Technical Report Documentation Page 1. Report No. 2 ... · Document is available to the public through the National Technical Information Service, Springfield, Virginia 22161 19

Technical Report Documentation Page 1. Report No. FHWA-HOP-04-011

2. Government Accession No.

3. Recipient's Catalog No.

5. Report Date March 2004

4. Title and Subtitle Monitoring Urban Roadways in 2002: Using Archived Operations Data for Reliability and Mobility Measurement

6. Performing Organization Code

7. Author(s) Tim Lomax, Shawn Turner and Richard Margiotta

8. Performing Organization Report No.

10. Work Unit No. (TRAIS)

9. Performing Organization Name and Address Texas Transportation Institute Cambridge Systematics, Inc. 3135 TAMU 1265 Kensington Drive College Station, TX 77843-3135 Knoxville, TN 37922

11. Contract or Grant No.

13. Type of Report and Period Covered Interim: January 2003 – December 2003

12. Sponsoring Agency Name and Address US Department of Transportation Federal Highway Administration Office of Operations Washington, DC 20590 14. Sponsoring Agency Code

HOTM-1 15. Supplementary Notes Project COTR: Mr. Dale Thompson Research Project Title: Expanded Mobility Monitoring Program

16. Abstract This report summarizes the processes and products from a study of archived traffic data generated from the transportation operations centers in 23 cities in 2002. Archived traffic data were analyzed for data quality and completeness and a database prepared for the sections of instrumented freeway. The database includes available traffic volume and speed records for every freeway section for each 5-minute period of the year. Other inventory information such as section length, number of lanes, detector type and other characteristics were also gathered. Several mobility and reliability measures were produced – all are based on travel time concepts. They examine the level of travel delay and mobility, as well as the variation in travel conditions from day-to-day through the year and between freeways in urban areas. A variety of illustration techniques and statistical analysis methods are included in the main report. An appendix was created for each city to show ways to provide detailed summaries of average and unusual conditions for corridors and the system (available at http://mobility.tamu.edu/mmp). Current practice emphasizes the real-time uses of the data for operations purposes but, in many cities, does not include extensive use of the data for other purposes. If the products from real-time databases assist operators and planners, there will be more interest in making the effort to create and improve the data collection process and the databases. 17. Key Word performance monitoring, archived data, operations data, databases, transportation planning, mobility, reliability

18. Distribution Statement Document is available to the public through the National Technical Information Service, Springfield, Virginia 22161

19. Security Classif. (of this report) Unclassified

20. Security Classif. (of this page) Unclassified

21. No. of Pages 72

22. Price

Form DOT F 1700.7 (8-72) Reproduction of completed page authorized

. . . . . . . . . . . . . . . Final Report MONITORING URBAN ROADWAYS IN 2002: USING ARCHIVED OPERATIONS DATA FOR RELIABILITY AND MOBILITY MEASUREMENT Prepared for U.S. Department of Transportation Federal Highway Administration Office of Operations Washington, DC 20590 Prepared by Texas Transportation Institute 3135 TAMU College Station, Texas 77843-3135

Under contract to Battelle 505 King Avenue Columbus, Ohio 43201

March 2004

v

CKNOWLEDGMENTS

In addition to the primary sponsorship of the Federal Highway Administration, the authors acknowledge the following agencies that participated in the third year of the Mobility Monitoring Program: Albany: New York State DOT Atlanta: Georgia DOT Austin: Texas DOT Charlotte: North Carolina DOT Cincinnati: ARTIMIS and Kentucky Transportation Cabinet Detroit: Michigan DOT Hampton Roads: Virginia DOT, Virginia Transportation Research Council and University of

Virginia Houston: Texas DOT and Houston office of Texas Transportation Institute Los Angeles: Caltrans and University of California at Berkeley (PeMS) Louisville: Kentucky Transportation Cabinet Milwaukee: Wisconsin DOT Minneapolis-St. Paul: Minnesota DOT and University of Minnesota-Duluth Northern Virginia: Virginia DOT, Virginia Transportation Research Council and University of

Virginia Orlando: Florida DOT Philadelphia: Mobility Technologies, Inc. Phoenix: Arizona DOT Pittsburgh: Mobility Technologies, Inc Portland: Oregon DOT Sacramento: Caltrans and University of California at Berkeley (PeMS) Salt Lake City: Utah DOT San Antonio: Texas DOT San Diego: Caltrans Seattle: Washington State DOT and Washington State Transportation Center (TRAC)

Quality Assurance Statement

The Federal Highway Administration provides high-quality information to serve Government, industry, and the public in a manner that promotes public understanding. Standards and policies are used to ensure and maximize the quality, objectivity, utility, and integrity of its information. FHWA periodically reviews quality issues and adjusts its programs and processes to ensure continuous quality improvement.

A

vi

vii

ABLE OF CONTENTS

Page Acknowledgments............................................................................................................................v Summary – The 2002 Mobility Monitoring Program.................................................................... xi Purpose of the Report............................................................................................................... xi Project Findings ...................................................................................................................... xii Performance Measure Development....................................................................................... xii Beyond the First Three Years – Summary of Future Issues .................................................. xiii Chapter 1 – Introduction ..................................................................................................................1 What’s New for This Year? .......................................................................................................2 Chapter 2 – The Issues.....................................................................................................................3 Mobility and Reliability.............................................................................................................3 The Mobility Monitoring Program Framework.........................................................................3 Why Collect and Analyze Such Large Datasets? ......................................................................3 System Measures and User Experience Measures.....................................................................5 Data Elements and Analytical Processes Used I the Mobility Monitoring Program.................5 Relating Archived Data and Estimates ......................................................................................5 Chapter 3 – The Data .......................................................................................................................7 Participating Cities and Their Archived Data............................................................................7 Overview of Data Processing...................................................................................................14 Data Quality Checking.............................................................................................................16 Mobility and Reliability Measure Calculations .......................................................................22 Chapter 4 – The Measures .............................................................................................................25 Mobility Measures ...................................................................................................................25 Reliability Measures ................................................................................................................26 Selection of Time Period .........................................................................................................28 Chapter 5 – The Results: What Do the Measures Show . . . ? ......................................................31 Performance Measure Observations ........................................................................................32 Trends in Mobility and Reliability...........................................................................................34 Daily and Monthly Patterns .....................................................................................................35 Corridor Observations..............................................................................................................47 Congested Travel .....................................................................................................................49 MMP, Trips and Census Information ......................................................................................50

T

viii

ABLE OF CONTENTS, CONTINUED

Page Chapter 6 – The Future: Additional Opportunities .......................................................................51 Validation of Travel Times from Multiple Sources.................................................................51 Expansion of the Program to Include Signalized Arterials......................................................51 More Sophisticated Quality Control Procedures .....................................................................52 Analyses Tailored to Local Areas............................................................................................52 Congestion Causes ...................................................................................................................53 Continue to Experiment with Measures...................................................................................54 Encourage the Development of Standardized Procedures for Data Archiving........................54 Long-Term Structure of the Mobility Monitoring Program ....................................................55 References......................................................................................................................................57

T

ix

IST OF EXHIBITS

Exhibit Page 3-1 Participating Cities and Agencies for 2002 Archived Data ..............................................8 3-2 Preferred Archived Data Formats for Mobility Monitoring Program ..............................9 3-3 Summary of Archived Data Characteristics for 2002.....................................................12 3-4 Overview of Data Processing within Mobility Monitoring Program .............................15 3-5 2002 Data Validity Checks in Mobility Monitoring Program........................................18 3-6 Summary of Archived Data Passing Validity Checks ....................................................20 3-7 Summary of Archived Data Completeness at Different Processing Steps .....................21 3-8 Estimating Route Travel Times and VMT from Spot Speeds and Volumes..................23 5-1 Summary of Instrumented Section Coverage in 2002 ....................................................32 5-2 2002 Average Peak Period Mobility and Reliability Statistic Summary........................33 5-3 Relationship between Average Peak-Period Mobility and Reliability in Each

Study City .......................................................................................................................34 5-4 Miles of Freeway Data, 2000 to 2002.............................................................................36 5-5 Average Peak Period Travel Time Index and Buffer Index, 2000 to 2002 ....................37 5-6 2002 Daily Mobility Summary .......................................................................................39 5-7 2002 Daily Reliability Summary ....................................................................................40 5-8 Mobility and Reliability Measures by Time of an Average Day, Albany

(Example)........................................................................................................................41 5-9 Mobility and Reliability during Daily Time Periods (Example) ....................................42 5-10 2002 Delay Summary by Time of Day...........................................................................43 5-11 Delay by Time of Day (Example)...................................................................................44 5-12 2002 Delay Summary by Day of Week ..........................................................................45 5-13 Delay by Day of Week (Example)..................................................................................46 5-14 Mobility and Reliability Measures by Day of the Year (Example) ................................47 5-15 Delay by Roadway (Example) ........................................................................................48 5-16 Traffic Speed Contour Map ............................................................................................49 5-17 Average Weekday Speed Variation ................................................................................49

L

x

xi

UMMARY – THE 2002 MOBILITY MONITORING PROGRAM

The Mobility Monitoring Program collected and analyzed data from 23 cities for 2002. The 2002 Program is the third year of an on-going effort (1, 2) to examine the usefulness of archived speed and volume data and develop techniques to improve the information available to transportation professionals. The 2002 effort broadens the study of this important but relatively scarce information source. Data collected directly from the transportation system—traffic volumes and speeds—for every minute of the year at hundreds of locations in each city will be indispensable for planning, design, management, operations and evaluation within the next decade. Right now, however, it is only being collected and archived (i.e., stored) in a limited number of cities. Some of the difficulty relates to the cost of equipment, communication details and limited staff resources. It must also be said, however, that some of the problem is that transportation professionals and decision makers have not “mined” this dataset or asked the questions that would encourage, fund or mandate that data of sufficient quality and completeness be available for at least the sections of road where monitoring equipment has been installed. PURPOSE OF THE REPORT This report serves two functions.

• The report increases the awareness of archived data. The information that can be gained from archived data is significant if the data limitations are understood and if the right data quality steps are performed. Using the data will also identify where improvements can be made. The Appendices and the Report seek to publicize the benefits of archived data in the areas of data quality, data analysis and information presentation.

• The data can be used to investigate issues. The effect of operational treatments and programs as well as local variations in travel demand can be studied in much greater detail than with any other data source at relatively low cost.

The archived databases have two notable limitations.

• It is not appropriate at this time to compare congestion or reliability levels in different cities. Between 9 percent and 100 percent of the freeway mileage in the study cities are covered in the archived databases—this wide range does not allow direct comparisons. The average coverage of freeway mileage in all cities is approximately 40 percent.

• The data cover only freeways and therefore cannot provide a complete picture of any event—normal or unusual. Until more information is compiled for the major street system and the transit system there will be an information gap.

S

xii

PROJECT FINDINGS The significant elements and findings of the project can be described in three areas:

• data collection and database development • performance measure development • future issues.

The following points stand out as those that either should be noted by other cities as they embark on such a program, or should be considered by the technical community as good practice. The full report provides more detail on each of these. DATA COLLECTION AND DATABASE DEVELOPMENT ISSUES There was a range of data collection technologies and practices, operating and archiving policies, and institutional arrangements in the 23 cities included in the third year. A few notable conclusions about the first three years are mentioned below.

• Areas that use the archived data for local transportation analyses typically had much higher quality data than those areas that simply archived and did not use their data.

• In most areas, local analysis of archived data has been a daunting task due to relatively difficult access to the data archives.

• There are no clear findings regarding the optimum type of traffic sensor for mobility monitoring but the need for initial calibration and continued maintenance is clear.

• The various data collection systems produce different patterns and statistics and the variations need more study.

• The staff in operating and planning agencies need more experience and better analysis tools to take full advantage of the data archives.

PERFORMANCE MEASURE DEVELOPMENT Using the data to create measures that transportation professionals and general audiences find valuable has only begun, but some issues were addressed in the study that should be recognized as the practice expands.

• Until more coverage is available, use the data to study local and national trends, but not to develop city-to-city comparisons.

• Mobility and reliability comparisons at the freeway corridor level can be greatly improved with archived data.

• Continuous data can be used to provide time-of-day, day-of-year, and seasonal trend analysis.

xiii

• The archived data can assist in monitoring congestion levels, programming improvements, scheduling maintenance operations, deploying staff and justifying investments in operations.

• Archived operations data should only be combined with other data sources when the differences in each type of data are well understood, and where the need for a combination of data is unavoidable.

• Traffic management center operators have different data requirements than other archived data users, which strongly indicates the need for a cooperative and integrated archiving program.

BEYOND THE FIRST THREE YEARS – SUMMARY OF FUTURE ISSUES The research has led to the development of performance measures and best practices for mobility monitoring. There are several improvement areas that are related to future improvements in archived data availability and use. The key points over the first three years have been:

• Significantly enlarge freeway and street sensor coverage and experiment with data sources to get a complete picture of the mobility provided by the roadway system.

• It will be worthwhile to identify conclusions and trends such as peak spreading, reliability changes, effect of ITS improvements, etc. to encourage faster use of archived data.

• Ensure that traffic monitoring data collected by roadway sensors are archived and made available in formats that a wide range of users can access.

• Improvement in data and measures will ultimately hinge on local developers and users benefiting from archived data systems for such uses as preparing congestion management system reports and other products.

• The percent of missing data (due to malfunctions, errors, or other outages) ranged from 7 percent to 94 percent in the 23 cities studied, indicating that maintenance of the data collection equipment should be improved.

• Incidents, weather and work zone locations have significant impacts on roadway travel times and can explain many of the unusual results, but their effect is difficult to understand without “event” databases tied to the traffic data.

• The collection of real-time traffic data by the private sector is growing with activities underway in a few cities. Some analysts optimistically project that the private sector will soon comprise the majority of real-time traffic data collection. If this does happen, data licensing agreements for archived data could be a significant stumbling block for performance monitoring efforts.

There are several methods to organize the data, calculate the measures, perform data quality assurance steps and interpret the results. The study team investigated several of these methods for each subject area. The information in this report represents the best approach identified to date. The authors and the sponsor hope that others will experiment with other methods and provide a useful dialogue.

xiv

1

HAPTER 1 - INTRODUCTION

The third year of the Mobility Monitoring Program saw an expansion in the number of cities, the miles of road and the amount of travel included in the statistics from the first two years (1,2). This report summarizes key elements of the process used to get the data into a useful form and the resulting information that can be derived. This report is one of several efforts designed to make archived data a more significant part of the data structure and decision-making process for a variety of purposes. As more areas develop operations centers and use the information to improve the service they provide, these procedures will allow professionals and non-technical users to have easy access to the information. This report summarizes the efforts undertaken on data collected in 23 U.S. cities during the year 2002. Additional information is available on the study website at http://mobility.tamu.edu/mmp.

The intent of the report is to provide traffic management center operators, planning and modeling professionals and other state and local agencies with some assurance that time invested in creating and maintaining archived operations databases will be well spent. Known as the archived data user service (ADUS), the data storage and analysis functions will be the foundation for future monitoring programs in the urban areas and on the roadways they cover.

The primary measures presented are those that use average travel time and variations in travel time to communicate system performance to users. This information can be used directly for a number of tasks and can be combined with surveys to estimate a variety of other trip experiences. The report also explores methods to present the measures and information in an easy-to-understand format.

The products presented in this report show that the information can be useful in a variety of ways. Technical and professional level staff can evaluate and “sell” the components of archived data systems that make the most sense for the public and decision-makers in their area. A common database format, discussions of the best practices for a variety of data archiving and analysis processes, the various measures that can be developed, and a framework for collecting and using the huge amount of information will help move data archiving systems forward in the studied areas as well as in other locations.

The report is oriented toward comparisons of mobility and reliability within individual cities; city-to-city comparisons are not appropriate with the data collection devices at such a low level of deployment. The research conducted as part of this study consistently showed that management center operators and staff from other state and local agencies value the ability to track changes within an area from year-to-year, and to use the information to evaluate programs. Comparisons between areas are less valuable as evaluation tools, although there will eventually be some value when a greater portion of the travel in an area is included in the database. Satisfying the local priorities will go a long way toward improving the quality of data available to the full range of users.

C

2

The report consists of five additional chapters.

• Chapter 2 – The Issues—a brief summary of key issues. • Chapter 3 – The Data—collection, processing, storage and analysis procedures. • Chapter 4 – The Measures—the measures that were calculated. • Chapter 5 – The Results: What do the Measures Show….?—a summary of findings from

the data. • Chapter 6 – The Future: Additional Opportunities—issues for study beyond the first

three years. WHAT’S NEW FOR THIS YEAR? Only a few changes have been made in this third year report to improve the database, the measures and the knowledge that can be gained from the archived data.

• More cities—Freeway data from 23 cities were used for the 2002 data year. • More miles—2,380 miles of freeway were used in this report. • More automation—With an increasing number of cities sending data, we continue to

fine-tune and automate the data processing and analysis. More effort was turned to analyzing and reporting the information for purposes of understanding the mobility and reliability issues, and to assisting others in the preparation and use of archived data systems.

3

HAPTER 2 - THE ISSUES

Extending the usefulness of archived operations data will require that evaluations of the transportation system and user experiences will be easier and more comprehensive. This project has investigated several issues related to those that will be faced by planners and operators. A brief summary is presented below to provide a framework for the report. MOBILITY AND RELIABILITY Mobility and reliability are the two key transportation system attributes that are being evaluated with archived data in this report. Travelers might ask questions such as “What is the average travel time?” or “How much time should I allow for this trip?” System evaluators ask questions such as “How easy is it to move around?” and “How much does that “ease of movement” vary?” Both of these users can benefit from the same measures. There are typically four components of mobility or congestion:

• Time of day mobility—the amount of time that the transportation system is congested or the mobility provided at various times of the day (e.g., duration).

• System mobility—the amount of the system that is congested or the level of mobility that the system provides (e.g., extent).

• Personal mobility or amount of people traveling in congested conditions—the level of mobility or congestion at the individual traveler level (e.g., intensity).

• The variation in those three—the amount of extra time that has to be built into trip planning so that travelers or goods will arrive on time (e.g., reliability).

THE MOBILITY MONITORING PROGRAM FRAMEWORK The benefits of developing, using and maintaining an operational data archiving system to support data analysis are a product of a long-term view. The framework of the Mobility Monitoring Program analytical process allows for local standards and issues to be incorporated, while benefiting from the cumulative experience of the range of users and to have a view of the broader applications for the information that can be developed. Having a view of the “market” for information not only provides structure to the program, it provides justification and motivation for improvements. WHY COLLECT AND ANALYZE SUCH LARGE DATASETS? Some in the profession have suggested that the amount of archived operations data is overwhelming. The procedures documented in this report are targeted to that audience. The procedures consist of an automated analytical process that provides information to a broad range of users and customers. The report and the other products are based on satisfying the needs of the range of potential audiences and users of the information, and the uses they have for data. In

C

4

addition to supporting the activities in centers on a daily basis, several other reasons exist for expanding performance monitoring activities by incorporating archived data. • Sound Business Practice. The private sector has been using performance measures as a way

to expand markets and expense efficiently for many years. “You have to know what’s happening before you can do anything about it” typifies the push for more and better data. While the goals of public agencies are quite different than private firms, elements of business-style discipline have direct relevance for the transportation profession. Performance monitoring is one of these elements and is key to providing better service and improved customer satisfaction for users of the highway system.

• Useful at All Levels of Transportation Decision-Making. Performance measures provide useful information for a variety of transportation personnel:

o Operations, where knowledge of how the system is performing now leads to specific actions.

o “Operational Planning”, where knowledge of how the system performed in the recent past lead to changes in Operational practices.

o Transportation Planning, where performance measures indicate long-term deficiencies in the transportation system.

o Policy, where system performance can help craft funding programs and emphasize research into specific areas.

• Multiple Uses for Data Already Collected. Operational strategies require knowledge of current system conditions in real time or near-real time. As a result, a wide variety of detailed data are collected by surveillance systems; these include traffic flow, weather, incidents, work zones, and special events. Once archived, these data provide a powerful resource for a variety of applications, of which performance monitoring is only one. Therefore, original data collection strictly for performance monitoring is not necessary—data already collected for operational strategies can be successfully leveraged for this purpose. In other words, the existence of operational data provides local, state and national agencies and interested parties with the means to conduct performance monitoring cost-effectively.

• Supplement Other Data Collection Efforts. Performance monitoring information can be combined with other data collection efforts—such as inventories of deployments and public opinion surveys—to provide a complete picture of how well the transportation system is meeting performance goals and the expectations of consumers.

5

SYSTEM MEASURES AND USER EXPERIENCE MEASURES Most operations-based data collection systems give relatively direct information about the four mobility/congestion components (i.e., duration, extent, intensity, variability). The data collection systems do not, however, give a direct indication about the trip-level experience of travelers. The trip level information can be estimated, however, through a combination of modeling, surveys and automatically collected data. The advantage of this approach is that the automated data collection can continuously monitor the areawide road network for many different uses and it can be calibrated to the user experiences with the surveys. Having a framework for integrating various data sources allows each source to be used according to its best application, and does not put undue pressure on data sources to provide information that they/it are not capable of supporting. Both system performance and user experience measures should be tracked because there are audiences for both types of measures and some of the statistics can be produced from the same database. DATA ELEMENTS AND ANALYTICAL PROCESSES USED IN THE MOBILITY MONITORING PROGRAM A summary of the data elements and analytical procedures is provided to orient the reader to the level of detail and the scope of the Program. Other data that might also be relevant and useful, but which was not collected (e.g., vehicle occupancy information) is also identified to indicate possible improvements in future reports. Detailed information on data validity checks used in archived data processing is also presented. RELATING ARCHIVED DATA AND ESTIMATES Archived data provides much more information than estimating techniques about the operation of freeway systems in normal time and during special or irregular events. There are some cautionary notes, however, and this report cites some of the main lessons. Until the technologies are more widely deployed, it will be difficult to compare performance characteristics from one area to another and it will be difficult to estimate the full effect of transportation improvement options. The data collected from continuous monitoring systems results in different performance measure values than those calculated from estimation techniques for a variety of reasons. Any estimation program will have difficulty replicating actual conditions, but the current real-time data collection devices suffer from the lack of coverage of the travel system elements. When events cause travelers to leave the monitored portion of the roadway, the performance measure statistics do not accurately reflect network conditions.

6

7

HAPTER 3 - THE DATA

This chapter summarizes aspects of the archived freeway operations data that were used in the mobility and reliability analyses described in this report. The chapter is organized as follows:

• Participating cities and their archived data—this section presents information on the cities that participated by submitting archived data, including the type of sensor technology used to collect the data, the level of detail of the archived data, and the data elements that were submitted.

• Overview of data processing—this section provides an overview of the data processing steps used to prepare the data for analysis, including pre-processing, data quality checking, and aggregation to a common data standard, and finally the mobility and reliability analysis.

• Data quality checking—this section describes the data quality checks used in preparing the archived data for analysis.

• Mobility and reliability measure calculations—this section introduces some of the steps that led to calculating the mobility and reliability statistics.

PARTICIPATING CITIES AND THEIR ARCHIVED DATA Representatives of a total of 23 cities participated in the Mobility Monitoring Program by submitting archived freeway traffic data from 2002 (Exhibit 3-1). Participants who submitted data were initially given basic guidelines about the type of traffic data needed and the preferred formats, with some variation being acceptable on a city-by-city basis. In the process of gathering archived traffic data from 23 cities and dealing with various data formats and organization, the project team developed written documentation on preferred data formats. These preferred data formats (Exhibit 3-2) were developed to clarify exactly what data was needed, as well as to “standardize” the archived data (with some minor variation) being submitted to the Program. Some of the preferred data formats and organization arose from how the data were to be processed in the SAS application software. Other details of organization were included because they were already present and consistent in the majority of cities submitting archived data. Note that the preferred data formats reference data elements contained in national ITS standards like the ITE/AASHTO Traffic Management Data Dictionary. In future years, the project team will encourage use of these preferred data formats to reduce our pre-processing burden and standardize our initial data processing software for all cities. Because participation and data submission is strictly voluntary, however, we are in a position to accept the data format and organization that is most convenient for cities to provide.

C

8

Exhibit 3-1. Participating Cities and Agencies for 2002 Archived Data

Participating City (Years of data in MMP) Contact Agency Albany, NY (2 years) New York State DOT Atlanta, GA (3 years) Georgia DOT Austin, TX (2 years) Texas DOT Charlotte, NC (2 years) North Carolina DOT Cincinnati, OH/KY (3 years) TRW, Inc./ARTIMIS/Kentucky Trans. Cabinet Detroit, MI (3 years) Michigan DOT Hampton Roads, VA (3 years) Univ. of Virginia/VTRC/VDOT Houston, TX (3 years) Texas DOT/TTI Los Angeles, CA (3 years) UC-Berkeley/Caltrans Louisville, KY (2 years) Kentucky Transportation Cabinet Milwaukee, WI (2 years) Wisconsin DOT Minneapolis-St. Paul, MN (3 years) Minnesota DOT and UM-Duluth Northern Virginia (1 year) Univ. of Virginia/VTRC/VDOT Orlando, FL (2 years) Florida DOT Philadelphia, PA (2 years) Mobility Technologies, Inc. Phoenix, AZ (3 years) Arizona DOT Pittsburgh, PA (2 years) Mobility Technologies, Inc. Portland, OR (2 years) Oregon DOT Sacramento, CA (1 year) UC-Berkeley/Caltrans Salt Lake City, UT (1 year) Utah DOT San Antonio, TX (3 years) Texas DOT San Diego, CA (2 years) Caltrans Seattle, WA (3 years) Washington State DOT/Washington State

Transportation Center (TRAC)

9

Exhibit 3-2. Preferred Archived Data Formats for Mobility Monitoring Program (Page 1 of 2)

PREFERRED DATA FORMATS FOR FHWA’S MOBILITY MONITORING PROGRAM

The following sections summarize the preferred formats for submitting data to FHWA’s Mobility Monitoring Program. While other formats are acceptable, the following formats are encouraged for unambiguous and efficient data exchange. The required data submissions should include 2 principal datasets: 1) actual traffic data records; and 2) traffic sensor location information. Many of the data elements have already been defined by national ITS standards (e.g., Traffic Management Data Dictionary, TMDD) and are indicated as such. FILE FORMATS • The traffic data records should be submitted in delimited ASCII-text files. Acceptable delimiting

characters include commas, tabs, or spaces. • Empty/blank fields or “null” values should be indicated by providing a blank space in the respective

field. Metadata should document particular error codes (e.g., “-1” or “255”) and meaning if these error codes are contained in the dataset.

• A separate text file should be submitted for each day for each city, with data from all sensor locations being included in a single daily file. The file should be named to include a location or agency code and a date stamp (YYYYMMDD format). For example, “msp_20020101.txt” contains data for Jan. 1, 2002 for Minneapolis-St. Paul, Minnesota.

• The text files should be compressed for transmission using industry standard PC (*.zip) or Unix (*.z or *.gz) compression software.

• The traffic monitoring data should be submitted by DVD, CD, or FTP. DATA ELEMENTS • The data should be aggregated to 5-minute time periods for each travel lane. Even 5-minute time

periods should be used (e.g., 12:00 am, 12:05 am, 12:10 am, etc.) • Each row of the text file should contain the following data elements:

1. Time (HH:MM with 24-hour clock) and date (MM/DD/YYYY) stamp. Documentation should indicate whether this is a start or ending time;

2. Detector identifier (DETECTOR_Identifier_identifier in TMDD); 3. Vehicle traffic volume count (DETECTOR_VehicleCount_quantity); 4. Average lane occupancy, if available (DETECTOR_Occupancy_percent); and 5. Average speed or travel time (LINK_SpeedAverage_rate or LINK_TravelTime_quantity).

• If the data have been aggregated from a shorter time period (e.g., 20 seconds or 1 minute), each 5-minute record should indicate how many sub-records were used in the summary statistic calculation. This “completeness” value is reported as the percentage of the total possible records that are included in the summary statistic.

10

Exhibit 3-2. Preferred Archived Data Formats for Mobility Monitoring Program (Page 2 of 2)

SENSOR LOCATION INFORMATION • Location information should be provided for each unique traffic sensor. The location information can

be provided in delimited text files, spreadsheets, or databases. • The location information should include the following for each traffic sensor that reports data at any

time during the year: 1. Detector identifier (same as used in the traffic data records, DETECTOR_Identifier_identifier); 2. Lane designation or code (DETECTOR_LaneNumber_code); 3. Number of directional through travel lanes at that location (LINK_LaneCount_quantity); 4. Roadway name and/or designation (LINK_RoadDesignator_number); 5. Roadway direction (DETECTOR_Direction_code); 6. Roadway facility type, such as mainlane, HOV, entrance ramp, etc. (LINK_Type_code); 7. A linear distance reference such as roadway milepost (in miles or kilometers); 8. Sensor activation date which indicates when the sensor began providing valid data; and 9. Sensor de-activation date which indicates when the sensor stopped providing valid data. If the

sensor is still active, this field could be blank or contain null values. ADDITIONAL DOCUMENTATION Additional documentation on the ITS data archives is encouraged. This documentation could include information on the following aspects: • Data collection technology and source; • Data quality control checks and summary results; • Data transformation or estimation processes (e.g., equations used to estimate speeds from single

loops); and • Other information that would help analysts better interpret the quality and content of the ITS data

archives.

11

The mobility and reliability analyses in the Mobility Monitoring Program are built around estimated travel times for freeway routes. As Exhibit 3-3 indicates, however, nearly all of the participating cities have traffic management centers that collect speeds and volumes at specific points along the freeway route. For 22 cities (all except Houston), the data were collected at point locations using a variety of traffic sensor technologies including single and double inductance loops, microwave radar, passive acoustic, and video image processing. For Houston, link travel times are collected via their automatic vehicle identification (AVI) system, and these link travel times are supplemented with volume trend data from a limited number of double inductance loops. In many cities, multiple sensor technologies were used to collect the traffic speed and volume data. All of these technologies use a small, fixed zone of detection, and the traffic speed and volume measurements are taken as vehicles pass through this zone. The last section in this chapter describes how these point speeds and volumes are transformed to travel time estimates for mobility and reliability performance measures. Exhibit 3-3 also indicates the level of detail at which the archived data is submitted to the Mobility Monitoring Program. The time aggregation level varies widely, from 20 seconds in San Antonio to 15 minutes in several areas. In some cases, the data are collected in smaller time intervals (e.g., 20 seconds to 2 minutes) but aggregated to larger time intervals for storage purposes. Nearly all of the archived data are provided on a lane-by-lane basis. The extent of freeway monitoring coverage is also presented in Exhibit 3-3, and ranges from 9 percent in Louisville, Kentucky to 100 percent in Milwaukee, Wisconsin and Salt Lake City, Utah. The average coverage is 40 percent, or slightly more than one-third of all freeway lane-miles in these urban areas. Note that the participating cities were not chosen based on their monitoring coverage, but on their ability to provide archived data. In many cities, this freeway monitoring coverage includes the most congested freeways as well as lightly congested freeway routes. In several cities, the monitoring coverage does not include very congested routes for a variety of reasons (e.g., reconstruction, upcoming deployment, etc.).

12

Exhibit 3-3. Summary of Archived Data Characteristics for 2002

Data Level of Detail Participating City Freeway System Monitored, % Traffic Sensor Technology Time Space

Albany, NY 10% (10 of 104 mi.) Single and double loop detectors

15 minutes by lane

Atlanta, GA 18% (73 of 300 mi.) Video imaging and microwave radar

15 minutes by lane

Austin, TX 22% (23 of 105 mi.) Double loop detectors 1 minute by lane Charlotte, NC 12% (13 of 92 mi.) Microwave radar 30 seconds by lane Cincinnati, OH/KY 27% (47 of 176 mi.) Double loop detectors, video

imaging, microwave radar 15 minute by direction

Detroit, MI 39% (110 of 282 mi.) Single and double loop detectors

1 minute by lane

Hampton Roads, VA 11% (19 of 181 mi.) Double loop detectors 2 minutes by lane Houston, TX 61% (298 of 368 mi.) Probe vehicle (AVI), limited

double loop detectors Anonymous individual probe vehicle travel times by link.

Loop data are 20 seconds by lane. Los Angeles, CA 86% (579 of 676 mi.) Single loop detectors 5 minutes by lane Louisville, KY 9% (12 of 137 mi.) Microwave radar, loop

detectors, video imaging 15 minutes by direction

Milwaukee, WI 100% (111+ of 111 mi.) Loop detectors, microwave radar

5 minutes by lane

Minneapolis-St. Paul, MN 60% (190 of 317 mi.) Single loop detectors 30 seconds by lane Northern Virginia 46% (59 mi of 127 mi.) Loop detectors 1 minute by lane Orlando, FL 20% (32 of 157 mi.) Double loop detectors 1 minute by lane Philadelphia, PA 37% (128 of 347 mi.) Microwave radar, passive

acoustic detectors 1 minute by lane

Phoenix, AZ 30% (53 of 179 mi.) Double loop detectors, passive acoustic detectors

5 minutes by lane

Pittsburgh, PA 27% (78 of 284 mi.) Microwave radar, passive acoustic sensors

1 minute by lane

Portland, OR 39% (54 of 137 mi.) Double loop detectors 15 minutes by lane Sacramento, CA 54% (57 of 105 mi) Loop detectors 5 minutes by lane Salt Lake City, UT 100% (80+ mi of 80 mi.) Double loops, microloops,

acoustic detectors 60 minutes by direction

San Antonio, TX 36 % (77 of 211 mi.) Double loop detectors, acoustic detectors

20 seconds by lane

San Diego, CA 66 % (163 of 248 mi.) Loop detectors 30 seconds by lane Seattle, WA 41 % (116 of 241 mi.) Mostly single loop detectors 5 minute by lane

13

Real-time traffic data collection and archiving processes have been developed independently in most of the cities and the details of these processes vary among the cities. As a general rule, TMCs at least have the capability to archive data from their surveillance systems. In a few cases, this capability is not used because of priorities elsewhere in the TMC, but it is clear that TMC software is being constructed with archiving as a function. However, the state of the practice in TMC archiving is still fairly primitive. The most common practice is to transfer the data to a storage device where they reside in simple file formats without an active information management system. Quality control is rarely performed at this level and access to the data is provided on a case-by-case basis without the benefit of a query or reporting structure – data are simply provided in whatever file formats are used to store them.

• Data are collected by traffic sensors and accumulated in roadside controllers. These field measurements are collected for each individual lane of traffic. At 20-second to 2-minute intervals, the roadside controllers transmit the data to a central location, typically a TMC.

• Some cities perform quality control on field-collected data, but this checking is simple and based on minimum and maximum range value thresholds.

• Cities that use single inductance loop detectors as sensors can measure only volumes and lane occupancies directly. In these cases, speed estimation algorithms are used to compute speeds from volumes and lane occupancies. These speed estimation algorithms vary among cities.

• Internal processes at the TMC aggregate the traffic data to specified time intervals for archival purposes. These time intervals vary from 20 seconds (no aggregation) to 15 minutes. In some cases, the data are also aggregated across all lanes in a given 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 offload some of the storage burden and to satisfy outside requests for the data.

Calibration and maintenance of field equipment and communications are nearly universal problems. The main impediment is lack of resources to devote to these tasks; TMC budgets are limited and must be used to address a multitude of issues. Calibration—at least to very tight tolerances—is not seen as a priority, given that operators focus on a broad range of operating conditions rather than precise volume and speed measurements. Or in some cases traffic managers may be willing to accept a certain level of data quality to satisfy only their current operations applications. This philosophy may be changing as a result of more stringent data requirements for traveler information purposes (e.g., travel time messages on variable message signs). However, we found the current data resolution used by TMCs to be quite coarse for supporting their traditional operations activities, such as incident detection and ramp meter control. Maintenance is a problem (due primarily to funding limitations) even when loops are known to be producing erroneous or no data. The problem is exacerbated where loops are used because most agencies are reluctant to shut down traffic on heavily traveled freeways just for loop repair.

14

This is not to say that faulty loops are never repaired, but maintenance is often postponed to coincide with other roadway activities, which helps spread the cost burden as well. Field checking of sensors is done periodically but no standardized procedures are used across all cities. If a detector is producing values that are clearly out of range, inspection and maintenance are usually performed. However, calibration to a known standard is rarely, if ever, performed. This means that more subtle errors may go undetected. Bearing in mind that TMCs typically do not require highly accurate data for most of their operations, this approach is reasonable and practical. Work zones exacerbate these problems and often contractors unknowingly sever communication lines or pave over inductance loops. OVERVIEW OF DATA PROCESSING This section presents a brief overview of the data processing steps used to transform the archived data into mobility and reliability statistics. The relatively mundane topic of data processing is included here because of its departure from traditional traffic data monitoring practices. In analyzing the archived freeway data from the 23 participating cities, the project team processed over 7 billion data records, with a total computer processing time best measured in days. Exhibit 3-4 shows an overview of the basic data processing steps used to prepare and analyze the archived data. Perhaps the greatest challenge in the data processing was “standardizing” the archived datasets from 23 different cities, or essentially 23 different legacy systems. In many cases, the lack of adequate metadata (i.e., descriptive information about the archived data) complicated the process of properly interpreting and analyzing the archived data. For example, each city’s dataset may use different data error codes to indicate various hardware or software failures. Or similar data error codes could be used by several cities to mean different types of data errors. In other cases, various flaws, nuances, or characteristics in the archived data may be known by the data collector but undocumented, and potentially go undetected by the project team unless careful study was initiated. The experience of the project team indicates that dealing with legacy system data is much more manageable when metadata is used to describe the origin, lineage, characteristics, and subtle nuances of the archived data. The data processing for the Mobility Monitoring Program is primarily accomplished using SAS software on a Microsoft Windows platform for 2 reasons: 1) the project team’s previous software programming experience with SAS; and 2) ability and flexibility of SAS to handle a wide range of complex computations on very large datasets. Many other relational database management systems (RDBMS) could also be used to accomplish the same data processing tasks as was performed in SAS. The data processing flows shown in Exhibit 3-4 have been optimized for the use of SAS in generating annual mobility and reliability reports. Some of the data processing steps, however, may be similar for other data archiving and analysis activities. For example, the first step that includes the “base code” is known as extraction, transformation and loading (ETL) in the data warehouse industry and is a common function for most data warehouse projects. The project team has attempted to standardize the software code as much as possible for ease and automation

15

Daily ASCII-Text Files• voluntary submission by cities• different formats for most cities• basic 1 record=1 obs format• up to 616 million records per city

“Standardized” Datasets• 1 to 10 GB SAS datasets• 20-100+M obs, 5 variables• each city in separate dataset• level of detail: 5-minute-by-lane

Summary Report Datasets• final datasets used in reports• Summary datasets, < 1MB total

Production Graphics• prepared in Excel• 30 minutes prep time per city once

in Excel• prefer charting capabilities and

chart appearance of Excel• graphics copied to Word using OLE

OLE copy fromSASView to Excel

“Non-Standard” Data• does not meet daily text input

format requirements (e.g., binaryfiles or thousands of separate filesper day per city)

• use batch DOS scripts, SAS, or 3rd

party software to “pre-process” todaily ASCII-text files

• “Base Code” • 5 to 6 SAS data steps• import (ETL), business rules, and

aggregation to “standard”• use macro to process day-by-day

due to hardware limitations• 6 to 24 hours run-time per city

• “Summary Code” • 20+ SAS data steps and PROCs• mostly summary and sorts• 1 to 6 hours run-time per city

of data processing. However, the software code is custom-tailored (mostly in the “base code”) to meet the different formats and organization of submitted archived data.

Exhibit 3-4. Overview of Data Processing within Mobility Monitoring Program

16

The data processing as shown in Exhibit 3-4 would ideally start with daily ASCII-text files that meet the preferred data formats indicated in Exhibit 3-2. However, many cities submit data in a form that requires pre-processing (e.g., binary file formats or thousands of separate files per city per day). Pre-processing this “non-standard” data requires extra steps and time at the beginning to prepare the archived data to be processed using the “base code.” Once the submitted archived data meets basic formatting and organization requirements, it is processed using the “base code.” This software code: 1) imports the data to SAS; 2) performs data quality checking; 3) aggregates detailed data to a common standard (currently 5-minute lane-by-lane); and 4) generates summary statistics on the data quality checking and processing steps. Some of these steps, such as the data quality checks, have been standardized for all cities. Other steps are unique to each city based on the aggregation level and other data characteristics. This step involves the longest amount of processing time, sometimes taking up to 24 hours for the largest cities with the most detailed data (e.g., 20-seconds, lane-by-lane). The “standardized” datasets are produced as a result of the “base code.” The data elements and table structure of these datasets are very similar with a few exceptions (e.g., some cities are 5-minute lane-by-lane, others may be 15-minute or by direction). Thus the “summary code,” which contains the mobility and reliability measure calculations described in Chapter 4, has largely been standardized for all cities. The “standardized” datasets are analogous to the database tables that would be kept on-line in an RDBMS environment. The “summary code” performs all mobility and reliability measure calculations, and produces relatively small datasets (less than 1 megabyte total) that are then used to produce the charts and tables shown throughout this report and the city report appendices. Microsoft Excel was selected for the ease of producing report-ready graphics. In summary, the data processing steps and software code used to analyze the archived data has developed in this way as a result of: 1) previous project team experience; and 2) the specific application of creating annual mobility and reliability reports. Different approaches are very likely given different implementation scenarios and development teams. Several of the data processing steps conducted in the Mobility Monitoring Program may be relevant to other data archiving or data warehouse activities. In particular, the “base code” contains data quality checking procedures and other steps that are most likely required in other data warehouse efforts. The “summary code” contains mobility and reliability measure calculations that are described in Chapter 4 and may be useful to others developing performance measure programs. DATA QUALITY CHECKING The topic of data quality is included here because of its overall importance in checking and evaluating the validity of archived data. Readers should note that the project team has not been able to systematically assess data accuracy. This means that the traffic speeds and volumes in the archived data could be systematically higher or lower (e.g., ± 10 to 20 percent) than true speeds and still be within the range of possible data values that pass quality control.

17

Exhibit 3-5 presents the data quality checks that were used in processing the 2002 archived data. The data quality checks have been developed from these sources:

• Current practices in other TMCs or data archiving systems; • Suggested practices recommended in the literature; and • Practices found to be necessary from project team analysis of the archived data.

These data quality checks can be characterized as basic validity checks and should detect major problems with data errors. More subtle erroneous or suspect data could potentially go undetected with these basic rules. The project team is reviewing the use of more sophisticated data quality checking, and we will continue to balance the sophistication of the data quality checking with the amount of available data processing time. The data quality checks shown in Exhibit 3-5 will likely evolve and further develop as the project team accumulates more experience with the archived data. More sophisticated quality checks could include tests like these:

• Rapid fluctuations in values across successive time periods; • Detectors in adjacent lanes at the same location reporting significantly different values or

trends; • Detectors in adjacent upstream or downstream locations reporting significantly different

values or trends; • Detectors from multiple locations reporting the same values (indicative of a system

problem); • Reported values that are significantly different from the location’s history for similar

days of the calendar. The results of the quality control checks are shown in Exhibit 3-6. This table reports the percent of the original dataset that passed the quality control checks. The table presents traffic volume and speed data quality separately, as some of the validity checks could have rejected one of the data values but not the other. Also note that Exhibit 3-6 only evaluates the validity of the data that was archived and submitted. This table does not reflect data that are missing and were never reported because of various hardware or software failures. Exhibit 3-7 summarizes information on data completeness or availability, another dimension of data quality. The data completeness measures the number of actual data values to the number of total possible values that one could expect (given the number of sensors and a polling rate). For example, if the data are reported by 5-minute time interval, 288 data values or records per day per detector are to be expected (i.e., 1,440 minutes per day divided by 5-minute periods equals 288 records). Exhibit 3-7 reports data completeness at three critical processing steps:

1. Original dataset as submitted by participating cities; 2. Dataset after quality control removes values failing the validity checks; and 3. Analysis dataset (after quality control and any imputation) that is used for mobility and

reliability performance measure calculations

18

Exhibit 3-5. 2002 Data Validity Checks in Mobility Monitoring Program Quality Control Test and Description Sample Code with Threshold Values Action

CONTROLLER ERROR CODES • Special numeric codes that indicate that

controller or system software has detected an error or a function has been disabled.

If VOLUME={code} or OCC={code} or SPEED={code} where {code} typically equals “-1” or “255”

• Set values with error codes to missing/null, assign missing value flag/code.

NO VEHICLES PRESENT • Speed values of zero when no vehicles present • Indicates that no vehicles passed the detection

zone during the detection time period.

If SPEED=0 and VOLUME=0 (and OCC=0) • Set SPEED to missing/null, assign missing value code

• No vehicles passed the detection zone during the time period.

CONSISTENCY OF ELAPSED TIME BETWEEN RECORDS • Polling period length may drift or controllers

may accumulate data if polling cycle is missed. • Data collection server may not have stable or

fixed communication time with field controllers.

Elapsed time between consecutive records exceeds a predefined limit or is not consistent

• Action varies. If polling period length is inconsistent, volume-based QC rules should use a volume flow rate, not absolute counts.

DUPLICATE RECORDS • Caused by errors in data archiving logic or

software process.

Detector and date/time stamp combination are identical.

• Remove/delete duplicate records.

QC1-QC3: Logical consistency tests • Typically used for date, time and location. • Caused by various types of failures.

If DATE={valid date value} (QC1) If TIME={valid time value} (QC2) If DET_ID={valid detector location value} (QC3)

• Write to off-line database and/or remove records with invalid date, time or location values.

QC4: MAXIMUM VOLUME • Traffic flow theory suggests a maximum traffic

capacity. •

If VOLUME > 17 (20 sec.) If VOLUME > 25 (30 sec.) If VOLUME > 250 (5 min.) If VPHPL > 3000 (any time period length)

• Assign QC flag to VOLUME, write failed record to off-line database, set VOLUME to missing/null.

QC5: MAXIMUM OCCUPANCY • Empirical evidence suggests that all data

values at high occupancy levels are suspect. • Caused by detectors that may be “stuck on.”

If OCC > 95% (20 to 30 sec.) If OCC > 80% (1 to 5 min.)

• Assign QC flag to VOLUME, OCCUPANCY and SPEED; write failed record to off-line database; set VOLUME, OCCUPANCY and SPEED to missing/null

QC6: MINIMUM SPEED • Empirical evidence suggests that actual speed

values at low speed levels are inaccurate.

If SPEED < 5 mph • Assign QC flag to SPEED, write failed record to off-line database, set SPEED value to missing/null

QC7: MAXIMUM SPEED • Empirical evidence suggests that actual speed

values at high speed levels are suspect.

If SPEED > 100 mph (20 to 30 sec.) If SPEED > 80 mph (1 to 5 min.)

• Assign QC flag to SPEED, write failed record to off-line database, set SPEED value to missing/null

19

Exhibit 3-5. 2002 Data Validity Checks in Mobility Monitoring Program (Continued)

Quality Control Test and Description Sample Code with Threshold Values Action MAXIMUM REDUCTION IN SPEED • Empirical evidence suggests that speed

reductions greater than some maximum value are suspect.

• Used only for AVI probe vehicle data that reports space mean speeds.

If SPEEDn+1 < (0.45 × SPEEDn) • Assign QC flag to SPEED, write failed record to off-line database, set SPEED value to missing/null

QC8: MULTI-VARIATE CONSISTENCY • Zero speed values when volume (and

occupancy) are non-zero • Speed trap not functioning properly

If SPEED = 0 and VOLUME > 0 (and OCC > 0) • Assign QC flag to SPEED, write failed record to off-line database, set SPEED value to missing/null

QC9: Multi-variate consistency • Zero volume values when speed is non-zero. • Unknown cause.

If VOLUME = 0 and SPEED > 0 • Assign QC flag to VOLUME, write failed record to off-line database, set VOLUME to missing/null

QC10: Multi-variate consistency • Zero speed and volume values when

occupancy is non-zero. • Unknown cause.

If SPEED = 0 and VOLUME = 0 and OCC > 0 • Assign QC flag to VOLUME, OCCUPANCY and SPEED; write failed record to off-line database; set VOLUME, OCCUPANCY and SPEED to missing/null

QC11: TRUNCATED OCCUPANCY VALUES OF ZERO • Caused when software truncates or rounds to

integer value • Calculate maximum possible volume

(MAXVOL) for an occupancy value of “1”:

If OCC = 0 and VOLUME > MAXVOL where MAXVOL=(2.932*ELAPTIME*SPEED)/600


QC12: MAXIMUM ESTIMATED DENSITY • Caused by improbable combinations of volume

and speed. • Traffic flow theory suggests that vehicle

density rarely exceeds 220 vehicles per lane per mile.

IF ((VOLUME*(3600/NOM_POLL))/SPEED) > 220 where NOM_POLL is the nominal polling cycle length in seconds.


QC13: CONSECUTIVE IDENTICAL VOLUME-OCCUPANCY-SPEED VALUES • Research and statistical probability indicates

that consecutive runs of identical data values are suspect.

• Typically caused by hardware failures.

No more than 8 consecutive identical volume-occupancy-speed values. That is, the volume AND occupancy AND speed values have more than 8 consecutive identical values, respectively.


20 20

Exhibit 3-6. Summary of Archived Data Passing Validity Checks

% of data passing checks City Volume Speed

Albany, NY 76% 75% Atlanta, GA 97% 94% Austin, TX 76% 44% Charlotte, NC 100% 100% Cincinnati, OH/KY 59% 63% Detroit, MI 69% 69% Hampton Roads, VA 65% 31% Houston, TX N/A 97% Los Angeles, CA 100% 97% Louisville, KY 85% 95% Milwaukee, WI 100% 83% Minneapolis-St. Paul, MN 100% 89% Northern Virginia 88% 71% Orlando, FL 49% 55% Philadelphia, PA 100% 99% Phoenix, AZ 70% 66% Pittsburgh, PA 100% 99% Portland, OR 77% 76% Sacramento, CA 99% 92% Salt Lake City, UT 90% 60% San Antonio, TX 95% 80% San Diego, CA 97% 94% Seattle, WA 98% 100%

21 21

Exhibit 3-7. Summary of Archived Data Completeness

at Different Processing Steps

% complete - original data

% complete - after QC

% complete- analysis data

City Volume Speed Volume Speed Volume Speed Albany, NY 74% 74% 50% 49% 50% 49% Atlanta, GA 46% 46% 43% 40% 43% 40% Austin, TX 96% 96% 78% 53% 80% 61% Charlotte, NC 20% 20% 26% 30% 26% 30% Cincinnati, OH/KY 51% 51% 10% 14% 10% 14% Detroit, MI 77% 77% 46% 46% 43% 43% Hampton Roads, VA 23% 23% 12% 7% 11% 6% Houston, TX N/A N/A N/A 93% N/A 52% Los Angeles, CA 57% 57% 57% 55% 57% 55% Louisville, KY 88% 88% 73% 83% 73% 83% Milwaukee, WI 79% 79% 79% 65% 79% 65% Minneapolis-St. Paul, MN 99% 90% 99% 85% 95% 85% Northern Virginia 23% 23% 21% 17% 22% 16% Orlando, FL 82% 82% 31% 37% 31% 37% Philadelphia, PA 95% 94% 95% 93% 95% 93% Phoenix, AZ 83% 83% 56% 53% 56% 53% Pittsburgh, PA 95% 89% 94% 87% 94% 87% Portland, OR 80% 80% 57% 56% 57% 56% Sacramento, CA 51% 51% 50% 47% 50% 47% Salt Lake City, UT 33% 33% 30% 19% 30% 19% San Antonio, TX 49% 49% 47% 40% 53% 50% San Diego, CA 94% 85% 88% 82% 92% 88% Seattle, WA 55% 55% 53% 55% 53% 55%

22 22

MOBILITY AND RELIABILITY MEASURE CALCULATIONS With the exception of Houston, which reported travel times collected with their AVI system, archived data from the participating cities consisted of traffic speeds and volumes collected at various points along the freeway routes. Because the mobility and reliability performance measures are based on travel time, the project team estimated freeway route travel times from the spot speeds. Exhibit 3-8 illustrates the process whereby lane-by-lane volumes and speeds are used as the basis for estimating freeway route travel times and vehicle-miles of travel (VMT). The steps are as follows:

1. If data are reported by lane, the lane-by-lane data are combined into a “station” (e.g., all lanes in a direction). Traffic volumes are summed across all lanes, and traffic speeds are a weighted average, with weighting based on respective traffic volumes.

2. Link properties were estimated from “station” data by assuming that each detector had a zone of influence equal to half the distance to the detectors immediately upstream and downstream from it. The measured speeds were then assumed to be constant within each zone of influence, and travel times were calculated using the equivalent link lengths. VMT were also computed in this way using traffic volume.

3. Freeway links were then grouped with other similar adjacent link into analysis sections, which were typically 5 to 10 miles in length. The beginning and end points of analysis sections were typically selected to coincide with major highway interchanges or other locations where traffic conditions were expected to change because of traffic or roadway characteristics.

Travel times for these analysis sections then served as the basis for all subsequent mobility and reliability measure calculations. The specifics of these performance measure calculations are contained in Chapter 4. Readers should note that equations using travel time refer to the analysis section travel times as described above. Several other aspects and definitions used in preparing the archived data for analysis were:

• Holidays were excluded from analysis. Future analyses may consider holidays separately or as part of weekends, but holidays were felt to be atypical of normal travel patterns.

• Consistent time periods for all cities were defined for analysis. These were:

o 12:00 am to 6:00 am – early morning o 6:00 am to 9:00 am – morning peak o 9:00 am to 4:00 pm – mid-day o 4:00 pm to 7:00 pm – afternoon peak o 7:00 pm to 12:00 am – late evening

• Only mainline freeway detectors were included. Some cities reported ramp data, but these were dropped to maintain consistency across the cities.

23 23

traffic sensors collect data in each lane at 0.5-mile nominal spacing

summary statistics computed across all lanes in a given direction

link travel time &vehicle-miles of travel

link travel time &vehicle-miles of travel

point-based properties extrapolated to roadway links 0.5 to 3 miles in length

directional roadway sectiontravel time & vehicle-miles of travel

directional roadway section travel time & vehicle-miles of travel

link properties summed to analysis sections 5 to 10 miles in length

Lane-by-LaneLevel

SectionLevel

LinkLevel

StationLevel

Exhibit 3-8. Estimating Route Travel Times and VMT from Spot Speeds and Volumes

24 24

Incident Data and Other Event Data

Archiving of incident data is becoming more prevalent at TMCs. However, the nature of the data collected and the structure of the storage formats are extremely diverse. This is a larger problem than for traffic data, where the basic measurements are fairly well known and understood. By comparison, even the definition of an “incident” is subject to interpretation. The resulting inconsistency in reporting formats for incidents limits, or at least complicates, analysis opportunities.

25

HAPTER 4 - THE MEASURES

Data for the mobility measures in this report were collected by freeway traffic sensor systems described in Chapter 3. The measures can also be estimated using widely available traffic count and road inventory data. The ultimate implementation of a set of travel time-based mobility measures in most urban areas will probably rely on some estimation procedures as well as an evolution toward a significant archived traffic operations database. This section describes the measures that form the basis for the mobility and reliability analyses. Included in the measures are the data items that can be gathered from real-time traffic data systems. MOBILITY MEASURES Four primary mobility measures were calculated with the 2002 data. The measures provide information about user experience as well as system operating condition. Any interpretation of the statistics should include recognition of the limited nature of the system mileage and travel coverage. This test phase also provides an opportunity to examine the measures, the calculation procedures and the conclusions and messages that can be supported. The travel time index (TTI) is a comparison between the travel conditions in the peak period to free-flow conditions. The measure can be averaged for streets, freeways, bus and carpool lanes, bus and rail transit, bicycle facilities and sidewalks. All of these system elements have a free-flow travel rate and when crowded, the travel rate (in minutes per mile) increases. Theoretically, the index could even be used to measure Internet service conditions. An average corridor value can be developed using the number of persons using each facility or mode to calculate the weighted average of the conditions on adjacent streets, freeways, HOV lanes, bus routes and/or rail transit lines. The corridor values can be computed for hourly periods and weighted by the number of travelers to estimate peak-period or daily index values. The index can be applied to various system elements with different free-flow speeds, although only freeways were analyzed in the 2002 MMP report. The travel time index in Equation 1 compares measured travel rates to free-flow conditions for any combination of freeways and streets. Index values can be related to the general public as an indicator of the length of extra time spent in the transportation system during a trip. Vehicle travel or person travel (measured in miles traveled on each part of the system) can be used as the weighting factor. The 2002 report included travel time index values for five periods of each day, as well as average peak and daily values.

C

26

+

×+

×

=

VMT PeriodPeak Street Arterial

Principal

VMTPeriodPeak

Freeway

VMTPeriodPeak

Street ArterialPrincipal

Rate flow-FreeStreet Arterial

PrincipalRate TravelStreet Arterial

Principal

VMTPeriodPeak

Freeway

Rateflow-Free

FreewayRate Travel

Freeway

Index TimeTravel (1)

The calculation procedure uses the units of travel rate due to the ease of mathematical calculation The peak-period value is calculated as a weighted average for all travel (vehicle-miles of travel) in the peaks. The percent of congested travel is primarily a system measure but can also estimate user experiences. A target speed is used as the benchmark and any travel on a road section for a time period that is at less than the target speed is determined to be congested. The 2000 MMP report (1) used a freeway speed of 60 mph as a congestion benchmark. Any 5-minute period with an average speed of less than 60 mph was recorded as congested and the travel in that time (measured in vehicle-miles traveled) was labeled as “congested.” In practice, the measure may over-report the amount of congestion with a 60 mph threshold so a 50 mph value has been used in subsequent reports (2). Unlike the other measures, the percentage of congested travel has an all-or-nothing characteristic. If the nighttime speed limit on the urban freeway system is 55 mph, a significant portion of travel could be categorized as congested, without a serious congestion problem being the cause. Spot speed detectors are also more likely to record lower speeds than longer distance travel time measurements, due to their frequent location near entrance ramps and the much greater variation in speed over short sections than long sections. These considerations might suggest that a lower speed is more appropriate for the congestion threshold when using point-based sensors. RELIABILITY MEASURES The mobility performance measures reflect the average level of congestion and mobility. However, a number of empirical studies and surveys 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. From a performance improvement standpoint, incident management and traveler information strategies target the atypical events that decrease reliability and evaluation of these programs should include data about these effects. This is important because it is usual for travel time savings to dominate the benefits assigned to major transportation improvement projects. Focusing only on average conditions would miss a large share of the benefits that accrue from these operations strategies.

27

The planning time index is one of two measures used to indicate reliability levels. The planning time index is statistically defined as the 95th percentile travel time index, this measure also represents the extra time most travelers include when planning peak period trips. For example, a value of 1.60 means that travelers plan for an additional 60% travel time above the off-peak travel times to ensure 95% on-time arrival. The planning time index is useful because it can be directly compared to the travel time index (essentially the average travel rate) on similar numeric scales. For example, assume that the peak period travel time index for a particular road section is 1.20, which means that average travel times are 20 percent longer in the peak period than during free-flow conditions. Now assume that the planning time index for that same road and time period is 1.60, which means that 95 percent of all travel times are less than 60 percent longer than during free-flow conditions. In other terms, the planning time index marks the upper limit for the nearly worst (95 percent of the time) travel conditions. Planning Time Index = 95th Percentile Travel Rate (2) The buffer index is the second measure used to estimate reliability levels in the Mobility Monitoring Program. It is similar in concept to a measure developed for the HOWLATE program by Mitretek (3). The buffer index expresses the amount of extra “buffer” time needed to be on-time 95 percent of the trips (for example, late for work on one day per month). Indexing the measure provides a time and distance neutral measure, but the actual minute values could be used by an individual traveler for a particular trip length or by an information provider for a set of typical travel routes. The index is calculated for each road segment and a weighted average is calculated using vehicle-miles of travel as the weighting factor (Equations 2 and 3).

For Each Section of Roadway:

×= 100%

mile)per minutes(in Rate Travel Average

mile)per minutes(in Rate Travel Average-mile)per minutes(in

Rate Travel Percentile95th

IndexBuffer (3)

( ) ( )

...VMTVMT...BIVMTBIVMT

Sections Severalfor BI ofAverage Weighted

2section 1section

2section 2section 1section 1section

+++×+×

= (4)

The calculations basically consist of calculating the average and 95th percentile travel time for each section of roadway (approximately five miles long and in some cases made up of several links of road) for each combination of days and time periods (Equation 2). The Buffer Index values of each five-mile road section can be calculated and then combined to calculate the Buffer Index for a corridor or area. Vehicle-miles (or person-miles) of travel would be used to weight each section buffer index value (Equation 4). Estimating the buffer index value for each situation is a process of identifying the time period and/or set of conditions that a traveler might use in trip planning, what the conditions are at the start of the trip and incorporating those into the calculation procedures so that the information from the measure matches the method used to create the statistics.

28

The current standard format is for the data to be stored in increments of 5 minutes for each of the lanes in one roadway direction. Sections of directional roadway of approximately 5 to 10 miles are used as the basis for calculations (see Exhibit 3-8). The travel time for these sections is created by adding travel times from each segment of road for each 5-minute time slice. Combining the data from each 5-minute time slice is accomplished using vehicle-miles or person-miles of travel to weight the values from each time slice. Measures that do not use specific origin-destination trips generally provide easier comparison methods because length-neutral measures can be applied to a wider variety of situations. The minutes of travel time can then be estimated by travelers, or by local agencies for particular trips. For many comparisons, travel rate (expressed in minutes per mile) may be better than travel time. In this application, however, the buffer index is already a unitless measure. The buffer index can be calculated for each road segment or particular system element using Equations 3 and 4. It seems appropriate to track several reliability performance measures. The 2000 MMP Report (1) used three measures (e.g., percent variation, misery index and buffer index) which described slightly different aspects of travel time variation. The messages and the numerical values that can be constructed from each measure are different, but the general relationships between congestion and reliability were similar for all three measures. There is no agreed-upon measure, and both customer/user market research and professional practice may develop additional measures. For each measure there are several levels of reliability or variability (e.g., 85 percent, 90 percent, 95 percent) that can be used. More research remains to be done on the level of reliability and the measures to communicate it to a range of audiences. SELECTION OF TIME PERIOD The time period over which the performance measures are computed must also be determined. No single time period will be correct for all analyses, but the considerations below are useful for all studies.

• Peak-hour or period—Transportation engineers have traditionally used a peak hour to describe congestion, but major urban areas now experience slow speeds for multiple hours in both the morning and the afternoon. Use of a single peak hour misses the congestion that occurs during other times, prompting many areas to define a multi-hour peak period.

• Urban area size—Using a 3- to 4-hour peak period for all area sizes, however, may mask congestion for the smaller urban areas. Smaller areas can probably develop useful statistics with only peak hour analyses.

• City-to-city comparison—A consistent peak-period length is necessary for any type of comparison between cities. Comparative studies between urbanized areas should probably use peak-period analyses, rather than only a peak hour.

29

• Daily or peak comparisons—For national comparisons of reliability trends, a day-to-day comparison is appropriate. For local purposes, where individual trip planning is also an issue, it will be useful to also include reliability in travel conditions within an hour or for several segments of the peak period.

The twin approach of both national and local focus is a strong point of the archived operations data analysis process, and strengthens the mobility and reliability information provided to a wide range of customers without a large incremental effort beyond a “basic approach.” Archived operations data is typically collected at a fine enough detail (both in time and space) to permit detailed local analyses (e.g., a 3-mile freeway section). National analyses at a city or areawide level is accomplished by aggregating this detailed data.

30

31

HAPTER 5 - THE RESULTS: WHAT DO THE MEASURES SHOW….?

A number of trends and observations about the data and measures were discovered as a result of analyzing the 23-city database. This section details some of the general findings as well as specific summaries of the urban area data. City Appendices (posted on the study website at: http://mobility.tamu.edu/mmp) are reports for each of the 23 cities included in the year 2002 data analysis. An important aspect of the study has been the relatively limited nature of the city list. There are only 23 cities in the database, and we received data for only a portion of the freeways in each city, meaning that the information in the report covers only a portion of travel in each area. With these limitations, the reader should be very careful about extending the conclusions too far beyond those freeways that have operations sensor coverage. But the detailed information about the day-to-day performance of the freeways is a unique window on traveler experiences and system operation for those roads that are instrumented with sensors. Exhibit 5-1 illustrates one of the “only” aspects: the amount of system coverage. At least 14 areas do not have coverage of more than half of the freeway system mileage. This is the principal reason for not using these data for city-to-city comparisons. Contrary to generally accepted ideas about where the monitoring equipment is deployed, it appears that the portion of the system that is covered by operations sensors are not always the most congested roadway sections. At least 11 areas have coverage on a higher percentage of the lane-miles than for the daily travel. This seems to indicate that the lane-miles covered in the archived database are not the heaviest traveled. Alternatively, the ITS sensors may be on the most heavily-traveled roadways but may be consistently undercounting the actual traffic volumes and VMT. A previous study of archived data in San Antonio (4) demonstrated that certain TransGuide detectors were consistently undercounting actual volumes. Some of the difference is caused by the methods used to construct the two databases used in Exhibit 5-1. The urban system statistics are based on FHWA’s Highway Performance Monitoring System (HPMS), which contains roadway system travel and road length data for the entire urban area. The traffic count information used to estimate roadway system travel within HPMS is based on statistical sampling. The archived system covers only parts of the freeway system and while the minute-by-minute counts are compiled for the entire year, rather than sampled, there are calibration and data outage issues that might affect the accuracy.

C

32

Exhibit 5-1. Summary of Instrumented Section Coverage in 2002

Instrumented Corridors Urban Freeway System1

Lane-Miles Annual Vehicle Miles of Travel2

City number percent million percent Lane-Miles

Annual Vehicle Miles of Travel

(million)

Albany, NY 83 15% 450 21% 550 2,124 Atlanta, GA 775 34% 6,360 40% 2,295 15,837 Austin, TX 146 25% 765 23% 585 3,360 Charlotte, NC 110 23% 830 29% 485 2,871 Cincinnati, OH/KY 410 40% 2,488 42% 1,015 5,887 Detroit, MI 855 47% 5,645 49% 1,810 11,538 Hampton Roads, VA 171 18% 780 17% 940 4,548 Houston, TX 1,900 77% N/A N/A 2,480 16,485 Los Angeles, CA 3,700 63% 29,940 61% 5,850 49,399 Louisville, OH/KY 84 13% 625 16% 670 3,811 Milwaukee, WI 695 100% 3,140 92% 605 3,396 Minneapolis-St. Paul, MN 999 63% 5,900 60% 1,590 9,877 Northern Virginia 533 64% 2,835 51% 827 5,549 Orlando, FL 230 30% 1,985 55% 760 3,626 Philadelphia, PA 688 33% 4,165 37% 2,085 11,231 Phoenix, AZ 565 50% 2,835 36% 1,140 7,885 Pittsburgh, PA 377 31% 1,790 42% 1,215 4,271 Portland, OR 295 42% 2,127 45% 710 4,710 Sacramento, CA 411 58% 2,525 52% 705 4,827 Salt Lake City, UT 530 100% 528 17% 530 3,028 San Antonio, TX 485 45% 2,465 42% 1,070 5,805 San Diego, CA 1,196 65% 7,875 64% 1,830 12,228 Seattle, WA 820 47% 5,285 48% 1,740 11,120 Note: 1Total freeway lane mileage obtained from Highway Performance Monitoring System (HPMS) and TTI

Annual Urban Mobility Report (5). 2Average VMT includes a process to estimate “missing” data points in archived data sets. PERFORMANCE MEASURE OBSERVATIONS The mobility and reliability measures presented in this report are a small subset of those that can be developed and that have been used to measure the performance of the transportation system or the user experience. The measures have been described as either mobility or reliability, but most non-technical readers and travelers see the two concepts as linked or even the same.

• The Travel Time Index has a much narrower range of areawide values than was initially expected. The peak-period conditions that are included in the values in Exhibit 5-2 do not have the range of values between cities or the same general ranking as those that are found in the Annual Urban Mobility Study (UMS) Annual Report (5). There are a number of possible reasons:

♦ The UMS Annual Report uses a relatively unsophisticated (when compared to the operations sensors) speed estimation process. There are a variety of potential inaccuracies.

33

♦ The freeways included in the MMP are not a representative sample. ♦ Ramp delay is not included in the MMP database. ♦ The sensors that collect operations data are not always calibrated or functioning.

Some inaccurate data may be present even after quality control. ♦ The off-peak direction travel has grown more rapidly than peak direction travel.

The high-speed operation that is typical of that direction is not properly accounted for in the UMS database.

♦ The incident management activities and other operational improvements have a beneficial effect that is not captured in the UMS procedures. Most of the MMP cities have an incident management program as part of the corridor operations.

♦ During incidents and serious vehicle breakdowns some traffic diverts from the freeway to adjacent streets. These roads are not included in the archived database, so any increased delay suffered by the diverted traffic is not recorded.

Exhibit 5-2. 2002 Average Peak Period Mobility and Reliability Statistic Summary

City Travel Time Index Buffer Index Albany, NY 1.10 25% Atlanta, GA 1.24 34% Austin, TX 1.10 23% Charlotte, NC 1.15 28% Cincinnati, OH/KY 1.29 33% Detroit, MI 1.11 31% Hampton Roads, VA 1.05 33% Houston, TX 1.22 36% Los Angeles, CA 1.42 49% Louisville, OH/KY 1.07 19% Milwaukee, WI 1.08 18% Minneapolis-St. Paul, MN 1.20 41% Northern Virginia 1.16 34% Orlando, FL 1.22 47% Philadelphia, PA 1.22 37% Phoenix, AZ 1.17 25% Pittsburgh, PA 1.23 33% Portland, OR 1.33 41% Sacramento, CA 1.08 21% Salt Lake City, UT 1.01 2% San Antonio, TX 1.10 24% San Diego, CA 1.18 32% Seattle, WA 1.26 31% Note: These values are for non-holiday, weekday peak-period conditions. Note: See website for more details: http://mobility.tamu.edu/mmp

34

• The Buffer Index values have a broader range than the Travel Time Index values. Until there is more experience with the reliability concept and with individual measures, it will be difficult to interpret the data with any degree of certainty.

• The average Buffer Index and Travel Time Index values for each city are graphed in Exhibit 5-3. They indicate a relatively good correlation between congestion and unreliability during the peak periods. In most cities this pattern also holds for the freeway section data. There are some sections where the two measures diverge, but these are more typically during the off-peak period than during a peak.

TRENDS IN MOBILITY AND RELIABILITY At most, the Mobility Monitoring Program has three years of archived data. With such a short time period, trends will always be difficult to identify (particularly with changes in the amount of data coverage and variations in the maintenance and data collection techniques in each area). The data are very useful, however, for examining several issues in data collection and system performance, rather than comparing cities. Exhibit 5-4 lists the amount of freeway miles covered and the size of the system in each area. Only five areas include more than half of the freeway miles in the data archives, but

Note: Each data point is the average for all the monitored freeway data in each study city, as shown in Exhibit 5-2.

Exhibit 5-3. Relationship between Average Peak-Period Mobility and Reliability in Each Study City

0%

10%

20%

30%

40%

50%

1.00 1.10 1.20 1.30 1.40 1.50

Travel Time Index

Buf

fer I

ndex

35

approximately one-third of the cities increased the amount of freeways covered. Areas that have a relatively low percentage of coverage, or changes from year to year make it difficult to identify general trends. Exhibit 5-5 illustrates the travel time index and the buffer index from 2000 to 2002. Several cities show trends that are either difficult to correlate to other data or to public perception. This does not diminish the usefulness of the data set. The variations in traffic congestion problems from day to day and between corridors can only be studied with this data source. DAILY AND MONTHLY PATTERNS Some of the findings confirm the common knowledge about urban roadway systems. There are some others that point toward an expansion of the list of things that should concern transportation professionals. The issues might suggest new or expanded programs to address reliability issues and congestion in areas and during times that have not been a large concern in many areas. There is also, however, some degree of skepticism associated with the statistics. The amount and intensity of congestion is not as significant or as widespread as many believe, and what other estimation processes and surveys indicate. Some of the differences are explainable, but the degree of difference is significant in some cases and will require some additional study. The differences could be related to:

• The data collection equipment and procedures.

• The amount of roadway included.

• Which sections of roadway are included.

• New discoveries about the level of congestion on urban roadways.

• Relationships between the causes of congestion and unreliability.

• Operational treatments designed to improve transportation systems.

36

Exhibit 5-4. Miles of Freeway Data, 2000 to 2002

Miles of Freeway with Archived Data City 2000 2001 2002

2002 Freeway System Mileage

Albany DNP 10 10 104 Atlanta 40 53 73 300 Austin DNP 23 23 102 Charlotte DNP 13 13 92 Cincinnati 46 47 47 176 Detroit 117 110 110 282 Hampton Roads 23 23 19 181 Houston 225 298 298 368 Los Angeles 192 577 579 676 Louisville DNP 12 12 137 Milwaukee DNP 122 122 111 Minneapolis 190 190 190 317 Northern Virginia DNP DNP 59 127 Orlando DNP 32 32 157 Philadelphia DNP 128 128 347 Phoenix 53 53 53 179 Pittsburgh DNP 79 78 284 Portland DNP 54 54 137 Sacramento DNP DNP 57 105 Salt Lake City DNP DNP 80+ 80 San Antonio 68 77 77 211 San Diego DNP 163 163 248 Seattle 68 116 116 241 DNP—did not participate in program in this year.

37

Exhibit 5-5. Average Peak Period Travel Time Index and Buffer Index, 2000 to 2002 Travel Time Index Buffer Index

City 2000 2001 2002 2000 2001 2002 Albany DNP 1.12 1.10 DNP 25% 25% Atlanta 1.13 1.22 1.24 25% 39% 34% Austin 1.39 1.11 1.10 111% 24% 23% Charlotte DNP 1.23 1.15 DNP 43% 28% Cincinnati 1.25 1.33 1.29 37% 38% 33% Detroit 1.12 1.12 1.11 31% 32% 31% Hampton Roads DNP 1.02 1.20 DNP 4% 33% Houston 1.26 1.22 1.22 50% 38% 36% Los Angeles 1.25 1.36 1.42 46% 44% 49% Louisville DNP 1.16 1.07 DNP 35% 19% Milwaukee DNP 1.07 1.08 DNP 19% 18% Minneapolis 1.04 1.27 1.20 72% 45% 41% Northern Virginia DNP DNP 1.16 DNP DNP 34% Orlando 1.07 1.33 1.22 30% 41% 47% Philadelphia DNP 1.21 1.22 DNP 37% 37% Phoenix 1.11 1.15 1.17 43% 24% 25% Pittsburgh DNP 1.17 1.23 DNP 21% 33% Portland DNP 1.37 1.33 DNP 42% 41% Sacramento DNP DNP 1.08 DNP DNP 21% Salt Lake City DNP DNP 1.01 DNP DNP 2% San Antonio 1.06 1.09 1.10 36% 24% 24% San Diego DNP 1.24 1.18 DNP 39% 32% Seattle 1.22 1.26 1.26 28% 29% 31% DNP—did not participate in program in this year. Other main findings about the variation of congestion and mobility across the day and year include the following features.

• Using average peak-period values to create the summary statistics provides a perspective more consistent with the user perspective than the average daily values. The average daily statistics indicate much lower congestion levels than the peak period due to the inclusion of the off peak-period.

• It seems clear that off-peak direction travel in most areas remains a) present—that is, there is an off-peak direction in most of the corridors studied, and b) reasonably good—that is, off-peak speeds are reasonably high. The growth in traffic volume in the off-peak direction has led to higher speed trips being a larger portion of travel, thus lowering the average TTI value and lowering the apparent level of congestion.

• When the off-peak direction reaches congested conditions, the corridor average TTI rises significantly. Off-peak speeds that decline precipitously in the beginning of congested conditions (as described in the Highway Capacity Manual) raise the peak-period TTI by a significant amount. And in several cities the evening average is much higher than the morning.

38

• The evening peak period shows equal or higher TTI values than the morning in 22 of the 23 areas studied (Exhibit 5-6). In several cities, the evening average is much higher than the morning.

• The midday period of most cities includes moderate congestion.

• Eight areas have midday TTI values of 1.10 or higher. In all but one city the midday period has the lowest of the three TTI averages.

• The reliability measures (Exhibit 5-7 and Exhibit 5-8) follow the same trend as the mobility statistics—unreliability is higher in the evening peak than in the morning, and the midday period is not a significant problem in most cities.

• Graphs like Exhibit 5-8 illustrate an apparent additional peak in some cities due to shift workers or slower travel during nighttime hours. The graph may also indicate a dip in congestion levels just before the morning peak, showing the effect of early commuters apparently trying to “beat the rush” by driving faster.

• The graph of Houston data similar to Exhibit 5-8 (Exhibit HOU-8 in the Appendix) is smoother than most other graphs due to the travel time data collection devices that collect section travel times, rather than the point data sources that estimate traffic speed at one spot on the road.

39

Exhibit 5-6. 2002 Daily Mobility Summary

Average Travel Time Index

City Morning Peak Period

(6 am to 9 am) Midday Period (9 am to 4 pm)

Evening Peak Period (4 pm to 7 pm)

Albany, NY 1.06 1.03 1.14 Atlanta, GA 1.17 1.08 1.30 Austin, TX 1.07 1.02 1.13 Charlotte, NC 1.13 1.08 1.17 Cincinnati, OH/KY 1.25 1.26 1.33 Detroit, MI 1.08 1.02 1.15 Hampton Roads, VA 1.03 1.01 1.07 Houston, TX 1.19 1.06 1.26 Los Angeles, CA 1.37 1.20 1.48 Louisville, OH/KY 1.05 1.04 1.08 Milwaukee, WI 1.08 1.03 1.08 Minneapolis-St. Paul, MN 1.15 1.06 1.25 Northern Virginia 1.15 1.06 1.17 Orlando, FL 1.17 1.12 1.28 Philadelphia, PA 1.20 1.11 1.24 Phoenix, AZ 1.14 1.08 1.19 Pittsburgh, PA 1.26 1.11 1.20 Portland, OR 1.24 1.19 1.42 Sacramento, CA 1.04 1.02 1.13 Salt Lake City, UT 1.01 1.00 1.01 San Antonio, TX 1.08 1.02 1.11 San Diego, CA 1.14 1.05 1.22 Seattle, WA 1.21 1.15 1.32 Note: See website for more details: http://mobility.tamu.edu/mmp

40

Exhibit 5-7. 2002 Daily Reliability Summary

Average Buffer Index (%)

City Morning Peak Period

(6 am to 9 am) Midday Period (9 am to 4 pm)

Evening Peak Period (4 pm to 7 pm)

Albany, NY 13 4 37 Atlanta, GA 26 21 42 Austin, TX 19 6 26 Charlotte, NC 22 18 33 Cincinnati, OH/KY 27 18 39 Detroit, MI 24 8 37 Hampton Roads, VA 9 3 12 Houston, TX 34 18 37 Los Angeles, CA 47 47 51 Louisville, OH/KY 15 9 23 Milwaukee, WI 18 9 18 Minneapolis-St. Paul, MN 36 25 46 Northern Virginia 30 20 39 Orlando, FL 39 39 56 Philadelphia, PA 31 25 42 Phoenix, AZ 22 10 27 Pittsburgh, PA 28 22 39 Portland, OR 33 36 49 Sacramento, CA 15 8 27 Salt Lake City, UT 2 1 2 San Antonio, TX 24 7 25 San Diego, CA 25 20 39 Seattle, WA 25 27 37 Note: See website for more details: http://mobility.tamu.edu/mmp

41

1.00

1.10

1.20

1.30

1.40

1.50

1.60

1.70

1.80

1.90

12 AM 2 AM 4 AM 6 AM 8 AM 10 AM 12 PM 2 PM 4 PM 6 PM 8 PM 10 PM 12 AM

Time of Day (weekdays, non-holidays only)

Inde

x Va

lue

Travel Time Planning Time

Exhibit 5-8. Mobility and Reliability Measures by Time of an Average Day, Albany (Example)

42

• Exhibit 5-9 presents a congestion and a reliability measure in a single graph. Travel Time is the index showing the peak-period travel time penalty. Planning Time is the index value that illustrates the time that should be allowed for an important trip. Both values show travel time relative to free-flow travel conditions. The Buffer Index is the difference in the Travel Time and Planning Time. Relationships between mobility and reliability for different times of the day in a city can be investigated with this type of chart.

1.00

1.10

1.20

1.30

1.40

1.50

1.60

Early AM AM Peak Mid-day PM Peak Late PM

Inde

x V

alu

e

Travel Time Planning Time

Exhibit 5-9. Mobility and Reliability during Daily Time Periods (Example)

Note: See website for more details: http://mobility.tamu.edu/mmp

43

• Evening peak period delay is the highest percentage in 15 of the 23 cities (Exhibit 5-10). In all but one of the other eight, midday delay is highest. While the midday congestion level is lower, there are more hours and the combination causes a significant number of hours of total delay.

• Fifteen of the 46 early morning or late evening periods account for 10 percent or more of daily delay.

• In four cities, the delay between 7 p.m. and 6 a.m. accounts for more than 10 percent of daily delay. In seven cities, the delay is 20 percent or more. While these may be the product of a lot of traffic flowing at only slightly slower speeds, they do point to a source of delay that could have operational treatment solutions rather than capacity additions.

• A portion of the delay in all periods is due to speeds between 50 mph and 60 mph. This seems to be particularly significant in the overnight period.

Exhibit 5-10. 2002 Delay Summary by Time of Day Delay by Time of Day (Percent)

City Early Morning

(12 to 6 am) AM Peak

(6 to 9 am) Midday

(9 am to 4 pm) PM Peak

(4 to 7 pm) Late Evening

(7 pm to 12 am) Albany, NY 1 20 23 49 7 Atlanta, GA 12 20 21 37 10 Austin, TX 4 21 30 33 12 Charlotte, NC 8 19 32 32 9 Cincinnati, OH/KY 12 16 34 24 14 Detroit, MI 2 28 18 49 3 Hampton Roads, VA 20 14 29 16 21 Houston, TX 1 28 26 41 4 Los Angeles, CA 3 26 32 33 6 Louisville, OH/KY 27 12 24 15 22 Milwaukee, WI 3 29 28 32 8 Minneapolis-St. Paul, MN 3 25 27 42 3 Northern Virginia 10 21 27 28 14 Orlando, FL 10 17 32 28 13 Philadelphia, PA 4 27 30 32 7 Phoenix, AZ 5 24 28 35 8 Pittsburgh, PA 6 30 31 25 8 Portland, OR 6 19 34 33 8 Sacramento, CA 1 21 20 57 1 Salt Lake City, UT 23 13 18 17 29 San Antonio, TX 2 28 21 42 7 San Diego, CA 1 26 24 47 2 Seattle 1 23 37 36 3 Note: See website for more details: http://mobility.tamu.edu/mmp

44

• Exhibit 5-11 is an effective method of illustrating the share of delay over time of day.

Early AM Off-Peak

(12a-6a)0%

Midday Off-Peak

(9a-4p)22%

Late PM Off-Peak

(7p-12a)1%

PM Peak Period(4p-7p)47%

AM Peak Period(6a-9a)

30%

Exhibit 5-11. Delay by Time of Day (Example)


45

• Exhibit 5-12 indicates weekend delay may be equal to the delay for one weekday in about half of the cities. The difference in data collection devices may be part of this, as indicated by the low Houston weekend delay. Although sweeping conclusions should be avoided, the weekend delay “problem” may be a subject for study, particularly in certain corridors.

• Monday delay is typically less than other weekdays; there is no city where Monday delay is highest.

• Friday is typically the days with the most delay. Traffic volumes are often similar to other weekdays but the evening peak is more congested as fewer people stay late at work.

Exhibit 5-12. 2002 Delay Summary by Day of Week

Delay by Day of Week (Percent) City Monday Tuesday Wednesday Thursday Friday Saturday Sunday

Albany, NY 17 14 17 18 22 6 5 Atlanta, GA 14 17 17 19 20 7 6 Austin, TX 15 18 17 17 17 9 7 Charlotte, NC 14 17 17 20 21 6 5 Cincinnati, OH/KY 14 17 18 18 18 8 7 Detroit, MI 17 17 19 20 21 4 2 Hampton Roads, VA 14 14 14 14 14 15 15 Houston, TX 15 18 18 19 22 6 2 Los Angeles, CA 14 16 17 19 20 9 5 Louisville, OH/KY 13 16 18 17 14 12 10 Milwaukee, WI 15 17 18 19 21 6 4 Minneapolis-St. Paul, MN 14 18 21 22 20 3 2 Northern Virginia 13 15 16 18 19 10 9 Orlando, FL 13 13 14 16 21 14 9 Philadelphia, PA 13 16 18 19 21 8 5 Phoenix, AZ 15 18 18 19 17 7 6 Pittsburgh, PA 14 17 18 20 19 7 5 Portland, OR 13 17 17 19 19 9 6 Sacramento, CA 12 18 19 23 23 4 1 Salt Lake City, UT 12 15 14 12 12 15 20 San Antonio, TX 16 18 17 18 20 6 5 San Diego, CA 13 17 18 21 24 5 2 Seattle, WA 13 18 19 20 21 6 3 Note: See website for more details: http://mobility.tamu.edu/mmp

46

• Exhibit 5-13 is an effective picture of delay distribution during the week.

Sunday2% Monday

15%

Tuesday18%

Wednesday20%

Thursday20%

Friday21%

Saturday4%

Exhibit 5-13. Delay by Day of Week (Example)


47

• Exhibit 5-14 can be used to illustrate the daily variation in the measures. It is particularly useful in identifying seasonal variations and “spike” days of unusually bad or good conditions. They identify special events, weather problems, and other irregular occurrences—some events can be planned for and others can only be dealt with. Because of the day-to-day fluctuations common in this chart, the research team added lines for monthly averages to better illustrate the monthly or season trend. It appears that these monthly trend lines provide a more easily interpretable indication of mobility trends throughout the year.

CORRIDOR OBSERVATIONS Some of the instrumented corridors illustrate issues that have been measured by travel time data collection in the past, but rarely as completely as is possible with full-time data collection abilities.

• In general, the most congested freeway sections are also the least reliable. • The most congested individual time periods and days are weekday peak periods, but

many of the least reliable sections are midday or late evenings.

1.001.10

1.201.30

1.401.50

1.601.701.80

1.902.00

2.102.20

2.302.40

2.502.60

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Day of Year (weekdays, non-holidays only)

Inde

x V

alu

e

Travel Time Planning Time Monthly Travel Time Monthly Planning Time

Figure 5-14. Mobility and Reliability Measures by Day of the Year (Example)


48

HOV lanes that are instrumented separately (some are included in the adjacent freeway data) show that they are much more reliable and have very low (i.e., desirable) travel time index values.

• Toll highways are more reliable and have lower travel time index values.

• If many miles of roadway are instrumented in a city, the presentation of the data becomes cumbersome. There can be many sections of roadway with many corresponding figures and charts. The information requires organization and highlights to point the readers to important elements. However, local analysts will be most interested in performance measures at the facility level or lower. Local elected officials and media may show more interest in an areawide measure.

• Most directional roadways in the study have a single peak. There are “double-peak” corridors, but many of the congested sections show very short periods of off-peak direction congestion. If these data are true, it could be the cause of overestimates of delay in procedures that assume an equal directional distribution.

• Exhibit 5-15 compares the percentage delay values for each corridor to the share of system capacity (measured in percent of lane-miles) and system travel (measured in percent of vehicle-miles traveled) in each city. (These values are only for the instrumented sections of freeway). The sections most in need of attention are those with percent delay values much higher than percent roadway or travel.

• Freeway section speeds can be displayed in a format similar to a topographic contour map. The speed ranges in Exhibit 5-16 show where and when problems occur in an easily understood graphical format. These plots can also be useful in identifying or confirming detector problems. Locations where speeds remain in the low speed ranges all day are probably the result of problem detectors.

Exhibit 5-15. Traffic Speed Contour Map

Other(85 mi)

8%

I-610 North Loop(9 mi)3%

US 59 Eastex(20 mi)

6%

US 290 Northwest(17 mi)

8%

I-45 Gulf(24 mi)10%

I-45 North(23 mi)11%

US 59 Southwest(16 mi)13%

I-610 West Loop(11 mi)14%

I-10 Katy(20 mi)27%

Exhibit 5-15. Delay by Roadway (Example)


49

CONGESTED TRAVEL Examining the number of days when travel falls below a target speed, and the amount of travel within the system that is lower than the target provides more detail on the system and traveler experiences. Exhibit 5-17 illustrates one approach to this analysis. The graph illustrates how conditions vary across the average day—as seen in the VMT-related lines—as well as how often conditions fall below a target speed—in this case the percentage of sensor locations and days with a travel time index value greater than 1.00 for each 5-minute time slice.

0:00

1:00

2:00

3:00

4:00

5:00

6:00

7:00

8:00

9:00

10:0

0

11:0

0

12:0

0

13:0

0

14:0

0

15:0

0

16:0

0

17:0

0

18:0

0

19:0

0

20:0

0

21:0

0

22:0

0

23:0

0

Fathom S

Technology

Pavilion N

Tweed Court

Thunder Cr Rd

Angus Rd S

Quarry Lk Pkwy

Braker Ln S

Great Hills S

Mopac North

Metric Blvd

Lazy Lane

Guadalupe St

Carver Ave

Time of Day

Loca

tio

n

2001 AVERAGE WEEKDAY SPEEDS (mph)

US 183 SB: Spicewood Springs Rd. to IH-35Austin, Texas

Average Speed Ranges (mph)

0 to 20 20 to 30 30 to 40 40 to 50 60+50 to 60

Exhibit 5-16. Traffic Speed Contour Map

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

12 AM 2 AM 4 AM 6 AM 8 AM 10 AM 12 PM 2 PM 4 PM 6 PM 8 PM 10 PM 12 AM

Time of Day (weekdays, non-holidays only)

Perc

enta

ge

Days below 60 mph VMT below 50 mph VMT below 60 mph

Exhibit 5-17. Average Weekday Speed Variation

50

The graph is fairly typical of the experiences in the three years of study. There is a significant amount of travel between 50 mph and 60 mph which does not significantly add to person-hours of delay, but pushes the apparent congestion percentage fairly high if a 60 mph target is used. The 50 mph line more closely mirrors the other performance measures, such as Travel Time Index or Buffer Index. There is a small increase in apparent congestion around 4 a.m., presumably due to slower nighttime driving. As the commuting population begins to travel, the percentage of VMT less than 50 mph decreases, until “real” congestion begins to slow travel again around 6 a.m. MMP, TRIPS AND CENSUS INFORMATION There are several techniques for compiling travel condition performance measures. In some situations, these techniques have an effect on the information and conclusions that can be derived. When interpreting the performance measures in the Mobility Monitoring report and other documents, it is useful to understand what is being measured, as well as what might be left out. MMP database is compiled from roadway sensors and quantify the operating condition at points along the roadway. Additional studies have been conducted to assess the usefulness of the point data for other uses. The type of statistics produced in this report are developed as compilations of the point data into corridor or system averages. The experiments conducted in the Mobility Monitoring Program show no difference in results between these relatively simple procedures, and more complicated combinations of archived data and computerized simulations. The complicated techniques would seek to replicate the travel path of vehicles along the road. Since a vehicle traveling long distances would occupy sections of the road during different 15-minute time periods, the computer simulations “start” trips through the roadway data at different times. The simulated experiences are used as the data for the analysis. The Mobility Monitoring Program experiments indicate these “trip simulations” are not needed for relatively accurate pictures of the situation, particularly at the peak period and corridor section levels of detail. U.S. Census data also provide a view of travel conditions on work trips. The information is gathered from survey reports of typical travel time. These reported values can be correlated with the archived data as one comparison, or as a way to calibrate the two sets of information. Until there is more experience with the archived data, however, it will be difficult to understand how the information in the two datasets should relate.

51

HAPTER 6 - THE FUTURE: ADDITIONAL OPPORTUNITIES

The third year of the Mobility Monitoring Program has extended the experience base and continues to show that archived operations data can greatly benefit performance monitoring. As the Program moves forward, a number of opportunities present themselves for improving and expanding the concepts demonstrated in previous years.

VALIDATION OF TRAVEL TIMES FROM MULTIPLE SOURCES The data from the 23 cities participating in the second year were generated primarily from roadway surveillance equipment that collects volumes and speeds at spot locations. Several issues associated with this form of data exist, and should be examined.

• A simple technique was used to extrapolate spot speeds to link travel times. The accuracy of these estimated travel times (as compared to probe vehicle travel times) is unknown.

• A variety of technologies are being used to collect spot speeds including single- and double-inductance loops, radar, passive acoustic, and video image processing. Tests by Minnesota DOT have shown that the technologies can produce comparable results, although testing continues and should be monitored. A specific concern of the Project Team is that speeds estimated from single inductance loops are significantly different from those that measure speeds directly. As agencies adopt the next generation of technologies this issue may take care of itself, but in the short-term it remains a concern.

• Because the Program relies solely on freeway detection, the results are viewed from a facility perspective. Of at least equal importance is the user perspective, i.e., how trips taken by travelers (from origin to destination) are affected by congestion. Areawide estimates of mobility may differ if measures are built up from trips rather than from facilities. Although both views are important—facility performance for operators and trip performance for travelers—it is important to know the relationship between the two approaches of measurement and whether the more easily obtained facility information can be used for both purposes.

• Comparison of the empirical results from the first three years of the Program with other analytic methods will be enlightening. The state-of-the-practice in performance monitoring is currently dominated by analytic methods such as the Highway Capacity Manual and the Highway Performance Monitoring System. How these methods compare to the results produced by this study—at both the corridor and areawide levels—is not known and should be tested.

EXPANSION OF THE PROGRAM TO INCLUDE SIGNALIZED ARTERIALS A significant portion of urban travel occurs on signalized highways and should be included in mobility estimates. However, estimating travel times on arterials using existing technologies is

C

52

problematic. Spot speeds are usually taken at mid-block locations but most of the delay occurs at the intersections (mid-block speeds are likely to be free-flow unless queues are excessive). Probe-based systems are clearly superior, but are not very common. Some combination of mid-block detection coupled with computer simulation could prove useful. In such an approach, the mid-block volumes are used as demand inputs to simulating the performance of a signal for very short time intervals (e.g., 1- to 5-minutes). This approach also requires details on signal operation: phasing and turning movements. If these data were available at the same time, results would be more accurate than if defaults were used. This is particularly the case where advanced signal control strategies are used to adjust phasing in real-time. MORE SOPHISTICATED QUALITY CONTROL PROCEDURES Common practices for examining the quality of archived operations data are still relatively unsophisticated and much work remains to be done. On-going research and practice should be investigated for their applicability nationwide. The advanced data quality checks that should be investigated include:

• Sequential Data Checks – will compare values in consecutive time periods for consistency (e.g., speeds cannot go from 60 mph to 20 mph and back to 60 mph in consecutive 5-minute time periods.

• Corridor Data Checks – will examine the relationship between data along a corridor (e.g., volume into an area should approximately equal volume out).

• Historical Data Checks – will examine the changes from one year to the next for reasonableness (e.g., high increases in volume or drastic changes in speeds).

Data quality checks are only the first step in the QC process—once suspicious or erroneous data are detected, an action must be taken. Possible actions include simply flagging or replacing the data. Methods for replacing QC-failed data, as well as for “imputing” missing data, offer the chance to improve data completeness. Such methods would be based on “good” data from surrounding locations for the same time period as well as using historical data. ANALYSES TAILORED TO LOCAL AREAS The field visits with state and local personnel revealed a mixed interest in performance monitoring. However, it was apparent that the local view of performance monitoring has a different focus than that of FHWA. Specifically, state and local personnel are more concerned with the geographic detail of mobility. Planners and operators both expressed this need, although their interests are at slightly different time and spatial scales: operators from the perspective of “what happened at a specific bottleneck yesterday and what can we expect today” and planners from the perspective of “how have travel trends in extended corridors changed over long periods of time”. In spite of their interest, however, the ability to perform these analyses on very large datasets is not common. An emerging trend appears to be the lead involvement of universities in data archiving activities. In several locations, state universities have been assigned the official role of maintaining a data archive for a region or state. The assignment of this role to a research group seems to indicate

53

that some agencies believe that further development of the data archive concept is necessary before being implemented in a DOT-based production environment. In some places, the universities’ research group is closely aligned with the goals and mission of the sponsoring state DOT and a university-based archive could prove to be an effective model. In many locations; however, university and state goals may differ slightly (universities exhibit more intellectual curiosity whereas state DOTs are focused on repetitive production) and a university-based archive may never become fully integrated into the DOT. As an example, Caltrans and the University of California-Berkeley have collaborated on the development of their Freeway Performance Measurement System (PeMS), which is currently housed at the University. According to the PeMS team, once a version of PeMS has proven to be reliable and stable, it will be implemented within the Caltrans network and organization. Thus the PeMS team plans to maintain a research-based PeMS at the University, which will test new features and designs. Once these new features are proven, they will be implemented with new releases in the Caltrans-based PeMS.

If local agencies are to take full advantage of archived operations data, additional resources will be needed for maintaining and analyzing archived operations data. The website contains information on the individual city reports developed for this study. (For more details see: http://mobility.tamu.edu/mmp). CONGESTION CAUSES The measures developed so far provide an overall picture of mobility. However, to be more useful for implementing operations strategies, the causes of congestion should be tracked. In other words, what factors (“events”) have contributed to overall mobility and what are their magnitude; factors include incidents, weather, work zones, changes in traffic demand and recurring bottlenecks. Ideally, the share of total congestion attributable to these sources would be estimated, which would allow strategies to be targeted at the root causes. Identifying the events that are restricting mobility is important at both the national level (development of overall programs) and the local level (development of specific actions). A research plan has been prepared for the Future Strategic Highway Research Program to address a variety of reliability related issues (6). It defines seven sources of travel time variability that should be adapted for any “cause” database effort.

• Incidents—collisions and vehicle breakdowns.

• Work Zones—both construction and maintenance activities.

• Weather—any environmental condition that changes driver behavior.

• Fluctuations in Demand—daily and seasonal variations.

• Special Events—exceptional demand or demand pattern changes.

• Traffic Control Devices—intermittent disruptions or poor signal timing.

54

• Inadequate Basic Capacity—the interaction of roadway limitations with the other six causes.

The first step in this process is to construct a comprehensive database that contains not only roadway surveillance data but also data on the external or causal factors. The experience of the Project Team has been that the archiving of external factors, such as incidents, is sporadic and even less standardized than roadway surveillance data. Once data have been archived, research is needed to link the surveillance data with the external factors. For example, delays in a corridor can be attributed to incidents on one day, weather on another, and high demand on another. CONTINUE TO EXPERIMENT WITH MEASURES The measures used in this report are useful and many have been presented to general audiences through other reports. They are not the only measures and while local agencies will experiment with their own measures, the national study should also investigate other measures. The range of uses, from real-time information to long-term planning will mean that a variety of measures will always be appropriate.

ENCOURAGE THE DEVELOPMENT OF STANDARDIZED PROCEDURES FOR DATA ARCHIVING Although it is apparent that many TMCs are now archiving data, the Project Team found considerable differences in how the archiving is performed. Although accounting for the differences can be done, it takes considerable effort to do so. As the number of participating cities grows, this effort will become nontrivial. Beyond the ease of analysis, a more important consideration is that standardized procedures for collecting and (especially) managing the data will allow more meaningful comparisons across cities. Standards for archived data will also promote use of the data among local agencies and the private sector, such as Advanced Traveler Information Services applications (e.g., historical patterns for short-term travel time prediction) and software vendors (e.g., TMC system integrators). Finally, if local processing and reporting of the data is the long-term goal of the Mobility Monitoring Program, then standards are necessary to ensure consistent results. Specifically, the areas where standardization would improve analysis and use of the data by local agencies are:

• File Formats. Individual file extraction and input procedures for each city must now be made. A common file structure and file storage/compression formats would greatly promote analysis.

• Aggregation Procedures. Data are currently submitted at various levels of time and spatial aggregation; a common aggregation definition would also ease the analysis burden. Also, internal procedures at the TMCs differ in how aggregation is performed. Treatment of missing values and the computation of average speeds are two such procedures that if standardized would allow more direct comparisons to be made.

55

• Quality Control. The degree of quality control varies substantially across the 23 cities. Application of different thresholds by the TMCs result in slightly inconsistent data. Standardized QC procedures would improve this situation and also would help TMCs get more closely acquainted with the details of their data.

• Metadata/Meta-Attributes. Documentation on how the data were collected and processed would allow analysts to determine the usefulness and accuracy of the data to a higher degree than now possible. One example would be documenting the number of observations that comprise an aggregated record; some of the systems supplying data for this study provide this function, but others do not.

LONG-TERM STRUCTURE OF THE MOBILITY MONITORING PROGRAM The long-term success of the Mobility Monitoring Program hinges on strong local involvement. The current process is based on the Project Team obtaining and processing the data for each area. This structure is necessary in the beginning to identify and resolve the many technical and institutional issues that have been uncovered. However, as the number of participating cities grows in future years, the amount of data processing needed to support the program will be substantial and has large cost implications. Further, local use of the data should be encouraged for quality purposes—problems can be quickly identified and fixed if local areas are actively engaged in applying the data to local applications. Therefore, the future structure of the Program should evolve toward more local control, with the Federal reports being just one of many uses of the data by local agencies. Standards and technical assistance are needed to support the transition to local control.

56

57

EFERENCES

1. Lomax, T.L., S.M. Turner, and R. Margiotta. Monitoring Urban Roadways in 2000: Using

Archived Operations Data for Reliability and Mobility Measurement. Report No. FHWA-OP-02-029, Federal Highway Administration, December 2001. Available: http://mobility.tamu.edu/mmp.

2. Lomax, Tim, Shawn Turner and Richard Margiotta. Texas Transportation Institute and

Cambridge Systematics, Inc. Monitoring Urban Roadways in 2001: Examining Reliability and Mobility with Archived Data. Report No. FHWA-OP-03-141, Federal Highway Administration. June 2003. Available: http://mobility.tamu.edu/mmp.

3. Wunderlich, K.E., Hardy, M.H., Larkin, J.J. and Shah, V.P. On-Time Reliability Impacts of

Advanced Traveler Information Services (ATIS): Washington, DC Case Study. Mitretek Systems, January 2001.

4. Turner, S.M., W.L. Eisele, B.J. Gajewski, L.P. Albert, and R.J. Benz. ITS Data Archiving: Case Study Analyses of San Antonio TransGuide® Data. Report No. FHWA-PL-99-024. Federal Highway Administration, Texas Transportation Institute, August 1999.

5. 2002 Urban Mobility Report, Texas Transportation Institute, June 2002. Available:

http://mobility.tamu.edu/ums.

6. Cambridge Systematics, et al., Providing a Highway System with Reliable Travel Times, Research Plan Prepared under Project 20-58(3), for the National Cooperative Highway Research Program, 2003.

R

To access an electronic version of this publication, visit: http://mobility.tamu.edu/mmp

FHWA web address: http://ops.fhwa.dot.gov

Toll-Free “Help Line” 866-367-7487 Or you can send e-mail to: [email protected]

Publication No.: FHWA-HOP-04-011

Documents

Technical Report Documentation Page 1. Report No. 2 ... · Document is available to the public through the National Technical Information Service, Springfield, Virginia 22161 19