federal Transit Administration (FTA) proposes a reliability performance standard of zero uncorrected Class 1 reliability failures and no more than two uncorrected Class 2 reliability failures at the completion of a 125 hour reliability test. Class 1 failures are described as malfunctions that could lead to a loss of bus control, serious injury, and/or property damage or loss due to a collision or fire. Class 2 failures are malfunctions that result in test interruption due to a lack of bus operation, but that could be repaired without serious safety risks. These are proposed standards, that may be changed and require formal adoption at various levels of government.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-129 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Wallis check, Weisenber ger, Berthaume, & Dinning, 2015 Wallischeck, E. Y., Weisenberger, T., Berthaume, A., & Dinning, M. G. (2015). TCRP Report 177: Preliminary Strategic Analysis of Next Generation Fare Payment Systems for Public Transportation. Washington, D.C.: Transportation Research Board of the National Academies. Retrieved from https://www.nap.edu/catalog /22158/preliminary-strategic- analysis-of-next-generation- fare-payment-systems-for- public-transportation Fare payment methods Researchers based at the U.S. Department of Transportation in Cambridge, Massachusetts, United States, Case studies of UTA and SEPTA None Fare payment methods, Reliability of fare equipment Flash passes and other contactless fare collection methods Several existing and emerging fare payment technologies are described in this report. More research should be conducted to compare the transaction times and reliability of various systems, to determine which is the most optimal at this time. Almost no discussion of the impacts of fare collection methods on service reliability. Wood, 2015 Wood, D. A. (2015). A Framework for Measuring Passenger-Experienced Transit Reliability Using Automated Data. Civil Engineering. Cambridge, MA: Massachusetts Institute of Technology. Retrieved from http://hdl.handle.net/1721.1/ 99539 Measuring service reliability from the passenger perspective Researcher based at Massachusetts Institute of Technology in the United States, Case study of Hong Kong MTR in Hong Kong, China Two new reliability metrics are proposed: Individual Reliability Buffer Time (IRBT) and Platform to Platform Reliability Buffer Time (PPRBT). Passenger demand, Incidents Monitoring reliability and working to manage and improve it periodically Passenger-focused reliability measures can benefit transit agencies by helping them to better manage overcrowding and its effects, track impacts of reliability improvements, improve customer communications, enhance passenger information systems, and engage in effective long-term planning. Design objectives for passenger-focused reliability metrics include that: they are inclusive of all sources of unreliability; distinguish between service variability and schedule adherence, as well as early and late arrivals; control for variation in passenger behavior and time of day; exclude extreme delays; calculated at the origin-destination pair level; and are unbiased with regard to passenger demographics. Metrics should be meaningful for passengers and non-experts, in that they are understandable, objective, and useful for planning journeys. Measures should ideally be comparable across different services and times of day, independent of schedules, and absolute (as opposed to relative). Time period flexibility, meaning that a metric is calculable for short time intervals and allows for exclusion of weekends, holidays, and specific events, and service scope flexibility are also desirable traits. Reliability Buffer Time is discussed at length, and other measures mentioned include Excess Reliability Buffer Time (ERBT), Excess Journey Time (EJT), and Passenger Journeys on Time. Two new metrics are Individual Reliability Buffer Time (IRBT) and Platform to Platform Reliability Buffer Time (PPRBT). The identified metrics can be used to help identify causes of unreliability, among other benefits. The IRBT alone is recommended for geographic equity analysis. Case study is based on a rail system. No bus system case studies are used. WSP | Parsons Brinckerhof f, 2015 WSP | Parsons Brinckerhoff. (2015). Bus Journey Time Variability in Urban Areas. Department for Transport. United Kingdom. Bus travel time variability Researchers based in the United Kingdom Number of services run versus scheduled, On-time performance (proportion of buses departing from time points between 1 minute before and 5 minutes after the published timetable) General traffic, Priority at intersections, Route length, Number of stops, Exclusive bus lanes, Vehicle spacing policies, Ticket type, Physical stop characteristics, Driver experience, Number of doors, Frequency of public transport use, Weather conditions Coefficient of journey time was estimated to improve with stop skipping and consolidation and less variation in passenger activity. Factors identified as relating to bus travel time variability in this study were assigned an implied importance, based in part on available literature. The factors with a "High" implied importance were traffic conditions, priority at intersections, route length, and number of stops. Factors assigned a "Medium" implied importance include exclusive bus lanes, vehicle spacing policies, and ticket type. Those assigned a "Low" implied importance were bus stop characteristics, bus driver experience, number of doors, frequency of public transport use, and weather conditions. Analysis based largely on literature review, simulation, and modeling.

Developing a Guide to Bus Transit Service Reliability Appendix A – Literature Review A-130 Citation Reference Focus of Paper Entities Involved Measures / Definitions Factors Improvement Strategies Findings and Results Study Limitations Yu, Chen, Liu, & Zhu, 2015 Yu, S., Chen, J., Liu, Z., & Zhu, L. (2015). Study on a Model and Demonstration of Public Transit Priority Technology Strategy at an Urban Arterial Road. American Society of Civil Engineers. Retrieved from http://ascelibrary.org/doi/10. 1061/9780784479292.134 Transit signal priority technology strategy for an urban arterial road Researchers based at Southeast University and the Jiangsu Province Collaborative Innovation Center of Modern Urban Traffic Technologies in China Headway consistency index (covariance of headways, which is the standard deviation divided by the mean) Illegal occupation of bus lanes, Poor design of bus stations, Lack of bus-only approaches, Poor convergence of multilevel bus lines, Sparse bus departure frequency Improvement of bus lanes (pavement color and markings), Bus station optimization design (A and B stations), Bus priority at intersections (bus-only approaches, priority signal control during peak hours) The improvement strategies had a beneficial effect overall for bus service operation efficiency and reliability, with minimal impacts on general traffic. Simulation- based results do not show impact of individual improvements, but rather the combined impact of all improvements. Gayah, Yu, & Wood, 2016 Gayah, V. V., Yu, Z., & Wood, J. S. (2016). Estimating Uncertainty of Bus Arrival Times and Passenger Occupancies. San Jose, CA: Mineta National Transit Research Consortium. Estimating uncertainty of real-time bus arrival information Researchers based at the University of San Jose in the United States, Data from Centre Area Transportation Authority (CATA) Uncertainty is measured as mean squared error. Uncertainty of bus arrival times, Passenger load, Weather, Headways Providing uncertainty information to qualify real-time bus arrival predictions The report used modeling of real data to estimate uncertainty of travel times and passenger loads. The researchers found that uncertainty in travel time increases with mean travel time, and can be predicted based on the travel time of the previous bus over a certain segment and passenger loads. Weather was not found to significantly impact travel times. For modeling passenger loads, "next-stop" modeling framework was found to be most accurate for 1-5 stops away, while "segment-based" modeling was preferred for predictions of buses that are more than 5 stops away. Small headways, precipitation, and low temperatures were associated with more uncertainty with regard to passenger loads. Snow, however, was correlated with a decrease in passenger load uncertainty. This report focused on uncertainty information for real-time bus arrival information predictions. NACTO 2017 NACTO, (2017) Better Boarding, Better Buses: Streamlining Buses and Fares Reducing fare payment time National Association of City Transportation Officials Reducing fare payment time Station type Implement off-board fare payment The report present case studies of bus routes that have implemented off-board fare payment. The treatment has been found to improve travel time and overall reliability of the routes. None