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.
24 C h a p t e r 4 The SHRP 2 Project L08 conceptual analysis framework (L08 framework) for predicting travel time reliability is designed for operations analysis and planning applications in which the analyst must estimate current reliability on the basis of a limited amount of data, or predict the impacts of demand changes, operational improvements, and design concepts on reliability. The L08 framework requires fewer analytical resources than the SHRP 2 Project L04 framework, which involves simulation models and demand models. And the L08 framework provides more detailed information on the sources of unreliability and the reliability effects of specific operational improvements than the SHRP 2 Project L03 regression equations. Thus, the L08 framework is designed to work with the existing HCM2010 methodologies for evaluat- ing freeway and urban street facilities. Overview of the L08 Conceptual analysis Framework The L08 framework employs scenarios to diagnose the causes of unreliable performance and to predict the impacts of spe- cific operational improvements on reliability, as shown in Figure 4.1. Each scenario is a specific combination of demand, weather, incidents, special events, and work zones. Each sce- nario presents a challenge to the operation of the facility. HCM2010 methodologies are used to predict how facility performance responds to each of the challenges. Special capacity, saturation flow, and free-flow speed adjustment fac- tors have been developed by the L08 team for use with the HCM2010 methods to account for the effects of weather and incidents on capacities and speeds. The HCM2010 methods themselves have also been selectively augmented to facilitate their application to reliability analysis. The results of the numerous challenges are then summed up and weighted according to their probability of occurrence to obtain statis- tics on the facilityâs reliability. Freeway and Urban Streets Methodologies While the overall reliability analysis framework is identical for freeway and urban street facilities, the specific implementa- tions of the L08 framework vary between freeway and urban street facilities applications to suit the specific characteristics of each facility analysis methodology in the HCM2010. The HCM2010 freeway analysis method deals with the study period and study section of the facility as a whole, breaking down the calculations to specific analysis periods and segments within the facility. The performance of all traffic movements sharing a mainline freeway segment is evaluated. Only one direction of flow on the freeway is evaluated at a time. The HCM2010 urban streets method also deals with the study section of the facility as a whole (with disaggregation for segments and intersections), but it focuses on a single 15-min analysis period within the peak hour. The method can be used to build up analysis periods into study period results, but that is not its common use. The method evaluates performance only for the through movement on the arterial (taking into account the effects of other movements on the through movement performance) and evaluates both direc- tions of travel at a time. These differences in the HCM2010 freeway and urban streets methods have resulted in two implementations of the L08 framework that take two slightly different approaches to estimating reliability. The L08 freeway method requires the analyst to provide a full study period of 15-min demands and then apply monthly and daily factors to obtain demand varia- tion over the reliability reporting period. The L08 urban streets method requires hourly demands, breaks those down into 15-min periods, and then extrapolates the hourly demand to the study period. Like the freeway method, the urban streets method also applies monthly and daily factors to obtain demand variation over the reliability reporting period. Both methods leave out some of the true day-to-day variabil- ity of demand by using average monthly and daily factors to generate their varying demands. Development of Freeway and Urban Streets Methodologies
25 A more significant difference in the two implementations of the L08 framework is the use of stochasticity within the methods. The L08 freeway method is primarily a deterministic approach that applies probabilities at the end of the process when tallying the reliability statistics. The L08 urban streets method is primarily a stochastic approach that applies stochas- tic methods (random numbers) within the scenario generation process to generate one set out of many possible sets of scenar- ios for the street. Knowledge of the probabilities of the scenarios at the end is not required because that is built into the steps used to create the scenarios in the urban streets method. Thus, the freeway analysis can identify specific scenarios with relative ease, while the number of potential scenarios for urban streets is very large (e.g., turning percentages, signalization, location of incidents). Therefore, the analysis for freeways is based on the travel time estimation for specific scenarios, while the analysis for urban streets identifies analysis scenarios stochastically. Treatment of Stochasticity in the L08 Framework The L08 research team decided to use the two different treat- ments of stochasticity in the L08 freeway and urban street implementations because each approach has advantages and disadvantages and neither is clearly superior to the other in all circumstances. Both implementations of the L08 framework are, in concept, interchangeable; later research by others may reveal which approach is preferred for practical applications. The primarily deterministic approach implemented in the L08 freeway method assuredly generates all significant prob- ability events, giving the same results each time it is run. How- ever, to keep the computations tractable, it sacrifices explicit consideration of extremely rare events, as well as more fre- quent events that are judged a priori to be unlikely to signifi- cantly affect demand or capacity. Extremely rare, high-impact events are unlikely to effect the overall annual distribution of travel times due to their rarity. They become important when one is concerned about travel times greater than the 95th percentile. The primarily stochastic approach implemented within the urban streets method does not, a priori, eliminate extremely rare events from consideration. However, given their low probability, they are unlikely to turn up in any given analysis. This strength of the primarily stochastic approach assures the analyst that all possibilities are considered, but the assurance comes at the cost of having to run the analysis several times and average the results. The need for replications ensures that a truly representative range of scenarios is considered. For alternatives analysis, detecting the effects of minor changes to the inputs (such as a modest demand increase or a control change) becomes more difficult because part of the computed difference in travel times may result from stochastic variation. To obtain some confidence in this difference, the analyst must apply a statistical hypothesis test to determine if the observed difference is significant and not the result primarily of chance. An example of the differences in the two approaches is the generation of incidents for scenarios. The primarily deter- ministic freeway method considers only three locations and two possible start times for incidents to keep the number of scenarios that have to be modeled to some value significantly below infinity. The primarily stochastic urban streets method considers all possible locations and all possible start times for incidents; but since the method is applied only a determinis- tic number of times within each scenario, it yields only one of many possible outcomes each time it is applied. The full urban streets analysis must be run several times to obtain the comprehensive power of the stochastic approach to consider incidents in all locations at all times. Introduction to the Freeway Facilities Methodology This section provides a high-level description of how travel time reliability can be incorporated in the Freeway Facilities chapter of the HCM2010. The HCM freeway facilities method enables the user to analyze the effect of recurring congestion over an extended facility (about 10 to 15 miles long) and study period (up to 6 hours in duration). This time-space domain allows for the analysis of queue formation and dissipation at bottlenecks, and produces performance measures at the Figure 4.1. Conceptual analysis framework.
26 freeway segment, analysis period (15 min), and overall facility levels. Details of the current methodology can be found in Chapter 10 of Volume 2 of the HCM2010 (TRB 2010a) and in Chapter 25 of Volume 4 (TRB 2010b). The computational engine for the methodology, FREEVAL, is also available for download in Volume 4. The objectives of the L08 project are twofold. The first objec- tive is to incorporate nonrecurring congestion effects into the HCM2010 procedure. The second objective is to expand the analysis horizon from a single study period (typically an a.m. or p.m. peak period) to an extended time horizon of several weeks or months, up to a 1-year reliability reporting period. Together, these objectives lead to a method that allows the analyst to assess the variability in the quality of service that a facility provides to its users. This expanded period, the reliabil- ity reporting period, can be thought of as a set of days, each one having its own set of demands and capacities that affect the facilityâs travel time. This study focused on weather, incidents, work zones, and special events on the supply side, and on vol- ume variability by time of day, day of week, and month of year on the demand side. Components of the Freeway Facilities Methodology At its highest level of representation, the freeway facilities methodology has three primary components: a data deposi- tory, a scenario generator, and a core computational proce- dure, which is an adapted and significantly revised version of FREEVAL for reliability, or FREEVAL-RL. These components are illustrated in Figure 4.2. The largest shaded oval and dotted line represent the cur- rent implementation of the freeway facilities method, with study period data specific to the facility being studied entered directly into FREEVAL for analysis of (predominantly) recur- ring congestion effects. The connection to reliability is enabled by the addition of a scenario generator. Each component and its interaction with the other two are explained in some details in the next sections. Data Depository The data depository can be viewed as the virtual space in which all the pertinent data elements needed to execute the method- ology reside. Some data are (indeed, some must be) specific to the freeway facility being studied. These data include, at a mini- mum, all segment geometrics, free-flow speeds, lane patterns, and segment types. Demands can be directly measured for a sample of days from field sensors on the facility, or estimated from projections of annual average daily traffic (AADT) and time-based factors. At a minimum, data must be available to execute one seed file in FREEVAL-RL, much like it is needed to run the current HCM2010 procedure. Complexities arise when the analyst incorporates sources of nonrecurring congestion effects. Several attributes are required to assess the impact of each source on the facility reliability, including the variations in source type, the probability of its occurrence during the reliability reporting period, and its potential impacts on segment free-flow speed, traffic demand, and segment capacity. An inventory of these attributes is shown in Table 4.1. All data elements are subsequently entered into the scenario generator to start the creation of detailed scenarios to run in FREEVAL-RL and thus estimate travel times. In summary, the reliability methodology relies on the avail- ability of both facility-specific data elements and default values when data are not available, nonexistent (future analysis), or too expensive to collect. This gives rise to the terms data-rich and data-poor analyses. However, in most cases, the analysis is a hybrid one, relying on both facility-specific and default data to generate and evaluate scenarios. Thus, each reliability prob- lem could be classified as X% data rich and 1 - X% data poor, depending on the data availability. Figure 4.2. Freeway facilities methodology components, including measures of effectiveness (MOEs).
27 Scenario Generator The purpose of the freeway scenario generator (FSG) is to enumerate a sufficiently complete set of operational scenarios that a freeway facility may experience during the reliability reporting period (RRP), along with their associated probabili- ties. Each scenario represents a single study period that is fully characterized in terms of demand and capacity profiles in time and space. The FSG is flexible, can operate with minimal input (i.e., uses defaults) when data are not available, and accepts facility-specific data when available. All entries are expressed as demand- and capacity-related parameters. Demand Variability Demand variations can be entered by time of day, day of week, and month of year (a maximum of 84 demand sce- narios for a given study period). The default used in this study is 12 demand scenarios encompassing three weekday types and four seasons. As stated earlier, segment flow rates can be entered directly into a seed file, or estimated on the basis of segment and ramp AADTs in combination with hourly fac- tors, to generate the study period demand. Daily and seasonal demand factors are applied to populate all other scenarios in the reliability reporting period. The only other place that demand patterns may be altered in the scenario generator is in the cases of work zones or special events. Demand in those cases is very much facility- and event-specific and therefore must be directly entered by the user as an input. Capacity Variability Much of the focused effort in the FSG is on estimating the probability and impact of nonrecurring congestion, including weather, incidents, work zones, and special events. Data ele- ments are explained in the following sections. Weather Frequency and eFFects The FSG generates the fraction of RRP time that the facility experiences a particular weather event, along with the impact of the event on capacity and free-flow speed (FFS). The proj- ect team extracted durations, for each hour of each month, of 11 HCM-defined weather event types for 101 metropolitan areas in the United States over the most recent 10-year period available. Depending on the application, an analyst can either use data from a specific year or estimate future-year weather probabilities on the basis of long-term historical averages. A screenshot of a weather probability table produced by the FSG is shown in Figure 4.3 along with mean duration and default adjustment factors for capacity and FFS. Each cell gives the fraction of time a weather event is present in the specified month. When entered in FREEVAL-RL, a weather event is assumed to occur either at the start of the study period or in the middle of the study period, with equal probability, thus generating a maximum of 11 (events) Ã 2 (start times) or 22 weather sce- narios. All the segments on the facility are affected equally by the weather event. When using the FSG to estimate weather probabilities, the analyst simply needs to select the metropoli- tan area closest to the study facility from a list of 101 national defaults. IncIdent Frequency and eFFects Similar to weather, two pieces of information are needed for modeling incidents: (1) the monthly probability of certain incident severities, and (2) the impact of each severity level on capacity. The first piece requires a significant effort to extract Table 4.1. Inventory of Nonrecurring Congestion Sources and Attributes Nonrecurring Congestion Source Elements of Variability Source for Estimating Probability of Occurrence Nonrecurring Event Duration Impact on Segment Free-Flow Speed, Demand, and Capacity Weather Nonsevere rain (low, medium, high), snow (light, medium, heavy), visibility (low, mini- mum), cold, fog Historical averages by hour and by montha; year- specific data From the same source or national defaults Extracted from the literature, including HCM2010 Incidents Shoulder closure; one, two, and three-plus lane closures Incident logs or rate pre- diction from crash rates Incident logs or national defaults Extracted from the literature, including HCM2010 Work zones Shoulder closure; one, two, and three-plus lane clo- sures; crossovers Detailed annual work zone schedules Detailed traffic control plans for each work zone Demands must be entered by analyst; capacity effects from literature, including HCM2010 Special events Shoulder closure; one, two, and three-plus lane clo- sures; crossovers; lane additions; lane reversals Detailed traffic control plans for each event Detailed traffic control plans for each event All free-flow speeds, demands, and capacities must be fully specified by the analyst a For this study, 10-year weather data for 101 metropolitan areas were extracted from Weather Underground (www.wunderground.com).
28 the number and duration of each incident from annual inci- dent logs in data-rich environments. Furthermore, the research teamâs experience with using incident logs revealed significant underreporting of certain incident types. Therefore, a recom- mended alternative approach is to estimate the facility incident rate from its predicted crash rate, and then use the Poisson pro- cess to estimate the likelihood of specific incident severities. Predictive models are available from the Highway Economic Requirements System (HERS) and the Highway Safety Manual (HSM) (AASHTO 2010). Capacity adjustments resulting from incidents are taken directly from the HSM2010. A sample inci- dent probability table generated by the FSG is illustrated in Figure 4.4. Additional incident details must also be generated before running incident scenarios in FREEVAL-RL. These include the following: â¢ Incident start time. Similar to weather, FREEVAL-RL assumes an incident start time either at the start or in the middle of the study period (SP). â¢ Incident location. Three possible locations at the first, mid- dle, and last segment on the facility are included in the FSG. â¢ Incident duration. Based on national averages for incident duration distribution by severity, three representative dura- tions at the 25th, 50th, and 75th percentile values of the distribution are included in the FSG. Month Med Rain Heavy Rain Light Snow LM Snow MH Snow Heavy Snow Severe Cold Low Vis Very Low Vis Min Vis Normal Weather January 1.970% 0.000% 5.911% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 92.1182% February 2.717% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 2.174% 0.000% 0.000% 95.1087% March 0.505% 0.000% 1.010% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 98.4848% April 0.000% 0.543% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 99.4565% May 1.951% 1.951% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 96.0976% June 0.505% 0.505% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 98.9899% July 0.500% 0.500% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 99.0000% August 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 100.0000% September 4.255% 0.532% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 95.2128% October 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 100.0000% November 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 100.0000% December 0.000% 0.000% 7.805% 0.488% 0.000% 0.000% 0.000% 0.000% 0.000% 0.000% 91.7073% Average Duration for Weather Type(min): 42.9 31.0 134.3 46.6 25.8 5.5 15.0 57.2 15.0 136 Default Capacity Adjustment Factor: 92.76% 85.87% 95.71% 91.34% 88.96% 77.57% 91.55% 90.33% 88.33% 89.51% 100.00% Default FFS Adjustment Factor: 93.00% 92.00% 87.00% 86.00% 84.00% 83.00% 93.00% 94.00% 92.00% 92.00% 100.00% Weather Categories (based on HCM2010 Chapter 10: Freeway Facilities) Figure 4.3. FSG-generated weather event probabilities, duration, and impact. Note: LM îµ light to medium; MH îµ medium to heavy; Vis îµ visibility. Figure 4.4. FSG-generated incident probability matrix. Month No Incident Shoulder Closure One Lane Closure Two Lane Closure Three Lane Closure Four Lane Closure January 82.80% 11.96% 3.60% 0.91% 0.73% 0.00% February 81.57% 12.74% 3.91% 0.99% 0.80% 0.00% March 81.61% 12.68% 3.91% 1.00% 0.80% 0.00% April 81.06% 13.06% 4.03% 1.03% 0.82% 0.00% May 78.50% 14.79% 4.60% 1.18% 0.94% 0.00% June 81.06% 13.05% 4.03% 1.03% 0.82% 0.00% July 80.20% 13.64% 4.22% 1.08% 0.86% 0.00% August 80.49% 13.45% 4.16% 1.06% 0.85% 0.00% September 81.87% 12.52% 3.85% 0.98% 0.78% 0.00% October 78.66% 14.67% 4.57% 1.17% 0.93% 0.00% November 82.51% 12.10% 3.70% 0.94% 0.75% 0.00% December 84.88% 10.52% 3.16% 0.80% 0.64% 0.00% Probability of Diï¬erent Incident TypesInsert Facility Specific
29 Thus, a maximum of 2 (start times) Ã 3 (locations) Ã 3 (durations) Ã 5 (severities) = 90 incident scenarios + 1 no- incident scenario = 91 total scenarios. When using the FSG to estimate incident probabilities, the analyst at a minimum needs to provide a facility-specific incident or crash rate. Work Zone and Special event Frequency and eFFectS Only significant scheduled work zones and special events are considered in the scenario generator. The user must provide the work zone schedule and characteristics (e.g., shoulder work, single-lane closure). In addition, if a significant change in demand is anticipated during the work zone or special event, the appropriate demand values must be entered. Capac- ity effects of work zones are taken primarily from the existing literature, including the HCM2010. Capacity effects of spe- cial events must be entered by the analyst, as those are highly facility- and event-specific. Generating Scenarios The FSG assumes that nonrecurring congestion events are independent of each other. Therefore, the probability of an event combination is equal to the product of their two proba- bilities. The total number of scenarios that will emerge cannot be predicted a priori because only a subset of combinations of demand and capacity variations resulting from the nonrecur- ring events will occur. However, an upper bound on the num- ber of scenarios can be estimated. Ignoring for the moment the presence of work zones and special events, there are 12 default demand scenarios, 22 weather scenarios, and 91 incident sce- narios. If all event combinations have a nonzero probability, then there are approximately 24,000 possible scenarios. In real- ity, many of the combinations do not exist (e.g., snow in the summerâin most places), and the actual number of generated scenarios is a fraction of the maximum. The generator com- putes the fractional number of study periods to which each scenario is applicable and divides that number by the reliability reporting period to estimate each scenarioâs probability. Fig- ure 4.5 shows an example allocation of scenarios, including their descriptions and probabilities. For this real-world facility, 2,508 scenarios are generated, slightly more than 10% of the theoretical maximum. Core Computation Engine FREEVAL-RL Interface with the FSG The FSG will create as many input files for execution in FREEVAL-RL as there are scenarios to analyze. Variations between scenarios result from three types of adjustment factors: â¢ Demand variability by time of day, day of week, and month/ season of year is expressed in terms of demand adjust- ment factors (DAFs) applied to the original demands in the seed file. â¢ Capacity variability resulting from weather, incidents, work zones, and special events is expressed in terms of capacity adjustment factors (CAFs) applied to seed file values; CAFs are applied to specific segments in the cases of incidents or work zones, and facilitywide in the case of weather. â¢ Free-flow speed variability resulting from weather condi- tions is expressed in terms of free-flow speed adjustment factors (SAFs) applied facilitywide for the duration of the weather event. Figure 4.6 illustrates a case in which the capacity adjust- ments for a weather event lasting for 30 min (i.e., two 15-min analysis periods) occurs in combination with an incident on Figure 4.5. FSG-generated detailed scenarios and their probabilities.
30 the last segment of the facility, also lasting 30 min. The seg- ment CAF reflects the combination of the two events. All adjustment factors by segment and analysis period are sent to the computational engine for processing. Core Procedural Enhancements Several enhancements to the original HCM2010 freeway facilities procedure were implemented in the course of the L08 study. These enhancements included (1) developing a method to incorporate FFS and capacity adjustments con- currently, (2) specifying a queue discharge rate less than the uninterrupted-flow capacity, and (3) reporting additional reliability-based outputs in FREEVAL-RL. Each enhance- ment is discussed in the following sections. Concurrent SAF and CAF Implementation To remain in general compliance with the HCM2010, the research team revised the original speed prediction model in HCM2010 Equation 25-1. For basic freeway segments, the new model simply replaces the base free-flow speed with the adjusted free-flow speed, using the appropriate SAF for the prevailing weather conditions. No free-flow speed effects were considered for incidents, as supporting data for this effect were not available in the literature. The revised Equa- tion 25-1 to predict speed (S) for any adjusted flow rate (vp) assuming a base free-flow speed FFS, base capacity C, and their adjustments SAF and CAF is presented as Equation 4.1. FFS SAF 1 (4.1) ln FFS SAF 1 CAF 45 CAFS e C v C p ( )= â + âï£®ï£°ï£¯ ï£¹ ï£»ï£º ( )( )â + â â As a rule, the estimated speed from Equation 4.1 can never exceed the speed at the adjusted capacity (i.e., the speed at a density of 45 passenger cars per mile per lane). This con- straint is always satisfied, guaranteeing that the predicted speed will always be at least 1 mph above the estimated speed at capacity. For ramp and weaving segments, the adjustments to capac- ity and speed are done independently, since speed estimation for those segment types is independent of capacity. In other words, the CAF is applied to reducing the segment capacity, while the SAF is applied to reducing the FFS and, by exten- sion, the estimated segment speed. The method multiplies c and FFS by CAF and SAF, respectively, throughout the HCM Chapters 12 and 13 methodologies. The implementation of SAF and CAF is detailed in Chapter 6 of this report. Incorporating Queue Discharge Flow To more realistically model queue propagation and dissipa- tion on congested freeway facilities, the core procedure now enables the analyst to specify a capacity âlossâ resulting from freeway breakdown. This feature was not provided in the HCM2010. However, Hu et al. (2012) found that it has a significant effect on the duration and severity of the con- gested region. That study found that, according to an exten- sive literature search, capacity loss averaged 7% during breakdown. In FREEVAL-ML (managed lanes), this value can be entered by the user, although it is restricted to an upper bound of 10% and a minimum of 0% (the current HCM2010 approach). Detailed information on incorporat- ing queue discharge flow in FREEVAL-RL is presented in Chapter 6. Figure 4.6. Illustration of CAF application for weather and incidents in FREEVAL input.
31 Enhancements to the FREEVAL-RL Outputs Some of the scenarios run in FREEVAL-RL can clearly generate severe congestion. In some cases, the congestion might be more than the model can handle (e.g., multiple interacting bottle- necks). In addition to providing flags for such occurrences, new performance measures were added to monitor these effects: â¢ Total number of vehicles denied entry into the facility when the first segment is fully queued; and â¢ Denied entry vehiclesâ queue length upstream of segment 1 in each time period. In addition, the team incorporated new reliability measures to enable comparisons across facilities. For example, (1) the TTI is now calculated and reported for each segment in each analysis period and (2) facility TTI is calculated for each anal- ysis period. Note that each 15-min analysis period contributes one data point to the overall facility travel time distribution. These measures are part of the standard FREEVAL-RL report, which characterizes the full TTI distribution, along with descriptions of the scenarios that generated the distribution. Introduction to the Urban Streets Methodology This section describes the development of a methodology for predicting travel time reliability for urban street facilities. Applying the methodology produces the facility travel time distribution for a specified reliability reporting period. This reliability methodology uses the HCM2010 urban streets methodology to compute facility travel time and other per- formance measures for each analysis period of interest within the reliability reporting period. Goals The research team established several goals to guide the devel- opment of a framework for the urban streets reliability meth- odology. The goals are described in the following list: â¢ The reliability methodology should use the HCM2010 urban streets methodology to estimate average travel time and other performance measures for a specified analysis period. â¢ The methodology should quantify the effect of the follow- ing sources of nonrecurring congestion: weather events, traffic demand variation, traffic incident occurrence, work zone presence, and special event presence. â¢ The reliability methodology should minimize the amount of new required input data, beyond that already needed to evaluate an urban street facility for one analysis period (using the HCM2010 urban streets methodology). â¢ The methodology should provide a default value for each calibration factor used in its component procedures. Stages Applying the reliability methodology involves the following three stages, which are implemented in the sequence listed. Each stage is summarized in the subsections below. â¢ Scenario generation; â¢ Facility evaluation; and â¢ Performance summary. Scenario Generation In the scenario generation stage, each analysis period in the reliability reporting period is identified. Then, the weather event, traffic demand level, traffic incident occurrence, work zone presence, and special event occurrence are defined for each analysis period. The effect of these factors on segment running speed, intersection saturation flow rate, or signal tim- ing is quantified. The traffic demand volume, speed, saturation flow rate, and signal timing established for each analysis period are assumed to be unique relative to the other analysis periods. Thus, each analysis period is considered to be one scenario for subsequent evaluation. This assumption recognizes that, in the urban street environment, analysis periods rarely have the same unique combination of demand volume, capacity, and traffic control characteristics for all segments and inter- sections that make up the facility. The likelihood of unique analysis periods increases when the analysis periods are sequential in a common study period and volumes are suffi- ciently high that residual queues from one analysis period become initial queues for the subsequent analysis period. Facility Evaluation In the facility evaluation stage, each analysis period is eval- uated using the computational engine that automates the HCM2010 urban streets methodology. This engine is referred to in this section as the urban streets engine. It is used to esti- mate the expected value of various performance measures for each intersection and segment, and the facility as a whole, for each analysis period. For any given performance measure, the estimate represents an average for the analysis period and is referred to as an analysis period average (APA). Performance Summary In the performance summary stage, the collective set of anal- ysis period results is used to describe a distribution of traffic
32 performance for the reliability reporting period. Facility travel time is the performance measure used to define reliability. However, other performance measures (e.g., intersection delay) can be examined in terms of their variation during the reliability reporting period. Regardless of the performance measure considered, the resulting distribution describes the variation in APA for the reliability reporting period. It does not describe the variation in performance experienced by individual travelers. As a result, some of the variability in per- formance experienced by individual travelers is not accounted for in this analysis. Work Flow The sequence of calculations in the reliability methodology is shown in Figure 4.7. The process is designed around the urban streets engine. It begins with one or more engine input data files. Each file is used to describe the traffic demand, geometry, and signal timing conditions for each intersection and segment on the subject urban street facility for one analysis period. Most reliability evaluations involve two or more input files. One input file describes base conditions (i.e., when work zones and special events are not present). It is called the base input file. Additional input files are used, as needed, to describe con- ditions when a specific work zone is present or when a special event occurs. These are called alternative input files. As a first step in the reliability evaluation, the analyst uses the urban streets engine to generate each of the desired input files. The analyst also identifies the range of dates to which each of the alternative input files is applicable. For example, if an analyst is interested in the travel time during weekday periods from 4:00 to 6:00 p.m. for the current year and the analysis period is 15-min long, then the base input file is used to describe conditions present for one analysis period (say, 4:00 to 4:15 p.m.) when no work zones or special events are present. The demand volumes represent a specified date (pro- vided by the analyst) and can be adjusted in the reliability methodology to estimate volumes for the other dates and times that occur during the reliability reporting period. If a work zone exists during a given month, then a second input file is used to describe average conditions for the analy- sis period during that month. As noted previously, the analyst develops this input file using the urban streets engine. The data in the input file reflect the analystâs knowledge of the lane closures and signal timing changes that result from the work zoneâs presence. The data also reflect the effect of work zone presence on volume, speed, and capacity. The means by which these effects are incorporated in the file is discussed in the Work Zones and Special Events subsection. Input Data Once the input files have been created, the data needed to use the reliability methodology are identified. These data are described in Table 4.2. To identify the typical weather conditions for the subject facility, analysts use the nearest city found in the National Climatic Data Centerâs (NCDC) publication on climatic data. Input Data Weather Event Procedure Traffic Demand Variation Procedure Traffic Incident Procedure Scenario File Generation Procedure Scenario Evaluation Scenario File Update (residual queue -> initial queue) Performance Measure Collection Performance Measure Summary Statistics End Scenario Generation Facility Evaluation Performance Summary Urban Streets Engine Input File Next Analysis Period Figure 4.7. Reliability methodology for urban street facilities.
33 The Center periodically publishes summaries from weather stations in each of 284 U.S. cities and territories (NCDC 2011a). The document contains 17 statistics related to tem- perature, wind, cloudiness, humidity, and precipitation. Each statistic is quantified by month of year and based on 10 or more years of data. Of interest to reliability evaluation are the following weather statistics from this document: â¢ Total normal precipitation; â¢ Total normal snowfall; â¢ Number of days with precipitation of 0.01 in. or more; and â¢ Normal daily mean temperature. The NCDC also provides storm event data for several thousand locations throughout the United States (including the aforementioned 284 locations). These data describe the average number of storms, average precipitation depth per storm, average storm duration, and average precipitation rate (i.e., intensity). Each statistic is quantified by month of year. Of interest to reliability evaluation is the average precipita- tion rate in the Rainfall Frequency Atlas (NCDC 2011b). The functional class of the subject facility is used to deter- mine the appropriate month-of-year and hour-of-day traffic volume adjustment factors. Hallenbeck et al. (1997) exam- ined continuous count station data from 19 states and found that these factors varied by functional class. They also noted some difference in factor values when comparing the coastal states with the Great Plains and Rocky Mountains. Their report was used as the basis for the default month-of-year, hour-of-day, and day-of-week adjustment factors described in Appendix H. The latter set of factors was not found to be sensitive to functional class, but the factors were sensitive to area type (i.e., urban or rural). The starting hour of the count is used to determine the hour- of-day adjustment factor. This factor, with the month-of-year and day-of-week factors, is then used to convert the volumes in the base input file into average-day-of-year volumes. A similar adjustment is made to the volumes in the alternative input files. During the scenario file generation, these averages are used to estimate the volume for specific hours and days of the year. The crash frequency data are used to estimate the frequency of non-crash-related incidents. The procedure for computing Table 4.2. Input Data Category Variable Description General Nearest city One of 284 U.S. cities and territories whose climatic conditions are summarized periodically by the National Climatic Data Center (www.ncdc.noaa.gov) Functional class Functional class of subject urban street facility Input file Date of traffic count Basis of traffic volumes in base file. Can be either 1. Traffic counts measured in the field (enter the date of the count) or 2. Planning estimates of volume during the average day of week and month of year (do not enter a date) Starting hour of the count Hour of the day that the traffic counts were measured or, if based on planning estimates, hour of the day to which the estimates apply Basis of traffic counts in the alternative input files Basis of traffic volumes in alternative file. Can be either 1. Adjusted traffic counts from base file (enter the date of the count) or 2. Planning estimates of volume when the work zone or special event is pres- ent (do not enter a date) Time period Analysis period Duration of analysis period (0.25 h or 1.0 h) Study period Starting hour of study period and its duration in hours Reliability reporting period Starting date of reliability reporting period and its duration in days Alternative file operating period Starting date of work zone or special event and its duration in days Days of week considered Days of week considered in reliability reporting period Crash Segment crash frequency The segment-related crash frequency for each segment, including all severities. The value entered represents the long-run average number of crashes each year when work zones and special events are not present. It is adjusted appropriately if the reliability reporting period is not 1 year in duration. Intersection crash frequency Same as for segments but based on intersection-related crashes Crash frequency adjustment factors This factor is multiplied by the segment or intersection crash frequency. The product represents the long-run crash frequency if the work zone or special event were in operation for 1 year.
34 this estimate is described in a subsequent section. For urban streets evaluation, crashes are categorized as â¢ Crashes related to the segment; or â¢ Crashes related to the intersection. The two categories are mutually exclusive. A technique for determining whether a crash is a segment- or intersection- related crash is described in Appendix A to Part C of the HSM (AASHTO 2010). The crash frequency that is input repre- sents an estimate of the expected crash frequency for base traffic demand volume, geometry, and signal timing condi- tions. The estimate should include all severity levels, includ- ing property-damage-only (PDO) crashes. It is provided in units of crashes per year, regardless of the duration of the reliability reporting period. The procedure uses the expected crash frequency to estimate the number of crashes that occur during the reliability reporting period. The expected crash frequency can be computed by using the predictive method in Chapter 12 of the HSM. If this method cannot be used, then a 3-year crash history for the subject facil- ity can be used to estimate the expected crash frequency. Crashes that occur when work zones and special events are present should be removed from the crash data. In this situation, the expected crash frequency is computed as the count of crashes during base conditions divided by the time period (in years) when base conditions are present. The crash frequency adjustment factor is used to estimate the expected crash frequency when a work zone or special event is present. This factor is multiplied by the expected crash frequency for base conditions. The product represents the expected crash frequency if the work zone or special event was in operation for 1 year. The factor value should include consideration of the effect of the work zone or special event on traffic volume (i.e., vol- ume may be reduced because of diversion) and on crash risk (i.e., the geometry and signal operation changes for the work zone or special event may increase the potential for a crash). For example, if a work zone is envisioned to increase crash risk by 100% (i.e., crash risk is doubled) and to decrease traf- fic volume by 50% (i.e., volume is halved), then the crash fre- quency adjustment factor is 1.0 (= 2.0 Ã 0.5). The analystâs experience with similar types of work zones or special events should be used to determine the appropriate adjustment fac- tor value for the subject facility. Scenario Generation The scenario generation stage consists of four sequential pro- cedures which are described in more detail in Chapter 5. Each procedure processes the set of analysis periods in chronologic order. The first procedure predicts weather event date, time, type (i.e., rain or snow), and duration. The second procedure iden- tifies the appropriate traffic volume adjustment factors for each date and time during the reliability reporting period. These factors are used during the scenario file generation pro- cedure to estimate the volume associated with each analysis period. The third procedure predicts incident event date, time, and duration. It also determines incident event type (i.e., crash or noncrash), severity level, and location on the facility. It uses weather event and demand variation information from the two previous procedures in the incident prediction process. The fourth procedure uses the results from the preceding three procedures to develop one urban streets engine input file for each analysis period in the reliability reporting period. As discussed previously, each analysis period is considered to be one scenario. Date and time represent a common basis for link- ing the events and conditions related to all four procedures. Each input file created in this procedure includes the appropri- ate adjustments to segment running speed and intersection saturation flow rate associated with the weather or incident events that occur during the corresponding analysis period. Similarly, the traffic demand volumes in each file are adjusted for monthly, weekly, and hourly variations. varIance control Weather events; traffic demand; and traffic incident occur- rence, type, and location have both systematic and random elements. To the extent practical, the reliability methodology accounts for the systematic variation component in its pre- dictive models. Specifically, it recognizes changes in weather and traffic demand depending on time during the year, month, and day. It also recognizes the influence of geographic location on weather and the influence of weather and traffic demand on incident occurrence. Models of the systematic influences are included in the meth- odology. They are used to predict average weather, demand, and incident conditions during each analysis period. However, the use of averages to describe weather events and incident occur- rence for such short time periods is counter to the objectives of reliability evaluation. The random element of weather events, demand variation, and traffic incident occurrence introduces a high degree of variability in the collective set of analysis periods that make up the reliability reporting period. Thus, it is impor- tant to replicate these random elements in any reliability evalu- ation. Monte Carlo methods are used for this purpose in the urban streets reliability method. A random number seed is used with the Monte Carlo meth- ods in the reliability methodology so that the sequence of ran- dom events can be reproduced. In fact, a unique seed number is separately established for weather events, demand variation, and incident occurrence. For a given set of three seed num- bers, a unique combination of weather events, demand levels,
35 and incidents is estimated for each analysis period in the reli- ability reporting period. One, two, or three of the seed numbers can be changed to generate a different set of conditions, if desired. For example, if the seed number for weather events is changed, then a new series of weather events is created and, to the extent that weather influences incident occurrence, a new series of inci- dents is created. Similarly, the seed number for demand varia- tion can be used to control whether a new series of demand levels is created. The seed number for incidents can be used to control whether a new series of incidents is created. When evaluating alternatives, analysts will likely use one set of seed numbers as a variance reduction technique. In this application, the same seed numbers are used for all evalua- tions. With this approach, the results from an evaluation of one alterative can be compared with those from an evaluation of the baseline condition. Any observed difference in the results can be attributed to the changes associated with the alternative (i.e., they do not result from random changes in weather or incident events among the evaluations). replIcatIons A complete exploration of reliability would likely entail the use of multiple, separate evaluations of the same reliability reporting period with each evaluation using a separate set of random number seeds. This approach may be particularly useful when the facility has infrequent severe weather events or incidents. With this approach, the evaluation is replicated multiple times and the performance measures from each rep- lication are averaged to produce a more reliable estimate of their long-run value. Facility Evaluation The facility evaluation stage consists of two tasks that are repeated in sequence for each analysis period. The analysis periods are evaluated in chronologic order. For the first task, the input file associated with an analysis period is submitted to the urban streets engine for evalua- tion. Then, the predicted performance measures for the sub- ject analy sis period are saved to an output file with a unique file name. During the second task, the performance measures are extracted from the output file and used to revise the input file associated with the next analysis period. Specifically, the input file for the next analysis period is read, modified, and saved before returning to the first task. The modification entails setting the initial queue input value for the next analysis period equal to the residual queue output from the current analysis period. Sampling Technique Typical combinations of reliability reporting period, analy- sis period, weather event occurrence, and incident event occurrence often produce a large number of scenarios. The collective evaluation of these scenarios could take an hour or more when the methodology is automated in software. This length of time may be considered too long for some reliability applications, in which case a sampling approach is available. A sampling technique can be used to minimize the total evaluation time. The analyst needs to input the scenario evalu- ation interval. The interval has units of days. The analyst can choose to evaluate every scenario for every day (i.e., input â1â). Alternatively, the analyst can chose to evaluate every scenario for every other day (i.e., input â2â). More generally, the analyst can input any integer number for the evaluation interval. The evaluation interval is checked to ensure that all days in the reliability reporting period are equally sampled. The check examines the pattern produced by the input âdays of week consideredâ D and the evaluation interval I. An interval factor F is computed as F = I - int[I/D] Ã D. If 5 or 7 days of the week are considered, then values of I that yield F > 0 provide the desired representative sample. If 2, 3, 4, or 6 days of the week are considered, then values of I that yield F = 1 or F = D - 1 provide the desired sample. Performance Summary The performance summary stage consists of two sequential tasks. The first task reads the output file for each analysis period and collects the desired performance measure. At the start of this task, the analyst identifies the specific direction of travel and the performance measure of interest, selecting from the following list: â¢ Travel time; â¢ Travel speed; â¢ Stop rate; â¢ Running time; and â¢ Through delay. The analyst also indicates whether the performance mea- sure of interest represents the entire facility or a specific seg- ment. The first three measures in the list are available for facility evaluation. All five measures are available for segment evaluation. At the conclusion of this task, the collected data represent observations of the performance measure for the each analysis period occurring during the reliability report- ing period (or a sampled subset). During the second task, the selected performance measure data are summarized using the following statistics: â¢ Average; â¢ Standard deviation; â¢ Skewness; â¢ Median;
36 â¢ 10th, 80th, 85th, and 95th percentiles; and â¢ Number of observations. In addition, the average base free-flow speed is always reported. It can be used with one or more of the distribution statistics to compute various reliability measures, such as the TTI. Work Zones and Special Events Work zones and special events influence traffic demand levels and travel patterns. To minimize the impact of work zones and special events on traffic operation, agencies responsible for traffic accommodation in the vicinity of the work zone or special event often reallocate some traffic lanes or alter the signal operation to increase the capacity of specific traffic movements. These characteristics make each work zone and special event unique, and their effect on facility performance equally unique. Different work zones and special events can occur during the reliability reporting period. The reliability methodology incorporates work zone and special event influences in the evaluation results. However, the analyst must describe each work zone and special event using an alternative input file. Each file describes the traffic demand, geometry, and signal timing conditions when the work zone is present or the special event is under way. A start date and duration are associated with each file. Work zone presence can have a significant effect on traffic demand levels. The extent of the effect depends partly on the availability of alternate routes, the number of days the work zone has been in operation, and the volume-to-capacity ratio of the segment or intersection approach with the work zone. Lee and Noyce (2007) evaluated motorist response to several freeway work zones in Wisconsin. They concluded that diver- sion resulted in a volume reduction of 40% to 50%. This con- clusion was based on their comparison of the observed queue forming upstream of the work zone with that predicted using volumes measured during normal-day operations. When using the reliability methodology, the analyst must provide an estimate of traffic demand volumes during the work zone or special event. These estimates should reflect the effect of diversion, and they can be based on field measure- ments, judgment, or area-wide traffic planning models. They are recorded by the analyst in the corresponding alternative input file. The analyst must have information about lane closures, alter- native lane assignments, and special signal timing that are pres- ent during the work zone or special event. This information can be based on agency policy or on experience with previous work zones or events. The available lanes, lane assignments, and sig- nal timing are recorded by the analyst in the corresponding alternative input file. A review of the literature indicates that work zone presence can affect intersection saturation flow rate. An adjustment fac- tor for this effect is described in Appendix I. It can be used with the saturation flow rate prediction procedure in Chapter 18 of HCM2010 to estimate the saturation flow rate when a work zone is present. The analyst then enters the adjusted saturation flow rate in the appropriate alternative input file. These adjust- ments are not made as part of the reliability methodology.