Broadening Integrated Corridor Management Stakeholders (2020)

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.

Performance Measure Guidance A P P E N D I X D D-1 NCHRP Web-Only Document 97: Guide to Effective Freeway Performance Measurement12 provides transportation engineers and planners assistance in developing and maintaining a comprehensive freeway performance-monitoring program. Multiple aspects of freeway performance were considered, including congestion, mobility, safety, operational efficiency, ride quality, environmental, and customer satisfaction. This report’s guidance can lend itself to ICM projects where freeway congestion is often the driver for these types of project implementations. Adapted from NCHRP Web-Only Document 97, TABLE D.1 provides a wide variety of potential performance measures that can be used to evaluate how well an ICM project meets its goals and objectives. The appropriate geographic and temporal scales are listed for each metric. In the table, “Quality of Service” denotes a more intuitive term for the outcome category of measures, while “Activity-Based” is more apt for the output category of measures. Quality of service (outcome) and activity-based (output) performance measures must be linked together and tied into the mission of the transportation agency. 12 NCHRP Web-Only Document 97: Guide to Effective Freeway Performance Measurement: Final Report and Guidebook, Document 97, 2006.

TABLE D.1. Recommended core freeway performance measures. Performance Metric Definition Units Geographic Scale Time Scale Average (Typical) Congestion Conditions (Quality of Service) Travel Time The average time consumed by vehicles traversing a fixed distance of freeway Minutes Specific points on a section or a representative trip only; separately for GP and HOV lanes Peak hour, a.m./p.m. peak- periods, midday, daily Travel Time Index The ratio of the actual travel rate to the ideal travel rate1 None; minimum value = 1.000 Section and areawide as a minimum; separately for GP and HOV lanes Peak hour, a.m./p.m. peak- periods, midday, daily Total Delay, Vehicles The excess travel time used on a trip, facility, or freeway segment beyond what would occur under ideal conditions2 Vehicle-hours Section and areawide as a minimum; separately for GP and HOV lanes Peak hour, a.m./p.m. peak- periods, midday, daily Total Delay, Persons The excess travel time used on a trip, facility, or freeway segment beyond what would occur under ideal conditions3 Person-hours Section and areawide as a minimum; separately for GP and HOV lanes Peak hour, a.m./p.m. peak- periods, midday, daily Delay per Vehicle Total freeway delay divided by the number of vehicles using the freeway Hours (vehicle- hours per vehicle) Section and areawide Peak hour, a.m./p.m. peak- periods; daily Spatial Extent of Congestion No. 1 Percent of Freeway VMT with Average Section Speeds <50 mph4 Percent Section and areawide Peak hour, a.m./p.m. peak- periods Spatial Extent of Congestion No. 2 Percent of Freeway VMT with Average Section Speeds <30 mph Percent Section and areawide Peak hour, a.m./p.m. peak- periods Temporal Extent of Congestion No. 1 Percent of Day with Average Freeway Section Speeds <50 mph Percent Section and areawide Daily Temporal Extent of Congestion No. 2 Percent of Day with Average Freeway Section Speeds <30 mph Percent Section and areawide Daily Density Number of vehicles occupying a length of freeway Vehicles per lane- mile Section Peak hour/periods for weekday/weekend Reliability (Quality of Service) Buffer Index The difference between the 95th percentile travel time and the average travel time, normalized by the average travel time Percent Section and areawide Peak hour, a.m./p.m. peak- periods, midday, daily Planning Time Index The 95th Percentile Travel Time Index None; minimum value = 1.000 Section and areawide Peak hour, a.m./p.m. peak- periods, midday, daily

TABLE D.1. Recommended core freeway performance measures (continued). Performance Metric Definition Units Geographic Scale Time Scale Capacity Bottlenecks (Activity-Based) Geometric Deficiencies Related to Traffic Flow (Potential Bottlenecks) Count of potential bottleneck locations by type5 Number Section and areawide N/A Major Traffic- Influencing Bottlenecks Count of locations that are the primary cause of traffic flow breakdown on a highway section, by type Number Section and areawide N/A Throughput (Quality of Service) Throughput— Vehicle Number of vehicles traversing a freeway Vehicles per unit time Section and areawide Peak hour, a.m./p.m. peak- periods, midday, daily Throughout— Persons Number of persons traversing a freeway Persons per unit time Section and areawide Peak hour, a.m./p.m. peak- periods, midday, daily Vehicle-Miles of Travel The product of the number of vehicles traveling over a length of freeway, times the length of the freeway Vehicle-miles Section and areawide Peak hour, a.m./p.m. peak- periods, midday, daily Truck Vehicle-Miles of Travel The product of the number of trucks traveling over a length of freeway,6 times the length of the freeway Vehicle-miles Section and areawide Peak hour, a.m./p.m. peak- periods, midday, daily Lost Highway Productivity Lost capacity due to flow breakdown—the difference between measured volumes on a freeway segment under congested flow versus the maximum capacity for that segment Vehicles per hour Section and areawide Peak hour, a.m./p.m. peak- periods, midday, daily Customer Satisfaction (Quality of Service) Worst Aspect of Freeway Congestion (Defined by question) 1) happens every work day; 2) incidents that are not cleared in time; and 3) encountering work zones Areawide or statewide Annually; tied to survey frequency Satisfaction with Time to Make Long- Distance Trips Using Freeways (Defined by question) 1) very satisfied; 2) somewhat satisfied; 3) neutral; 4) somewhat dissatisfied; 5) very dissatisfied; and 6) do not know Areawide or statewide Annually; tied to survey frequency

TABLE D.1. Recommended core freeway performance measures (continued). Performance Metric Definition Units Geographic Scale Time Scale Safety (Quality of Service) Total Crashes Freeway crashes as defined by the State, i.e., those for which a police accident report form is generated Number All safety measures computed areawide; section level may be computed if multiple years are used All safety measures computed annually Fatal Crashes Freeway crashes as defined by the State, i.e., those for which a police accident report form is generated, where at least one fatality occurred Number All safety measures computed areawide; section level may be computed if multiple years are used All safety measures computed annually Overall Crash Rate Total freeway crashes divided by freeway VMT for the time period considered Number per 100 million vehicle-miles All safety measures computed areawide; section level may be computed if multiple years are used All safety measures computed annually Fatality Crash Rate Total freeway fatal crashes divided by freeway VMT for the time period considered Number per 100 million vehicle-miles All safety measures computed areawide; section level may be computed if multiple years are used All safety measures computed annually Secondary Crashes A police-reported crash that occurs in the presence of an earlier crash7 Number All safety measures computed areawide; section level may be computed if multiple years are used All safety measures computed annually Ride Quality (Quality of Service) Present Serviceability Rating (PSR) The general indicator of ride quality on pavement surfaces8 (Internal scale) Section and areawide Annually International Roughness Index (IRI) Cumulative deviation from a smooth surface Inches per mile Section and areawide Annually Environment (Quality of Service) Nitrous Oxides (NOx) Emission Rate Modeled NOx attributable to freeways divided by freeway VMT Number Annually Volatile Organic Compound (VOC) Emission Rate Modeled VOC attributable to freeways divided by freeway VMT Number Annually Carbon Monoxide (CO) Emission Rate Modeled CO attributable to freeways divided by freeway VMT Number Annually Fuel Consumption per VMT Modeled gallons of fuel consumed on a freeway divided by freeway VMT Number Annually Section and areawide Section and areawide Section and areawide Section and areawide

TABLE D.1. Recommended core freeway performance measures (continued). Performance Metric Definition Units Geographic Scale Time Scale Incident Characteristics (Activity-Based) No. of Incidents by Type and Extent of Blockage Self-explanatory Type: 1) crash; 2) vehicle breakdown; 3) spill; and 4) other. Blockage: Actual number of lanes blocked; separate code for shoulder blockage Section and areawide a.m./p.m. peak- periods, daily Incident Duration9 The time elapsed from the notification of an incident to when the last responder has left the incident scene Minutes (median) Section and areawide a.m./p.m. peak- periods, daily Blockage Duration The time elapsed from the notification of an incident to when all evidence of the incident (including responders’ vehicles) has been removed from the travel lanes Minutes (median) Section and areawide a.m./p.m. peak- periods, daily Lane-Hours Loss Due to Incidents The number of whole or partial freeway lanes blocked by the incident and its responders, multiplied by the number of hours the lanes are blocked Lane-hours Section and areawide a.m./p.m. peak- periods, daily Work Zones (Activity-Based) No. of Work Zones by Type of Activity The underlying reason why the work zone was initiated: 1) resurfacing only; 2) RRR; 3) lane addition w/o interchanges; 4) lane additions w/interchanges; 5) minor cross-section; 6) grade flattening; 7) curve flattening; 8) bridge deck; 9) bridge superstructure; 10) bridge replacement; and 11) sign related Number Section and areawide Daily Lane-Hours Lost Due to Work Zones The number of whole or partial freeway lanes blocked by the work zone, multiplied by the number of hours the lanes are blocked Lane-hours Section and areawide a.m./p.m. peak- periods; midday; night; daily Average Work Zone Duration by Type of Activity The elapsed time that work zone activities are in effect Hours Section and areawide Daily Lane-Miles Lost Due to Work Zones The number of whole or partial freeway lanes blocked by the work zone, multiplied by the length of the work zone Lane-miles Section and areawide a.m./p.m. peak- periods, daily

TABLE D.1. Recommended core freeway performance measures (continued). Performance Metric Definition Units Geographic Scale Time Scale Weather (Activity-Based) Extent of highways affected by snow or ice Highway centerline mileage under the influence of uncleared snow or ice multiplied by the length of time of the influence Centerline-Mile-Hours Section and areawide Daily Extent of highways affected by rain Highway centerline mileage under the influence of rain multiplied by the length of time of the influence Centerline-Mile-Hours Section and areawide Daily Extent of highways affected by fog Highway centerline mileage under the influence of fog multiplied by the length of time of the influence Centerline-Mile-Hours Section and areawide Daily Operational Efficiency (Activity-Based) Percent Freeway Directional Miles with traffic sensors, surveillance cameras, DMS, service patrol coverage One measure for each type of equipment deployed in an area Percentage (xxx.x%) Section and areawide Annually Percent of Equipment (DMS, surveillance cameras, traffic sensors, ramp meters, RWIS) in “Good” or Better Condition One measure for each type of equipment deployed in an area Percentage (xxx.x%) Section and areawide Annually Percent of total device-days out-of-service (by type of device) One measure for each type of equipment deployed in an area Percentage (xxx.x%) Section and areawide Annually Service patrol assists Self-explanatory Number Section and areawide Annually Source: NCHRP, Guide to Effective Freeway Performance Measurement: Final Report and Guidebook, Document 97, 2006. 1 Travel rate is the inverse of speed, measured in minutes per mile. The “ideal travel rate” is the rate that occurs at the free-flow speed of a facility, or a fixed value set for all facilities that is meant to indicate ideal conditions or “unconstrained” (see text for discussion of the ideal/unconstrained/free-flow speed). 2 See text above for definition of “ideal.” 3 See text above for definition of “ideal.” 4 A freeway “section” is the length of freeway that represents a relatively homogenous trip by users. Logical break points are major interchanges (especially freeway-to-freeway) and destinations (e.g., Central Business District). The term “section” is sometimes used to describe this, but it usually implies additional parallel freeways and/or transit routes. 5 Bottleneck types are: Types A-C weaving areas (see HCM, Section 7.0); left exits; freeway-to-freeway merge areas; surface street on-ramp merge areas; acceleration lanes at merge areas <300 feet; lane drops; lane width drops >= 1 foot; directional miles with left shoulders <6 feet; directional miles with right shoulders <6 feet; steep grades; substandard horizontal curves. The shoulder categories are included because of the ability of more than 6-foot shoulders to shelter vehicles during traffic incidents. 6 Trucks are defined as vehicles with at least six tires, i.e., FHWA Classes 5 through 13 plus any larger vehicles as defined by a state. 7 See report text for discussion. 8 See: http://www.fhwa.dot.gov/policy/1999cpr/ch_03/cpg03_2.htm. 9 Since in many cases the actual time the incident occurred is unknown, the notification time is used to indicate the official “start” of the incident. On most urban freeways, with use of cell phones by the public, the time between when the incident occurs and when it is first reported is very small.

Performance Measure Guidance D-7 Once the necessary data are collected and key performance metrics have been calculated, it is recommended that the results be tested for statistical significance. This helps to determine whether the changes observed (e.g., increase or decrease in travel times, speeds, throughput) are “real,” or whether they are simply due to chance. TABLE D.2 provides several examples of relevant statistical tests to use for a variety of performance metrics, while TABLE D.3 provides additional detail on the test themselves. TABLE D.2. Performance measures and statistical tests. Performance Metric Relevant Tests for Evaluating Statistical Significance Vehicle-hours of delay Mann-Whitney U Test or Wald Test Corridor travel times Mann-Whitney U Test (Wilcoxon Rank Sum Test) Carpool rates Chi-Square Test Vehicle throughput Student’s Two-Sample (Unpaired) t test Percent of time that speed > 45 mph1 Fisher’s Test HOV violation rates Student’s Two-Sample (Unpaired) t test 1. The federal performance standard for Express Lanes requires the facility to meet a minimum average operating speed of 45 mph for 90 percent of the time. These are several guidelines regarding distributions of traffic-related data: Traffic crash counts follow Poisson distributions.13 Queue lengths and delays vary depending on operating conditions (e.g., the reason for the queue) and traffic arrivals (e.g., platoons, Poisson). Travel times on arterials may follow an overly complex distribution.14 though there is limited evidence for Lognormal or Gamma distributions.15 16 Gamma distributions can be summed to result in a new Gamma distribution; Lognormal distributions do not have this property. Bottleneck discharge, when considered as an average for the full time the bottleneck is active from day to day, follows a normal distribution.17 13 Wasserman. Page 119. 14 Bayen. Probability distributions of travel times on arterial networks: a traffic flow and horizontal queuing theory approach. http://bayen.eecs.berkeley.edu/sites/default/files/conferences/TRB2012_stat_traffic.pdf 15 Wu. Standard Deviation of Travel Time in a Freeway Network – a Mathematical Quantifying Tool for Reliability Analysis. https://www.researchgate.net/profile/Ning_Wu7/publication/264690319_Standard_Deviation_of_Travel_T ime_in_a_Freeway_Network__a_Mathematical_Quantifying_Tool_for_Reliability_Analysis/links/53eb62a 70cf2593ba7087b33.pdf. 16 Li, Chai, Tang. Empirical Study of Travel Time Estimation and Reliability. http://www.hindawi.com/journals/mpe/2013/504579/. 17 Zhang and Levinson. Some Properties of Flows at Freeway Bottlenecks. https://wiki.cecs.pdx.edu/pub/Main/SirishaKothuri/Zhang.pdf. Travel times in general are challenging to characterize by a parameterized distribution,

TABLE D.3. Statistical tests for comparison. Statistical Test Description Assumptions Applications Comments Wilcoxon Rank- Sum (or Mann- Whitney U) Test Used as an alternative to the two-sample t-test without the assumption of normally distributed data. Data can be ordered. Data distribution of each sample has same shape and variance.2 Tests whether the medians (not the means) of two independent samples are equal. The Median Test is another, less powerful non-parametric test alternative to this. Wald Test3 Used to compare two estimates for a distribution scalar parameter (e.g., mean, median). The estimate for the scalar parameter is asymptotically normal (e.g., mean, median). Can be used to test whether two samples have the same mean, without making assumptions about the sample distributions. Can be used to test whether two samples have the same median, without making assumptions about the sample distributions. Requires estimating standard error using bootstrap. The Likelihood Ratio Test is another similar method with greater flexibility. Chi-Square Test4 Measures how well the observed distribution of data fits with the distribution that is expected if the variables are independent. Datasets are large enough. Data is not correlated. Can be used to test how likely it is that an observed distribution is due to chance. It is called a “goodness of fit” statistic. Fisher’s test may be more appropriate when sample sizes are small.

TABLE D.3. Statistical tests for comparison (continued). Statistical Test Description Assumptions Applications Comments Student’s Two- Sample Test5 Used to determine if two population means are equal. Data may be paired or unpaired. The variances of two samples may be assumed to be equal or unequal. Testing if a process or treatment is equivalent to another process or treatment. Testing if a new process or treatment is superior to a current process or treatment. Testing if a new process or treatment is superior to a current process or treatment by at least some predetermined threshold amount. When the paired data does not have normally distributed differences, use the Wilcoxon Signed- Rank Test. Fisher’s Test6,7 Used to test whether one categorical variable is independent from another categorical variable. Assumes row and column totals in the contingency table are known in advance. This is generally not the case for the outcome totals (e.g., the total number of travelers that will arrive on time is not known in advance). When this assumption is violated, the test becomes more conservative. Testing a treatment (i.e., the first categorical variable is with/without treatment) and evaluating a categorical outcome (e.g., traveler arrives on time or not). There can be more than two categorical outcomes (e.g., different service rating classifications). When the sample size is larger than 1,000, the Pearson’s Chi-Squared test is recommended instead. If using a Pearson’s Chi-Squared Test, ensure no expected value is less than 5. For paired testing, use McNemar’s test. 1 http://www.mathworks.com/help/stats/ranksum.html#bti4z5t. 2 http://support.minitab.com/en-us/minitab/17/topic-library/basic-statistics-and-graphs/hypothesis-tests/nonparametrics-tests/which-test-should-i- use/#choosing-between-the-two-sample-mann-whitney-test-and-the-pooled-t-test. 3 Wasserman. 4 http://www.ling.upenn.edu/~clight/chisquared.htm. 5 http://www.itl.nist.gov/div898/handbook/eda/section3/eda353.htm. 6 http://www.biostathandbook.com/fishers.html. 7 Weisstein, Eric W. "Fisher's Exact Test." From MathWorld--A Wolfram Web Resource. http://mathworld.wolfram.com/FishersExactTest.html.