Incorporating Travel Time Reliability into the Highway Capacity Manual (2014)

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.

1This final report documents the activities performed during SHRP 2 Reliability Project L08: Incorporating Travel Time Reliability into the Highway Capacity Manual. It serves as a supple- ment to the proposed chapters for incorporating travel time reliability into the Highway Capacity Manual prepared for this same project. The proposed chapters demonstrate how to apply travel time reliability methods to the analysis of freeways, urban streets, and corridors. The final report summarizes the work activities conducted during the course of the Phase 1 and Phase 2 research by memorializing the activities, the processes, and the findings of the L08 project. In this way, the final report articulates the how and why of key decisions made and key activities under- taken during the project so that the logic and rationale are not lost to future researchers and practitioners who aim to build on the work completed in this effort. This report addresses the following topics: • The project’s purpose, objectives, and work tasks (Chapter 1); • The research team’s proposed definition of reliability, along with means for measuring reliability (Chapter 2); • A state-of-the-art and state-of-the-practice literature review (Chapter 3); • An overview of the methodologies for calculating reliability for freeways and urban streets (Chapter 4); • A description of the development of freeway and urban street scenario generators (Chapter 5); • Enhancements to the Highway Capacity Manual (HCM) freeway facilities methodology and its computational engine FREEVAL-RL (FREeway EVALuation—ReLiability), and enhance- ments to the urban streets methodology and its computational engine STREETVAL (STREET eVALuation) (Chapter 6 and Appendices A and B); • A procedure for conducting a corridor application (Chapter 7); and • A method for future consideration to define levels of service using reliability as a service measure and a discussion of future research needs (Chapter 8). The remainder of this executive summary provides a brief overview of the research results. Definition for Travel Time Reliability Travel time reliability aims to quantify the variation of travel time. It is defined using the entire range of travel times for a given trip, for a selected time period (e.g., the p.m. peak period on weekdays) over a selected horizon (e.g., a year). For the purpose of measuring reliability, a trip can be defined as occurring on a specific segment, facility (combination of multiple consecutive segments), or any subset of the transportation network; or the definition can be broadened to include a traveler’s initial origin and final destination. Measuring travel time reliability requires Executive Summary

2that a sufficient history of travel times be present to track travel time performance. This history is described by the travel time distribution for a given trip. Once the travel time distribution is established, performance measures can be established to capture reliability. The two general types of reliability performance measures are the following: 1. Those that capture the variability in travel times that occurs for a trip over the course of time; and 2. Those that reflect the number of trips that fail or succeed according to a predetermined performance standard or schedule. In both cases, reliability (more appropriately, unreliability) is caused by the interaction of the factors that influence travel times: fluctuations in demand (which may result from daily or sea- sonal variation, or special events), traffic control devices, traffic incidents, inclement weather, work zones, and physical capacity (based on prevailing geometrics and traffic patterns). These factors produce travel times that vary from day to day for the same trip. The following terms, illustrated in Figure ES.1, are used throughout this report: 1. Analysis period is defined as the smallest time unit for which the HCM analysis procedure is applied. In the case of freeway and urban street facility analysis, the HCM analysis period is 15 min, although it can be of greater duration at the discretion of the analyst. Alternative tools may define different analysis period lengths. 2. Study period is defined as the sum of the sequential analysis periods for which the HCM facil- ity analysis procedure is applied (e.g., a 4-hour peak period). The study period is defined by the analyst for each specific application, on the basis of the guidance provided in the HCM. 3. Reliability reporting period is defined as the period over which reliability is to be estimated (e.g., the 250 nonholiday weekdays in a year). In essence, the reliability reporting period specifies the number of days for which the reliability analysis is to be performed. Reliability Metrics (for Use as Performance Measures) Travel time reliability relates to how travel times for a given trip and time period perform over time. From a measurement perspective, reliability is quantified from the distribution of travel times—for a given facility or trip and the time period (e.g., weekday peak period)—which occurs over a sig- nificant span of time. One year is generally long enough to capture nearly all of the variability caused by disruptions. A variety of metrics can be computed once the travel time distribution has HCM Study Period HCM Facility 66 66 69 70 63 66 66 66 66 68 68 65 69 63 63 63 68 66 60 67 63 39 64 64 64 70 70 65 38 39 67 67 62 64 68 40 18 37 69 69 64 70 37 14 14 40 65 65 69 39 25 21 16 37 69 69 66 65 38 13 11 37 70 70 68 63 62 40 18 38 67 67 63 63 62 68 40 37 68 68 64 61 65 62 61 39 61 61 63 63 60 65 67 63 63 63 65 70 64 63 67 64 64 6415:00 18:00 Reliability Reporting Period Da ily Re pe titi on s HCM Analysis Period HCM Analysis Segment Figure ES.1. Reliability terms.

3been established, including standard statistical measures (e.g., standard deviation, kurtosis), percentile-based measures (e.g., 95th percentile travel time, buffer index), on-time measures (e.g., percentage of trips completed within a travel time threshold), and failure measures (e.g., percentage of trips exceeding a travel time threshold). Some of these metrics are shown in Figure ES.2. The set of performance measure metrics listed in Table ES.1 is recommended for Project L08. Both variability- and failure-based metrics are included. Which metric should be highlighted as the primary reliability metric is difficult to say. Much depends on the specific application being used. When interpreting Table ES.1, it should be noted that many of the selected performance measures are defined relative to the free-flow travel time, rather than the average travel time. This is deliberate because the average travel time (a) is not known before the analysis is conducted, (b) varies between different facilities, and (c) varies between different scenarios for the same Figure ES.2. Travel time distribution as the basis for defining reliability metrics. Table ES.1. Recommended Reliability Performance Measure Metrics for SHRP 2 Project L08 Reliability Performance Measure Definition Core Measure Reliability rating Percentage of trips serviced at or below a threshold travel time index (TTI) (1.33 for freeways, 2.50 for urban streets) Planning time index (PTI) 95th percentile TTI (95th percentile travel time divided by the free-flow travel time) 80th percentile TTI 80th percentile TTI (80th percentile travel time divided by the free-flow travel time) Semistandard deviation The standard deviation of travel time pegged to free-flow travel time rather than the mean travel time (variation is measured relative to free-flow travel time) Failure or on-time measures Percentage of trips with space mean speed less than 50, 45, and/or 30 mph Supplemental Measure Standard deviation Usual statistical definition Misery index (modified) The average of the highest 5% of travel times divided by the free-flow travel time

4facility. Performance measures based on the average travel time are therefore deemed to be less appropriate for HCM analysis. The distribution of travel times is the starting point for measuring reliability. In a statistical sense, the distribution is continuous only if it is based on measuring travel times from indi- vidual vehicles. As of this writing, the data used to monitor travel times—as well as modeling methods—rarely are managed in this way. For example, consider roadway detectors of spot speeds, which measure every vehicle that crosses their detection zone. These systems are designed to aggregate measurements in the field to 20- or 30-s summaries before transmission. So, in its lowest form, the speed “measurement” is really an average. The data are sometimes further aggre- gated to 1-, 5-, or 15-min summaries for archiving. At each aggregation, variability in the mea- surements is reduced. (When aggregating travel times over analysis periods, it is extremely important to weight the travel time averages by volume or vehicle miles traveled, rather than taking just the arithmetic mean.) Similarly, Bluetooth-based vehicle reidentification has a sam- pling rate well below 100%. Methodology for Calculating Reliability The objectives of SHRP 2 Project L08 are twofold. The first objective is to incorporate nonrecur- ring congestion effects into the HCM procedure. The second objective is to expand the analysis horizon from a single study period (typically an a.m. or p.m. peak period) to an extended time horizon of several weeks or months to assess the variability and the quality of service the facility provides to its users. This expanded period—referred to as the reliability reporting period—can be thought of as a series of consecutive days, each one having its own set of demands and capaci- ties that affect the facility travel time. This study focused on weather, incidents, work zones, and special events on the supply side, and on volume variability by time of day, day of week, and month of year on the demand side. Separate methodologies are used to evaluate reliability for freeway facilities and for urban streets, although many parallels exist between the two methods. Freeway Facilities Methodology At its highest level of representation, the freeway facilities methodology has three primary com- ponents: a data depository, a scenario generator, and a core computational procedure, which is an adapted and significantly revised version of the FREEVAL computational engine for reliabil- ity, or FREEVAL-RL. These components are illustrated in Figure ES.3. Figure ES.3. Freeway facilities methodology components, including measures of effectiveness (MOEs).

5The largest shaded oval and dotted line represent the current implementation of the HCM freeway facilities chapter, with study period data specific to the facility being studied entered directly into FREEVAL-RL for analysis of (predominantly) recurring congestion effects. The connection to reliability is enabled by the addition of a scenario generator. Each component and its interaction with the other two are explained in some detail in the following sections. The freeway scenario generator (FSG) developed by the L08 research team assigns initial prob- abilities to a number of base scenarios. A base scenario probability is expressed as the fraction of time a particular combination of events takes place during the study period (SP) of interest (e.g., a.m. or p.m. peak periods). In this project, a scenario is akin to a study period, which may or may not contain a given combination of weather or incident events. Base scenario probabilities are computed assuming independence between the events, and at that initial stage do not take into account the actual duration of the event (weather or incident) in question. They only take into account the categories of weather and/or incidents. (See Appendices C through G for informa- tion about recurring demand for the freeway scenario generator, weather and incident-related crash frequencies, and weather modeling.) Urban Street Facilities Methodology The reliability methodology for urban streets consists of the following three components: • Scenario generation; • Facility evaluation; and • Performance summary. These components are used in sequence to generate, evaluate, and summarize the various scenarios that make up the reliability reporting period. The HCM2010 urban streets methodol- ogy (implemented in a computational engine) is used to estimate the travel time and other performance measures associated with each scenario (TRB 2010a). (For information about default factors for urban streets reliability methodology, see Appendix H.) The sequence of calculations in the reliability methodology is shown in Figure ES.4. The pro- cess is based on the urban streets engine. It begins with one or more engine input data files. An input file is modified during the scenario generation stage to reflect demand variation and the effect of other causes of nonrecurring congestion on running speed and saturation flow rate, as they occur during the reliability reporting period. Input Data Scenario Generation Stage Facility Evaluation Stage Performance Summary Stage End Urban Streets EngineInput File Figure ES.4. Urban streets methodology components.

6Once all of the scenarios associated with the reliability reporting period have been generated, they are evaluated during the facility evaluation stage. The urban streets computational engine is used to automate the calculations. The evaluation results are then summarized during the performance summary stage. Various travel time distribution statistics and reliability perfor- mance measures are calculated for through vehicles traveling along the facility. The freeway facility and urban street facility reliability models use different methods to develop a travel time distribution for the reliability reporting period: • The freeway facility method develops scenarios on the basis of their probability of occurrence during the reliability reporting period. Some highly unlikely scenarios may be dropped from the analysis. • The urban streets method randomly assigns demand, weather, and incident conditions to each day, on the basis of distributions of conditions likely to occur within a month. Some highly unlikely combinations may be included by random chance; therefore, multiple runs of the method may be needed to establish a representative travel time distribution. Importantly, no direct link exists between the two methods. The weather pattern generated by the urban streets method may produce more or less severe conditions over a given model run compared with the 10-year average weather conditions used by the freeway method. An incident scenario for the freeway does not generate a corresponding high-demand scenario for the urban street. When local data are used to generate demand patterns, traffic diversion effects will appear in individual days’ demands used to create month-of-day factors; but the effects of days with diversion will likely be washed out by demands from all of the days without diversion. When default demand pattern data are used, there is no diversion effect at all beyond that resulting from bad weather (and associated higher incident rates) occurring more often in some months of the year than in others. Development of Scenario Generators Scenario Generation for Freeway Facilities A deterministic approach to scenario generation is proposed for freeway facilities. This determin- istic approach enumerates different operational conditions of a freeway facility on the basis of different combinations of factors that affect travel time. These operational conditions are expressed as operational scenarios or, simply, scenarios. Four principal steps explain the construction of the scenario generation process for freeway facility analysis, as depicted in Figure ES.5. The three main contributors to travel time variability on a freeway facility are variable demand level, weather, and incidents. Further, user-defined effects of work zones and special events can be incorporated as scenarios in the reliability analysis. These factors introduce stochasticity to Figure ES.5. Process flow overview for freeway scenario generation.

7travel time. In other words, they generate a travel time distribution instead of a deterministic and fixed travel time, as would be obtained by running a single study period. The reliability of a freeway facility is expressed as the portion of time in which the facility operates at or above the reliability standard set by the implementing agency. The freeway scenario generation process uses a deterministic approach to model these varia- tions. It categorizes different sources of variability (e.g., demand patterns or incident types) into different subcategories. For instance, weather—which is one of the main contributors to travel time variability—is defined in 11 weather categories (e.g., normal weather, medium rain, snow). Each category has a time-wise probability of occurrence and an impact on facility capacity, speed, and possibly demand. Thus, while the resulting distribution of travel times is stochastic, the process for generating scenarios is not; rather, it takes the approach of enumerating (nearly) all viable scenarios, each associated with varying probabilities of occurrence. The mathematical performance model starts from the development of base, study period, and detailed scenarios. The latter are forwarded to the computational engine FREEVAL-RL for esti- mating analysis period facility travel times. While full automation has yet to be accomplished, the process readily lends itself to automation. Scenario Generation for Urban Street Facilities The scenario generation for urban streets consists of four sequential procedures. Each procedure processes the set of analysis periods in chronologic order. • The first procedure predicts weather event date, time, type (i.e., rain or snow), and duration. • The second procedure identifies the appropriate traffic volume adjustment factors for each date and time during the reliability reporting period. • The third procedure predicts incident event date, time, and duration. It also determines inci- dent event type (i.e., crash or noncrash), severity level, and location on the facility. • The fourth procedure uses the results from the preceding three procedures to develop one urban streets engine input file for each scenario in the reliability reporting period. Enhancements to the HCM Base Methodologies Freeway Facilities Enhancements The adaptation of the freeway facilities method developed by SHRP 2 Project L08 for performing a reliability analysis required several changes and enhancements to make the HCM methodology and associated computational engine “reliability ready.” The enhanced computational engine is named FREEVAL-RL. The list of enhancements is as follows: • Incorporation of the two-capacity phenomenon under queue discharge conditions; • Improved modeling of capacity adjustment factors (CAFs) and speed adjustment factors (SAFs) for basic, merge, diverge, and weaving segments; • New default values for CAF and SAF for incident and weather events on freeways; • Enhanced performance measures for congested conditions; and • Automation of computations. The freeway reliability methodology can generate several thousand scenarios, many of which may have exceptionally low or exactly zero probability. In addition, some scenarios may be infea- sible. The infeasible scenarios are automatically filtered out by the freeway scenario generation procedure. The scenarios with extremely low probability are not expected to be observed in the field in a single year; however, they are included in the predicted travel time index (TTI) distribu- tion. This makes the comparison of predicted and observed distributions hard to interpret. In

8addition, these scenarios tend to have exceptionally large TTI values that significantly shift the tail of the cumulative distribution to the right (i.e., toward higher TTI values). Finally, these scenarios may also result in demand shifts in the real world that are not directly accounted for in the freeway reliability method. To address these differences between predicted and observed distributions, the procedure allows the user to specify an “inclusion threshold” to include only scenarios with probabilities larger than the threshold specified for the analysis. For instance, an inclusion threshold of 1.0% means that only the scenarios with probabilities larger than 0.01 are considered in the analysis. Figure ES.6 presents the TTI cumulative distributions for four different inclusion threshold values for the case study of I-40 in Raleigh, North Carolina, and compares them with the observed TTI distribution obtained from the INRIX.com data warehouse. Urban Street Enhancements Three enhancements were made to the HCM2010 urban streets methodology. The first is a pro- cedure for adjusting the discharge rate from a signalized intersection when a downstream inci- dent or work zone blocks one or more lanes on the segment. The second is a procedure for computing the effective average vehicle spacing on a segment with spillback. The third is a meth- odology for using the HCM methodology to evaluate urban street facilities with spillback in one or both travel directions on one or more segments. Validation of the enhanced methodology was based on a comparison of performance esti- mates obtained from a traffic simulation model. Three street segments were selected for the evaluation. The findings from this activity indicated that the enhanced methodology was able to provide accurate estimates of delay during congested and uncongested conditions. Corridor Applications A corridor study, by definition, goes beyond the single-facility focus of a typical HCM facility analysis. The purpose of a corridor study is to assess the ability of a subsystem of interrelated facilities to achieve a set of transportation performance objectives. For the purposes of a reli- ability analysis, corridor is defined as a freeway facility and one or more parallel urban street facilities. When traffic diversion occurs between the facilities in a corridor, the freeways, Figure ES.6. Predicted freeway HCM TTI distribution with different inclusion thresholds versus INRIX for 2010.

9highways, and urban streets that cross the corridor and provide connections between the corri- dor’s facilities will also be affected; however, those effects are beyond the scope of a reliability analysis. The focus of a corridor evaluation is on the parallel facilities. An analysis of overall corridor reliability involves comparing selected reliability performance measures (e.g., travel time index, planning time index, percentage of on-time arrivals) generated for the individual facilities against either an established standard or against comparative national values of reliability. Because different agencies may be responsible for different facilities within a corridor (or, in the case of urban streets, different portions of the facility), and because corridor analysis focuses on longer-distance travel, a regional standard might be most appropriate. In the absence of such a standard, a percentile threshold could be used. In that case, unacceptable per- formance could be defined in terms of, say, a facility’s planning time index (PTI) (e.g., among the worst 20% of U.S. facilities). Potential Methods for Defining Level of Service by Using Reliability as a Service Measure The research team initially considered four options as potential methods for defining level of service (LOS) by using reliability. Briefly, the options are as follows: • Reliability LOS based on current LOS ranges. This option is the most consistent with current LOS concepts in the HCM. Inherently, a reliability analysis captures a range of operating con- ditions on the same facility, which are attributed to the various sources of (un)reliability. Using a distribution of LOS values therefore intrinsically mirrors the variability of traffic conditions on the facility. • Freeway reliability LOS based on travel speed ranges. This option makes freeway reliability LOS conceptually consistent with urban streets and urban street segments. The problem of pre- senting a distribution rather than a single LOS value is still present. • Freeway reliability LOS based on most-restrictive conditions. This option avoids the problem of presenting a distribution and assigns a single LOS value. It is more complicated to apply and explain in that two values must be set: a percentage threshold for the trips that fail to meet the LOS criteria and the ranges for each LOS category. • Reliability LOS based on the value of travel. This option is the most complicated both to develop and explain. It has the advantage of being based on travelers’ perception of reliability, but it relies on a factor (the reliability ratio, used to measure how travelers value reliability) that has not been precisely identified and will likely change with new research. Not only is this option complex, but establishing LOS ranges based on travel time equivalents is highly problematic. Testing the four options with field data failed to reveal a clear choice on which to base reli- ability LOS. Furthermore, the four options were thought to be difficult to communicate to the profession, the public, and decision makers. As a result, the research team decided to develop an on-time–based measure, similar to Option 2. This measure, termed the reliability rating, is the percentage of trips serviced at or below a threshold TTI (the ratio of the actual travel time to the free-flow travel time). The TTI thresholds selected were 1.33 for freeways and 2.50 for urban streets. These thresholds approximate the points at which most travelers would consider a facility congested; thus, the measure roughly reflects the percentage of trips on a facility that experience conditions better than level of service F (LOS F). The difference in threshold TTI values results from differences in how free-flow speed is defined for freeways compared with urban streets, as TTI is measured relative to free-flow speed. The research team did not define a service measure for travel time reliability. Because travel time reliability is a new concept for the transportation profession, the research team recommends that performance measures be used to describe the travel time reliability performance on freeways and urban streets. Subsequently, consideration can be given to using travel time reliability to

10 define level of service. When reliability is considered as a service measure, the research team recommends that the reliability rating (now a performance measure) be the basis. Other considerations for future reliability LOS deliberations are as follows: • Urban streets. Figure 16-4 of HCM2010 defines LOS F as either (1) where the travel speed is 30% or less of the base free-flow speed or (2) where the subject through movement at one or more intersections has a volume-to-capacity ratio greater than 1.0 (TRB 2010a). Because the LOS definition is based on travel speed, which is a derivative of travel time, no changes in the LOS concept for urban streets is needed. • Freeways. For freeway reliability, the research team first recommends that the existing density- based LOS definition be replaced with a travel speed–based definition. Density should be maintained as the indicator of general freeway performance, especially for rural facilities. The research team recommends that, in the future, travel speed be considered as a replacement to density even for general performance on urban facilities. The use of travel speed as the indica- tor of both general and reliability performance on freeways also provides consistency with the urban streets method. (See Appendix I for an example of existing freeway reliability.) Future Research Needs Supporting Implementation of the SHRP 2 Project L08 Research The proposed HCM reliability chapters and the FREEVAL and STREETVAL software computa- tional engines were completed in 2012 and reviewed by the TRB Highway Capacity and Quality of Service Committee in conjunction with the 2013 TRB annual meeting. The computational engines consist of spreadsheets with embedded Visual Basic code. Separate Excel spreadsheet tools are used for generating the scenarios and then running the FREEVAL and STREETVAL engines to execute the HCM calculations in an automated fashion and process the results for reliability reporting purposes. While not part of the L08 project, a natural extension of the computational engines and other tools would be the development of a more user-friendly, inte- grated software tool that would execute the files faster than the Excel-based computational engines. Such a software tool could be hosted on a fast server and could be located in any secure environment, including a cloud-based environment. At present, the updated FREEVAL and new STREETVAL computational engines are hosted in the developer’s environment at the contractor’s site. Freeway Facility Research Needs Research needs in the freeway facilities methodology incorporate improvements to the core HCM methodology and to the submodels developed in the course of this study. Research to Overcome Methodology Limitations Although the research team was able to improve and expand the freeway facility methodology significantly during the course of this study, additional research is still needed to fill some gaps: • The oversaturated flow-density relationship has not been calibrated since its inception in HCM2000. • The spillback from off-ramps is not considered in the current methodology, significantly weak- ening its ability to model congested corridors. • The free-flow speed and capacity adjustment factors used throughout the methodology to account for nonrecurring congestion effects have been adopted from the most recent and relevant literature and have not been locally calibrated or validated. • The methodology does not include the effect of managed lanes on reliability.

11 Research to Improve the Reliability Submodels Research is needed to understand and quantify the impact of weather, work zones, or special events on traffic demand: • The method assumes that incident rates and weather conditions are independent. Research is needed to develop models that can explain the relationship. • The current methodology does not account for weather events that have a small effect on seg- ment capacity reduction (<4%). In addition, a given weather event (e.g., rain, snow) is always assumed to occur at its mean duration value. Furthermore, only two possible start times for weather events are considered. • To consider the average effect of incidents on a facility, an incident is modeled only on three possible segments: the first segment, the segment at the facility midpoint, and the last segment. The timing of the incident is either at the start of a study period or at its midpoint. Finally, only three possible incident durations are considered: the 25th, 50th, and 75th percentiles of the incident duration distribution. Urban Streets Research Needs Future urban streets research is divided into two categories. The first category describes the research needed to overcome known limitations in the scope of the urban streets reliability methodology. The second category describes research needed to improve specific models within the reliability methodology. Research to Overcome Methodology Limitations In general, the urban streets reliability methodology can be used to evaluate the performance of most urban street facilities. However, the methodology does not address some events or condi- tions that occur on some streets and influence their operation. These events and conditions are identified as follows: • Facilitywide performance measures; • Truck loading and delivery; • Signal malfunction; • Railroad crossing and preemption; and • Adverse weather conditions. Research to Improve Specific Models The urban streets reliability methodology was developed using currently available data and research publications. The data were used to calibrate the various models that make up the meth- odology. Calibration data were also collected in the field when existing data were not available. In some instances, the research team noted that a model’s reliability could be improved if addi- tional data were collected or made available through subsequent research. The following list identifies the specific models that would benefit from additional research: • Wet-pavement duration; • Effect of weather on signalized intersection saturation flow rate; • Effect of incident length on segment operation; and • Incident distribution.