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.

18 C h a p t e r 3 Domestic and International agency Usage California The California Department of Transportation (Caltrans) has produced performance measures for the entire multimodal system (Downey 2000). The measures are intended to do the following: • Monitor and evaluate system performance. • Share existing data and forecast future performance information. • Develop mode-neutral customer and decision information. • Build consensus using performance measures information. • Improve accountability of system development and operations. Caltrans tested the measures on corridors in four metro- politan counties in 2000. The peak period travel time varied by 10% to 50% on all corridors. The agency also found that reliability may not be directly correlated with delay; some areas that had high delay also had low travel time variability, partly because of the difficulty in deviating from slow speeds. Travel time reliability depended on several factors, including distance between interchanges and roadway geometrics. Caltrans began to measure travel time reliability in January 2011. Travel time reliability is defined as the predicted mean travel time compared with the actual travel time. The geo- graphic coverage of the measurement is evolving, and recent changes have focused on selected corridors. For each corri- dor, division and district transportation professionals calcu- late one or more reliability measures. Florida The Florida DOT (FDOT) has developed a method and tools for estimating travel time reliability for the freeway portion of its Strategic Intermodal System (SIS). The method is illustrated in Figure 3.1. As shown in the figure, the freeway network to be analyzed is first segmented at a section level (interchange to interchange). Interchanges and beginning and ending milepost numbers are obtained from FDOT’s Roadway Characteristics Inventory. As a first step in the travel time reliability estimation, the methodology considers a variety of possible scenarios that may occur on any given freeway section. These scenarios are based on the presence of congestion, rain, incidents, and work zones. For example, one scenario may be that the section is congested and an incident is occurring along its length. Another scenario may be that the section is not congested, but it has a work zone along its length. In the second step, the method estimates the travel time for each scenario identified in Step 1. The travel time estimation is based on a combination of previously developed models (HCM2000; Elefteriadou et al. 2010a). In the third step, the method obtains the probability of occurrence for each sce- nario identified in Step 1. The fourth step develops the travel time distribution for the section and estimates selected travel time reliability measures on the basis of this distribution. Finally, the travel time reliability for the entire freeway net- work is estimated by aggregating the respective measures for each of the sections analyzed. The travel times for each of the segments within a given route are summed for each hour to obtain facility travel times. From these, travel time reliability measures are calculated in a similar way to those for segments. FDOT is currently proceeding with obtaining metrics for both categories of travel time reliability definitions (i.e., based on the traditional concept of reliability as nonfailure over time and based on the concept of variability of travel time). The FDOT Traffic Operations Office is interested in travel time variability, which it would ultimately like to report to travelers on a real-time basis; the Systems Planning and Policy Planning Offices are interested in the on-time arrival estimation and in the evaluation of the performance of the State of the Art and State of the Practice

19 SIS so that improvements can be prioritized on the basis of this measure and reported to decision makers. Additional information regarding this method is provided in a series of reports (Elefteriadou and Xu 2007; Elefteriadou et al. 2008, 2010b, 2010c). Current research work by FDOT focuses on the develop- ment of models for arterial sections of the SIS and on the refinement of existing freeway models by comparing their output with field data from instrumented sections. Nevada Nevada DOT’s Integrated Transportation Reliability Pro- gram (ITRP) aims to implement new and innovative pro- grams to prevent congestion and improve reliability. As part of the program, the agency will coordinate with statewide stakeholders to develop strategies to improve travel time reli- ability in Nevada. More than 15,000 traffic crashes occur each year in the Las Vegas valley. The Las Vegas Traffic Incident Management Coali- tion brought southern Nevada emergency response and trans- portation agencies together to enhance emergency response. The group established collision clearance time goals to restore road travel following traffic crashes (Kimley-Horn and Associ- ates, Inc. 2010). International Research This section draws from a range of experiences by interna- tional transportation agencies. Some of the information is drawn from an FHWA international scan of transportation performance measurement practices. The scan included vis- its to transportation agencies with mature performance man- agement systems in Australia, Great Britain, New Zealand, and Sweden. It focused on how these organizations demon- strate accountability to elected officials and the public. One of the interests of the scan team was how transportation agencies used reliability performance measures and practices to meet their goals (Braceras et al. 2010). All of the agencies reported that their reliability measures were evolving and they were not entirely satisfied with their measurement tools. However, the more urbanized agencies in Great Britain, Australia, and Sweden had clearly invested con- siderable effort in measuring real-time highway, transit, and rail operations to improve travel time reliability, enhance transportation choices, and reduce greenhouse gas emissions. This section describes several findings from the international scan, along with Japanese and Dutch research. Great Britain National The British have invested considerable effort in measuring reli- ability on high-volume national routes. The Highways Agency (HA) of Great Britain has identified a Strategic Road Network of 2,700 km (1,678 mi) of motorways and 4,350 km (2,703 mi) of other trunk routes. These routes are analyzed in 103 sections with 2,500 total links. The HA actively tracks reliability perfor- mance on a daily basis across this network and defines travel time reliability as the average vehicle delay on the slowest 10% of the journeys (Cambridge Systematics, Inc. et al. 2013). The network reliability program has improved British offi- cials’ understanding of system performance, and the HA has increased its use of reliability analysis in evaluating improve- ment strategies. The HA identified several difficulties in mea- suring reliability, including shortcomings in data and varying definitions. They also noted difficulties explaining the results to the public because the performance measures are not very sensitive to the improvements. For example, improvements reduced the average of the worst 10% of trips making a 16-km journey from 3.9 min to 3.4 min of delay. For a slow trip, an improvement of half a minute is marginal. Additionally, the HA could not be sure whether the improvement created the travel time reliability benefit or whether it was a function of changes in economic conditions. London Research found that travel time varies in three ways: interday variability (caused by seasonal and day-to-day variations in travel times), interperiod variability (caused by different depar- ture times and consequent changes in congestion) and Step 1: Identification of Possible Scenarios Step 2: Estimation of Travel Time for Each Section Step 3: Estimation of the Probability of Occurrence for Each Scenario Step 4: Development of the Travel Time Distribution for Each Section Freeway Network Segmentation Aggregation of Performance Measures for the Selected Route and for the Network Tr av el Ti m e R el ia bi lit y fo rE ac h Se ct io n Source: Elefteriadou et al. (2010a). Figure 3.1. FDOT reliability methodology overview.

20 intervehicle variability (caused by personal driving styles and behavior of traffic signals along a certain route) (Bates et al. 1987). The authors measured travel time reliability using the mean-variance approach (based on variance or standard devia- tion of travel times), the scheduling approach (based on disutil- ity incurred because of late arrivals), or the probabilistic/mean lateness approach (based on mean lateness at departure/arrival). Sweden The Swedish Road Administration (SRA) includes travel reli- ability among a large set of transportation performance mea- sures. Travel times and speeds are tracked on major routes in the three major cities (Stockholm, Malmö, and Göteborg) and on routes to towns for rural residents (Franklin 2009). The SRA reports are designed to connect the performance of the system with “the steps taken in each area to improve traffic flow and reliability and report on planned improvement strate- gies for the next year.” Rural reporting includes the effect of seasonal weather problems and summarizes the number of residents who saw increases or improvements in travel times between towns. Japan Use of predicted reliability within project benefit-cost analysis is in its nascent stages in Japan. Higatani et al. (2009) examined the characteristics of travel time reliability measures using traffic flow data from the Hanshin Expressway, an urban toll expressway network that stretches from Osaka to Kobe. For this study, travel time reliability indices were calculated for five radial routes connected to the downtown loop route in Osaka City. Several measures were calculated for one radial route (Route 11 Ikeda Line), including average travel time, 95th per- centile travel time, standard deviation, coefficient of variation, buffer time, and buffer index. The buffer time and buffer index showed tendencies similar to the standard deviation and coef- ficient of variation, respectively. The time-of-day variation of traffic flow was also investigated for all five radial routes, and the effect of traffic incidents on travel time reliability measures was analyzed for one radial route (Route 14 Matsubara Line). The Netherlands Research found that travel time variance accounts for only a portion of the delay effects from unreliability. The studies recommend including the skew of travel time distribution (e.g., the amount of extra travel time for the worst 5% of trips) to measure the remaining effects of unreliable travel times (van Lint et al. 2008; Tu 2008). U.S. research SHRP 2 Project L03 SHRP 2 Project L03 examined the potential performance measures used to describe travel time reliability. Table 3.1 summarizes the recommended reliability performance met- rics from that study. The recommendations were based on an examination of measures in use in the United States and in Table 3.1. Reliability Performance Metrics from SHRP 2 Project L03 Reliability Performance Metric Definition Units Buffer index (BI) The difference between the 95th percentile travel time and the average travel time, normalized by the average travel time The difference between the 95th percentile travel time and the median travel time, normalized by the median travel time Percent Failure or on-time measures Percentage of trips with travel times less than 1.1 × median travel time and/or 1.25 × median travel time Percentage of trips with space mean speed less than 50, 45, and/or 30 mph Percent 80th percentile TTI 80th percentile travel time divided by the free-flow travel time None Planning time index 95th percentile TTI (95th percentile travel time divided by the free-flow travel time) None Skew statistic The ratio of (90th percentile travel time minus the median) divided by (the median minus the 10th percentile) None Misery index (modified) The average of the highest 5% of travel times divided by the free-flow travel time None Standard deviation of travel time or travel ratea Standard statistical definition None a Not included in the L03 recommendations, but added here. See text. Source: Cambridge Systematics, Inc. et al. (2013).

21 other parts of the world. The table also includes the skew statistic proposed by European researchers. In addition, the researchers added the 80th percentile TTI because analysis indicated that this measure is especially sensitive to opera- tions improvements, and it has been used in previous studies on the valuation of reliability. All of these measures can be easily created once the travel time distribution is established, as illustrated in Figure 3.2. Because of the need to normalize travel time, the TTI was used as the variable of interest in this research. Therefore, the base distribution is actually based on the distribution of the TTI, rather than raw travel times. The L03 research also demonstrated that the buffer index can be an unstable measurement for tracking trends over time in part because of its linkage to two factors that change (aver- age and 95th percentile travel times); if one changes more in relation to the other, counterintuitive results can appear. Note that standard deviation of travel time or travel rate appears in Table 3.1 and Figure 3.2. Project L03 did not define this as a reliability performance metric, but it has been added because several other SHRP 2 research projects have indi- cated that it is useful in both costing reliability and in model- ing traveler choices. Project L03 included predictive methods for the standard deviation, even though it was not formally identified as a useful performance measure because of the difficulty in explaining it to nontechnical audiences. NCHRP Project 3-97 NCHRP Project 3-97, Traffic Signal Analysis with Varying Demands and Capacities, developed a recommended set of performance measures for evaluating the robustness of signal timing plans when challenged with varying demand and capac- ity conditions (Dowling et al. 2011). Robustness is defined as the ability of the signal system to continue to provide satis- factory performance under varying demand and capacity conditions. Three measures were recommended for evaluating signal system performance. One relates to average performance, tak- ing into account expected fluctuations in demand and satura- tion flow rates for the analysis period over an extended period. The other two relate to the robustness of the timing plan when challenged with demand and saturation flow rate fluctuations. The first measure of effectiveness (MOE) is the weighted- average whole-year performance for the subject peak period. Performance can be measured using any one of many com- monly used signal performance measures (e.g., delay, stops, performance index). Differences in the average performance between two peak period timing plans can be used to compute differences in total performance. For example, the difference in the average vehicle hours traveled multiplied by the number of nonholiday weekdays per year can be used to estimate total annual vehicle hours saved for one plan versus the other. The second MOE is the 95th percentile performance (or a similar high-percentile performance). The analyst can inter- pret this MOE to mean that 95% of the time the performance of the timing plan will not be worse than the values associated with the 95% scenario. The third recommended MOE is the probability of break- down. The probability is that demand will exceed the timing plan capacity for the duration of the analysis period (typically Figure 3.2. Travel time distribution as the basis for defining reliability metrics.

22 1 to 2 hours) somewhere in the system. When demand exceeds capacity for extended periods, queues and delays build rap- idly. The probability is a quick, intuitive measure of the likeli- hood of capacity failure with the current plan. FHWA ATDM Evaluation Guidebook The FHWA’s Guide for Highway Capacity and Operations Analy- sis of Active Transportation and Demand Management Strategies, which was in publication at the time of writing, recommends a basic set of reliability performance measures from which vari- ous statistics can be computed (Dowling and Margiotta 2013). The guidebook then recommends a specific set of measures of effectiveness that may be useful for comparing performance across different Active Transportation and Demand Manage- ment (ATDM) strategies and might eventually serve as a foun- dation for a level of service measure of reliability. The basic performance measures are useful for most eco- nomic and environmental analyses. In addition, the basic performance measures are key components of the recom- mended measures of effectiveness for evaluating ATDM. The recommended MOEs are designed to address two key objectives of ATDM: to improve facility/system efficiency and to improve reliability. In addition, two of the recommended MOEs provide measures that individuals can relate to: aver- age speed and average delay per trip. The recommended basic performance measures and mea- sures of effectiveness for evaluating the performance benefits of ATDM measures are • Basic performance measures useful for computing MOEs: 44 Vehicle miles traveled demand (VMT-Demand) 44 Vehicle miles traveled served (VMT-Served) 44 Vehicle hours traveled (VHT) 44 Vehicle hours delay (VHD). • Measures of effectiveness: 44 System efficiency: average system speed (mph) 44 Traveler perspective: vehicle hours delay per vehicle trip (VHD/VT) 44 Reliability: planning time index (PTI). The VMT-Demand is the sum of the products of the input origin–destination (O–D) table vehicle trips and the shortest- path distance between each origin and destination. Although not traditionally a performance measure for highway improve- ment projects, demand is a measure of the success of ATDM at managing the demand for the facility. The VMT-Served is the sum of the products of the total link volumes for the peak period and the link lengths. VMT-Served is a measure of the productivity of the facility, the improvement of which is one of the key objectives of ATDM. VHT is the sum of the products of the total link volumes and the average link travel times. Delays to vehicles prevented from entering the facility during each time slice (vehicle hours of entry delay, VHED) (either by controls, such as ramp meter- ing, or by congestion) are added to and included in the reported VHT total. VHD is the difference between the VHT (including vehicle entry delay) and the theoretical VHT if all links could be tra- versed at the free-flow speed with no entry delays. VHD is summed over all time slices within the scenario. VHD is useful in determining the economic costs and benefits of ATDM mea- sures. VHD highlights the delay component of system VHT. ( )= −VHD VHT VHT FF (3.1) where VHD = vehicle hours delay; VHT = vehicle hours traveled, including vehicle entry delay; and VHT(FF) = vehicle hours traveled, recomputed with seg- ment free-flow speeds. VHED for any given scenario is the number of vehicles prevented from entering the system during each time slice, multiplied by the duration of the time slice and summed over all time slices. VHED should be included in the computed VHD and VHT for each scenario. Average system speed (mph) is a measure of the efficiency of the highway system. It is computed by summing the VMT- Served for each scenario, then dividing by the sum of the sce- nario VHTs (including any vehicle entry delay). One of the key objectives of ATDM is to maximize the productivity of the system, serving the greatest number of VMT at the least cost to travelers in terms of VHT. Thus, changes in the average system speed are a good overall indicator of the relative suc- cess of the ATDM strategy at achieving its objective of improving efficiency. Vehicle hours delay per vehicle trip (VHD/VT) is the vehi- cle hours delay summed over all scenarios divided by the sum of the number of vehicle trips in the origin–destination (O-D) tables for all scenarios. This gives the average delay per vehi- cle, which is useful for conveying the results in a manner that can be related to personal experience. The travel time index (TTI) is a measure of congestion on the facility. It is the ratio of the mean travel time to the free- flow travel time. For example, a TTI of 1.20 can be interpreted as meaning that the traveler must allow 20% extra time over free-flow travel time to get to his or her destination on time. When a percentile greater than 50% is used, then the TTI becomes a reliability measure. For example, an 80th percen- tile TTI of 1.20 can be interpreted as meaning that, over the course of a year for a given trip leaving at a given time, 80% of the trips will take no more than 20% longer than the free- flow travel time. A 95th percentile TTI is also referred to as the planning time index (PTI).

23 While various travel time percentiles historically have been used for the TTI, is the L08 team recommends that the 80th per- centile highest travel time be used for the predicted travel time. The 80th percentile travel time has a more stable relationship to the mean travel time than the 90th, 95th, or 99th percentiles, so it is useful in predicting changes in reliability that are based on changes in the mean travel time. The formula for computing a systemwide TTI follows: ( ) ( ) ( ) ( )=80%TTI VHT 80% VMT 80% VHT FF VMT FF (3.2) where 80%TTI = 80th percentile travel time index; VHT(80%) = 80th percentile highest vehicle hours trav- eled among scenarios evaluated; VMT(80%) = vehicle miles traveled for scenario with 80th percentile highest vehicle hours traveled among scenarios evaluated; VHT(FF) = vehicle hours computed with segment free- flow speeds; and VMT(FF) = vehicle miles traveled with segment free- flow speeds.