Lessons Learned: Monitoring Highway Congestion and Reliability Using Archived Traffic Detector Data
2.0 Overview of Current Practice
Development and use of performance measures by the transportation profession has grown substantially in the past five years. At the Federal level, the requirements of the Government Performance Results Act of 1993 drove much of the initial activity at the U.S. DOT. At the same time, a general recognition of the importance of performance measurement and its potential for improving investments and policy actions grew independently at the Federal, State, and local levels. As a result, most transportation agencies have adopted - or are adopting - performance measurement as part of their routine processes. Several reasons can be noted for the use of performance measurement by transportation agencies:
Sound Business Practice. The private sector has adopted performance measures as a way to better serve customers and assess return on investment. Transportation agencies have been moving in this direction also - a strong customer focus is now one of the primary stated goals of most agencies. Performance measurement is the mechanism by which progress toward this goal is assessed: "Know where you are before you decide where to go" is one way to describe this new direction. In addition, service-oriented measures increasingly being used in State and MPO Long Range Plans.
Use of Performance Measures Becoming More Widespread and Accepted as Best Practice within Transportation Agencies. Performance measurement has been well established in pavement and bridge management for several years. Routine reporting of performance has become a powerful tool for showing progress and competing internally for funding. Other areas - such as congestion management, operations, and safety management - are finding that the use of performance measures by pavement and bridge management has "set the bar" for upper management review of programs and funding requests.
Accountability. Transportation agencies are increasingly being asked by legislatures and executives for an assessment of how well problems are being addressed. The public is also more interested in "how are we doing?" kind of questions. Without the hard evidence provided by performance measures, transportation professionals are left with skimpy anecdotes that are insufficient at best and embarrassing at worst.
Performance Measurement is Becoming Easier To Do With New Technologies. Performance measurement is based on the availability of the right kinds of data. As the deployment ITS technologies becomes more widespread, the data to undertake performance measurement becomes available.
Several challenges still exist, however, before performance measurement by transportation agencies realizes its full potential. These challenges may be categorized around the steps involved in establishing a performance measurement process for congestion.
- What metrics (measures) should be used?
- What data are needed to support the metrics and how are they obtained?
- How are performance measures presented internally and to decision-makers and the public?
- How are performance measures used to make better investment decisions and to change policies and daily activities?
There has been a substantial amount of activity addressing what metrics should be used. (Section 2.4 discusses some the advances made in this area.) A significant challenge is in the area of data to support congestion performance measurement. Table 1 describes what challenges are created by these issues.
|Issue||Why Is It a Problem?|
|Availability||Continuous streams of data are not readily available in many regions. The snapshot nature of data availability makes it difficult to analyze conditions during unique events or over time.|
|Coverage||Data is only available for a portion of the transportation network. Therefore, it is difficult to accurately assess the entire impacts of congestion.|
|Quality||Datasets often contain erroneous data or have gaps of missing data. The datasets need significant cleaning before they can be used. Accuracy may be compromised because of little or no calibration/validation.|
|Standards||Data is not consistently collected, analyzed, and stored across different regions, and often times within the same region. Standardization is needed to provide for the meaningful comparison of conditions in different regions.|
Presentation of performance measures to various audiences is still an emerging field, but many agencies are nonetheless engaged in it. As agencies try different presentation methods (for example, "dashboards"), a consensus may be reached on the most effective methods. Graphs and exhibits are possibly the easiest method to convey performance measures. These might be stand-alone illustrations of measures or data that are not presented or addressed in other forms, or pictures that provide additional support to numerical presentations. Exhibits have been used to present a variety of concepts. The following are some that can be developed from the real-time traffic center data:
- Average number of days or peaks of "unacceptably bad" congestion;
- Relationship between traffic volume and travel time;
- Effect of major traffic incidents on volume and travel time in special circumstances;
- Time of day that congestion more frequently occurs;
- Maps of average speed by 30-minute time period; and
- Contour diagrams of travel speed along a roadway.
The final challenge - how to use performance measurement to guide activities - is a particularly daunting task. Put another way, how do we move beyond the simple reporting of performance and trends so that agencies' processes are affected?
Table 2 presents an overview of potential uses of freeway performance measures. As shown, a variety of transportation applications can make use of performance measures and significant overlap exists in the requirements of each application.
|Potential Uses of Performance Measures||Specific Applications||Requirements of Performance Measures|
|Roadway Operations - Real Time Applications||Incident Management|
|Traveler Information/ Diversion||Current and expected traffic states due to traffic flow breakdowns (travel time-based); throughput; diversion volumes|
|Coordinated Freeway-Arterial Contro|
|Special Event Management|
|Roadway Operations - Operational Planning||Incident Management||Detail on detection, verification, on-scene, and response times|
|Traveler Information/ Diversion||Trip- and corridor-based performance
Effects of information content and timeliness
|Coordinated Freeway-Arterial Control||Effects of improved ramp and signal timing plans|
|Evaluations of Operational Improvements||Consistent before/after measurements (travel time performance)|
|Safety Countermeasures||Consistent before/after measurements (crash histories and profiles)|
|Transportation Planning||Travel demand forecasting||Ability to identify and rank deficiencies; inputs to assignment process; volumes and speeds for calibration|
|Demand management||Trip- and corridor-based performance|
|Air quality analysis||Inputs to emission models|
|National Performance||Corridor-based and area-wide performance|
|Truck travel estimation; parking utilization and facility planning; high-occupancy vehicles (HOV), paratransit, and multimodal demand estimation; congestion pricing policy||Trip- and corridor-based and area-wide performance|
|Freight and Intermodal Planning||Trip- and corridor-based performance|
|Transportation Programming||Investment analysis; programmatic funding levels||Corridor-based and area-wide performance|
|Homeland Security||Evacuation Planning||Trip- and corridor-based performance|
|Transportation Research||Traffic flow model development||Highly detailed (time/space) performance measures|
|Emergency Response||Route planning||Trip- and corridor-based performance|
|Freight Carriers||Resource requirements|
Table 3 presents several principles that would help guide the development of congestion performance monitoring programs.
|1||Mobility performance measures must be based on the measurement of travel time.|
|2||Multiple metrics should be used to report congestion performance.|
|3||Traditional HCM-based performance measures (V/C ratio and level of service) should not be ignored but should serve as supplementary, not primary measures of performance in most cases.|
|4||Both vehicle-based and person-based performance measures are useful and should be developed, depending on the application. Person-based measures provide a "mode-neutral" way of comparing alternatives.|
|5||Both mobility (outcome) and efficiency (output) performance measures are required for congestion performance monitoring. Efficiency measures should be chosen so that improvements in their values can be linked to positive changes in mobility measures.|
|6||Customer satisfaction measures should be included with quantitative mobility measures for monitoring congestion "outcomes".|
|7||Three dimensions of congestion should be tracked with congestion-related performance measures: source of congestion, temporal aspects, and spatial detail.|
|8||The measurement of reliability is a key aspect of roadway performance measurement and reliability metrics should be developed and applied. Use of continuous data is the best method for developing reliability metrics, but abbreviated methods should also be explored.|
Foremost among these is the notion that congestion performance measures must be based on the measurement of travel time. Travel times are easily understood by practitioners and the public, and are applicable to both the user and facility perspectives of performance. Figure 1 shows how travel times can be developed from data, analytic methods, or a combination. Clearly, the best methods are based on direct measurement of travel times, either through probe vehicles or the more traditional "floating car" method. However, both of these have drawbacks: probe vehicles currently are not widely deployed and the floating car method suffers from extremely small samples. Further, since many performance measures require traffic volumes as well, additional collection effort is required to develop the full suite of performance measures. Use of ITS roadway equipment addresses these issues, but this equipment does not measure travel time directly; ITS spot speeds must be converted to travel times first. Other indirect methods of travel time estimation use traffic volumes as a basis, either those that are directly measured or developed with travel demand forecasting models.
Figure 1. Travel Time is the Basis for Defining Mobility-Based Performance Measures
Figure 1 also shows how basic travel times can then be converted into a variety performance measures using a few fundamental prices of information about the environment where travel times were measured (roadway characteristics, "ideal" travel speeds, and traffic volumes). This implies that travel time-based performance measures are extremely similar in their basic nature, although some researchers have tended to exaggerate the differences. Travel time-based performance measures can be thought of as two types: 1) absolute measures, and 2) relative measures. Relative measures require comparison to some base conditions, usually "ideal" or "free flow" conditions.
Another principle highly relevant to the use of archived traffic data is the measurement of travel time reliability (usually referred to as just "reliability"). Travel time reliability is growing in significance and use in the transportation profession. The F-SHRP Reliability Research Program1 defined reliability this way:
".. from a practical standpoint, travel time reliability can be defined in terms of how travel times vary over time (e.g., hour-to-hour, day-to-day). This concept of variability can be extended to any other travel time-based metrics such as average speeds and delay. For the purpose of this study, travel time variability and reliability are used interchangeably."
Because reliability is defined as the variability in travel times (or travel time-based metrics), measurement of reliability requires a distribution (i.e., a history) of congestion. The distribution can only be developed by using continuously collected data, such as those generated by ITS. The recently initiated NCHRP Project 7-15 (Cost-Effective Measures and Planning Procedures for Travel Time Variation, Delay, and Reliability) recognizes that complete data may not always be available for this purpose and is developing methods to estimate reliability from limited data.
A number of empirical studies have demonstrated that travelers value not only the time it usually takes to complete a trip, but also the reliability in travel times. For example, many commuters will plan their departure times based on an assumed travel time that is greater than the average to account for this unreliability. Also, because reliability is directly related to the different sources of congestion, its measurement can provide insight into how much of an influence these events have on congestion. This insight can lead to crafting specific strategies for improving roadway performance.
The use of congestion performance measures has been growing in recent years, and ranges from site-specific operations analysis to corridor-level alternative investments analysis and to area-wide planning and public information studies.
In the short term, some combination of surveillance data, planning data, and modeling must be used to support congestion performance measurement. Since surveillance coverage is not complete and data problems will cause gaps in existing coverage, other means must be used to fill in the freeway performance picture. However, the system performance data derived from surveillance data may be significantly different from other estimates or modeling efforts. Combining freeway surveillance data with other data sources should be conducted only where the differences in each type of data are well understood, and where the need for a combination of data is unavoidable.
As indicated in Figure 2, archived data from traffic operations systems currently is one of the most promising sources of data for freeway performance monitoring. This data source typically includes traffic volumes, spot speeds, and estimated or measured travel times. Archived operations data also can include causal information about freeway performance, such as traffic incidents and special events, work zones, or weather. When integrated, these archived data sources can provide significantly better performance information than the transportation profession has ever had. However, the research team recognizes that archived data sources are not perfect and do not represent the "silver bullet" of performance data. Several issues, such as accuracy, consistency, completeness, and coverage, must be addressed before archived data is a reliable source of performance information.
Even after equipment maintenance and quality assurance procedures are established, computer modeling or estimation techniques (left side of Figure 2) may be required for performance information. For example, the historical performance (e.g., what happened 15 minutes ago? one day ago? one month ago?) may indeed be largely and adequately measured by operations sensors on the roadway. How can transportation managers gauge the effects of alternative operating strategies without actually implementing them? How can planners determine the most effective transportation investments at the 20-year future time horizon? In these cases, freeway performance measures will have to be estimated or predicted using computer models or simulation. Similarly, manual data collection (e.g., special studies) may also be required. For example, vehicle occupancy (i.e., number of persons per vehicle) may be important data to have for measuring the performance of multimodal corridors. Vehicle occupancy, however, is not routinely collected by traffic operations as an essential data element; thus, additional data collection may be necessary.
Figure 2. Getting Performance Data
Source: Measuring and Communicating the Effects of Traffic Incident Management Improvements, NCHRP Research Results Digest Number 289, Transportation Research Board, May 2004.
For most cities, the data are collected at point locations using a variety of technologies, including single- and double-inductance loops, radar, passive acoustic, and video image processing. These technologies establish a small and fixed "zone of detection" and the measurements are taken as vehicles pass through this zone. Data collection and processing procedures have been developed individually and the details of the archiving vary from site to site. However, there are several procedures that are common to all sites. In general, the process works as follows for each city:
Data are collected by field sensors and accumulated in roadside controllers. These field measurements are by individual lane of traffic. At 20-second to two-minute intervals, the roadside controllers transmit the data to the traffic management center (TMC).
Some areas perform quality control on original data, but this screening is typically simple and based on minimum and maximum value thresholds. These steps eliminate obviously incorrect data, but do not identify all of the problems.
Areas that use single inductance loops as sensors can only directly measure traffic volume and lane occupancy. In these cases, algorithms are used to estimate speeds for the combinations of volume and occupancy. The algorithms vary from site to site.
Internal processes at the TMC aggregate the data to specified time intervals for archival purposes. These intervals vary from 20 seconds (no aggregation) to 15 minutes. In some cases, the data are also aggregated across all lanes in one direction at a sensor location.
The aggregated data are then stored in text files or databases unique to each TMC. CDs are routinely created at the TMCs to reduce some of the storage burden and to satisfy outside requests for the data.
Three programs currently monitor congestion performance trends nationally using archived traffic detector data. The first of these is the Urban Congestion Report (UCR) effort, which yields a monthly snapshot of roadway congestion in 10 urban areas and three national composite measures. UCR utilizes efficient, automated data collection procedures (colloquially known as "screen scraping" or "web mining") to obtain travel time directly from traveler information web sites and archives them at five-minute intervals on the weekdays when these services are available. Concurrent with the travel time data collection, other UCR acquisition programs obtain web-based data on weather conditions, traffic incidents, and work zone activity. The UCR produces monthly "snapshots" of congestion in the reporting cities. An example UCR snapshot is presented in Section 3 of this report.
The Mobility Monitoring Program (MMP) calculates system performance metrics based on data archived at traffic management centers (TMCs). These data are highly detailed measurements from roadway surveillance equipment installed for operational purposes; data from spot locations (volumes and speeds) are used as well as travel time estimates from probe vehicles (where available). For each participating city, the MMP develops congestion metrics at both the corridor and area levels; 23 cities participated in 2002 and 29 are reporting data for 2003. The concepts, performance measures, and data analysis techniques developed and used in the MMP are being considered for adoption and implementation by several State and local agencies. The MMP produces an annual report which presents a standard set of congestion graphics for each city; Figure 3 is one of the many graphics used.
The Intelligent Transportation Infrastructure Program (ITIP) is an ongoing program designed to enhance regional surveillance and traffic management capabilities in up to 21 metropolitan areas while developing an ability to measure operating performance and expanding traveler information through a public/private partnership involving the FHWA, participating State and local transportation agencies, and Mobility Technologies. Under this partnership, Mobility Technologies is responsible for deploying and maintaining traffic surveillance devices, and integrating data from these devices with existing traffic data to provide a source of consolidated real-time and archived data for the participating metropolitan areas. Deployment has been completed in Philadelphia, Pittsburgh, Chicago, and Providence, and is under way in Boston, Tampa, San Diego, Washington, DC, Phoenix, Los Angeles and San Francisco. Negotiations are currently active in 10 additional cities.
Figure 3. Congestion Trends on Minneapolis-St. Paul Freeways, 2000-2002
Source: Mobility Monitoring Program, http://mobility.tamu.edu/mmp/.
State and Local Activities
In addition to archiving ITS-generated data, many States and MPOs have embraced the concept of performance measurement. This trend is developed a substantial amount of inertia and can no longer be seen as theoretical - transportation agencies are imbedding performance measurement into their day-to-day activities. Examples include:
Arizona — Both the Arizona Department of Transportation (ADOT) and the Maricopa Association of Governments (MAG) are supporting performance monitoring programs. ADOT has folded the implementation of a scaled down CMS (based on the MMP's reliability index) into the Arizona State transportation plan (MoveAZ Plan).
Minnesota — The Minnesota Department of Transportation (MnDOT) is studying adoption of the primary performance measures in MMP, namely the travel time index as a mobility measure and the buffer time index as a reliability measure.
Oregon — The Oregon Department of Transportation (ODOT) Traffic Management Section is studying several of the measures used in the MMP reports. The goal is to use the archived data in combination with other, more widely available data to construct a method to evaluate operations on the entire roadway network.
California — The California Department of Transportation (Caltrans), with the technical support of the University of California-Berkeley, is in the process of developing and integrating a statewide data archive and performance monitoring system called PeMS (Freeway Performance Measurement System).
Virginia — The University of Virginia and the Virginia Transportation Research Council (VTRC), the research arm of Virginia DOT, have been developing a statewide data archive managers for the Virginia DOT. Performance measurement is one of the many functions provided by the archive. Additionally, they have conducted several feasibility studies of the performance measures used in the MMP reports and are considering adoption of some of the mobility and reliability measures.
Michigan — The Southeast Michigan Council of Governments (SEMCOG), the MPO for the Detroit region, is exploring the use of archived ITS data for performance monitoring.
Washington — The Washington State DOT has a research and implementation effort to develop a set of mobility performance measures. The University of Washington, the primary analyst of archived data for WsDOT, is conducting the research and produces one of the premier annual congestion performance reports in the country. Figure 4 is one of the exemplary graphics from this report.
Figure 4. Interstate 405 South Traffic Profile: General Purposes Lanes, 1999 Weekday Average
Source: Ishimaru, J.M., Nee, J. and Hallenbeck, M.E., Central Puget Sound Freeway Network Usage and Performance, 1999 Update, Volume 1, Washington State Transportation Center, Seattle, Washington, September 2000.
1 Future Strategic Highway Research Program (F-SHRP) reports, including the report on reliability, can be accessed at http://www4.trb.org/trb/newshrp.nsf/web/progress_reports?OpenDocument.