Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
17 KEY QUESTIONS ⢠What measures are available for monitoring reliability performance? ⢠How should the measures be tailored to reflect the reliability needs of the system? ⢠What is the best way to communicate performance measures to various audiences? SELECTING A PERFORMANCE MEASURE It is critical to select a performance measure that can help users understand how reli- ability affects them on an intuitive level and to help planners and operators through- out the agency understand why reliability is important. Fundamentally, reliability 2 MEASURING AND TRACKING RELIABILITY Color versions of the figures in this chapter are available online: www.trb.org/Main/Blurbs/168855.aspx. Performance measures provide the technical basis for monitoring performance, setting program funding levels, and prioritizing projects. Performance measures can support goal setting by demonstrating the significance of a given need and can be used to help set program funding levels or prioritize projectsâthe key steps of a performance-based process. Performance measures provide an opportunity to âlevel the fieldâ or allow comparison of unlike programs or benefits (e.g., comparing capacity addition to operational or other programs) for the purposes of finding the right package of strategies to address transportation needs.
18 GUIDE TO INCORPORATING RELIABILITY PERFORMANCE MEASURES INTO THE TRANSPORTATION PLANNING AND PROGRAMMING PROCESSES measures variability in travel times. There are several ways to capture this variability, and this chapter describes the meaning of these measures. Figure 2.1 and Table 2.1 define, describe, and illustrate the calculation of common measures used to describe travel time reliability. As they indicate, the measures are all based on the travel time distribution. Typically, travel time data used to calculate these distributions are captured at a fine-grained level (e.g., travel times on a facility every 5 min). Chapter 2 of the technical reference provides additional details on how to use travel time data to calculate reliability performance measures. This guidance has been developed before the FHWA has issued regulations on performance measures that will be required as part of MAP-21 implementation. When they become available, agencies should consult the regulations when selecting an appropriate performance measure. Agencies are encouraged to estimate multiple reliability performance measures to provide a robust perspective on reliability. Individual measures capture different slices of the travel time distribution and may suggest different strategies to employ. Figure 2.2 illustrates this point, providing three points on the travel time distribu- tion (average TTI, 80th percentile TTI, 95th percentile TTI) for several real corridors. Looking at these three points together provides additional perspective on the specific challenges each corridor faces and potentially some of the strategies to address these challenges. Figure 2.1. Travel time distribution is the basis for defining reliability metrics. 0 50 100 150 200 250 300 350 400 4.5 9.5 14.5 19.5 24.5 29.5 Travel Time (in Minutes) Thousands of Trips Free Flow Mean 99th Percentile Misery Time Buffer Time Planning Time Failure Measure Standard Deviation 95th Percentile
19 GUIDE TO INCORPORATING RELIABILITY PERFORMANCE MEASURES INTO THE TRANSPORTATION PLANNING AND PROGRAMMING PROCESSES TABLE 2.1. TRAVEL TIME RELIABILITY MEASURES DESCRIBED Measure Calculation Description Planning- Time Indexa (PTI) 95th Percentile of TT Free Flow TT The extra time required to arrive at a destination on time 95% of the time. Can be calculated for trips, corridors, or segments. The PTI is the recommended measure because it gives intuitive and consistent results. Buffer-Time Indexb (BI) 95th Percentile of TT Average TT Average TT â (could replace Average with Median TT) The extra time required to arrive at a destination on time 95% of the time, compared with average or median travel time. A BI of 1.5 indicates that, 95% of the time, it will take you 50% more time to arrive at your destination than it would under average conditions. Standard Deviation N TT Average TT 1 ( )i i N 2 1 â â = The variation in travel time compared with the average. A standard deviation of 5 min indicates that it is not unlikely for it to take 5 min more to travel than it would during average congestion. Semi- Standard Deviation N TT Free Flow TT 1 ( )i i N 2 1 â â = The variation in travel time compared with free flow. A semi- standard deviation of 5 min indicates that it is not unlikely for it to take 5 min more to travel than it would during uncongested conditions. Failure Measure Trips with TT Median Total Trips 1.1< â The percentage of trips arriving on time. A failure measure of 85% indicates that 85% of trips are arriving on time. Misery Index Average of the Highest 5 Percent of TT Free Flow TT How much longer it takes to travel on the worst 5% of all trips. A misery index of 4 indicates that the worst trips take 4 times as long as they would if it were uncongested. a The travel time index (TTI) is the travel time for a point on the travel time distribution divided by the free flow travel time. The PTI is a specific instance of the travel time index, calculated at the 95th percentile. A TTI value can be calculated at any percentile of the travel time distribution. b Research has raised questions about the consistency and intuitiveness of the buffer-time Index. This is explained in more detail in the technical reference. ⢠When the TTImean, TTI80, and TTI95 are all clustered and low, congestion is limited and travel is generally reliable. ⢠When TTI80 and TTI95 are higher than the TTImean, but close together, the corri- dor experiences reliability challenges but sees limited outliers (i.e., extremely long travel times). Work conducted under SHRP 2 L03 has demonstrated that routine operations strategies, such as incident management, may be effective in addressing congestion in these corridors. ⢠When TTI95 is higher than TTI80, a corridor experiences significant influence of outliers. These may be caused by extreme weather, special events, or major inci- dents that require closing the road. Challenges such as extreme weather and special events may require specialized planning efforts. Planners will need to experiment with these measures to determine which combi- nation of measures best helps them understand the reliability of the system and helps them evaluate strategies. Understanding the travel time distribution for individual
20 GUIDE TO INCORPORATING RELIABILITY PERFORMANCE MEASURES INTO THE TRANSPORTATION PLANNING AND PROGRAMMING PROCESSES Fi gu re 0 .1 . V ar ia tio n in r el ia bi lit y m ea su re s f or e xa m pl e co rr id or s ( 1) . 1 .0 0 2 .0 0 3 .0 0 4 .0 0 5 .0 0 6 .0 0 7 .0 0 8 .0 0 Tr av el T im e In de x (M ul tip le o f F re e Fl ow T im e R eq ui re d to C om pl et e a Tr ip ) Av er ag e tri p 80 % o f t rip s 95 % o f t rip s Li m ite d co ng es ti on a nd re lia bl e tr av el Ex tr em el y un re lia bl e, m aj or in ï¬u en ce o f o ut lie rs Si m ila r a ve ra ge co nd iti on s, d iï¬ er in g re lia bi lit y M ul tip le o f f re e flo w ti m e re qu ire d to c om pl et e ... Fe w o ut lie rs , p ot en ti al ly re sp on siv e to op er ati on s t re at m en ts Ex tr em el y un re lia bl e, m aj or in ï¬u en ce o f ou tli er s, b ut a lso si gn iï¬ ca nt a ve ra ge co ng es ti on . Si gn iï¬ ca nt c on ge sti on . M ay re qu ire c ap ac ity . Fi g u re 2 .2 . Va ria tio n in r el ia bi lit y m ea su re s fo r ex am pl e co rr id or s (1 ).
21 GUIDE TO INCORPORATING RELIABILITY PERFORMANCE MEASURES INTO THE TRANSPORTATION PLANNING AND PROGRAMMING PROCESSES corridors will help planners understand what they are planning forâday-to-day chal- lenges, extreme events and outliers, or both. Chapters 4 and 5 identify potential strate- gies to evaluate and methods for evaluating those strategies. Examples of Reliability Performance Measures in Use at Transportation Agencies Knoxville Regional Transportation Planning Organization (TPO) CMP. In their con- gestion management process (CMP), the Knoxville TPO measures the PTI for all users on freeways and major arterials in the region and plans to narrow the time period to a specific time period of the day. In addition, the TPO has developed an incident- management-specific measure to support the overall reliability statistic: clearance time of traffic incidents on freeways and major arterials in the region. Madison MPO CMP. The Madison MPO developed guidelines for the reliability measures that they will include in their CMP. They will include both peak and off- peak measures because, while congestion often focuses on peak period commutes, off-peak measures can identify different system problems, including those that can be important to freight movement efficiency. They will also include measures for the region and key sub-areas and corridors that reflect primary modal travel patterns. SELECTING A METHOD TO ESTIMATE RELIABILITY Selecting a measure is important, but estimating reliability performance often requires tools and methods. This chapter describes the ways agencies can estimate reliability using several methods. Chapter 5 of the technical reference provides more details and examples of each of these analysis methods. Monitoring Reliability The simplest way to measure reliability is to monitor travel time. Because reliability monitoring measures the variability of travel times, it has significant data require- ments. Unlike average travel time, which can be calculated using a relatively small sample of travel times over a few days, accurately monitoring reliability requires cap- turing travel time data across a wide range of conditionsâdays of the week, times of day, seasons, weather conditions, and the presence or absence of incidents. Data for monitoring reliability can come from a variety of sources, including tradi tional travel monitoring sensors, intelligent transportation systems (ITS) sensors ( Bluetooth, cameras, induction loops, etc.), instrumented vehicles, and others. In addi- tion to collecting data directly, several third party vendors use instrumented vehi- cles and other methods to provide data for purchase to agencies (e.g., INRIX and NAVTEQ). These data can support both operations and planning. Besides travel time data, other factors influence reliability, including crashes and other incidents, weather, variations in demand (i.e., travel volumes), special events, and others. Collecting data on these factors can help measure the impact of circum- stances on reliability.
22 GUIDE TO INCORPORATING RELIABILITY PERFORMANCE MEASURES INTO THE TRANSPORTATION PLANNING AND PROGRAMMING PROCESSES As part of a reliability monitoring program, agencies should also keep track of the investments that have been made in the transportation system, both those that are specifically intended to improve reliability and those that have been implemented for other reasons. Tracking reliability of the system over time allows for a before-and-after comparison of investments in the transportation system. With sufficient system cover- age, agencies can examine both localized improvements from individual investments and system improvements from packages of improvements over time. In the technical reference, Chapter 5 contains a detailed description of different travel time data resources, how to set up a travel time monitoring system, and how to estimate reliability using various sketch-planning methods. SHRP 2 L02 provides guidance for developing a travel time reliability monitor- ing system (TTRMS) to monitor, assess, and communicate reliability to end users. SHRP 2 L02 discusses the various technologies available for collecting travel times and the foundation of a TTRMS; in addition, it distinguishes between roadway-based and vehicle-based equipment. Travel time data should preferably be collected continuously so that travel time density functions can be developed. These data are used to describe the reliability characteristics of a corridor or a trip. Augmenting travel times are data on nonrecurring disruptions: incidents, weather, work zones, and special events. Modeling Reliability When travel time data are limited or when agencies need to forecast reliability (not just estimate current conditions), agencies can use tools that can help estimate reli- ability. Many of these tools can also be used to evaluate the impact of strategies on reliability. Because reliability is a function of the variability of travel times, the ideal tools for estimating reliability can estimate variability. Typical planning analysis tools, such as the standard four-step travel demand model, produce static estimates of travel times (potentially varying by time of day), making them a poor fit for estimating reliability. However, these are among the most common tools in use at transportation agencies; bridging the gap to more sophisti- cated tools will require using techniques to translate static estimates into reliability impacts, including the following methods. ⢠Sketch-planning methods. Sketch-planning methods provide a quick assessment of reliability using readily available data (travel times, volumes, etc.) as inputs. They are the least resource-intensive of the analysis methods and produce order- of- magnitude results. It is typical to use a spreadsheet to build a sketch-planning model. ⢠Model post-processing methods. These methods focus on applying customized analysis routines to more robust network supply-and-demand condition data from travel demand models to generate more specific estimates of travel time reliability. Common tools to post-process model results include the FHWAâs ITS Deployment Analysis System (IDAS) and the Florida ITS Evaluation (FITSEval) tool. Figure 2.3 presents an example of the output developed by the Florida Department of Trans- portation (DOT) using FITSEval and real travel time data to evaluate the TTI for all users on key segments of its Strategic Intermodal System.
23 GUIDE TO INCORPORATING RELIABILITY PERFORMANCE MEASURES INTO THE TRANSPORTATION PLANNING AND PROGRAMMING PROCESSES More sophisticated tools include ⢠Simulation. These methods make use of advanced analytical models to assess driver behavior and driversâ reactions to unpredictable circumstances. Simulation models can give modeled travel time distributions from which reliability perfor- mance measures can be built. ⢠Multiresolution methods. These methods combine several other analysis methods to assess reliability through different lenses. Multiresolution methods take advantage of the integration of several standard analysis tools (e.g., microsimulation and travel demand models), combining different toolsâ ability to assess shorter- and longer- range impacts of various congestion mitigation strategies. Figure 2.4 describes the resources required to use each of these methods. Sketch- planning methods require the fewest resources while simulation, multiresolution, and monitoring methods require the most. COMMUNICATING RELIABILITY PERFORMANCE This section presents thoughts on how to communicate reliability performance mea- sures to the public and stakeholders. SHRP 2 L14 is also developing advice on how to communicate reliability performance measures. Figure 2.3. Florida Department of Transportationâs performance measures annual report example. 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 9/30/11 12/31/11 3/31/12 6/30/12 12-month Average Travel Time Index - District 7 I-275 (SR 60 to I-4) Northbound p.m. Southbound p.m. Southbound a.m. Northbound a.m.
24 GUIDE TO INCORPORATING RELIABILITY PERFORMANCE MEASURES INTO THE TRANSPORTATION PLANNING AND PROGRAMMING PROCESSES Figure 2.4. Resources required for applying different tools and methods to evaluate reliability performance. Focus Reliability Measures on Key Issues When crafting reliability measures, it can be useful to focus them on specific issues, including ⢠Time periods. Typical time periods include the a.m. or p.m. peak hour or period. The measure should reflect the userâs experience. For example, a reliability analy- sis focused on special events may select various evening and/ or weekend midday periods to capture when issues are anticipated. ⢠Travel patterns. Reliability performance can be considered for trips or for seg- ments, and the selection can affect the choice of measurement. (In much of the reliability literature, segments are referred to as facilities.) Travel time data ven- dors are beginning to release data on individual trip-based travel times. These data can help identify key commuter patterns and their reliability traits. ⢠Roadway types. Appropriate thresholds (or measures) may vary by roadway types (i.e., functional class, levels of vehicle-miles-traveled, statewide roadway designa- tions, and so forth). ⢠Users. System users perceive reliability differently depending on their circum- stances. When presenting reliability performance measures, it is important to consider these perceptions and to incorporate them into the measures. Examples include â Freight carriers balance the need to pick up loads and the need to arrive on time to avoid a penalty for being late. These users will likely be interested in the PTI or the 99th percentile TTI. For freight-heavy segments (e.g., the road- way from the Miami Airport to the flower distribution center to the west), travel may be unreliable if the carrier is late once out of 1,000 times (i.e., the 99.9th percentile TTI).
25 GUIDE TO INCORPORATING RELIABILITY PERFORMANCE MEASURES INTO THE TRANSPORTATION PLANNING AND PROGRAMMING PROCESSES â Visitors and tourists making a one-time pass through an area without time constraints will perceive travel time to be reliable if they are on time 6 times out of 10 (i.e., 60th percentile TTI). â Commuters will perceive travel time to be reliable if they are on time 95 times out of 100 (e.g., late to work no more than once per month). Developing Corridor-Level Measures Because reliability measures variability in travel times, corridors and roadway seg- ments are a natural level to present information to users. However, presenting reli- ability at a corridor level requires developing thresholds that make reliability mea- sures meaningful to system users. One simple way to do this is to convert reliability performance into good/ fair/ poor categories. This style of presentation is common for infrastructure performance measures (i.e., percentage of pavement in good condition). Appropriate thresholds will depend on the characteristics of the corridor or region. Chapter 4 of this guide indicates a thorough explanation for how to tailor thresholds for the agency. Potential examples include ⢠Good: Good performance is when the PTI is less than 1.3 (PTI < 1.3); ⢠Fair: Fair performance is when the PTI is between 1.3 and 2 (1.3 < PTI < 2); and ⢠Poor: Poor performance is when the PTI is greater than 2 (2 < PTI). Examples of Corridor-Level Measures in Use at Transportation Agencies Figure 2.5 provides examples of maps to communicate reliability performance. The first example, from the Capital District Transportation Committee (CDTC) in Albany, N.Y., presents the PTI (2). In this example, the width of the line represents free-flow (base) travel time and the dark line represents the 95th percentile travel time. The sec- ond map example, from the Georgia Regional Transportation Authority, illustrates the segments that experience the worst reliability using the PTI (3). Red and purple seg- ments have poor reliability, yellow segments have fair reliability, and green segments have good reliability, according to the above definitions. In the 2011 Annual Congestion Report, the Washington State Department of Transportation (DOT) reports that 17 of the 36 high-demand commutes in Puget Sound saw modest changes (less than or equal to 2 min) in 95% reliable travel time between 2008 and 2010. Fourteen commutes saw reliable travel times worsen between 3 and 10 min, while reliable travel times improved on five commutes ranging from 3 min to 11 min. The Washington State DOT uses âstamp graphsâ to help illustrate cur- rent performance and how performance is changing from 2009 (light gray lines) and 2011 (dark gray lines) (Figure 2.6). Developing System-Level Measures Many agencies use performance measures to present a summary of overall system performance. Tracking system performance over time can be a useful tool for com- municating whether performance is improving or worsening. Reducing reliability, a measure of variability, into a single number that can be tracked over time can be
26 GUIDE TO INCORPORATING RELIABILITY PERFORMANCE MEASURES INTO THE TRANSPORTATION PLANNING AND PROGRAMMING PROCESSES Figure 2.5. Examples of communicating travel-time reliability at the corridor level. Color version of both parts of this figure: www.trb.org/Main/Blurbs/168855.aspx. challenging. The simplest approach is to find a way to combine data from multiple corridors. Two basic ways to present such a measure are 1. A weighted average of the reliability measure. For example, the PTI for several corridors could be weighted by volume or another factor to generate a single PTI measure for the system; and 2. The percentage of travel that occurs at various reliability conditions. This type of measure examines all corridors (or a subset of corridors) and calculates the per- centage in good, fair, and poor conditions. Examples of System-Level Measures in Use at Transportation Agencies In the 2011 Annual Congestion Report, the Washington State DOT measured perfor- mance, described trends, and communicated reliability using the 95th percentile travel time (the numerator in the PTI) for segments along high-demand commute routes
27 GUIDE TO INCORPORATING RELIABILITY PERFORMANCE MEASURES INTO THE TRANSPORTATION PLANNING AND PROGRAMMING PROCESSES Figure 2.6. Washington State DOTâs tracking of travel times in general purpose (GP) lanes (4). (Table 2.2). To convey reliability trends, DOT staff members categorized how much the 95th percentile of travel time had changed in the most recent two-year period. They report that 17 of the 36 high-demand commutes in Puget Sound saw modest changes (less than or equal to 2 min) in 95% reliable travel time between 2008 and 2010. Fourteen commutes saw reliable travel times worsen between 3 and 10 min, while reliable travel times improved on five commutes ranging from 3 to 11 min.
28 GUIDE TO INCORPORATING RELIABILITY PERFORMANCE MEASURES INTO THE TRANSPORTATION PLANNING AND PROGRAMMING PROCESSES TABLE 2.2. WASHINGTON STATE DOT RELIABILITY PERFORMANCE MEASURES Performance Measure Definition 95% reliable travel time Travel time with 95% certainty (i.e., on time 19 out of 20 work days). Maximum throughput travel time index (MT³I) The ratio of average peak travel time compared with maximum throughput speed travel time. Percentage of days when speeds are less than 36 mph Percentage of days annually that observed speed for one or more 5-min intervals is less than 36 mph (severe congestion) on key highway segments. HOV lane reliability An HOV lane is deemed âreliableâ as long as it maintains an average speed of 45 mph for 90% of the peak hour. REFERENCES 1. Texas Transportation Institute. 2011 Congested Corridors Report and Appendices. http://mobility.tamu.edu/corridors/report/. Accessed July 2, 2013. 2. Federal Highway Administration, U.S. Department of Transportation. Congestion Management Process. Capital District Transportation Committee (CDTC), Albany, New York. http://www.fhwa.dot.gov/planning/congestion_management_ process/ case_studies/ cdtc.cfm. Accessed July 2, 2013. 3. Georgia Regional Transportation Authority. 2010 Transportation Metropolitan Atlanta Performance Report. http://www.grta.org/tran_map/2010_ Transportation_ MAP_Report.pdf. Accessed July 2, 2013. 4. Washington State Department of Transportation. 2011 Annual Congestion Report. http://www.wsdot.wa.gov/Accountability/Congestion/2011. Accessed July 2, 2013.