Analytical Procedures for Determining the Impacts of Reliability Mitigation Strategies (2012)

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 C h a p t e r 2 Introduction The project was organized in three phases: foundational research, data collection and preliminary analyses, and reli- ability prediction models. Phase 1: Foundational Research The foundational research effort included • Conducing a literature review; • Identifying the reliability metrics to be used in the research; • Defining the improvement strategies that affect travel time reliability; • Specifying an experimental design for the research; • Identifying the types of data that were needed to conduct the research; and • Defining an analysis plan for conducting the research, including the model forms to be investigated. Phase 2: Data Collection and Preliminary Analyses The data collection effort and preliminary analyses were doc- umented and included • A description of the data sets that were assembled; and • Exploratory analyses of the data to establish fundamental concepts for the detailed analyses. Phase 3: Reliability Prediction Models The Phase 3 effort is documented for the first time here in the final report. Literature review Reliability Performance Metrics The recognition that travel time reliability is a problem is reflected in changes to traditional monitoring programs that examine average or typical congestion. Increasingly, traffic monitoring agencies understand that those traditional studies must be supplemented with tracking efforts that include day- to-day measures, as well (1). The National Transportation Operations Coalition Performance Measurement Initiative, for example, identified travel time reliability (buffer time) as one of the 14 key measures for operations programs (2). Data and analysis procedures, however, are not being developed as fast as the recognition of the problem. Table 2.1 displays several transportation agencies that have included travel time reliability as a portion of mobility mea- surement in their performance evaluations. Some of the eval- uations are performed on a corridor basis, and others are done on a systemwide or statewide basis. Table 2.1 includes only those cases in which reliability measures have been endorsed or adopted by a public entity responsible for oper- ating and/or maintaining transportation systems, such as a state department of transportation (DOT). The table does not include recommendation or use of performance mea- sures by academic or research groups. NCHRP Project 3-68 identified several measures of travel time reliability that provide a basis for selecting measures for the research: • Buffer Index—Difference between the 95th percentile travel time and the average travel time, divided by the aver- age travel time; • Planning Time Index—95th percentile Travel Time Index (TTI). A TTI of 1.2 indicates that a trip takes 20% longer than it would under ideal conditions; Preparatory Analyses

18 • Percentage of trips with space mean speeds ≤50 mph; and • Percentage of trips (section or origin–destination) with space mean speeds ≤30 mph (4). Tu et al. classified travel time reliability measures into five types: (a) statistical range methods, (b) buffer time methods, (c) tardy-trip measures, (d) probabilistic measures, and (e) skew- width methods (10). The first three measures were first defined by Lomax et al. (11). Probabilistic measures, which are in the same category as failure-based or on-time measures, have been proposed for use in Florida, in combination with a buf- fer time measure (12). Skew-width methods are based on the observation that most travel time distributions are skewed to the right, as shown with example measures in Figure 2.1. It has been suggested that travel times follow either a lognormal distribution or gamma distribution with an adequately scaled shape parameter (13). In traditional statistics, two standard measures are used to express the unevenness of distributions: • Skewness is a measure of symmetry, or more precisely, the lack of symmetry. A distribution, or data set, is symmetric if it looks the same to the left and right of the center point; and • Kurtosis is a measure of whether the data are peaked or flat relative to a normal distribution. That is, data sets with high kurtosis tend to have a distinct peak near the mean, decline rather rapidly, and have heavy tails. Data sets with low kurto- sis tend to have a flat top near the mean rather than a sharp peak. A uniform distribution would be the extreme case (14). Van Lint and van Zuylen noted that buffer time and Misery Index measures based on the mean may not be appropriate because of the underlying skewed distribution (15). They also defined two measures that describe the size and shape of the travel time distribution: 1. A skewness statistic, defined as (90th percentile - median)/ (median - 10th percentile); and 2. A width statistic, defined as (90th percentile - 10th percentile)/median. Total number of trips, shown in Figure 2.1, for the time period = 3,485 million (each point on the line represents the number of trips grouped by 30-second travel time intervals). Note that about 8% of trips (275,000 out of 3.485 million) occurred at free flow during this period. Table 2.1. Reliability Measures in Selected Transportation Agencies Agency Reliability Metrics Used Data Source Coverage Freeway Georgia Regional Transportation Authority (for annual mobility performance in Atlanta) and Georgia DOT (3, 4) Buffer Index, Planning Time Index Georgia DOT and local agencies Facilities Florida DOT (5) Buffer Index, on-time arrival Florida DOT and local agencies Facility Statewide Southern California Association of Governments (for goods movement study) (6) Buffer Index Caltrans and local agencies Facility Washington State DOT (WSDOT; for performance monitoring and traveler information) (7) 95th percentile travel time WSDOT and local agencies Facility (time is the sum of link times) National Transportation Operations Coalition (for performance measure initiative); potential case study with I-95 Corridor Coalition (2) Buffer Index Various agencies To be determined Arterials NCHRP 3-68 Buffer Index Various agencies Facilities PRUEVIIN (process for regional understanding and evaluation of integrated ITS networks) Coefficient of trip time variation WSDOT Facilities Private companies—Inrix and Traffic.com Private Facilities Maryland State Highway Administration and Delcan-NET Private Facilities Freight American Transportation Research Institute (ATRI) (FHWA freight performance measurement) (8) Buffer Index Private State- and national-level Interstates Missouri DOT ATRI I-70 across state Note: ITS = intelligent transportation system.

19 Freight Efforts In terms of economic value, reliability is probably more important to freight carriers and shippers than to personal travelers. With the rise in just-in-time deliveries (largely as a replacement to extensive warehousing), providing depend- able (reliable) service has become extremely valuable. Con- versely, failure to provide dependable service can increase costs significantly. The chemicals supply chain provides an example of how reliability affects truck freight operations. Increases in trans- portation reliability play an important role in reducing inven- tory in the chemicals supply chain. Because of the many nodes, up to one-third of chemical inventory is in transit at any point. Inventory managers keep safety or buffer stock to cushion against the variability of inbound arrivals, and the amount of safety stock increases with the degree of unreliability and the number of stocking locations. However, capacity to receive chemical stocks is limited by the size of the liquid storage silos. Balancing capacity with demand is a challenge. As one indus- try consultant explains: “If the tank is full, there’s no place to put it [incoming chemicals] and you pay demurrage [storage charges] on the railcar. But if the vessel is early, you have wait time or dead freight.” As transportation reliability decreases, wait time, dead freight, and cost increase (16). Conceptually, reliability for trucks is no different than for personal travel; that is, it is measured the same way (the travel time distribution) with the same metrics (e.g., Buffer Index). Also, all roadway, demand management, and operations improvement types (except for those that specifically target trucks, such as lane and service restrictions) affect both truck and personal travel. A practical difference is the length of the trip. Much truck travel is intercity, and therefore occurs on long sections of rural highways that are not routinely con- gested. This means that only a small portion of the entire trip is within urban areas, where most of the delay and associated unreliability occur. In 2002, the American Transportation Research Institute (ATRI) partnered with the Federal Highway Administration (FHWA) to develop methods for measuring freight perfor- mance on U.S. highways (8). With the freight performance measurement project, ATRI demonstrated that it is possible to collect roadway operational data for trucks using satellite technology and that individual truck data could be rendered unidentifiable through a cleansing process. The trucking companies wanted some assurance (primarily caused by safety and security concerns) that their trucks could not be tracked once the identity cleansing process had been per- formed. The freight performance measurement results were deemed successful in identifying freight-significant corridors and developing measures for evaluating the performance of full highway corridors, as well as providing information on individual segments within these corridors. Missouri Department of Transportation The Missouri Department of Transportation (Missouri DOT) has developed a set of performance measures to grade its activities and system performance. The measures are housed in Tracker, a report that includes average truck speed as one of its freight performance measures (17). The average Note: Analysis of NavGAtor data from I-75 northbound from I-285 to Wade Green Road (13.33 miles), Atlanta, Georgia, from 5:00 to 7:00 p.m. on weekdays, 2004. Figure 2.1. Travel time reliability is determined by the distribution.

20 truck speed is updated monthly for the entire length of I-70 across Missouri as well as I-70 nationwide. This speed esti- mate is supplied as a monthly average to the Missouri DOT by ATRI and the freight performance measurement database described above. Washington State Department of Transportation A research project by the Washington State Transportation Center analyzed options for collecting travel time data for trucks to determine the benefits provided by freight mobility projects in Washington State (18). The report identifies two types of travel time data that need to be collected for trucks: first, the average travel time experienced while making rou- tine trips; and second, travel time data that demonstrate what happens when trucks experience severe, unexpected delay. The report states that collecting truck travel times using float- ing car techniques is not practical to gather enough data to show truck trip reliability. In addition, travel times must be collected for trip lengths longer than just the affected portion of a corridor where improvements have been made. Since some trucks would change their travel patterns to make use of the improved roadway, the travel time between truck origin– destination pairs should be used to determine the effect of the improvement on delay reduction for the area. Texas Department of Transportation Work Zone Studies The Texas Transportation Institute developed two case studies using archived speed data and more detailed work zone data from Houston and San Antonio in an ongoing TxDOT research project (19). This study related detailed information on work zone start–stop times, weather information, and crash information to determine the delay that is caused by the work zone. PRUEVIIN A research effort in the Seattle, Washington, area developed a technique to combine regional travel demand models and commercially available traffic simulation software into a scenario-based framework (20). The process for regional understanding and evaluation of integrated intelligent trans- portation systems (ITS) networks (PRUEVIIN) has two main features. First, it uses state-of-the-art traffic simulation mod- els to identify the impacts of ITS on a transportation system under average conditions. Second, it provides a method to incorporate system variability, which links the simulation analysis to the travel demand modeling framework. This second feature allows the evaluations to include realistic conditions (e.g., inclement weather, collisions, vehicle break- downs, work zones) rather than to model the expected or best-day conditions. In one analysis, the coefficient of trip time variation was calculated by examining the variation in travel times across each of the different modeled scenarios for a specific trip. Results showed that as the coefficient gets larger, the variability of trip times increases, and reliability for the trip decreases. PRUEVIIN demonstrates that reliability measures can be generated without enormous amounts of travel time data collection and may provide a means of obtaining travel time reliability measures on arterial streets, where data can be scarce. Inrix and Traffic.com Several private companies have been collecting travel time data on freeways and arterial streets in many U.S. cities for several years. Inrix (21) and Traffic.com (22) collect travel time data by tracking fleets of probe vehicles in each area using global positioning system (GPS) tracking. They also obtain data from state DOT web sites and other sources of speed data to supplement the probe vehicle data. They pro- duce real-time travel speed estimates that are posted to web sites and provided to the media in the majority of these areas. These real-time data are generally archived and could be used to calculate travel time reliability on arterial streets. Few independent analyses have been performed on the GPS- tracked travel time data from these two sources, so there is a great deal of uncertainty as to the composition of the data. The Maricopa Association of Governments (the metropoli- tan planning organization for the Phoenix, Arizona, urban area) compared private vendor travel time data from two firms with their own sources (freeway detectors and floating car runs). The evaluation indicated that on freeways, both companies’ historic average speeds compared favorably with data from eight accurate loop detector freeway locations maintained by Arizona. The evaluation also found that on arterial streets, both companies’ historic average speeds compared favorably with the Maricopa Association’s traffic speed data. Beyond Reliability: The Seven Sources Reliability metrics provide an understanding of how depend- able or variable travel conditions are, but they do not iden- tify the cause of the variability. In this sense, reliability measures are top-level outcome measures. A deeper under- standing of what causes unreliable travel (and congestion, in general) is useful because it indicates which general areas or specific strategies should be emphasized. The original research plan for the SHRP 2 Reliability areas recognized the need for this deeper understanding and identified seven

21 sources of congestion. Figure 2.2 shows how these seven sources interact to produce total congestion. Reliability is an aspect of total congestion that is greatly influenced by the complex interactions of traffic demand, physical capacity, and roadway events. An understanding of how each source contributes to total congestion (as well as reliability) is limited, although the current research attempted to determine these contribu- tions analytically. National estimates have been produced by FHWA (Figure 2.3), but these were determined by consensus rather than analysis. FHWA estimates also are meant to be a national snapshot, not indicative of individual corridors or highways. For example, in rural conditions, delays are nearly always a function of events rather than a bottleneck. In urban conditions, especially on a facility with a dominant bottleneck, most of the delay will be determined by the bottleneck. Improvements that affect reliability Tables 2.2, 2.3, and 2.4 show an effects matrix for the three major categories of improvement: capacity additions, opera- tional improvements, and demand management, respectively. The list is illustrative rather than exhaustive. The assessment listed in the far-right column (Significance of Expected Effect on Reliability) of each of the three tables is based on the team’s initial subjective judgment about the magnitude of the strategy’s effect on reliability; it does not reflect the results of any of the research conducted for the project. experimental Design Types of Analyses Conducted Three main forms of analysis were undertaken, as described below. In addition, a large set of exploratory analyses were conducted before the primary analyses as part of Phase 2 (see No. Figure 2.2. A model of congestion and its seven sources. Figure 2.3. FHWA national estimates of delay by source (23).

22 Chapter 4) to identify the parameters necessary to conduct the primary analyses. 1. Before-and-after analysis—Since the major objective of the research was the development of models that could predict the change in reliability due to improvements, before-and-after analysis was the most appropriate experi- mental design. Here, before is a period of time prior to implementing the improvement, and after is a period of time after the improvement has been implemented. Ideally, before-and-after analysis is applied with a control group to help account for the influence of background factors. In this approach, the same highway section or network is studied with and without the improvement. However, it was recognized early in the research that it would be impos- sible to study all the possible improvement types in the field due to data limitations. Therefore, a second approach was developed that could handle reliability prediction. Table 2.2. Congestion Strategy Effects Matrix: Add Capacity Strategy Expected Effect on Reliability Existing Methodology to Calculate Effects Significance of Expected Effect on Reliability Add Capacity—Freeways New freeways Add new system capacity, reduce demand on adjacent freeways and arterials, and reduce level of incident impacts HCM, planning model Medium Widen freeways Add new system capacity, reduce demand on adjacent freeways and arterials, and reduce level of incident impacts HCM, planning model Medium New toll roads Add new system capacity, reduce demand on adjacent freeways and arterials, and reduce level of incident impacts HCM, planning model Medium New toll lanes on existing roads Add new system capacity, reduce demand on adjacent freeways and arterials, and reduce level of incident impacts HCM, simulation Medium Interchange improvements Add capacity at bottleneck, reduce potential for secondary incidents HCM, simulation Medium New HOV–managed lanes Add new system capacity, reduce demand on adjacent freeways and arterials, and reduce level of incident impacts HCM, simulation Medium Truck-only lanes Add new system capacity, reduce demand on adjacent freeways and arterials, reduce level of incident impacts, and reduce crash potential by eliminating auto–truck speed and braking differential HCM, simulation Medium Add Capacity—Arterials New arterials Add new system capacity, reduce demand on adjacent freeways and arterials, and reduce level of incident impacts HCM, planning model Medium Widen arterials Add new system capacity, reduce demand on adjacent freeways and arterials, and reduce level of incident impacts HCM, planning model Medium Street connectivity Add new system capacity, reduce demand on adjacent freeways and arterials, and reduce level of incident impacts Simulation Medium Grade separations Reduce delay at intersections and reduce crash potential HCM, simulation Medium HOV–managed lanes Add new system capacity, reduce demand on adjacent freeways and arterials, and reduce level of incident impacts HCM, simulation Medium Note: HCM = Highway Capacity Manual; HOV = high-occupancy vehicle.

23 Table 2.3. Congestion Strategy Effects Matrix: Operational Improvements Strategy Substrategies Included Effect on Congestion Sources Factors Affecting Reliability Strategy Implementation Existing Methodology to Calculate Effects Significance of Expected Effect on Reliability Operational Improvements—Freeways TMC Operations Integrated real-time incident manage- ment, verification, detection, and traveler information Reduces delay due to incidents, weather, special events, work zones, and bottlenecks Geographic coverage, equipment density, con- gestion level, and program aggressiveness IDAS High Service patrols Must include incident scene manage- ment methods Reduces delay due to incidents Geographic coverage, vehicle route density, congestion level, and program aggressiveness IDAS High On-scene incident management improvements Response agency coordination and training Reduces delay due to incidents Program aggressiveness IDAS Medium Remote verifica- tion (CCTV) Camera views available to multiple agencies and in TMC Reduces delay due to incidents Geographic coverage, equipment density, and program aggressiveness IDAS High Event management Incident management coordination among agencies and event ingress– egress planning and coordination Reduces delay due to special events Geographic coverage, equipment density, con- gestion level, and program aggressiveness IDAS Medium Ramp metering Ramp meter algorithms based on real-time traffic information Reduces delay due to incidents, weather, special events, work zones, and bottlenecks Geographic coverage, equipment density, con- gestion level, and program aggressiveness IDAS, simulation High Lane controls DMS over lanes to close lanes in advance of incidents Reduces delay to incidents, special events, and work zones Geographic coverage, equipment density, con- gestion level, and program aggressiveness IDAS, simulation High Managed lanes HOV lanes, HOT lanes, truck-only lanes, and TOT lanes Reduces delay due to incidents and bottlenecks Geographic coverage, equipment density, con- gestion level, and program aggressiveness Simulation High Electronic toll collection Toll payment by electronic toll tags Reduces or eliminates delay at toll booths Geographic coverage, equipment density, con- gestion level, and program aggressiveness Simulation High Real-time traveler information Pretrip information by 511, web sites, subscription alerts; en route infor- mation on DMS, 511, and real-time navigation systems Reduces delay due to inci- dents, weather, special events, work zones, and bottlenecks Geographic coverage, equipment density, con- gestion level, and program aggressiveness IDAS High Work zone management Active management in TMC coverage areas, real-time information from portable equipment in non-ITS areas Reduces delay in work zones Geographic coverage, equipment density, con- gestion level, and program aggressiveness IDAS, simulation, QuickZone High Road weather information systems Weather information supplied to TMCs from roadside weather stations Reduces delay due to inci- dents and weather Geographic coverage, equipment density, con- gestion level, and program aggressiveness IDAS High Road weather pretreatment Application of anti-icing chemicals on defined road segments to pre- vent or retard icing Reduces delay to incidents and weather Geographic coverage, equipment density, and program aggressiveness IDAS Medium (continued on next page)

24 Table 2.3. Congestion Strategy Effects Matrix: Operational Improvements (continued) Strategy Substrategies Included Effect on Congestion Sources Factors Affecting Reliability Strategy Implementation Existing Methodology to Calculate Effects Significance of Expected Effect on Reliability Variable speed limits DMS to change speed limits based on current conditions Reduces delay due to incidents, weather, special events, and work zones Geographic coverage, equipment density, and program aggressiveness Simulation High Ramp improvements Construct additional ramp lanes and lengthen ramps to provide longer acceleration space Reduces delay due to bottlenecks Extent of improvement Simulation Medium Ramp closures Close entrance ramps in areas with closely spaced ramps Reduces delay due to bottlenecks Extent of closures and ramp spacing Simulation Medium Bottleneck removal Add auxiliary lanes and improve road geometrics Reduces delay due to bottlenecks Geographic coverage and congestion level Travel demand models, simulation High Integrated multi- modal corridors Integrated control of freeways and arterials within a corridor Reduces delay due to incidents, weather, special events, work zones, and bottlenecks Geographic coverage, equipment density, and program aggressiveness Travel demand models, simulation High Advanced technol- ogy for freight management Fleet management, advanced vehicle location, real-time truck traveler information, roadside permitting– inspection, and weigh-in-motion Reduces truck delay Geographic coverage, equipment density, and program aggressiveness IDAS Medium Operational Improvements—Arterials Geometric improvements Reduce grade and curvature Reduces delay due to incidents and bottlenecks Geographic coverage and congestion level HCM, HERS Low Intersection improvements Add turn lanes, improve intersection geometrics Reduces delay due to bottlenecks Geographic coverage and congestion level Simulation, HCM Low One-way streets Convert two-way streets to one-way Reduces delay due to bottlenecks Geographic coverage and congestion level Travel demand models, simulation Medium (continued on next page)

25 Access management Reduce driveways on arterials, provide interparcel access Reduces delay due to bottlenecks Geographic coverage and congestion level Travel demand models Medium Advanced signal systems Centrally controlled signals, advanced detection, and advanced signal con- trol strategies Reduces delay due to poor sig- nal timing Geographic coverage, equipment specifications, and program aggressiveness Simulation High Signal retiming and optimization Regularly scheduled signal optimiza- tion programs Reduces delay due to poor sig- nal timing Geographic coverage, equipment specifications, and program aggressiveness Simulation High Changeable lane assignments Reversible lanes Reduces delay due to bottlenecks Geographic coverage and congestion level Simulation Medium HOV by-pass ramp Provide by-pass lanes for HOVs and buses at entrance ramps Reduces delay due to ramp bottlenecks Congestion level Simulation Medium Parking restrictions Restrict parking on arterial streets dur- ing peak hours Reduces delay due to bottlenecks Geographic coverage and congestion level Simulation Medium Incident management Incident management coordination among agencies focused on arterials Reduces delay due to incidents Geographic coverage, vehicle route density, congestion level, and program aggressiveness IDAS Medium Event management Incident management coordination among agencies and event ingress– egress planning and coordination Reduces delay due to special events Geographic coverage, equipment density, con- gestion level, and program aggressiveness IDAS Medium Road weather information systems Weather information supplied to TMCs from roadside weather stations Reduces delay due to incidents and weather Geographic coverage, equipment density, con- gestion level, and program aggressiveness IDAS High Remote verification (CCTV) Camera views available to multiple agencies and in TMC Reduces delay due to incidents Geographic coverage, equipment density, and program aggressiveness IDAS High Real-time traveler information Pretrip information by 511, web sites, subscription alerts; en route informa- tion on DMS, 511, and real-time navigation systems Reduces delay due to incidents, weather, special events, work zones, and bottlenecks Geographic coverage, equipment density, con- gestion level, and program aggressiveness IDAS High Note: TMC = traffic management center; IDAS = ITS deployment analysis system; HOT = high-occupancy toll; TOT = truck-only toll; DMS = dynamic message sign; HERS = Highway Economic Requirements System. Table 2.3. Congestion Strategy Effects Matrix: Operational Improvements (continued) Strategy Substrategies Included Effect on Congestion Sources Factors Affecting Reliability Strategy Implementation Existing Methodology to Calculate Effects Significance of Expected Effect on Reliability

26 2. Cross-sectional analysis—Patterned after classical exper- imental design, this approach establishes a matrix of fac- tors and their levels. Ideally, observations are taken for each combination of factors. But as noted, strict control of all factors was not achievable; consequently, there were missing combinations, which precluded studying interac- tions directly from the field data. Statisticians refer to this situation as a quasi-experimental design. In this approach, experimental design is used to ensure that a range of con- ditions is represented in the data. 3. Congestion by source analysis—Identifying the contribut- ing factors (the seven sources) to congestion and reliability is a major concern for the transportation profession. Table 2.5 shows how several previous studies identified congestion by source. The primary issue is how to split up delay so that each contributing source gets a share. The analyst must decide how much delay would have occurred in the absence of the event and how to reasonably split the delay when multiple sources are at work. These decisions are further complicated if a crash occurs when congestion sources such as inclement weather and work zones, which can increase the likelihood of a crash, are present. Should the resulting delay be charged completely to the weather or work zone category, or shared with the incident category? Table 2.4. Congestion Strategy Effects Matrix: Demand Management Category Strategy Substrategies Included Expected Effect on Reliability Existing Methodology to Calculate Effects Significance of Expected Effect on Reliability Travel alternatives Public education on aggressive driving Public service announce- ments, driver training, and brochures Reduce crashes due to aggressive driving, fewer incidents None Low Travel alternatives Reduction in trips, diversion to other modes and/or times Transit trip itinerary plan- ning, real-time transit information, and com- mercial vehicle fleet scheduling Reduce trips and reduced congestion Travel demand modeling Medium Land use Smart growth policies Transit-oriented design, access management, street connectivity, bike–pedestrian facilities, and mixed use development Reduce trips and reduced congestion Travel demand modeling Medium Pricing Reduction in trips or time shift due to pricing Toll roads, HOT lanes, time-of-day pricing, cor- don pricing, parking pric- ing, and HOV parking Reduce trips and reduced congestion Travel demand modeling Medium HOV Rideshare programs Vanpool and carpool pro- grams, transportation management associations Reduce trips and reduced congestion Travel demand modeling Medium Freight Truck-only toll lanes Toll lanes exclusively for trucks and time-of-day pricing Removes trucks from gen- eral purpose lanes, reduces truck–auto conflicts, reduces crashes, and reduces congestion in general- purpose lanes by removing slower trucks Simulation Low Freight Lane restrictions Restrict left lanes from use by trucks Reduces truck conflicts in restricted lanes, reduces crashes, reduces congestion in restricted lanes Simulation Low Freight Delivery restrictions Restrictions on deliveries in peak hours Reduces congestion in restricted areas during peak hours Travel demand modeling Low

27 Factors Considered The experimental design is detailed in Table 2.6. The top-level design in Table 2.6 shows the overarching factors that were studied. The experimental design does not specify a classic factorial experiment because the number of locations needed to cover all possible factorial combinations was prohibitive. Rather, the experimental design was used to ensure that a range of conditions was covered by the data and to identify the important factors and levels of those factors that were desirable, but not necessarily achievable. The combinations of factors that resulted, therefore, were dependent on the data that could be assembled. However, it was useful to document what the experimental design matrix looked like after the data were assembled, as it provided a basis for seeing what interactions could be studied. The approach outlined in Table 2.6 is obviously a compro- mise, but it was decided early in the study that if empirical data were used, then for cost control the team would have to access data already being collected by transportation agen- cies. A long history of travel time data is needed to establish reliability, and the cost of undertaking special instrumenta- tion to collect these data would have been exorbitant. Instead, team members identified areas in which their past experience indicated that data were of sufficiently high quality to under- take the research. Originally, it was thought that rural two- lane highways could be studied, but data availability at the time of the study was nonexistent, and the team wanted to focus new data collection efforts on signalized highways, where reliability and congestion are greater issues. One key factor common to all improvement types and any predictive relationship of reliability is traffic pressure or demand level. In Table 2.6, the AADT/C ratio is used as a gen- eral measure of congestion level to ensure that roadways at all levels are considered in the analysis. AADT/C also may be used directly as an independent (predictor) variable in reliability relationships, but doing so masks the peaking characteristics of the facility. Other indicators of traffic pressure may include single- or multiple-hour volume-to-capacity ratios. Variations in traffic demand variability also influence traffic pressure. Accurately characterizing traffic demand was a critical part of the research. The data collection plan was clearly oriented to facility-level rather than corridor- or system-level analysis. Existing continuous data collection activities by public agen- cies, on which the research heavily relied, were concentrated on major facilities, usually freeways; data on parallel nonfree- ways were scarce to nonexistent. During times of severe con- gestion, traffic demand can be suppressed by travelers switching to alternative routes or delaying their trips. Controlling for this diversion effect was handled by carefully measuring traffic demand on the test facilities; original data collection to cap- ture diversion was cost prohibitive for this study, given the wide range of conditions that needed to be addressed. The entry in Table 2.6 for proximity to a major bottleneck requires elaboration. If a major bottleneck (e.g., a freeway- to-freeway interchange) is immediately downstream of a study segment, then it will tend to dominate congestion on it (i.e., queues will routinely form on the study segment). It is, there- fore, important to note both the presence and characteristics Table 2.5. Results from Previous Studies Identifying Congestion by Source Study Statistics Dowling Associates et al. (24) Kopf et al. (25) Kwon et al. (26) CDTC (27) Metro area Los Angeles Seattle San Francisco Albany Route I-10 I-405, I-90, SR 520 I-880 I-87, I-90 Freeway (mi) 10 42 45 15 Amount of data 7 days 4 months 6 months 1 year Total Delay Recurrent delay 69% 71% 80% 72% Nonrecurrent delay 31% 29% 20% 28% Nonrecurrent Sources Incident 31% 16% 13% 28% Work zone Not studied Not studied Not studied Not studied Weather Not studied 9% 2% Not studied Special events Not studied Not studied 5% Not studied High volume Not studied 4% Not studied Not studied

28 Table 2.6. Experimental Design Factors Levels Highway Type Urban Rural Freeways Signalized Arterials Freeways Area size Small, medium l l Large, very large l l Base congestion Low (AADT/Ca <7) l Moderate (AADT/C ~9) l l Severe (AADT/C ~12) l l Number of lanes 4 l l l 6 l l 8+ l l Base crash rateb Low l l l High l l l Trucks (%) <10% l l l >10% l l l Traffic variabilityc Low l l l High l l l Traffic signal density <2/mile l 2–5/mile l >5/mile l Proximity to major bottleneck <1 mile downstream from segment l >5 miles downstream from segment l Improvement type Incident management l l l Work zone management l l l Weather managementd l l Traffic device controle l l Demand management l l Special event management l l Traveler information l l l Physical expansion and/or changes l l l a AADT/C is annual average daily traffic-to-capacity ratio (specifically, two-way hourly capacity). b Categories were based on comparison to each state’s average crash rate by type of highway. c For urban highways, traffic variability was determined based on the coefficient of variation (CV) of weekday peak period travel. For rural highways, the CV of the 24-hour volume was used. d Weather management depended on what was being covered in other research activities, such as FHWA’s Road Weather Research and Development Program. e Ramp meter control on freeways; signal control on signalized arterials.

29 (e.g., capacity) of a nearby downstream bottleneck. If the bottleneck is upstream of the study segment, then flow onto the study segment will be limited or metered as a result of the lower discharge rate from the oversaturated bottleneck. This is a potential problem because the study segment may not ever receive enough demand to cause recurring congestion. Additional subfactors varied by type of improvement or type of source delay. The key was ensuring that a spread of conditions was represented: • Incidents—Presence of a usable shoulder on each side of the highway; levels of incident management that lead to low, medium, or high average incident durations; • Work zones—Nature of geometric change, translated into Highway Capacity Manual–based capacity loss to account for multiple combinations (such as lane narrowing with and without shoulder loss): <5%, 5% to 15%, 15% to 30%, 30% to 50%, and 50% to 75%; and • Traffic signals—Type of progression: actuated, central control, or adaptive. Facility-Based Spatial Measurement Scale Because nearly all the data were based on measurements taken at the roadway (not the trip) level, the focus of the work was to define reliability at the facility level. This focus pro- vided the most practical results for implementation, at least in the short run. Several spatial levels were investigated: • Urban links (distance between signalized intersections and freeway interchanges); • Urban facility segments (distance between multiple signal- ized intersections and multiple freeway interchanges): 44 2 to 5 miles for freeways, and 44 1 to 3 miles for arterials; and • Rural extended sections (long stretches of rural highways, probably 30 to 200 miles in length). Temporal Measurement Scale Reliability measurements for the following time periods were captured and used in the analysis: • Peak hour and peak direction (based on maximum volume); • Peak period (to encompass typical commuting times that include most delay, broken down by a.m. and p.m. and directionality); • Midday or overnight; • Daily (to encompass all delay); and • Weekday versus weekend. references 1. National Cooperative Highway Research Program. Guide to Effec- tive Freeway Performance Measurement. Research Results Digest 312, Transportation Research Board of the National Academies, Wash- ington, D.C., February 2007. http://onlinepubs.trb.org/onlinepubs/ nchrp/nchrp_rrd_312.pdf. 2. National Transportation Operations Coalition (NTOC) Performance Measurement Initiative. Final report. 2005. http://downloads .transportation.org/2005am/bod/A%20Few%20Good%20Ops% 20PM%20Report.pdf. 3. Georgia Regional Transportation Authority. 2007 Transportation Metropolitan Atlanta Performance Report. Atlanta, 2007. 4. Georgia Department of Transportation. Strategic Planning. www .opb.state.ga.us/strategic-planning/strategic-planning%282%29 .aspx. Accessed January 2005. 5. Florida Statewide Operations Performance Measures and Data Col- lection. Traffic Engineering and Operations Office, Florida Depart- ment of Transportation, Tallahassee, October 2008. www.dot.state .fl.us/trafficoperations/ITS/Projects_Deploy/PerfMeas/081028%20 FDOT%20ITS%20PM%20Annual%20Report%202008%20_ final__CamSys.pdf. Accessed Aug. 13, 2012. 6. Southern California Association of Governments. Goods Move- ment in Southern California: The Challenge, the Opportunity, and the Solution. 2005. www.scag.ca.gov/goodsmove/pdf/Goodsmove Paper0905.pdf. 7. Washington State Department of Transportation. WSDOT Accountability and Performance Information. www.wsdot.wa.gov/ accountability/. Accessed May 1, 2012. 8. Jones, C., D. Murray, and J. Short. Methods of Travel Time Measure- ment in Freight-Significant Corridors. American Transportation Research Institute, Alexandria, Va., 2005. www.atri-online.org/ research/results/Freight%20Performance%20Measures%20 TRB%20for%20atri-online.pdf. 9. Cambridge Systematics, Inc., Texas Transportation Institute, Uni- versity of Washington, and Dowling Associates. Guide to Effective Freeway Performance Measurement: Final Report and Guidebook. NCHRP Web-Only Document No. 97. Transportation Research Board of the National Academies, Washington, D.C., 2006. http:// onlinepubs.trb.org/onlinepubs/nchrp/nchrp_w97.pdf. 10. Tu, H., J. W. C. van Lint, and H. J. van Zuylen. Travel Time Reli- ability Model on Freeways. Presented at 87th Annual Meeting of the Transportation Research Board, Washington, D.C., 2008. 11. Lomax, T., D. Schrank, S. Turner, and R. Margiotta. Selecting Travel Reliability Measures. Texas Transportation Institute and Cambridge Systematics, Inc., 2003. http://tti.tamu.edu/documents/TTI-2003-3 .pdf. Accessed Aug. 12, 2012. 12. University of Florida. Travel Time Reliability and Truck Level of Ser- vice on the Strategic Intermodal System, Part A: Travel Time Reliabil- ity. Final report. Florida Department of Transportation, Tallahassee, April 2007. www.dot.state.fl.us/research-center/Completed_Proj/ Summary_PL/FDOT_BD545_48_Part_A.pdf. 13. Rakha, H., I. El-Shawarby, and M. Arafeh. Trip Travel Time Reli- ability: Issues and Proposed Solutions. Journal of Intelligent Trans- portation Systems, Vol. 14, No. 4, 2010, pp. 232–250. 14. Engineering Statistics Handbook. National Institute of Standards and Technology, U.S. Department of Commerce, 2007. www.itl .nist.gov/div898/handbook/. Accessed May 11, 2012. 15. van Lint, J. W. C., and H. J. van Zuylen. Monitoring and Predicting Freeway Travel Time Reliability: Using Width and Skew of Day-to- Day Travel Time Distribution. In Transportation Research Record:

30 Journal of the Transportation Research Board, No. 1917, Transportation Research Board of the National Academies, Washington, D.C., 2005, pp. 54–62. 16. Cambridge Systematics, Inc., Boston Logistics Group, and Global Insight, Inc. Freight Transportation Demand and Logistics: Bottom Line Report. AASHTO, 2006. http://downloads.transportation.org/ DR_3%20Freight%20Demand_Report-12-07.pdf. 17. Tracker: Measures of Departmental Performance. Report. Missouri Department of Transportation, Hannibal, July 2007. www.modot .org/about/tracker_archive/documents/Tracker_PDF_July07/ Tracker_July07.pdf. 18. McCormack, E., and M. E. Hallenbeck. Options for Benchmarking Performance Improvements Achieved from Construction of Freight Mobility Projects. Final report. Washington State Transportation Center, Seattle, 2005. www.wsdot.wa.gov/research/reports/full reports/607.1.pdf. 19. Ullam, G. L., R. J. Porter, and G. J. Karkee. Monitoring Work Zone Safety and Mobility Impacts in Texas. Report 0-5771-1. Texas Trans- portation Institute, College Station, May 2009. http://tti.tamu.edu/ documents/0-5771-1.pdf. Accessed Aug. 13, 2012. 20. Incorporating Intelligent Transportation Systems into Planning Analysis: Summary of Key Findings from a Seattle 2020 Case Study: Improving Travel Time Reliability with ITS. Intelligent Transpor- tation Systems, U.S. Department of Transportation, May 2002. http://ntl.bts.gov/lib/jpodocs/repts_te/13605.html. Accessed May 21, 2012. 21. Inrix. www.inrix.com. Accessed May 2, 2012. 22. NAVTEQ Traffic. www.traffic.com. Accessed May 2, 2012. 23. Cambridge Systematics, Inc., and Texas Transportation Institute. Traffic Congestion and Reliability: Trends and Advanced Strategies for Congestion Mitigation. Federal Highway Administration, U.S. Department of Transportation, September 2005. www.ops.fhwa.dot .gov/congestion_report/index.htm. 24. Dowling Associates, Berkeley Transportation Systems, and Systems Metrics Group. Measuring Non-Recurrent Traffic Congestion. Cali- fornia Department of Transportation, Sacramento, 2003. 25. Kopf, J. M., J. Nee, J. M. Ishimaru, and M. E. Hallenbeck. Measure- ment of Recurring and Non-Recurring Congestion: Phase IIB. Wash- ington State Transportation Center, Seattle, June 2005. 26. Kwon, J., M. Mauch, and P. Varaiya. Components of Congestion: Delay from Incidents, Special Events, Lane Closures, Potential Ramp Metering Gain, and Excess Demand. In Transportation Research Record: Journal of the Transportation Research Board, No. 1959, Transportation Research Board of the National Academies, Wash- ington, D.C., 2006, pp. 84–91. 27. Capital District Planning Commission. Working Group B Draft Report: Expressway System Options. August 2005. http://www.cdtc mpo.org/rtp2030/b-expressway.pdf.