Identification and Evaluation of the Cost-Effectiveness of Highway Design Features to Reduce Nonrecurrent Congestion (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.

71 This chapter presents both general conclusions of the research and recommendations for the implementation of the research results. The conclusions are discussed as basic summaries of what was learned through literature reviews, interviews with highway agencies, careful examination of research methods and findings from SHRP 2 Project L03 (a key foundation of this research), and methods that were developed by this research team to meet the project objectives. The recommendations presented in this chapter are geared toward highway agency decision makers, including planners, traffic and operational engineers, and managers, who seek to maximize the potential operational benefits of their freeway design decisions within their resource constraints. Conclusions of the research Geometric Design Treatments and Nonrecurrent Congestion The research team found that highway agencies tend to address recurrent congestion issues with infrastructure treatments and nonrecurrent congestion with intelligent transportation sys- tem treatments. That is, daily demand peaks that cause peak hour congestion are often treated by adding base capacity. Con- gestion caused by incidents, special events, work zones, and other infrequent and unpredictable events are typically addressed by providing travelers with real-time information through traffic management centers. These centers monitor freeways and post information about travel time, lane blockages, and alternate routes to drivers in real time via radio, websites, and message boards. Geometric design treatments that address base capac- ity issues have been investigated and evaluated thoroughly in the literature, and more recently, operations-based treatments such as real-time traveler information and motorist-assist patrols have been evaluated for their effectiveness at alleviat- ing nonrecurrent congestion. However, the use of geometric design treatments to help reduce nonrecurrent congestion is not well documented in the literature. Through interviews with highway agencies, the research team identified instances of agencies using design elements to help manage nonrecurrent congestion; however, in most cases these treatments had not been designed specifically for this purpose. Instead, treatments designed to manage recurrent congestion were manipulated to apply to nonrecurrent con- gestion events, frequently in an ad hoc fashion. When major incidents occurred, agencies would use whatever tools were at their disposal to minimize the disruption to traffic. Typi- cally, the facility was not “designed” to function as a treatment for nonrecurrent congestion, and usually there was no policy in place to implement the treatment under certain defined conditions. For example, some agencies will open a shoulder as a driving lane to bypass an incident causing congestion, even though having this option available was not specifically considered during the design of the shoulder, and the decision to implement shoulder driving is made by on-site responders, rather than defined in policy. This research fills an important gap in the literature by doc- umenting the benefits of using design treatments to reduce nonrecurrent congestion and by encouraging the consider- ation of these benefits during the planning and design phases of highway projects. By more accurately predicting the ben- efits of these types of treatments, decision makers are better informed of the available options for addressing nonrecurrent congestion, and greater benefits to the traveling public can be achieved. Relationship Between Nonrecurrent Congestion and Reliability The literature contains a great deal of research on transpor- tation reliability, but there is no consensus on the definition of reliability for roadway segments. Reliability is often dis- cussed in the literature in terms of trip reliability (sometimes for a specific subset of vehicles, such as freight deliveries or commuters), measuring the percentage of on-time trips or C h a p t e r 8 Conclusions and Recommendations

72 the variation between actual trip time and ideal trip time. This research explores the reliability of a specific segment of roadway and includes the travel times of all vehicles travel- ing across the segment. The segment travel time is only one part of each of the various trips made by the drivers on that segment, so little can be known about the reliability of any driver’s trip. However, this analysis can help highway agencies evaluate how well a certain segment of roadway is operating and if it is contributing to trip delay and reliability issues for the drivers using it. This approach makes sense when evalu- ating geometric design treatments that are applied at specific locations on the roadway. This measure of reliability can be used to evaluate how improvements to a section of roadway reduce delay and improve reliability along that section. This research adopted the definition of segment reliability used in SHRP 2 Project L03. For the purposes of the present research, reliability is a measure of the variation in travel times across the segment over a long period of time (here, 1 year). Reliability describes only one characteristic of freeway opera- tions: the predictability of travel times. Delay is another char- acteristic of freeway operations. Both recurrent congestion (resulting from inadequate base capacity for daily demand) and nonrecurrent congestion (resulting from crashes, inci- dents, weather, work zones, and special events) cause delay, and roadway users incur costs from either type of delay. Recurrent congestion alone is generally predictable, and therefore famil- iar drivers can estimate their travel time accurately, factoring in the expected amount of delay, when only recurrent conges- tion is present. However, roadway users incur additional costs when they experience nonrecurrent congestion and their travel times vary from one day to the next. On roadway segments with substantial nonrecurrent congestion, drivers must plan for a longer-than-average trip every day to accommodate the possibility of unexpected congestion, which leads to wasted time. This travel time variability can be described in terms of reliability. Reliability is evaluated separately for each hour of the day, so that the analyst may find a road to be highly reliable during off-peak hours and not very reliable during peak hours. The research team found that events that cause nonrecur- rent congestion have a much bigger impact on reliability during hours of recurrent congestion (i.e., hours with high delay). That is, a crash or work zone will have a bigger impact on travel time when traffic is already congested. For this reason, treatments that reduce recurrent congestion will have a positive impact on reliability. In addition, design treatments that address non- recurrent congestion will have greater benefit on roadways that experience congestion or regularly operate with a demand that approaches capacity (where even a minor disturbance could cause congestion). The primary causes of nonrecurrent congestion on freeways are traffic crashes and other incidents, special events, work zones, weather, demand surges, and sometimes traffic control devices (such as malfunctioning ramp meters). These events cause congestion either by reducing the effective capacity of the roadway or by increasing demand. For example, snow storms often reduce the capacity of a four-lane freeway segment to two lanes, and crashes often block one or more lanes. Special events, such as sporting events and concerts, can substantially increase the demand on a freeway segment before and after the event. Design treatments that can help increase capacity (or decrease the lost capacity) or decrease demand will help to reduce the impact of these events on congestion, and therefore improve reliability. Reliability can be a good measure of the impact of non- recurrent congestion on the operation of a roadway, especially for roadways that experience incidents that cause nonrecur- rent congestion fairly regularly. However, very infrequent major incidents that last for several hours and block several lanes of traffic or shut down a road entirely are not well cap- tured in a reliability measure. Because reliability captures the day-to-day variation of travel times on a segment of roadway, a major incident occurring on a roadway that rarely experi- ences any congestion (either because incidents are infrequent or because traffic demand is low enough that incidents have a very minor impact) may not have much of an impact on reliability. If the roadway operates smoothly 364 days of the year, but is shut down for 1 day, it is highly reliable, despite having serious impacts on the motorists trying to use the roadway on that particular day. And because reliability is measured individually for certain hours of the day, the impact of a catastrophic event is typically spread over several hours. So, although there are treatments that may help alleviate the consequences of major catastrophic incidents (such as using a median opening to allow trapped traffic to turn around), the benefits of these treatments may be more appropriately measured in terms of delay reduction for individual incidents rather than in terms of reliability improvement. Evaluating Treatment Impacts on Reliability Project L03, which preceded this research effort, developed models for predicting a travel time index (TTI) at various percentiles. The input variables to the models were a measure of lane hours lost due to incidents and work zones, the num- ber of hours during the year during which more than a trace amount of rain fell, and the critical demand-to-capacity (d/c) ratio for the roadway segment, all during the particular time- slice being evaluated (e.g., 5:00 to 6:00 p.m.). As explained in detail in Chapter 1, these TTI percentiles can be used to estimate a cumulative distribution of TTIs, from which many observations and measurements can be made. As part of the Project L07 research effort, the research team improved on these models in two important ways. First, the Project L03 models were found to be based on data from cities

73 that did not experience significant snowfall, so this research incorporated a snowfall variable in addition to the rainfall variable in the models. Second, the Project L03 models were developed for peak hours in large metropolitan areas. This research developed additional models to be used for facilities or hours of the day (or both) with lower d/c ratios. Models were needed that could be applied to all 24 h of the day so that the full benefit of treatments that could potentially be used during any hour of the day could be accounted for. The result- ing set of models estimates the distribution of TTIs for a given freeway segment for each hour of the day by using four input variables: rainfall, snowfall, d/c ratio, and lane hours lost. As explained in Chapter 1, the shape of the cumulative TTI curve provides a great deal of information about delay and reliability. A curve with a nearly vertical line at TTI = 1.0 indi- cates that almost every trip on that segment is made at free- flow speed, which means that the roadway is reliable and that drivers experience very little delay. A hypothetical curve with a steeply vertical line at a higher TTI would indicate reliability (very little variance in TTI), but that most drivers experience delay because their trip takes longer than it would at free-flow speed. A curve with a strong “lean forward” indicates a high variability in TTI and, therefore, lower reliability. To measure the impact that a specific design treatment has on reliability, the research team developed a method of mea- suring the difference between a TTI curve for a roadway in an untreated condition and a TTI curve for the treated condition. To develop the curve for the treated condition, the impact of the design treatment must be described in terms of the four model input variables. In general, most treatments have an effect on the lane hours lost variable by minimizing the number of incidents that occur, reducing the time that lanes are closed or blocked by traffic incidents or work zones, or providing extra capacity dur- ing events that close lanes. Hours of rain or snowfall cannot be affected by design treatments, but their impacts on lane capacity can be affected by treatments such as snow fences and anti-icing treatments. Some treatments also affect the d/c ratio. Once the impacts on these variables are determined for a given treatment, the delay reduction and improvement in reliability can be mea- sured by analyzing the difference between the two TTI curves. The degree to which treatments affect the lane hours lost or d/c ratio input variables is highly dependent on site-spe- cific characteristics, as well as implementation and policy decisions. For example, a jurisdiction that provides easily accessible, well-signed crash investigation sites and enforces a policy that all crashes must be moved to one of them if pos- sible will see a greater impact on the lane hours lost variable than an agency that implements only a few sites that are hid- den from view of the public and that law enforcement rarely uses. Therefore, it is only possible to estimate the potential impact of a design treatment when information is known about the likelihood and frequency with which it will be used. Relationship Between Nonrecurrent Congestion and Safety This research explored the relationship between congestion and safety—specifically the relationship between level of service (LOS) and crash frequency—and developed a math- ematical model to quantify the increase in crash frequency at all severity levels as LOS worsens. Crash frequency is lowest at LOS B and into LOS C, but then begins increasing through LOS D, E, and F. This relationship indicates that if LOS can be improved (by implementing design treatments that decrease congestion) in the range from LOS C to LOS F, then crash frequency will fall. Therefore, treatments that reduce conges- tion also improve safety. Benefit–Cost Analysis of Design Treatments for Nonrecurrent Congestion One of the objectives of this research was to conduct a benefit– cost evaluation for the various design treatments that were evaluated. Because both the benefits and the implementa- tion and maintenance costs of the treatments are dependent on existing site characteristics, specific implementation plans, and accompanying policies for use, a spreadsheet-based analy- sis tool was developed to allow agencies to estimate the poten- tial benefit of a specific implementation of a treatment in a specific location. This tool also allows agencies to compare the benefits of various treatments as they might be implemented in a given location. In the tool, both construction and annual maintenance costs are entirely user defined. Initially, the research team consid- ered providing default values for treatment costs. However, the team received feedback from potential tool users that agencies can easily estimate these costs, and that as construction and materials costs vary greatly from location to location, as well as over time, any defaults provided would likely be inappropriate for many users. To calculate treatment benefits, three main components are considered: delay savings, reliability improvement, and safety improvement. Using the untreated (base condition) TTI curve and the treated (after treatment implementation) TTI curve, a reduction in delay due to treatment implementa- tion can be calculated. This measurement is in terms of vehicle hours, which is converted to dollars by assigning a monetary value to travel time. Many agencies have a default value that is typically used to convert delay hours to economic cost in dollars. A change in reliability can also be determined on the basis of the shift in TTI cumulative curves from untreated to treated. In this project, reliability was quantified as the stan- dard deviation of the travel time distribution, converted into units of hours. There is no consensus in the literature on how this measure should be valued in economic terms, but one common method is to use a reliability ratio. A reliability ratio

74 specific local conditions as much as possible by replacing default values with local information. The impact a given treatment may have will be highly dependent on the spe- cific site characteristics, implementation choices, and poli- cies governing treatment use. • Reliability should be considered not only when planning for the design of new facilities or major reconstruction, but when highways are being reconstructed to add capacity for recurrent congestion concerns. Although reducing recur- rent congestion often reduces nonrecurrent congestion, this positive benefit can be negated when storage areas for vehi- cles involved in crashes or other incidents are removed. In these cases, lane-blocking time for a crash-involved vehi- cle may increase substantially, making the roadway sig- nificantly less reliable despite the additional capacity. The procedures and Analysis Tool developed in this project allow decision makers to weigh the costs of decreased reli- ability against the estimated costs of delay reduction from the capacity increase. Potential research needs related to reliability analysis for nonrecurrent congestion and potential enhancements to the tool include the following: • Developing the capability for the tool to import data from, or export data to, other software packages or databases to promote more efficient data analysis and reduce redundant data entry. • Adding calibration and comparison features for users who have detailed TTI data for existing conditions. • Developing methodologies for considering multiple treat- ments applied simultaneously. • Extending the tool to allow analysis of facilities and corri- dors, not just segments. • Improving the file and scenario management capabilities of the tool to make analysis of multiple sites easier. • Expanding the tool to explicitly compare nondesign (oper- ational or technology) treatments and recurrent congestion enhancements to the base (no treatment) case. • Expanding the tool to explicitly evaluate the operational and safety effects of removing a treatment (e.g., converting a drivable shoulder to a driving lane). • Incorporating into the benefit–cost methodology and Analysis Tool additional treatment benefits, such as fuel and other vehicle operating costs savings and emissions reduction. • Including the capability to specify traffic growth over the design life of the treatment in the benefit–cost methodology and the Analysis Tool. • Refining the safety–congestion relationship with data from additional cities or regions. is the ratio of the value of reliability to the value of time. By defining this ratio as a fixed number, the value assigned to reliability is always a multiple of the value of time. Just as the value of time may vary from one user group to the next (such as freight or peak hour commuters), so too can the reliabil- ity ratio vary from one group to the next. The research team defined the reliability ratio to be 0.8 for all travelers at all times of day in this research, which fell within the range of most values presented in the literature. The results of this research provide a method for incor- porating both the economic savings due to delay reduction and the economic savings due to reliability improvement for a design treatment over its life cycle. Treatments commonly used to address recurrent congestion can be analyzed using the approach developed in this research, which takes into account not only the delay improvements associated with the treat- ment, but the potential improvements to reliability, as well. Taking these benefits into account results in a more accurate valuation of a treatment’s net present benefit and benefit– cost ratio. In addition, agencies considering removing roadway features beneficial to nonrecurrent congestion in order to alle- viate recurrent congestion (such as by converting a shoulder to a driving lane) can use the methods presented in this report and the Analysis Tool to calculate the expected increase in non- recurrent congestion and decrease in reliability that might be expected due to the change and compare these costs with the benefits achieved for recurrent congestion by adding addi- tional capacity. recommendations for Implementation of research results and Future research Needs On the basis of the conclusions of the research effort described above, the L07 research team recommends the following: • Reliability is an important measure of highway operations and has a value beyond delay savings. Design choices should be evaluated for the full range of benefits they may provide. Even design elements aimed at reducing recurrent conges- tion may affect nonrecurrent congestion and reliability. • Improving reliability should be a goal for all highway design projects in the planning phase. Often, designs can be altered slightly to serve as or accommodate nonrecur- rent congestion treatments at a minimal or negligible cost. Considering reliability impacts in the planning process will help maximize treatment benefits while minimizing imple- mentation costs. • Methods and procedures documented in this report and applied in the Analysis Tool should be adjusted to reflect