Understanding the Contributions of Operations, Technology, and Design to Meeting Highway Capacity Needs (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.

1Executive Summary Continuing growth in urban travel demand will inevitably require more physical capacity in the transportation system. However, because of limited financial resources, high construction costs, environmental considerations, long timelines, and an increasingly complex regulatory process, capacity-adding projects have become actions of last resort. It therefore behooves decision makers, planners, and engineers to evaluate operational improvement strategies that can—singly or in combination—eliminate or mitigate the need for a more traditional highway construction project. Effectively evaluating the wide range of operational improvement strategies that are available is not a trivial matter, particularly when their performance is to be compared against the con- struction of new lanes. Traditional travel demand forecasting models are not effective for this kind of comparative analysis for several reasons: • They assume that all drivers have perfect knowledge of the travel time on each of the travel paths available to them, an assumption that masks the effectiveness of operational improve- ment strategies aimed at improving driver awareness. • They assume that the capacity of a freeway link or an arterial segment is a constant value, whereas an emerging body of research indicates that such capacity is better represented as a random variable (1–3). This limitation reduces the effectiveness of traditional tools for com- paring alternatives because fluctuating capacity introduces variability that measurably affects vehicle assignments and network performance characteristics. • They are not usually sensitive to the effects that upstream bottlenecks and blockages can have on downstream service rates. As an example, the models do not generally recognize that when the upstream queue of a separate turn lane extends into the adjoining through lane, this block- age prevents through traffic from reaching the downstream intersection for as long as the blockage exists, even when the downstream signal is green. • They assume that all vehicle trips identified in the origin–destination (O-D) matrix will be completed by the end of the time period being analyzed, regardless of whether there is actually sufficient capacity to accommodate these trips within the specified time window. Thus, each vehicle trip is assigned to an entire travel path from origin to destination, even if some bottle- necks along that path operate with a volume/capacity ratio greater than 1.0. Some modeling advancements are beginning to address these issues, but the advancements have not yet reached the point of practical and regular application, nor do they address all of the issues simultaneously. Ideally, the analysis methods should enable evaluation of improvement strategies that cut across the full spectrum of operations, technology, and design. They should also provide multiple performance measures that can be used to evaluate different strategies according to their impacts at the point, link, corridor, and network levels.

2This report summarizes the results of a capacity project undertaken through the second Strategic Highway Research Program (SHRP 2) to advance the state of practice in this area. The objectives of this project were to (a) quantify the capacity benefits, individually and in combi- nation, of operations, design, and technology improvements at the network level for both new and existing facilities; (b) provide information and tools to analyze operational improve- ments as an alternative to traditional construction; and (c) develop guidelines for sustained service rates (SSRs) to be used in planning networks for limited access highways and urban arterials (2). Strategies Selected for Testing Table ES.1 lists 25 operational strategies that were selected from an initial list of more than 100 as particularly effective in enhancing the performance characteristics of links, corridors, and networks. Some of the strategies are applicable only to freeways, some only to arterials, and some to both. These strategies stood out from others because of the following characteristics: • Their ability to reduce recurring congestion effects during peak periods; • Their ability to be implemented rather quickly by agency decision makers, compared with major capital improvement projects; • The feasibility of implementing them considering economic, social, political, and environ- mental factors; • Their general capacity-enhancing effects; and • The number, location, and characteristics of known successful applications. Network Operations Modeling Approach The effectiveness of each operational strategy listed in Table ES.1 was found to vary according to the context in which it is applied. Physical factors such as network structure as well as the exis- tence and relative proximity of freeway/arterial alternatives have an important influence on a particular strategy’s effectiveness, as do travel desire lines and overall demand levels. It is thus not possible to reliably estimate the effectiveness of a particular operational strategy in a particular network and demand setting from static, location-blind lookup tables. Instead, some form of a travel demand forecasting model is necessary. Table ES.1. Non-Lane-Widening Strategies to Improve Capacity Freeway Arterial Both HOV lanes Signal retiming Narrow lanes Ramp metering Signal coordination Reversible lanes Ramp closures Adaptive signals Variable lanes Congestion pricing Queue management Truck-only lanes Pricing by distance Raised medians Truck restrictions HOT lanes Access points Pretrip information Weaving section improvements Right and left turn channelization In-vehicle information Frontage road Alternate left turn treatments VMS/DMS Interchange modifications Note: HOV = high-occupancy vehicle; HOT = high-occupancy toll; VMS = variable message sign; DMS = dynamic message sign.

3Dynamic traffic assignment (DTA) models are especially advantageous for evaluating strategy effectiveness because they provide a realistic assignment of traffic in oversaturated networks: • They recognize that drivers have varying levels of knowledge about the travel time on each of the travel paths available to them. • They recognize that the effects of congestion and queues can prevent drivers from reaching their destinations in a timely manner and therefore do not assume that all vehicle trips identi- fied in the OD matrix will be completed by the end of the time period being analyzed. The capabilities of DTA models overcome some but not all of the limitations associated with traditional travel demand models. A review was conducted of available DTA models, and none included all the modeling capabilities desired. Even so, several DTA models use a common and generally accessible graphical user interface (GUI) for input data; among these are DYNASMART-P (Dynamic Network Assignment Simulation Model for Advanced Roadway Telematics: Plan- ning version) (4), DynusT (5), and DTALite (6). For this research project, the internal logic of DYNASMART-P was modified to incorporate several analytic enhancements (described in the following subsections), and the new version served as the test engine for the validation and dem- onstration activities. The modified version of DYNASMART-P developed in this project is not available for general use, but other DTA models can incorporate these enhancements and at least one open source model (DTALite) has already done so. A summary of the modeling enhancements incorporated into the updated DTA model follows. Stochastic Capacity for Freeway Bottlenecks Traditional DTA models assume a constant capacity for freeway bottlenecks, which are generally located at merge points, lane drops, and weaving areas. However, empirical data reveal that breakdowns occur across a range of volumes, even at the same location. Therefore, a probabilistic approach was developed wherein random capacity values are generated at fixed time intervals (15 minutes) during queue-free time periods based on an empirically derived distribution of pre-breakdown headways as shown in Figure ES.1. This approach results in freeway bottleneck Note: pc/h/ln = passenger cars per hour per lane. 0.00% 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% 90.00% 100.00% 0 50 100 150 200 250 300 Fr e qu e n c y Time Headway (s) Frequency Fitted_Log-Normal Cumulative % Fitted_Cumulative 85th Flow = 1778 pc/h/ln 50th Flow = 1976 pc/h/ln 15th Flow = 2114 pc/h/ln Figure ES.1. Stochastic headway/capacity flow distribution for freeway bottlenecks.

4activity that randomly varies from day to day. During simulation periods when these random freeway queues are present, the queue discharge rate at active bottlenecks is also represented with random variation by modeling it as a time-correlated stochastic process. The simulated free- way environment created by the joint action of the random queue-free capacity and the time- correlated queue discharge provides a realistic representation of day-to-day variation in recurring freeway network congestion. Details on stochastic freeway models are available in Jia, Williams, and Rouphail (3). Stochastic Capacity for Arterials The bottleneck points for signalized arterials are usually very easy to identify—they are most often located at intersections—and traditional DTA models once again assume a constant satu- ration flow rate during the green time either for the approaching links or for individual turn movements at these locations. However, it is well known that the saturation flow rate for indi- vidual links and turn movements varies significantly according to the behavior of individual drivers. At signalized intersections this is revealed by saturation flow rates that vary from cycle to cycle. Therefore, a stochastic approach was developed that allows the capacity of a signalized intersection to vary during each time interval (typically 15 minutes) according to an empirically observed distribution such as that illustrated in Figure ES.2. Improved Arterial Bottleneck Representation The approaches to signalized intersections along arterial roadways often include left and right turn pockets as a way of separating turn movements and increasing capacity. But when the queue length of through and/or turning vehicles extends beyond the length of the turn pocket, the result is a demand blockage that prevents upstream vehicles from taking advantage of the capac- ity available at the intersection (Figure ES.3). This is an important phenomenon to model in oversaturated networks because it directly affects the efficiency and productivity (or SSRs) Figure ES.2. Stochastic headway/capacity flow distribution for arterials. All Sites, meanlog=0.76, sdlog=0.20 0.00 % 10.00% 20.00% 30.00% 40.00% 50.00% 60.00% 70.00% 80.00% 90.00% 100.00% 0 20 40 60 80 100 120 140 160 180 1. 1 1. 4 1. 7 2 2. 3 2. 6 2. 9 3. 2 3. 5 3. 8 4. 1 4. 4 4. 7 5 5. 3 Fr eq ue nc y Average Headway (Seconds) Frequency Fitted Log Normal Cumulative% Fitted Cumulative

5of individual links and turn movements. Therefore, an enhanced model was developed that recognizes when queue lengths exceed available storage lengths at these locations and adjusts the downstream discharge rate accordingly. Day-to-Day Learning The consequence of stochastic freeway and arterial bottlenecks is that traffic flow and the pres- ence or absence of breakdown conditions at a specific time on any particular link vary from day to day. This, combined with day-to-day variations in travel demand, means that real-world driv- ers take into account the conditions they have encountered across multiple days to make route choices based on the expected travel time of their available options. This more realistic represen- tation of drivers’ route selection processes was incorporated into the enhanced model by simu- lating multiple consecutive days and allowing a user-defined fraction of randomly selected drivers in each O-D pair the ability to remember their travel time experiences over the most recent days when making their route choices. As illustrated in Figure ES.4, a simulation-based evaluation framework was used in this study to estimate the system performance for a multiday planning horizon under stochastic link capacity. In addition to stochastic road capacity, day-to-day travel time variability is affected by the manner in which travelers obtain, process, and react to traveler information. To account for infor- mation uncertainty and cognitive limitations of individual travelers, the theory of “bounded rationality” (4, 7)—the concept that decision-making abilities are constrained by the informa- tion at hand and the available time to make a decision—is adapted in this study to describe route switching and departure time choice behavior. One of the aims of this study was to enhance a mesoscopic dynamic traffic simulator by incor- porating stochastic road capacity for both freeway and arterial links and by developing a new set of day-to-day learning and route updating models under stochastic travel time variations (intro- duced by variable capacity). The introduction of stochastic capacity at critical points in the net- work that suffer from queue and congestion more frequently, such as freeway bottlenecks and signalized intersections, enables reasonable and realistic modeling of travel time variability and sustainable flow rates. Figure ES.3. Blockage effects of short turn pockets.

6Under stochastic capacity, the travel time experience on a single day can be dramatically affected by the underlying capacity on that particular day. (In this study, “day” is defined as any regular weekday; i.e., a day other than a weekend or a public holiday.) This study therefore devel- oped a set of reliability-oriented system performance measures that consider multiple days’ per- formance in order to systematically evaluate medium-term benefits of traveler information provision strategies. In addition to the previously described model enhancements, the new network diagnostic features in the following list facilitate network evaluation as well as identification of points and corridors where the potential benefits of implementing one or more operational strategies appear to be high. • Active bottleneck identification. Active bottlenecks are defined here as locations (on both free- ways and arterials) where one or more actual breakdowns occur during the simulated days and time periods. They are represented with red circles on the network map. The diameter of each circle is proportional to the number of breakdowns observed during a particular analysis day. The analyst can thus easily identify the network points most susceptible to breakdown, revealing patterns and locations that can suggest specific operational strategies for congestion mitigation. • Movement-specific intersection delay. The delay experienced by individual links and move- ments can be displayed both visually and in tabular format, allowing the analyst to quantify this important performance measure. • Stochastic link performance and breakdown probability. Travel time variability is arguably at least as important to drivers as delay because variability dictates the amount of buffer time they must build into their travel schedule. This measure (defined as the difference between the 5th and 95th percentile travel times) is reported for each link, corridor, and/or O-D pair and provides further insight to the analyst on vulnerable links and corridors where one or more operational improvement strategies might be appropriate. Calibrated Stochastic Capacity Models Enhanced Dynamic Traffic Simulation Engine Day-to-day Travel Learning / Route Switching Module Obtain stable results? Or reach the last day of planning horizon? No Yes Active Bottleneck Identification Day d=d+1 Capacity Enhancing Operational Design and Technological Strategies Stochastic Travel Time Performance Measures Module Developed in SHRP2 C05 Project Figure ES.4. Module developed in SHRP 2 C05 project as a capacity-enhancing strategy evaluation framework.

7• Queue growth and dissipation. A sliding time bar can be used along with a visual representation of the network to observe the start, growth, and dissipation of queues. This allows the analyst to easily trace link breakdowns to their point of origin, again for purposes of identifying one or more operational improvement strategies that might be appropriate. Baseline Models Two separate networks were used to test the enhanced models and demonstrate both the useful- ness and the usability of the new methodology. The first network was a very small subarea of the Dallas–Fort Worth, Texas, area (Figure ES.5). The small size of this network produced great efficiencies in testing and debugging the enhanced DTA models, and it was also a good platform for implementing and evaluating each of the 25 operational strategies presented in Table ES.1. The second network was a subarea of the Portland, Oregon, metropolitan area, encompassing approximately 210 traffic analysis zones, 860 nodes, 2,000 links, and more than 200,000 vehicle trips initiated during a 4-hour weekday time interval between 3 p.m. and 7 p.m. Both DTALite and DYNASMART-P were used in this network application; DTALite provided a regional base- line equilibrium as the starting point for the subarea analysis, which was conducted by using DYNASMART-P. This network clearly demonstrated the usefulness of the procedure in a real- world environment and featured the analytic elements of network diagnosis; identification and evaluation of treatment options; and interpretation of the results. Strategy Testing A straightforward method was developed to test the effectiveness of one or more operational strate- gies either as stand-alone projects or as alternatives to traditional new construction projects. • First, the location of the operational strategy and/or new construction project to be tested is identified and a subarea or network that appropriately surrounds the location is established. Figure ES.5. Fort Worth area. Source: © 2010 Google Maps.

8• Next, geometric, volume, and operational characteristics of each link in the subarea are identi- fied and provided as inputs to the DTA model, including stochastic capacity distributions at the geometric or operational bottlenecks. Appropriate link, corridor, and/or network perfor- mance measures are also established for subsequent evaluation purposes. • To effectively use the day-to-day learning process and generate results that can be usefully compared, the DTA model must be run under three separate regimes as shown in Figure ES.6. During the baseline stabilization period (Regime I), the DTA model is run for a period of simulated days to achieve equilibrium (i.e., without any of the operational strategies or new construction projects that are to be evaluated). The number of simulated days necessary to achieve equilibrium will vary according to the characteristics of the network and/or subarea being investigated. Figure ES.6 shows that for the Dallas–Fort Worth network the 200-day baseline stabilization period was longer than necessary. This was not a problem because the subarea was small and the runtime for each simulated day was very short. For larger networks, a baseline stabilization period of 50 days may be more appropriate. After baseline stabilization has been achieved, the operational strategies and/or new construction projects to be evaluated are introduced into the network and the DTA model is run for an additional period of simulated days to allow driver adjustments and achieve stable conditions under the new scenario. This strategy stabilization period is illustrated as Regime II in Figure ES.6. A period of 30 simulated days is generally sufficient to achieve stabilization under Regime II conditions. After conditions have stabilized in Regime II, the DTA model is run for an additional 20 simu- lated days (Regime III in Figure ES.6). The results of this period are compared with those of Regime I for the purpose of evaluating the effectiveness of the strategies being tested. This methodology provides useful information about the effectiveness of the operational strategies and new construction projects being evaluated. As an example, consider the capacity addition scenarios that were tested for a southbound freeway corridor section in the Dallas–Fort Worth subarea network. Figure ES.7 illustrates the existing (baseline) condition and three lane addition projects (A, B, and C) that were contemplated and tested. In addition to these lane addition projects, four operational strategy alternatives to the proj- ects were evaluated: • An advanced traveler information system (ATIS) strategy in which the fraction of drivers with access to pretrip information (e.g., via radio, television, or the Internet) increased from 1% to 10%; I II III Figure ES.6. Overview of strategy testing plan under stochastic capacity conditions.

9• Another ATIS strategy in which the fraction of drivers with access to en route information (e.g., through in-vehicle navigation systems) increased from 1% to 10%; • An operational modification to the existing baseline condition in which the width of the free- way lanes and shoulder in a critical 3.1-mi (5-km) section of the southbound (SB) freeway corridor was narrowed so that a fifth lane could be introduced; and • An operational modification to the existing baseline condition in which one northbound (NB) lane was reversed in the same 3.1-mi (5-km) section during the peak hour so that a fifth lane could be added in the southbound direction. Figure ES.8 summarizes travel time results taken from a subarea test network for the southbound direction along an 8.5-mi freeway study section. The analysis was performed for a peak-period condition across a 20-day time horizon. The study section consists of three segments that have four, five, and four lanes, respectively. The gray bar represents performance for baseline conditions. The black bars represent the effects of individual non-lane-widening strategies. The white bars repre- sent three lane-widening scenarios: (1) five lanes across all three segments; (2) one additional lane across all three segments; and (3) six lanes across all three segments. Figure ES.7. Capacity-enhancing scenarios for a southbound freeway corridor. A Baseline # of Lanes B C 4 5 5 5 5 6 6 4 +16.0 (+46%)+8.5 (+24%) Lane-Mile Addition +7.6 (+22%) 8.5 miles 1.9 miles1.9 miles 0.9 miles0.9 miles 5.7 miles5.7 miles 8.5 miles Figure ES.8. 20-day model results at corridor level. SB Travel Times Average Travel Time 5th Percentile to 95th Percentile Travel Time Range LEGEND Tr a v e l T im e (M in u te s ) 16 14 12 10 8 6 4 2 0 B a s e lin e AT IS Pr e - tr ip AT IS En - ro u te N a rr o w La n e s R e v e rs ib le La n e s A (+ 22 % ) B (+ 24 % ) C (+ 46 % )

10 Figure ES.8 demonstrates how trade-offs for improvement strategies can be examined in terms of their impact on average travel time (expressed as minutes of travel time) and travel time reliability (expressed as the range between the 5th and 95th percentile travel times). In some cases (e.g., the provision of pretrip information), travel time reliability associated with the tested option is significantly improved in relation to the base condition, even though the average travel time is largely unaffected. In other cases, such as the narrow-lanes strategy and each of the new construction projects, both travel time and travel time reliability are significantly improved by the tested option, although the narrow-lanes strategy may have negative safety impacts that were not considered in this analysis. Without the examination and assessment of reliability as a per- formance measure, a primary benefit of the strategies would go unrecognized, particularly for the non-lane-widening strategies. These results were taken from a test network and should not be considered representative of outcomes that can be expected in other applications, because they are dependent on the particu- lar characteristics of the network and the travel demand levels that are being modeled. Model Portability Considerations The enhanced DTA model described in this report can be applied in virtually any local network environment with only a few relatively straightforward adaptations: • The stochastic capacity distribution functions for arterials and freeways should be modified to reflect local driving and car-following characteristics. This can be easily done by collecting and analyzing discharge headway distribution data at signalized intersections as well as pre- breakdown, breakdown, and post-breakdown speed-flow characteristics at freeway bottle- necks. In both cases, care should be taken to ensure that data are collected at locations not influenced by upstream or downstream intersections or bottlenecks. • Before testing alternative operational strategies, the network structure and O-D patterns should be examined and calibrated under known existing conditions to ensure that the model adequately replicates them. This is identical to the calibration process used for many years with traditional travel demand forecasting models. Even so, it is likely that the network struc- ture associated with a well-calibrated traditional travel demand forecasting model will need some modification before it can be used effectively by the enhanced DTA model. This is because the performance of the latter is more sensitive to certain network characteristics (e.g., the number, location, and length of centroid connectors; the type of intersection control; and the length of intersection turn lanes). • The capacity adjustments that necessarily accompany some of the operational improvements strategies may need to be modified to better reflect local driving habits. For example, the capacity reduction that can be expected from the use of narrow lanes could differ by region or county. Conclusions and Next Steps This report presents an enhanced DTA model; new link, corridor, and network diagnostic tools; and an analytic methodology that can significantly improve the information available to decision makers and thus the robustness of their investment decisions. In today’s environment where financial resources for new transportation construction projects are scarce and where environ- mental, regulatory, and policy constraints make such projects very difficult and time-consuming, it is incumbent upon both transportation professionals and decision makers to consider all viable options to new construction before making a final decision. The new analytic tools made available through SHRP 2 and reviewed in this report represent a significant new capability in this regard. With regard to next steps, incorporating the ability to simulate the additional effects of events that cause nonrecurring congestion (e.g., crashes, other incidents, severe weather) will further

11 enhance the usefulness and usability of these tools. In the meantime, they can still be effective in significantly improving investment decision making in the transportation industry. References 1. Brilon, W., J. Geistefeldt, and M. Regler. Reliability of Freeway Traffic Flow: A Stochastic Concept of Capacity. Proc., 16th International Symposium on Transportation and Traffic Theory, College Park, Md., 2005, pp. 125–144. 2. Lorenz, M., and L. Elefteriadou. A Probabilistic Approach to Defining Freeway Capacity and Breakdown. Transportation Research Circular E-C018: Proceedings of the Fourth International Symposium on Highway Capacity. TRB, National Research Council, Washington, D.C., 2000, pp. 84–95. 3. Jia, A., B. M. Williams, and N. M. Rouphail. Identification and Calibration of Site-Specific Stochastic Freeway Breakdown and Queue Discharge. Transportation Research Record: Journal of the Transportation Research Board, No. 2188, Transportation Research Board of the National Academies, Washington, D.C., 2010, pp. 148–155. 4. Mahmassani, H. S. Dynamic Network Traffic Assignment and Simulation Methodology for Advanced System Management Application. Networks and Spatial Economics, Vol. 1, 2001, pp. 267–292. 5. Chiu, Y.-C., H. Zheng, J. A. Villalobos, W. Peacock, and R. Henk. Evaluating Regional Contra-Flow and Phased Evacuation Strategies for Central Texas Using a Large-Scale Dynamic Traffic Simulation and Assignment Approach. Journal of Homeland Security and Emergency Management, Vol. 5, No. 1, Article 24, 2008. 6. Zhou, X., and C. C. Lu. Traffic Flow Models in DTALite: White Paper. https://sites.google.com/site/dtalite/ traffic-flow-model. Accessed July 27, 2010. 7. Simon, H. A. A Behavioral Model of Rational Choice. Quarterly Journal of Economics, Vol. 69, No. 1, 1955, pp. 99–118.