<%BANNER%>

Exploiting the Operational and Research Potential of Archived Transportation Management Data

Permanent Link: http://ufdc.ufl.edu/UFE0041179/00001

Material Information

Title: Exploiting the Operational and Research Potential of Archived Transportation Management Data
Physical Description: 1 online resource (197 p.)
Language: english
Creator: Lee, Seok
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: florida, steward
Civil and Coastal Engineering -- Dissertations, Academic -- UF
Genre: Civil Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: This dissertation focuses on the challenges involved in the development of a central data warehouse that archives data from transportation management centers in Florida and on the use of the archived data for operational and research purposes. The first challenge was the development of a system to meet requirements set forth by its stakeholders. The components of the system are described in this document. The next challenge was to ensure that the system performs useful functions. This challenge was addressed by applying the principles of traffic flow theory to the analysis of archived data to reconstruct the operation of a facility in terms of performance measures, several of which are described and demonstrated using actual archived data. It was demonstrated that the archive data characteristics are consistent with the principles of traffic flow theory. Relationships between the macroscopic descriptors of traffic flow produced good agreement with those found in the literature, and with the empirical data presented in the Highway Capacity Manual. Specific examples are provided for traffic count extraction, travel time reliability reporting, incident analysis and the analysis of managed lanes. Additional quality control tests based on station and system level analysis were developed to supplement the set of individual lane detector tests found in the literature. The project described in this dissertation has created the Statewide Traffic Engineering Warehouse for Archived Regional Data (STEWARD), which provides an important resource for a wide variety of traffic data users in Florida, including both practitioners and researchers. The web site described in this dissertation provides the capability to download several reports summarized over a range of temporal and spatial requirements. The data can serve a variety of operational, administrative and research purposes.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Seok Lee.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Yin, Yafeng.
Local: Co-adviser: Courage, Ken G.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-06-30

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0041179:00001

Permanent Link: http://ufdc.ufl.edu/UFE0041179/00001

Material Information

Title: Exploiting the Operational and Research Potential of Archived Transportation Management Data
Physical Description: 1 online resource (197 p.)
Language: english
Creator: Lee, Seok
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: florida, steward
Civil and Coastal Engineering -- Dissertations, Academic -- UF
Genre: Civil Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: This dissertation focuses on the challenges involved in the development of a central data warehouse that archives data from transportation management centers in Florida and on the use of the archived data for operational and research purposes. The first challenge was the development of a system to meet requirements set forth by its stakeholders. The components of the system are described in this document. The next challenge was to ensure that the system performs useful functions. This challenge was addressed by applying the principles of traffic flow theory to the analysis of archived data to reconstruct the operation of a facility in terms of performance measures, several of which are described and demonstrated using actual archived data. It was demonstrated that the archive data characteristics are consistent with the principles of traffic flow theory. Relationships between the macroscopic descriptors of traffic flow produced good agreement with those found in the literature, and with the empirical data presented in the Highway Capacity Manual. Specific examples are provided for traffic count extraction, travel time reliability reporting, incident analysis and the analysis of managed lanes. Additional quality control tests based on station and system level analysis were developed to supplement the set of individual lane detector tests found in the literature. The project described in this dissertation has created the Statewide Traffic Engineering Warehouse for Archived Regional Data (STEWARD), which provides an important resource for a wide variety of traffic data users in Florida, including both practitioners and researchers. The web site described in this dissertation provides the capability to download several reports summarized over a range of temporal and spatial requirements. The data can serve a variety of operational, administrative and research purposes.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Seok Lee.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Yin, Yafeng.
Local: Co-adviser: Courage, Ken G.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-06-30

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0041179:00001


This item has the following downloads:


Full Text

PAGE 1

1 EXPLOITING THE OPERATIONAL AND RESEARCH POTENTIAL OF ARCHIVED TRANSPORTATION MANAGEMENT DATA By SEOKJOO LEE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIRE MENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009

PAGE 2

2 2009 Seokjoo Lee

PAGE 3

3 To my wife Jinam Kim

PAGE 4

4 ACKNOWLEDGMENTS I would like to thank my advisor, Professor Ken G. Courage for his e ncouragement, guidance and knowledge. W ithout that, I should not have finished my study here. I would like to thank Dr. Yafeng Yin, Dr. Lily Elefteriadou, Dr. Scott S. Washburn, and Dr. Ruth L. Steiner for their invaluab le comments

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 9 LIST OF FIGURES ................................ ................................ ................................ ........ 11 LIST OF ABBREVIATIONS ................................ ................................ ........................... 14 CHAPTER 1 INTRODUCTION AND SUMMARY ................................ ................................ ........ 19 Sta tement of the Problem ................................ ................................ ....................... 19 Project Objectives ................................ ................................ ................................ ... 20 Summary of Project Tasks ................................ ................................ ...................... 20 Literature Review ................................ ................................ ............................. 20 Analysis of the Archived Traffic Data ................................ ................................ 20 Data Management System Development ................................ ......................... 20 Statewide Deployment ................................ ................................ ..................... 21 Quality Assurance ................................ ................................ ............................ 21 Organization of the Dissertation ................................ ................................ .............. 21 2 LITERATURE REVIEW ................................ ................................ .......................... 23 Traffic Data Warehouses in Other States ................................ ............................... 23 California PeMS System ................................ ................................ ................ 24 System design ................................ ................................ ........................... 24 Data quality control ................................ ................................ .................... 25 Performance measures ................................ ................................ .............. 26 Oregon PORTAL System ................................ ................................ .............. 26 System design ................................ ................................ ........................... 27 Data quality control ................................ ................................ .................... 27 Performance measures ................................ ................................ .............. 28 Maryland CHART System ................................ ................................ .............. 28 System design ................................ ................................ ........................... 28 CHART applications ................................ ................................ ................... 28 Summary of Traffic Archive Systems ................................ ............................... 29 The SunGuide Traffic Management System in Florida ................................ ........... 30 Operational Features ................................ ................................ ........................ 30 Archive Functions ................................ ................................ ............................. 31 Quality Control of Traffic Data ................................ ................................ ................. 31 3 FREEWAY TRAFFIC FLOW PRINCIPLES AND ARCHIVED DATA ...................... 38

PAGE 6

6 Speed, Flow Rate, and Density Relationships ................................ ........................ 38 The Fundamental Diagram ................................ ................................ ............... 39 Other Speed Flow Density Models ................................ ................................ ... 40 Highway Capacity Manual Treatment of Speed, Flow and Density .................. 41 Level of Service (LOS) ................................ ................................ ............................ 43 Platoon Propagation ................................ ................................ ............................... 43 Maximum Flow Rates ................................ ................................ ............................. 44 Effective Vehicle Lengths ................................ ................................ ........................ 44 Lane Volume Balance Ratio ................................ ................................ ................... 44 I/O Volume Balance ................................ ................................ ................................ 44 4 CENTRAL DATA WAREHOUSE REQUI REMENTS ................................ .............. 49 Functional Requirements ................................ ................................ ........................ 49 Raw SunGuide Archive Data ................................ ................................ ............ 49 TMC Configuration Requirements ................................ ................................ .... 50 The station data spreadsheet ................................ ................................ ..... 50 The Lane Data Spreadsheet ................................ ................................ ............ 51 Steps in configuring the TSS data ................................ .............................. 52 ETL requirements ................................ ................................ ...................... 53 Data Transfer Automation Requirements ................................ ................................ 54 Reporting Requirements ................................ ................................ ......................... 55 5 CENTRAL DATA WAREHOUSE DEVELOPMENT ................................ ................ 57 STEWARD System Description ................................ ................................ .............. 57 System Overview ................................ ................................ ............................. 57 Database Design and Architecture ................................ ................................ ... 58 Data Flow ................................ ................................ ................................ ......... 58 STEWARD Operation ................................ ................................ ............................. 59 Oracle Database Program Installation ................................ .............................. 59 STEWARD Deployment ................................ ................................ ................... 60 STEWARD Web Installation ................................ ................................ ............. 60 STEWARD Management ................................ ................................ .................. 60 Internet Access Features ................................ ................................ ........................ 61 Overview of the STEWARD Web Interface ................................ ...................... 61 STEWARD Web Ar chitecture ................................ ................................ ........... 61 Maps ................................ ................................ ................................ ................ 62 Report Levels ................................ ................................ ................................ ... 62 Report Types ................................ ................................ ................................ .... 63 ETL Operations ................................ ................................ ................................ ....... 63 Current Status of the System ................................ ................................ .................. 64 6 DEVELOPMENT OF QUALIT Y ASSURANCE PROCEDURE ............................... 75 Level 1 Completeness Test ................................ ................................ .................... 76 Availability of the District Data ................................ ................................ .......... 76

PAGE 7

7 Missing Detector Data ................................ ................................ ...................... 77 All Zero or Stuck Detectors ................................ ................................ ............... 78 Duplicate or Negative scan Data ................................ ................................ ...... 79 Level 1 Test Summary ................................ ................................ ...................... 80 Level 2 Data Validity Test ................................ ................................ ....................... 80 Maximum Volume Test ................................ ................................ ..................... 80 Maximum Occupancy Test ................................ ................................ ............... 81 Minimum and Maximum Speed Tests ................................ .............................. 81 Multivariate Consistency Test ................................ ................................ ........... 81 Truncated Occupancy Values Test ................................ ................................ .. 82 Maximum Estimated Density ................................ ................................ ............ 82 Summary of Level 2 Tests ................................ ................................ ................ 82 Level 3 Station Data Validation ................................ ................................ ............... 83 Maximum Flow Rates ................................ ................................ ....................... 84 EVL Characteristics ................................ ................................ .......................... 85 Lane Volume Balance Ratio ................................ ................................ ............. 88 Daily Volume Variation ................................ ................................ ..................... 90 Annual Volume Variation ................................ ................................ .................. 91 Level 4: System Level Tests ................................ ................................ ................... 92 Co ntinuity of EVL ................................ ................................ .............................. 92 Continuity of Volume ................................ ................................ ........................ 93 7 OPERATIONAL APPLICATIONS FOR ARCHIVED DATA ................................ ... 112 Summary of Available Reports ................................ ................................ ............. 112 Diagnostic Procedures ................................ ................................ ................... 112 Station Level Reports ................................ ................................ ..................... 113 Section Level Reports ................................ ................................ .................... 113 Facility Level Reports ................................ ................................ ..................... 114 Traffic Volume Data for Tr affic Counting Programs ................................ .............. 114 Converting ITS Data to FDOT Counts ................................ ............................ 115 Reporting of Traffic Volume Trends ................................ ................................ 115 Integration with Statewide Crash Data Records ................................ ................... 115 Integration with the Roadway Characteristics Inventory ................................ ....... 117 General Support for Periodic Reporting Requirements ................................ ......... 118 Diagnostic Support for TMC Detector Operation and Maintenance ...................... 118 Other Applications for the STEWARD Reports ................................ ..................... 119 Work Zone Crash Analysis ................................ ................................ ............. 119 Support for Identification of Recur ring Congestion ................................ ......... 120 Travel Time Reliability Reporting ................................ ................................ .... 120 8 RESEARCH APPLICATIONS FOR ARCHIVED DATA ................................ ........ 140 Analysis of Breakdown at a Freeway Ramp ................................ ......................... 140 Simulation Support for SunGuide ................................ ................................ ......... 141 Current Managed Lane Applications ................................ .............................. 142 Other TRC Research Applications ................................ ................................ ........ 142

PAGE 8

8 Modeling the Location of Crashes within Work Zones ................................ .... 142 A Case Study in Spatial Misclassification of Work Zone Crashes .................. 143 Analyzing the Effectiveness of Enhanced Penalty Zones and Police Enf orcement as Freeway Speed Control Measures ................................ .... 143 Capacity of Florida Freeways, FDOT Project BDK 75 977 08 ....................... 143 Travel Time Reliabil ity ................................ ................................ .................... 144 Freeway Work Zone Capacity ................................ ................................ ........ 144 NCHRP 3 Flow Breakdown ................................ ................................ .......................... 144 Doctoral Dissertation Project: ................................ ................................ ......... 144 Other STEWARD Users ................................ ................................ ........................ 145 9 DATA AN ALYSIS EXAMPLES ................................ ................................ ............. 146 Speed Flow Density Relationships ................................ ................................ ....... 146 Speed ................................ ................................ ................................ ............. 147 Density ................................ ................................ ................................ ........... 147 Flow Rate ................................ ................................ ................................ ....... 149 Examples of Relationships ................................ ................................ ............. 149 Crash and Incident Analysis Applications ................................ ............................. 152 Overall Crash Characteristics ................................ ................................ ......... 152 Sample Crash Analysis ................................ ................................ .................. 153 Hourly flow rates ................................ ................................ ...................... 154 Occupancy ................................ ................................ ............................... 155 Speed ................................ ................................ ................................ ...... 156 Estimation of delay from the travel time reliability report .......................... 157 Comparison of delay estimation methods ................................ ................ 158 Lan e volume balance ratio ................................ ................................ ....... 158 Speed variance ................................ ................................ ........................ 159 Kinetic energy ................................ ................................ .......................... 160 Analysis of Managed Lanes ................................ ................................ .................. 160 Managed Lane Performance Measures ................................ ......................... 161 Example Results ................................ ................................ ............................ 164 Travel Time Reliability ................................ ................................ ........................... 166 10 CONCLUSIONS AND RECOMMENDATIONS ................................ ..................... 189 Conclusions ................................ ................................ ................................ .......... 189 Recommendations ................................ ................................ ................................ 192 LIST OF REFERENCES ................................ ................................ ............................. 195 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 197

PAGE 9

9 LIST OF TABLES Table page 2 1 Evaluation criteria for the PeMS loop detector data ................................ ............ 33 2 2 PeMS traffic data quali ty statistics ................................ ................................ ...... 33 2 3 Evaluation criteria for the PORTAL detector data ................................ ............... 33 2 4 Comparison of traffic data warehouses in other stat es ................................ ....... 34 2 5 Basic rules of the QC criteria ................................ ................................ .............. 34 2 6 Quality control criteria ................................ ................................ ......................... 35 3 1 Level of service thresholds (Source Exhibit 11 5 of the 2010 HCM) ................... 45 5 1 Status of STEWARD facilities and stations ................................ ........................ 66 5 2 TSS data availability in STEWARD ................................ ................................ .... 66 5 3 STEWARD web statistics for June July, and August, 2009 .............................. 66 6 1 Availability of the D istrict 2,4, and 6 data (7/1/08~12/31/08) ............................... 95 6 2 All zero or stuck data for the D istrict 2,4, and 6 d ata (7/1/08~12/31/08) ............. 95 6 3 Duplica te or negative scan records in D istrict 2, 4, and 6 (10/1/08~10/30/08) .... 95 6 4 Summary of Lev el 1 test ................................ ................................ ..................... 95 6 5 Maximum volume test for the D istrict 2,4, and 6 data (7/1/08~12/31/08) ............ 95 6 6 Maximum occupancy test for the D i strict 2,4, and 6 data (7/1/08~12/31/08) ...... 95 6 7 Speed test for the D istrict 2,4, and 6 data (7/1/08~12/31/08) ............................. 95 6 8 Multivariat e consistency test for the D istrict 2,4, and 6 data (7/1/08~12/31/08) .. 96 6 9 Truncated occupancy values test for the D istrict 2,4, and 6 data (7/1/08~12/31/08) ................................ ................................ ............................... 96 6 10 Maximum estimated density for the D istrict 2,4, and 6 data (7/1/08~12/31/08) .. 96 6 11 Error code for level 2 Tests ................................ ................................ ................. 96 6 12 Summary of Level 2 test for the D istrict 2,4, and 6 data (7/1/08~12/31/08) ........ 96

PAGE 10

10 6 13 Configuration of four selected station s ................................ ............................... 97 6 14 Maximum flow rate frequencies ................................ ................................ .......... 98 6 15 Histogram of lane volume balance ratio ................................ ............................. 99 6 16 Freque ncy of lane volume balance ratios > 10.0 ................................ ................ 99 6 17 Detector configuration problem ................................ ................................ ......... 100 6 18 Lane configuration problem ................................ ................................ .............. 100 7 1 Column description for the all data fields report ................................ ................ 121 7 2 Column descriptions for the station traffic co unts report ................................ ... 122 7 3 Column description for the maximum flow report ................................ .............. 123 7 4 Column description for the effective vehicle length r eport ................................ 124 7 5 Column descriptions for the performance measures report .............................. 126 7 6 Column descriptions for the performance measures r eport baseline ................ 128 7 7 Sample performance measures report ................................ ............................. 1 29 7 8 Column description for the travel time reliability report ................................ ..... 130 7 9 Column descriptions for the travel time reliability report baseline ..................... 131 7 10 Column descriptions for the frequency table in the travel time reliability report 132 7 11 Sample performance measures report ................................ ............................. 133 7 12 Column descriptions for the all data fie lds report ................................ .............. 134 7 13 Column descriptions for the volume map and i/o balance report ...................... 135 7 14 Column descriptions for the facility traffic counts report ................................ ... 136 9 1 Results for travel time reliability example ................................ ......................... 169

PAGE 11

11 LIST OF FIGURES Figure page 2 1 Demonstration of the Fervor application ................................ ............................. 36 2 2 SunGuide system architecture ( Dellenbeck, 2008 ) ................................ ............. 37 3 1 Fundamental diagram relating speed, flow rate and density .............................. 46 3 2 Determination of parameters from the fundamental diagram ............................. 46 3 3 Current concept of the speed flow density relationship ................................ ...... 47 3 4 Typical speed flow relationship (Source: Exhibit 11 1 of the 2010 HCM) ........... 47 3 5 Speed flow relationships for basic freeway segments (Source: Exhibit 11 6 of the 2010 HCM) ................................ ................................ ................................ ... 48 3 6 Highway Capacity Manual assumed f low density relationship (Source: Exhibit A22 5 of the 2010 HCM) ................................ ................................ ......... 48 4 1 Summary of required functions and data flow ................................ ..................... 56 4 2 Examples of the raw data from the SunGuide TSS archive ................................ 56 5 1 STEWARD system configuration ................................ ................................ ........ 67 5 2 Automated data flow diagram for the SunGuide archive data ............................ 67 5 3 Example screen capture from the installation process ................................ ....... 68 5 4 STEWARD web architecture ................................ ................................ .............. 69 5 5 STEWARD overview Page ................................ ................................ ................. 70 5 6 Example of an interactive satellite photo map ................................ .................... 71 5 7 GIS map example ................................ ................................ ............................... 72 5 8 ETL utility data flow ................................ ................................ ............................ 73 5 9 Sample daily report from the ETL Process ................................ ......................... 74 6 1 Location of four selected station s ................................ ................................ ..... 101 6 2 Flow rate hist ogram from four selected station s ................................ ............... 102 6 3 Cumulative percentage over 2400 veh/ln/h from I 95 stations .......................... 102

PAGE 12

12 6 4 EVL with flow rate s at Station 210471 ................................ .............................. 103 6 5 EVL histogram from four selected station s ................................ ....................... 103 6 6 EVL histogram from all stations ................................ ................................ ........ 104 6 7 Flow occupancy diagram for low EVLs (Station 210192) ................................ 104 6 8 Changes of lane volume balance ratio with flow rates at station 210511 ......... 105 6 9 Lane volume balance ratio histogram from four selected station s .................... 105 6 10 Location of Station 210032 ................................ ................................ ............... 106 6 11 Flow rates of lane1 and lane 3 from station 210451 ................................ ......... 106 6 12 Location of Station 200201 ................................ ................................ ............... 107 6 13 Average flow rates of station 210471 for year of 2008 data ............................. 108 6 14 Average flow rates of station 210231 for year of 2008 data ............................. 108 6 15 Total day volume for year of 2008 at station 210471 ................................ ........ 109 6 16 Total day volume for year of 2008 at station 210531 ................................ ........ 109 6 17 EVLs over mileposts during 10/3/09 morning peak ................................ .......... 110 6 18 EVLs over mileposts over five weekday data ................................ ................... 110 6 19 H ourly station volume over system milepost ................................ ..................... 111 7 1 Example SunGuide TSS and statistics office count comparison ...................... 137 7 2 Overview of the ITSCounts data flow ................................ ............................... 138 7 3 Sample page from the facility count analysis report for District 6 ..................... 139 9 1 Traffic flow rate speed graph at Station 210511 (Oct., 2008 weekdays, morning peak: 7:00 AM~10:00 AM) ................................ ................................ .. 170 9 2 Density flow rates graph at station 210511 (Oct., 2008 weekdays, morning peak: 7:00 AM~10:00 AM) ................................ ................................ ................ 171 9 3 Density speed graph at station 210511 (Oct., 2008 weekdays, morning peak: 7:00 AM~10:00 AM) ................................ ................................ .......................... 172 9 4 Comparison of density computation methods ................................ ................... 173

PAGE 13

13 9 5 Number of crashes by day of the week ................................ ............................ 174 9 6 Number of crashes by month of the year ................................ .......................... 174 9 7 Number of crashes by county milepost ................................ ............................. 175 9 8 Number of crashes by time of day ................................ ................................ .... 175 9 9 Incident location in aerial map ................................ ................................ .......... 176 9 10 Hourly flow rates changes with the incident ................................ ...................... 177 9 11 Hourly flow rates changes without incident ................................ ....................... 178 9 12 Comparison of five minute volume counts with incident and non incident case ................................ ................................ ................................ .................. 179 9 13 Cumulative differences between incident and non incident volume counts ...... 180 9 14 Queuing diagram for the incident ................................ ................................ ...... 181 9 15 Occupancy changes on the incident day ................................ .......................... 182 9 16 Occupancy contour graph with milepost ................................ ........................... 183 9 17 Speed changes at incident ................................ ................................ ............... 184 9 18 Speed contour graph during the incident ................................ .......................... 185 9 19 Time space diagram for lane volume bala nce during the incident .................... 185 9 20 Time space diagram for lane occupancy during the incident ............................ 186 9 21 Changes of speed coefficient of variance (CV) during the incident .................. 187 9 22 Flow rate and kinetic energy at the point of the incident by time of day ............ 188

PAGE 14

14 LIST OF ABBREVIATION S AADT Ann ual Average Daily Traffic ADMS Archived Data Management Subsystem ASP Active Server Pages ATMS advanced traffic management system ATRI American Transportation Research Institute AVL Automatic Vehicle Location Caltrans California Department of Transportation C ARS Crash Analysis Reporting System CCTV Closed Circuit Television CDW Central Data Warehouse CHART Coordinated Highways Action Response Team CHP California Highway Patrol CMS Center for Multimodal Solutions for Congestion Mitigation CSS Cascading Style Sh eets CSV Comma Separated Values CV Coefficient of variation DB Database DMS Dynamic Message Signs ETL Extraction Transformation and Loading ETL Extract, Transform and Load EVL Effective Vehicle Length FDOT Florida Department of Transportation FFS Free Flo w Speed FHWA Federal Highway Administration

PAGE 15

15 FTP File Transfer Protocol HAR Highway Advisory Radios HICOMP Highway Congestion Monitoring Program HCM Highway Capacity Manual HO T High Occupancy Toll HOTTER High Occupancy Toll and high occupancy vehicle Traff ic Evaluation Report HOV High Occupancy Vehicle HTTP Hypertext Transfer Protocol ID I d entification IIS Internet Information Services ITS Intelligent Transportation System LOS Level of Service LVB R Lane Volume Balance Ratio MM Mobility Monitoring MPH Mile p er Hour NCHRP National Cooperative Highway Research Program OWB Oracle Warehouse Builder PORTAL P ortland O regon R egional T ransportation A rchive L isting PVM Price per Vehicle Mile PeMS freeway Performance Measurement System PATH C alifornia P artners for A dva nced T ransit and H ighways QA Quality Assurance RCI Roadway Characteristics Inventory RITIS Regional Integrated Transportation Information System RTMC Regional Traffic Management Center

PAGE 16

16 RTMS Road Traffic Microwave Sensor RWIS Roadway We ather Information Sys tems SIS Strategic Intermodal System SITSA Florida Statewide Intelligent Transportation System Architecture SQL Structured Query Language SR State Road STEWARD Statewide Transportation Engineering Warehouse for Archived Regional Data SWRI Southwest Researc h Institute TASAS Traffic Accident Surveillance and Analysis System TCP Transmission Control Protocol TERL Traffic Engineering Research Laboratory TMC Traffic Management Center TriMet Tri County Metropolitan Transportation District of Oreg on TSS T raffic Se nsor Subsystem TTI Texas Transportation Institute TVT Travel Time Subsystem XML Extensible Markup Language TSS Traffic Sensor Subsystem TVT Travel Time Subsystem UF University of Florida VHT Vehicle Hours of Travel t ime VMS Variable Message Signs VMT /VMT T Vehicle Miles of Travel

PAGE 17

17 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy EXPLOITING THE OPERATIONAL AND RESEARCH POTENTIAL OF ARCH IVED TRANSPORTATION MANAGEMENT DATA By Seokjoo Lee December 200 9 Chair: Yafeng Yin Cochair: Ken Courage Major: Civil Engineering This dissertation focuses on the challenges involved in the development of a central data warehouse that archives data fro m transportation management centers in Florida and on the use of the archived data for operational and research purposes. The first challenge was the development of a system to meet requirements set forth by its stakeholders. The components of the syste m are described in this document. The next challenge was to ensure that the system performs useful functions. This challenge was addressed by applying the principles of traffic flow theory to the analysis of archived data to reconstruct the operation of a facility in terms of performance measures, several of which are described and demonstrated using actual archived data. It was demonstrated that the archive data characteristics are consistent with the principles of traffic flow theory. Relationships betw een the macroscopic descriptors of traffic flow produced good agreement with those found in the literature, and with the empirical data presented in the Highway Capacity Manual.

PAGE 18

18 Specific examples are provided for traffic count extraction, travel time r eliability reporting, incident analysis and the analysis of managed lanes. Additional quality control tests based on station and system level analysis were developed to supplement the set of individual lane detector tests found in the literature. The pro ject described in this dissertation has created the Statewide Traffic Engineering Warehouse for Archived Regional Data (STEWARD), which provides an important resource for a wide variety of traffic data users in Florida, including both practitioners and res earchers. The web site described in this dissertation provides the capability to download several reports summarized over a range of temporal and spatial requirements. The data can serve a variety of operational, administrative and research purposes.

PAGE 19

19 CHAPTER 1 INTRODUCTION AND SUMMARY The potential benefits of maintaining an archive of data produced by transportation management centers (TMCs) is well recognized (Bertini and Makler, 2008) With that in mind, the University of Florida is developing a pro totype central data warehouse (CDW) to demonstrate that data from TMCs around the state can be centrally archived in a practical manner and that a variety of useful reports and other products (Courage and Lee, 2008) This dissertation focuses on the challe nges involved in the development of a CDW and on the use of the archived data for various operational and research purposes. The product of this research is known as the Statewide Traffic Engineering Warehouse for Archived Regional Data (STEWARD). Stateme nt of the Problem and software that was developed specifically for the Florida Department of Transportation. SunGuide includes a rudimentary archive element that creates a daily text file containing the basic data produced by each of its sensors during each reporting interval (usually 20 sec onds). While the data themselves are numerically accurate, the information is not useful until it is organized geographically within the syst em, stored in a database that can be interrogated and presented in the form of useful reports. The problem addressed by this dissertation is the design, implementation and operation of a storage and retrieval system that uses the basic archive files from SunGuide to generate reports that meet the requirements of a wide range of users and to provide researchers with a rich supply of data for various purposes.

PAGE 20

20 Project Objectives The following specific objectives have been formulated to address the problem as described above: Review the literature as it pertains to traffic management and data archiving. Establish the basis for analyzing archived data in terms of freeway traffic flow principles. Design and implement a data management scheme to accommodate the a rchived data. Collect and archive data from participating TMCs throughout Florida. Develop a quality assurance methodology that makes maximum use of the system aspects of the archived data. Identify and explore potential operational and research applicatio ns for the archived data. Summary of Project Tasks The following tasks were carried out in support of the stated objectives: Literature Review The literature was reviewed with respect to data archiving activities in other states, the characteristics of the Florida SunGuide traffic management system, and quality control of traffic data. Analysis of the Archive d Traffic Data A number of analysis tools have been developed to verify the archived data characteristics. Traffic flow principles have also been incor porated into the diverse research applications. Data Management System Development A data management system was designed, established and verified with two years of traffic data warehouse operations.

PAGE 21

21 Statewide Deployment The system was designed to accommo date statewide traffic data. At this point, data from District 2, 4 5, 6 and 7 have been incorporated Quality Assurance Current quality assurance methods focus on individual lane detectors. A systematic approach has been developed to improve the qualit y of traffic data by adding additional data quality tests based on relationships among individual lane data at a detector station and consistency of data between adjacent stations. Organization of the Dissertation There are ten chapter s in this dissertat ion Chapter 1 is the introduction Chapter 2 presents the literature review on traffic data warehouses in other states, the SunGuide system in Florida and traffic data quality. Chapter 3 describes the traffic flow theory and performance measures to be app lied in the development of summary reports Chapter 4 describes the Central Data Warehouse requirements, including functional requirements, data transfer requirements and reporting requirements. Chapter 5 summarizes the system design, operation, web interf ace, ETL process and operational status. Chapter 6 describes the development and application of q uality a ssurance p rocedures including additional data quality tests proposed to supplement those that are found in the literature Chapter 7 describes the op erational features and available reports It also summarizes the current and potential operational applications for the archived data. Chapter 8 describes current and future research a pplications that are supportable by the archived data. Chapter 9 presen ts some interesting examples that demonstrate the use of the data for traffic count extraction, travel time reliability reporting, managed lane

PAGE 22

22 analysis and incident analysis. Chapter 10 presents conclusions and recommendations generated by the project.

PAGE 23

23 CHAPTER 2 LITERATURE REVIEW To meet the challenges of developing a central data warehouse and demonstrating its capability to perform useful functions, it is first necessary to review the body of past research. The main areas that must be covered include similar traffic data archive systems implemented in other states, the SunGuide traffic management system that will provide the raw data and the quality control concepts that are typically applied in existing systems. This chapter will cover the state of the practice in each of those areas. It is noted that each traffic data archive system has its own architecture to satisfy its diverse requirements and interface. The details of each system will be identified and described. There is a potential ambiguity in the terminology found in the literature as it relates to the definition of characteristics. The first definition refers to the proportion of a given time interval that a detector sense s the presence of a vehicle. It is generally expressed as a percentage value. This characteristic has been shown to be an indicator of the density of traffic on a roadway and therefore as a measure of traffic congestion. The second definition relates to the number of persons per vehicle ( PPV ) within the traffic stream. Vehicles with a specified minimum number of occupants occupancy context i n which they are used. Traffic Data Warehouse s in Other States Hranac presented the progress of Archived Data User Service s (ADUS) as the status of traffic data warehouses ( Hranac 2009 ) He defined five stages as follows:

PAGE 24

24 Data Reports (for decision sup port system) Application (web 2.0) Prediction Control Automation The reported stages represent the traditional outputs derived from established data archive systems. The application stage is the principal focus of applications that take advantage of traffic data warehouse technology. A complete and comprehensive review of all data archiving activities is beyond the scope of this dissertation. Instead, three traffic data warehouses were selected to provide an overview of the state of the practice i n different locations within the USA. These systems include the PeMS system in California, the PORTAL System in Oregon and the Chart system in Maryland. These systems are fully deployed and are active in operation and research. The following reviews indic ate the current status and future direction of each system. California PeMS S ystem The Freeway Performance Measurement System (PeMS) in California uses the traffic data collection, processing and analysis software developed by Caltrans, the University of California, Berkeley, and PATH ( PATH 2009 ) System d esign PeMS collects data from 8100 detector stations in nine districts in California. D istrict TMCs send loop detector data directly into PeMS. These data are processed, archived and posted in real ti me within PeMS. All of the processed traffic information is available to Internet users It also archives the incident data from the California Highway Patrol (CHP) and Traffic Accident Surveillance and Analysis System (TASAS) It also archives the lane c losure data from Caltrans in real time ( PATH 2009 )

PAGE 25

25 Data q uality c ontrol PeMS developed its own error detection algorithm using a time series method for the detect ion of errors ( Chen 2003 ) When this algorithm was developed, PeMS collect ed data from 16,000 loop detector s at 30 second resolution T he validity checks are performed on the daily data from each loop detector If it fails, the detector is identif ied as invalid for that day. Therefore, PeMS marks a daily error flag for each detector as good or bad. T o exclude the low traffic volume condition s at night, the diagnosis is execute d on the data collected from 5 AM to 10 PM (2,041 samples per day). This serves to prevent the misinterpretation of very low volume s as a case of detector malfunction PeMS checks four types of detector errors, stuck off, two types of hanging on (non zero occupancy and zero flow case or very high occupancy case), and stuck on/off. A summary of the error detect ion algorithms is presented in Table 2 1 Stuck off: Detector data are bad if 1,200 or more observations ha ve zero occupancy per day Hanging on : (case 1) If the non zero occupancy and zero flow case happen s more than 50 times per day Hanging on: (case 2) If the occupancy values are mor e than 35% more than 200 times Stuck on/off: If e ntropy of the occupancy samples is less than 4 The entropy of the occupancy is defined as follow (2 1) wh ere p(x) is the probability of variable that has the value x Data is invalid if the entropy of variables is less than 4. A low entropy value indicates that data values are not changing much over time. Originally implemented in

PAGE 26

26 PeMS, the entropy criterion has since been replaced with a "consecutive identical values" criterion for easy understanding ( Turner, 2007 ) If the test results of the detector data are not valid, PeMS discard the entire daily sample and imputes the contents using the neighborho od As a single day example, the statistics for 6/ 4/200 9 show that detector errors accounted for 24 .3% of the total errors that occur red on that day Controller error and communications malfunction problems accounted for the remainder. Table 2 2 summarizes the data quality statistics Performance measures PeMS has developed several performance reports and congestion analysis tools ( PATH 2009 ) This section will summarize the main performance measures provided by PeMS. Vehicle m iles traveled and vehicle hours traveled: Vehicle miles traveled (VMT) and vehicle hours traveled (VHT) are provided for every roadway and facility. These are fundamental measures used to evaluat e the movement of goods and people in a transportation system The Highway Congestion Monitoring Program (HICOMP) report is provided for each year to measure freeway system performance. This report measures delay by county and district and identifies the most congested locations. The delay used in this report is de fined by Caltrans as the difference between actual travel time and the travel time of the same trip at a constant speed of 35 mph. Negative values are set to zero. The Annual Average Daily Traffic (AADT) report is provided each year to evaluate freeway sys tem performance. It includes the average daily volumes at 751 locations along the freeway in 2007 Oregon PORTAL System The Portland Transportation Archive Listing ( PORTAL ) system is a traffic data archive system in Portland, OR. It was developed by Po rtland State University with the

PAGE 27

27 support of Oregon DOT, Metro, the City of Portland, and the TriMet (transit agency) ( Bertini 2009 ) System design The PORTAL system archives data from approximately 600 loop detector s in Portland, OR and Vancouver, WA. It receives traffic detector data every 20 second from the Oregon DOT and processes and archives them into the PORTAL database server ( Bertini 2009 ) Besides the traffic detector data, PORTAL archives incident data, bus data, weather data, VMS data and tru ck weigh in motion records. All of these data are available via the PORTAL web site and some of the data (traffic detector data, incident and weather data) are available in real time. Data quality control The PORTAL system applies two types of detector da ta tests, including a detector configuration test and communication failure test ( Tufte et al., 2007 ) The detector configuration test was adapted from the test sets used in an urban freeway monitoring project. This test verifies that the detector data a re within an acceptable operational range. Six conditions are used for the PORTAL system. These rules are developed for 20sec loop detector data. Table 2 3 summarizes the evaluation criteria for the PORTAL detector data. The comm unication failure test checks the percentage of communication failures or zero traffic during the peak period. These two test results are used to create the detector status report for Oregon DOT and provide the maintenance requests on the suspicious detect ors. The PORTAL system policy is to flag and filter out the erroneous data but not to impute the data. PORTAL provides the processed data for various uses

PAGE 28

28 Performance measure s Basic performance measures, such as vehicle miles traveled, vehicle hours tra veled, travel time, and delay are provided in the PORTAL web site. These measures are aggregated over time (five, fifteen and sixty minutes) and over lanes (station, corridor and system) ( Bertini 2009 ) Other performance measures, such as the green measur es, emissions, energy consumption, delay cost and person mobility are under development ( Bertini 2005 ) Maryland CHART System The Coordinated Highways Action Response Team ( CHART ) program is a traffic management system in Maryland. It was developed by The Maryland State Highway Administration, Maryland Transportation Authority, Maryland State Police, Federal Highway Administration, and University of Maryland ( COMPUTER SCIENCE CORPORATION 2009 ) System design The CHART system supports 155 miles of ro adway traffic speed sensors 70 Dynamic Message Signs (DMS) 30 Highway Advisory Radios (HARs) 220 closed circuit television and 55 Roadway Weather Information Systems (RWIS) ( MDOT 2009 ) The CHART web site provides most of its traffic reports in real time. Incident Reports Route Restrictions/Lane Closures Live Traffic Cameras Local Weather Station Images Local Weather Station Data Speed Sensor Data and Highway Message Signs are available from CHART ( MDOT 2009 ) CHART applications After the succe ssful deployment of the CHART system in Maryland, two applications were developed from this traffic data management system: Regional

PAGE 29

29 Integrated Transportation Information System (RITIS) and an incident data visualization tool (Fervor). RITIS is the traffi c management system that integrates the existing transit and transportation management data in Virginia, Maryland, and Washington D.C. It receives regional data from multiple agencies and then it fuses, translates and standardizes the data to achieve integ rated results. In this project, participating agencies are able to view the entire regional traffic information and use it to improve their operations and emergency preparedness. The traveler information system uses RITIS to provide regional standardized d ata for traveler information, including web sites, paging systems, and 511 ( Pack 2009 ) Incident data visualization (Fervor) is another example of an application developed in the Maryland traffic data management system. The existing incident data analysi s tools are defined in the traffic data management system and generate incident reports from the pre developed reports. The user may also generate an offline data file, download it, and then perform graphing and statistical processes independent of the we bsite application ( Pack 2009 ) Fervor also provides w eb based visual analytics applications with an interactive user interface, geo spatial analysis, statistical ranking functions, and multi dimensional data exploration capabilities ( Wongsuphasawat et al ., 2009 ) A screen capture of the internet interface is shown in Figure 2 1 Summary of T raffic A rchive S ystems Table 2 4 presents a summary of the functionality of the three traffic data warehouses di scussed previously.

PAGE 30

30 The SunGuide Traffic Management System in Florida The SunGuide system is a traffic management system developed by Southwest Research Institute (SwRI) for the Florida Department of Transportation (FDOT). Its design goals are to provide the most technically comprehensive advanced traffic management system (ATMS) software available and to establish the standard traffic management center throughout the State of Florida ( Dellenb a ck and Duncan 2008 ) The SunGuide software is comprised of v arious subsystems that interact with each other in a cooperative environment. Each subsystem allows the control of roadway devices as well as information exchange across a variety of transportation agencies. This software provides a common base that utiliz es a common communication interface and standard data format (Extensible Markup Language [XML]). Data are stored in an underlying Oracle database ( Duncan and Halbert, 2008 ) Operational Features Figure 2 2 shows the overall archit ecture of the SunGuide system. A data bus, which provides the common communication interface for the entire system is located in the middle of the figure. Below this data bus, there are eleven subsystems for the external equipment, which provide the interf ace between the external equipment and the SunGuide system. Each traffic equipment component has its own subsystem. For example, the traffic detection subsystem takes care of all the roadway traffic detectors connected to SunGuide. The external subsystem s include DMS, CCTV control, video switching, v ideo wall, t raffic detection, h ighway a dvisory r adio, r amp meters, s afety barriers, RWIS, AVL/Road ranger, and i ncident detection.

PAGE 31

31 The SunGuide database archives all of the data from these eleven subsystems as a repository of real time configuration and historical data for the system ( Duncan and Halbert, 2008 ) Other SunGuide subsystems include administrative, operative, and informative subsystems. In Figure 2 2 these subsystems are located on the top of the Data Bus. As one of informative subsystem, the Data Archive subsystem allows the administrator to query the database and retrieve traffic data in the database. This subsystem is the data source for STEWARD ( Dellenb a ck and Duncan 2008 ) Archive Functions Florida TMCs in District 2, 4, 5, 6, and 7 have deployed the SunGuide system as their main operating system for traffic management and control. The Traffic Sensor Subsystem (TSS) and Travel Time (TVT) data from SunGuide system a re the main data source for STEWARD. Their characteristics are as follows: TSS data are delivered from the external equipment into the SunGuide traffic detection subsystem every 20 or 30 seconds. These data are archived into the SunGuide database via the data bus. Once a day, they are retrieved from the database and saved as an archive data file via the data archive subsystem. This data file is transferred into STEWARD everyday. TVT data are calculated from TSS data and archived in the SunGuide database. Once a day, these data are retrieved from the database and saved as an archive data file via the data archive subsystem. This data file is transferred into STEWARD everyday. Quality C ontrol of T raffic D ata The Quality Control (QC) methods offered in the F assessment. Proposed QC methods were developed by the Texas Transportation Institute (TTI) and have been widely applied to traffic data in 30 cities with a bout 3,000

PAGE 32

32 miles of freeway ( Turner et al., 2004 ) One of the main reasons to apply this method to STEWARD is to make the performance measures comply with those of TTI urban mobility program. The suggested QC criteria are divided into three categories ( Tur ner et al., 2004 ) : Completeness testing: The data completeness (availability) measures the number of available data values to the number of total possible values that one could expect Basic Rules: Table 2 5 shows t he detail of basic rules. These data quality checks can be characterized as basic validity checks and should detect major problems with data errors. The following table show the basic rules Quality Control Criteria: Table 2 6 shows the detail of the criteria. These data quality checks are designed as quality control criteria to detect more subtle erroneous or suspect data that could potentially go undetected with these basic rules

PAGE 33

3 3 Table 2 1 Evaluation criteria for the PeMS l oop d etector d ata Rules Description Parameter 1 Number of zero occupancy > P 1 P 1 = 1200 2 Number of (zero volume and non zero occupancy) > P 2 P 2 = 50 3 Number of (occupancy > 0.35) > P 3 P 3 = 200 4 Entropy of occup ancy < P 4 P 4 = 4 Table 2 2 Pe MS t raffic d ata q uality s tatistics # of Detectors % Good % Bad 26,865 75.7 24.3 = Line Down (2.0%) + Controller Down (6.8%) + No Data (5.0%) + Insufficient Data (1.4%) + Card Off (6.3%) + Hig h Value (2.1%) + Intermittent (0.6%) + Constant (0.0%) + Feed unstable (0.0%) Based on PeMS traffic data on 6/4/2009 Table 2 3 Evaluation criteria for the PORTAL d etector d ata Rules Description 1 Count > 17 2 Occupancy > 95% 3 Speed > 100 MPH 4 Speed < 5 MPH 5 Low (maximum occupancy per day) 6 Low (average occupancy in peak period per day)

PAGE 34

34 Table 2 4 C omparison of traffic data warehouses in other states PeMS PORTAL CHART Coverage 9 out of 12 districts in California Portland, OR and Vancouver, WA Northern parts of Washington D.C. and Baltimore, MD Data sources 8,100 detectors Incident data Lane closure data 600 loop detectors Incident data Bus data Weather data VMS da ta Truck weight in motion records Detectors at 155 mi of freeways Dynamic message signs Highway advisory radios Closed circuit television Roadway weather information systems Main reports Vehicle miles traveled Vehicle hours traveled H ighway congestion monitoring program report Annual average daily traffic report Vehicle miles traveled Vehicle hours traveled Travel time Delay Incident reports Route Restrictions/lane closures Live traffic cameras Local weather Station images Local weather station data Speed sensor data Highway message signs System software Oracle, PHP, and Google map Linux, PostgreSQL, Apache, Adobe flash and Google map Oracle, CORBA, Apache, Javascript Applications PeMS 10.1 PORTAL 2.0 Chart Release 3 Regional Integrated Transportation Information System (RITIS) Incident data visualization tool (Fervor). Table 2 5 B asic rules of the QC criteria Quality control rules Sample code with threshold values Contro ller error codes If VOLUME={code} or OCC={code} or No vehicles present If SPEED=0 and VOLUME=0 (and OCC=0) Consistency of elapsed time between records Elapsed time between consecutive records excee ds a predefined limit or is not consistent Duplicate records Detector and date/time stamp combination are identical.

PAGE 35

35 Table 2 6 Q uality control criteria Quality control test Sample code with threshold values QC1 QC3: Logica l consistency tests If DATE={valid date value} (QC1) If TIME={valid time value} (QC2) If DET_ID={valid detector location value} (QC3) QC4: Maximum volume If VOLUME > 17 (20 sec.) If VOLUME > 25 (30 sec.) If VOLUME > 250 (5 min.) If VPHPL > 3000 (any time period length) QC5: Maximum occupancy If OCC > 95% (20 to 30 sec.) If OCC > 80% (1 to 5 min.) QC6: Minimum speed If SPEED < 5 mph QC7: Maximum speed If SPEED > 100 mph (20 to 30 sec.) If SPEED > 80 mph (1 to 5 min.) QC8: Multi variate consistency If SPEED = 0 and VOLUME > 0 (and OCC > 0) QC9: Multi variate consistency If VOLUME = 0 and SPEED > 0 QC10: Multi variate consistency If SPEED = 0 and VOLUME = 0 and OCC > 0 QC11: Truncated occupancy values of zero If OCC = 0 and VOLUME > MAXVOL where MAXVO L=(2.932*ELAPTIME*SPEED)/600 QC12: Maximum estimated density 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 No more than 8 consecutive ide ntical volume occupancy speed values. That is, the volume AND occupancy AND speed values have more than 8 consecutive identical values,

PAGE 36

36 Figure 2 1 Demonstratio n of the Fervor application

PAGE 37

37 Figure 2 2 SunGuide system architecture ( Dellenb a ck and Duncan 2008 )

PAGE 38

38 CHAPTER 3 FREEWAY TRAFFIC FLOW PRINCIPLES AND ARCHIVED DATA The daily archive data from SunGuide for each polling interval (typically 20 seconds) includes only the volume (number of vehicles), average vehicle speed and the proportion of time that each detector was occupied by a vehicle (denoted as ndard traffic flow principles and measures to provide information that is of value to the end user. This chapter will review the quantitative measures that are commonly used to describe the flow of traffic. It will identify the relationships that have be en developed in the past between the various measures. It will also identify the ways in which the descriptive measures may be incorporated into measures that evaluate the performance of a freeway facility. Later chapters will describe the application of the principles put forth in this chapter for the following purposes: Development of requirements for performance measures from the central data warehouse Development of the computational methodology by which those requirements can be met Application of t raffic flow principles to the evaluation of the quality of the archived data Investigation of the archived data to determine how well their internal relationships conform to established principles Speed Flow Rate and Density Relationships The macroscopic descriptors of traffic flow are flow rate, speed and density. Mathematically, these descriptors are related by a simple equation: Q = K x U (3 1)

PAGE 39

39 Where (u sing the common symbols from the literature and commonly applied dimensions): Where Q = Flow rate ( v eh /hr ) K = Density ( v eh /mi ) U = Speed ( m i /hr ) Thus, any of the three parameters may be computed deterministically given the other two. The nature of traffic flow creates certain internal dependencies between the parameters based on the widely observed phenomenon that speed drops as density increases. These internal relationships have been incorporated into several empirical models that make it possible to compute the value of any two parameters given the third. The Fundamental Diagram The original model of these relationships was developed by Greenshields in 1935 ( May, 1990 ) Greenshields proposed a linear relationship between s peed and density, thereby creating parabolic speed flow and flow density relationships. Known at the time t he Greenshields relationships endured for many years. The fundamental diagram is illustrated in Figure 3 1 Some other important parameters can be derived from the individual relationships in the fundamental diagram shown in Figure 3 2 The density at any point on the speed flow curve may be determined as the slope of the radius vector from the origin to that point. The speed of a backward wave during a shift in the operating point of the flow density curve may be obtained as the rate of change of flow with respect to density or dQ/dK Backward wave speed speeds are g enerally computed numerically from a shift in operation from Point 1 to Point 2 points as ( Q 2 Q 1 )/( K 2 K 1 )

PAGE 40

40 Other Speed Flow Density Models Several other models have been proposed since the Greenshields classic paper was published. is one of the nonlinear relationships for speed and density ( May, 1990 ) (3 2) where K j = Jam Density ( v eh /mi ) u 0 = Optimum Speed ( mi/hr ) Optimum speed is defined as the speed wh en the traffic flow is at capacity level and the jam density is defined as the density when vehicles are bumper to bumper and stopped. Current thinking, based on empirical observation is that it is necessary to divide the model space into two regimes repre senting oversaturated and undersaturated separates these regimes and that this region is difficult to quantify mathematically. The current concept of the speed fl ow density relationships is illustrated in Figure 3 3 The speed flow density relationships from the archive data will be investigated in a later task. Note that several of the research projects identified later used the STEWARD data to investigate speed flow relationships. A typical empirical s peed f low relationship is illustrated in Figure 3 4 which shows a number of individual observations taken over short intervals of time. It is i mportant to note that the observations in the bottom (oversaturated) part of Figure 3 4 are associated with backup from a downstream bottleneck. The conditions reflected in

PAGE 41

41 these observations will not occur in a basic freeway seg ment with no downstream bottleneck. Highway Capacity Manual Treatment of Speed, Flow and Density Recognizing this, the speed flow curves are presented in the Basic Freeway Segments chapter of the 2010 HCM as shown in Figure 3 5 Note that these relationships do not extend past the point of capacity. The free flow speed is an important parameter in this relationship. Lower free flow speeds result in lower maximum flow rates. From the fundamental relationship (Q=K*U), it is possib le to compute the density at any point on the curves and to represent the density graphically as the slope of the radius vector as illustrated previously. Density is important because it is the measure used to determine the level of service on a basic fre eway segment. The density thresholds for each level of service are shown in this manner in Figure 3 5 Note that the speed flow graphics terminate for each value of free flow speed at a density of 45 veh/mi/ln This density lev el defines the capacity of a segment. One important difference between the Greenshields fundamental diagram and the 2010 HCM is the speed at which the capacity is determined to occur. Because of the symmetry in the Greenshields parabolic relationship, the capacity occurs at a speed equal to of the free flow speed. As indicated in Figure 3 4 the 2010 HCM places the capacity at speeds between 50 and 54 mph. Considering an example of 75 mph FFS, .the Greenshields relationship wo uld place the maximum flow rate at 37.5 mph, while the 2010 HCM would place it at 54 mph. The difference is probably due to a combination of changes in car following behavior in the last 75 years and advances in the modeling of traffic flow.

PAGE 42

42 The HCM flow density model is also of interest, especially when dealing with oversaturated operation. The segment oriented chapters of the HCM (Basic Freeway Segments, Weaving Segments and Ramps) do not deal with oversaturation explicitly. In these chapters, Level of Service F is declared whenever the demand exceeds the capacity. The Freeway Facilities chapter does recognize oversaturation and queues are propagated upstream from bottlenecks and released when the bottleneck situations are cleared. Figure 3 6 shows the assumed flow density relationship in the 2010 Highway Capacity Manual for undersaturated and oversaturated conditions. Note that this relationship conforms generally to the shape shown in Figure 3 3 except that there is no region of unstable flow shown here. The s pecific values indicated in Figure 3 6 apply to a free flow speed of 75 mph. The left (undersaturated) side conforms to the 75 mph curve presented in Figure 3 5 The right (oversaturated) side assumes a linear decline in flow between the capacity of 2400 veh/ln/h at a density of 45 veh/mi/ln and the jam density, which defaults in the HCM procedure to 190 veh/mi/ln The value of 190 veh/m i/ln is equivalent to a spacing of 27.8 feet (5280/190) from front bumper to front bumper Assuming an average vehicle length of 16 ft, the space gap between vehicles (front bumper to rear bumper) would be slightly less than 12 ft at jam density. The rela tionships shown in Figure 3 6 are used to project queues in both directions from bottlenecks. The linear relationship between speed and flow introduces the very convenient approximation of a constant backward wave speed independ ent of

PAGE 43

43 the operating point. For the given assumptions, the calculated speed of the backward wave would be Level of Service (LOS) Level of service is defined by the Highway Capacity Manual (HCM) in terms of a six letter grade syste m (A representing conditions in which demand exceeds capacity. The LOS criteria are specified in the HCM for each type of facility, based on threshold values of a selected performance measure. The sel ected performance measure for a freeway segment is the average density within the segment. The LOS thresholds for density are given in Table 3 1 (Source : Exhibit 11 5 of the 2010 HCM) Density is not an explicit measure provide d by the archive data; however, reasonable approximations can be obtained from the speed flow density equation by dividing the flow rate by the speed. Thus, it is possible to estimate the level of service for each freeway segment represented in the archiv e data. Platoon Propagation All traffic flow models and theories must satisfy the law of conservation of the number of vehicles on the road. Assuming that the vehicles are flowing from left to right, the continuity equation can be written as (3 3)

PAGE 44

44 where x denotes the spatial coordinate in the direction of traffic flow, t is the time, k is the density and q denotes the flow. This relationship will be used in a later task to explore the research potential for archived data applications to congestion modeling Maximum Flow Rates A very high flow rate (e.g., greater than the capacity of any lane) could be an indication of a detector calibration proble m. The maximum flow rate for any interval was one of the QC criteria mentioned previously. Effective Vehicle Lengths The effective vehicle length is defined as the length of the vehicle plus the length of the detection zone. It may be calculated from the volume, speed and occupancy values for each time interval. The consistency of effective vehicle length provides another quality assessment indicator that will be discussed later in this dissertation Lane Volume Balance Ratio The lane volume balance rati o (LVBR) is expressed as the ratio of the highest to lowest lane volume at each station. If all lane volumes at a given station were identical, then the lane balance value would be 1.0. During periods of moderately heavy flow, lane balance values above 2 .0 might indicate detection problems unless a reasonable explanation such as a downstream lane closure, can be found. I/O V olume B alance T he total volume entering and leaving each link in the system including freeway and ramp inputs and outputs should balance, except for short intervals in which congestion is either building or dissipating. Over reasonable time periods, an unbalance between inputs and outputs would suggest volume counting errors unless there are entrance or exit ramps without detector s.

PAGE 45

45 Table 3 1 Level of s ervice t hresholds (Source Exhibit 11 5 of the 2010 HCM) Level of Service Density (pc/mi/ln) A < 11 B > 11 C > 18 D > 26 E > 35 F > 45

PAGE 46

46 Figure 3 1 Fundamental diagram relating speed, flow rate and density Figure 3 2 Determination of parameters from the fundamental diagram

PAGE 47

47 Figure 3 3 Current concept of the speed flow density relationship Figure 3 4 Typical speed flow relationship (Source: Exhibit 11 1 of the 2010 HCM)

PAGE 48

48 Figure 3 5 Speed flow relationshi ps for basic freeway segments (Source: Exhibit 11 6 of the 2010 HCM) Figure 3 6 Highway Capacity Manual assumed flow density relationship (Source: Exhibit A22 5 of the 2010 HCM)

PAGE 49

49 CHAPTER 4 CENTRAL DATA WAREHOUSE REQUIREMENTS Functional Requirements The first step in the development of an information storage and retrieval system is to determine what the system must do. To this end, it was determined in consultation with various stakeholders that the system must provide TMC ma nagers, District ITS program managers, traffic operation engineers, and management with the following useful functions: Identify detector malfunctions Provide calibration guidance for detectors Perform quality assessment data reliability tests on data Prov ide daily performance measures for system, and statewide performance measures Facilitate periodic reporting requirements Provide data for research and special studies A summary of the required functions and data flow is presented in Figure 4 1 Raw SunGuide Archive Data The Traffic Sensor Subsystem (TSS) data are stored in comma delimited flat files, with each file representing a 24 hour day. Zipped versions of these files will be posted periodically by the TMC staff The TSS dat TSS mmddyyyy 1.dat one lane over a single 20 second period. An example of the format is shown in Figure 4 2 The detect or_id and lane_id fields contain the station and lane detector names assigned by the TMC. Each district uses its own naming convention.

PAGE 50

50 TMC Configuration Requirements Not all of the information required to convert the raw data to the STEWARD database is c ontained in the raw data. Some additional information is required for three purposes: To ensure that each record in the STEWARD database represents a globally unique time and location To support the analysis and reporting of system based measures and qual ity assessment To relate the measures obtained from a specific location to other forms of data, such as RCI, Statistics O ffice counts, crash records etc. Two facility information databases must be created for each facility to be included in the STEWARD dat abase. This information must be presented in two Excel spreadsheets: The s tation d ata s preadsheet The station data spreadsheet must include the following fields for each station on the facility. Station_Index : This is a number assigned sequentially to all stations in a facility. It is required for internal processing purposes and does not appear in the database or the reports. Stationcdw_Num : This is a 6 character of the form dfnn where: d represents the district number f represents the facility number within the district (0 9) n n n represents the station number within the facility (0 9 9 9) x represents the direction (1 or 2) 0 0 0 Description : A physical description of the station (Example: I 95 NB at Forest St). Some districts embed the description in their station ID and lane ID. Status : This indicates the known status of the station (0=Normal, 1=Offline, 2=Undetected). The offline stations will not be reported as defective. The

PAGE 51

51 undetected station locations are required for the input/output analysis to indicate that the inputs and outputs for a specific link should not be expected to balance. Road : The name given to the facility (Either I 95N or I 95S in District 2) Latitude : Expressed in d egrees and decimal degrees (Example 30.32339) Longitude : Expressed in degrees and decimal degrees (Example 81.6807) State_Milepost : Required for sequential ordering of stations (Example 351.451) Roadway_Id : Required for correlation with RCI and crash data and for identifying the county number for generating traffic count files (Example 72020000) Roadway_Milepost : Required for correlation with RCI and crash data and to identify the county number for generating traffic counts (Example 2.7) The Lane D ata S pre adsheet The lane data spreadsheet must include the following fields for each detected lane on the facility: CDW S tation : This is the same 5 digit station number as in the station data spreadsheet. It is used as a key to relate the station and lane data (Ex ample 20 00 01). Lane : The lane number reference in the STEWARD database (Example: 2001131). The compositors of the lane number are CDW Station number ( 6 characters) Function code (1 character) 1. Left entrance ramp 2. Left exit ramp 3. Freeway main lane 4. Right entra nce ramp 5. Right exit ramp 6. Auxiliary lane 7. HOV Lane Tmc_Id : The lane ID used by the archive file generated by the TMC (Example: R95N001_01Lane_01) data file Archive data records in which the la ne_id is not found in the lane data

PAGE 52

52 Det_Type : ny analysis at present, but is provided for future use. Direction : The direction of the traffic detected on this lane (1=Increasing mileposts, 2=Decreasing mileposts) Status : This indicates the known status of the station (0=Normal, 1=Offline, 2=Undetected ). The offline stations will not be reported as defective. The undetected station locations are required for the input/output analysis to indicate that the inputs and outputs for a specific link should not be expected to balance. Roadway_Id : Required for correlation with RCI and crash data (Example 72020000). Also required to obtain the county number for generating traffic count files compatible with the FDOT Statistics Office files. Roadway_Milepost : Required for correlation with RCI and crash data. No te that these fields are also in the station data file. They are required here because stations that detect traffic in both directions may have different roadways assigned. Max_Speed : Normally the speed limit. This information is required for travel time limit. Count_Station : The number assigned by the FDOT Statistics Office or District Planning Office for generating traffic count data files from the SunGuide detectors. S teps in c onfiguring the TSS d ata The following steps are involved in configuring the TSS data for STEWARD. Develop a list of stations and lanes from sample TSS archive files. This is done hich reads the Assign each station to a facility or geographical subsystem. Each district may have up to 10 facilities, numbered 0 through 9. The facility assignment must be carried out by district personnel Assign station ID numbers to each station. When the facility numbers have been assigned, the station number can be added sequentially. The order is not especially important as long as each number represents a unique s tation within the facility.

PAGE 53

53 Establish the position of each station on the facility. The coordinates, state milepost, RCI road number and county milepost must be determined. Most districts have this information compiled in separate records. Establish the station status. The station status is now assigned as (0=normal, 1=0ffline). Other values might occur in the future Assign lane ID numbers to each lane. This is the most detailed part of the process. The station number determined in Step 3 provides the first four characters of the lane ID. Three more characters are needed to complete the lane ID: The direction of traffic in the lane (1 = increasing mileposts, 2 = decreasing mileposts) The function of the lane: 1 = left entrance ramp 2 = left exit ramp 3 = normal freeway mainline 4 = right entrance ramp 5 = right exit ramp 6 = auxiliary lane 7 = HOV lane The lane number (left to right) Assign lane detector operating parameters. The operating parameters for each detector include the status (same definit ion as the station status), the detector type (RTMS, Loop, etc) and the speed limit. ETL r equirements The Extraction Transformation and Loading (ETL) process must accept the raw archive data, combine it with the facility data that describes the properties of each detector in the system and load the combined data into the CDW database. Three summary intervals will be required: 5 minutes, for compatibility with the analysis of short term phenomena and perturbations 15 minutes, for compatibility with general traffic engineering analyses

PAGE 54

54 60 minutes, for compatibility with statewide traffic counting program data. In addition, it will be necessary to summarize the data by one minute intervals to provide a resource for researchers. The one minute data will be stored on a separate medium and will not be included in the database. Data Transfer Automation Requirements Productive operation of the CDW requires that the daily archive files be transferred automatically from the TMCs to be loaded into the STEWARD Data base. The key to this scheme is the implementation of a scheduled task by the district s in their SunGuide systems. STEWARD will process the re ceived the data file from each d istrict, create the summary reports and load the traffic data into STEWARD databas e.

PAGE 55

55 Reporting Requirements The following reports will be required from the CDW. Each report represents commonly used performance measures that can be derived from the basic traffic flow theory relationships described earlier: Vehicle Miles of Travel (VMT ): This is a measure of productivity of the freeway, typically accrued over a peak period or longer. Vehicle Hours of Travel Time (VHTT): This is the accumulated travel time of all vehicles in the system over the analysis period. Average speed: A figure re presenting the average speed of all vehicles in the system is computed by dividing the VMT by the VHTT. Delay: There are several definitions of delay, each with its own method of computation. For a freeway system, the most appropriate delay measure is o btained by subtracting the VHTT that would have accrued at some desired speed from the measured VHTT. The result is expressed in vehicle hours of delay. Kinetic Energy: Kinetic energy is proportional to the product of speed and volume. Higher values of kinetic energy are obtained when heavy volumes are carried at high values of kinetic energ y could be associated with safety hazards. This measure is produced to support future research. T hree performance measures derived from the travel times will be investigated in this report : Congestion Delay: based on a travel time index of 1.5. The trav el time index is defined as the ratio of the actual travel time to the travel time at the free flow speed. The speed limit will be used to represent the free flow speed. The unit of On Time Delay: referenc ed to a travel speed of 10 mph below the speed limit. This threshold has been specified for purposes of travel time reliability reporting in Percent of on time trips : defined as t he percent of trips that were made at a speed no less than ten mph below the speed limit.

PAGE 56

56 Figure 4 1 Summary of required functions and data flow timestamp detector_id lane_id speed volume occupancy 00.00.04 RTMS 9 5N003 R95N003_01Lane_01 55 1 1 00.00.09 RTMS 95S004A R95S004A_01Lane_03 55 1 1 00.00.09 RTMS 95N006 R95N006_01Ramp_01 0 0 0 00.00.09 RTMS 95N026 R95N026_01LaneN_01 0 0 0 00.00.09 RTMS 95N026 R95N026_04LaneS_01 55 2 6 Figure 4 2 Examples of the raw data from the SunGuide TSS archive

PAGE 57

57 CHAPTER 5 CENTRAL DATA WAREHOUSE DEVELOPMENT This chapter summarizes the main features of the system that w ere developed to meet the requirements outlined in the previous chapter I t provides a high level overview of the system description, operation and Internet access features. It also discusses the extraction transformation and loading (ETL) process by which the archived data from regional traffic management centers is processed and incorporated into the STEWARD database. Considerably more detail on each of these topics is presented in technical appendices of STEWARD Final Report (Courage and Lee, 2009 ) STEWARD System Description STEWARD consists of three main elements, includ ing the (ETL) process, the database (DB) and the web user interface. This chapter will describe the details of this system architecture and implementation with respect to those elements. STEWARD was developed using a variety of tools to design, deploy and maintain the system efficiently. The Oracle database was selected as a basic requirement from FDOT at the beginning of the project. The Windows 2003 Server and Microsoft Internet Information Services were selected as the operating system and web server, r espectively. Based on this decision, the Oracle Warehouse Builder 10g2, Oracle Enterprise manager and ASP/JavaScript were selected for the integrated ETL processes, the database management and the web development by TRC. System Overview Figure 5 1 shows the overall STEWARD configuration. The front end is the FTP server, which collects the traffic data from each district, processes it and archives it into the backup storage. The STEWARD DB server retrieves and loads these data into the

PAGE 58

58 STEWARD DB server. These data are archived into the STEWARD database and used to update the materialized views. STEWARD users can access the data via the web site or retrieve the data from the data back up on request. Database Design and Architecture The STEWARD database design, development and management were carried out using the Oracle Warehouse Builder program, which is an integrated tool with a graphic user interface. This program includes predefined rules that are generally required in the wareho use design. The database consists of several types of tables, the functions of which are integrated be the database manager. The following table types are involved: External tables Dimensions (Dimension tables) Cubes (Fact tables) Materialized views Func tions The relationships between these tables are quite complex. The entire database schema is described in detail in STEWARD Final Report Appendix 1 (Courage and Lee, 2009 ) Data Flow Compressed (zipped) archive data files are obtained daily from each of the SunGuide TMCs by FTP file transfer. An overview of the data flow for the required task that transfers the daily archives to an FTP site on the UF Campus is illustrated in Fig ure 5 2 The required ETL functions previously s ummarized in Fig ure 5 2 are performed and the data are added to the STEWARD database. The database may then be

PAGE 59

59 queried through the Internet to select locations by station, section and facility to produce several reports that will be described later in this document. STEWARD Operation The STEWARD operation has been documented in detail in the STEWARD Final Report Appendix 2 (Courage and Lee, 2009 ) which was developed for personnel who must install, operate and maintain the system. The topics include Oracle database program installation, STEWARD deployment, the STEWARD web site installation and the STEWARD web site management. A working knowledge of the Oracle data base manager and internet site management is assumed in the discus sion contained in the STEWAD Final Report Appendix 2 (Courage and Lee, 2009 ) Oracle Database Program Installation Oracle Version 10gR2 and the Oracle Workflow Server 2.6.4 are required for the STEWARD database installation. The Oracle Workflow Server is included in the Oracle Database 10g Companion CD. Detailed instructions are provided for installing and configuring the following components: Oracle 10g Release 2 Oracle Workflow Install Oracle Warehouse Builder Oracle Database Configuration Assistant Ora cle Net Configuration Assistant Oracle Enterprise Manager The instructions are presented as a step by step process with screen captures displayed to describe each step. A sample screen capture from the documentation of the installation process is include d as Figure 5 3 A total of 139 screen images of this type are presented in STEWARD Final Report Appendix 2 to guide the reader smoothly through the process (Courage and Lee, 2009 )

PAGE 60

60 STEWARD Deployment Deployment instructions are also presented in detail in STEWARD Final Report Appendix 2 (Courage and Lee, 2009 ) The topics Include: First Step: Login Prerequisites: Create a Target User Prerequisites: Uploading Files Importing Metadata Registration of the Control Center Manager Dat a Deployment process Data loading process STEWARD Web Installation To communicate with the Oracle database, the Net configuration for Oracle data needs to be set up. This is accomplished through the Net Configuration Assistant for Oracle. A chapter of S TEWARD Final Report Appendix 2 guides the installer through the following steps (Courage and Lee, 2009 ) : Net Configuration Assistant setup STEWARD Web Program Installation System configuration Firewall setting Permission for file sharing Web Program Config uration STEWARD Management STEWARD receives archive data from SunGuide systems in each district everyday. All data are processed and loaded into the STEWARD database for users to access the various reports. The process by which this operation is managed is detailed in STEWARD Final Report Appendix 2 (Courage and Lee, 2009 ) The topics include: Data transfer from d istrict SunGuide systems Data backup and transformation in the STEWARD FTP server Data loading into the STEWARD Database Refresh configuratio n for materialized views in the STEWARD Database

PAGE 61

61 Updating the materialized views in the STEWARD Database Backup plan and procedure for the STEWARD Database and web Adding a new district or facility Internet Access Features While some use of STEWARD will b e made within FDOT by accessing the databases directly, most users in the future will gain access to the archived data via the I nternet. This section summarizes the I nternet based features of STEWARD from the perspective of a user who seeks to query the d atabase and produce reports via the Internet. A more detailed description of the various reports and features and instructions for web site use is provided in STEWARD Final Report Appendix 3 (Courage and Lee, 2009 ) The A ppendix material is also incorpor ated into a user manual, which is accessible from the web site. The following topics are covered: Overview of the STEWARD Web Interface The STEWARD web site has been developed for an audience of general users to access and retrieve the data. The web inte rface allows users to access the database remotely, to retrieve the specific data easily and to download the data to the local computer for further analysis. All data are downloaded in comma delimited (CSV) format to facilitate presentation with office pro ductivity software. At this time, the web site can be accessed from the following Internet address: http://cdwserver.ce.ufl.edu/steward/index.html STEWARD Web Architecture The STEWARD web sit e consists of four main categories: Overview, resources, maps and reports. The overall architecture is shown in the site map of Figure 5 4

PAGE 62

62 The Overview page, which is shown in Figure 5 5 provides a gene ral description of the STEWARD project. This page includes two panes as shown in the figure. The right pane displays a brief description of the STEWARD project, objectives and tasks. The left pane is used to navigate to the STEWARD Overview, Resources an d District Data/Reports sections. The resource page provides access to reports, desktop utilities and traffic volume summaries. The report section includes the Phase II final report, progress reports and presentation materials. The utility section inclu des several utility programs: SunETLUtility, MPConverter, ITSCounts, SunVol, Hotter, SimTMC, and FTP Scripts. The traffic volume report section has the links to the traffic volume reports for all the detectors in 2008. The utility programs are described in detail in STEWARD Final Report Appendix 4 (Courage and Lee, 2009 ) Maps Two graphics based maps can be accessed for each district. The first, as shown in Figure 5 6 presents an interactive map superimposed on a Google Maps sat ellite photo. The second, as shown in Figure 5 7 presents an overview of the facilities in the district with detector locations shown on a GIS map. Report Levels Reports are available at the facility, section and station levels Facility level reports apply to the entire facility, covering all stations. Section level reports apply to a user defined section that includes all stations between a specified beginning and ending point. Station level reports apply to a single statio n. At all levels it is possible to specify several selection criteria, including: The facility and direction within the district;

PAGE 63

63 A date and time range; day of week or combination of days and The desired aggregation level (5 minutes, 15 minutes or 1 hour ) The selected report will be downloaded in coma delimited (CSV) format conforming to these selection criteria. Report Types Several report types are available from STEWARD, including: All Data Fields in the TSS Facility Level Report Volume Map and I/O Ba lance in the TSS Facility Level Report Traffic Counts in the TSS Facility Level Report Performance Measure in the TSS Section Level Report Travel Time Reliability in the TSS Section Level Report All Data Fields in TSS Station the Level Report Traffic Count s in TSS Station the Level Report Maximum Flow Rates in TSS Station Level Report Effective Vehicle Lengths in the TSS Station Level Report Link Travel Times in the TVT Station Level Report Maximum Delay in TVT the Station Level Report Maximum Travel time i n the TVT Station Level Report STEWARD Final Report Appendix 3 (Courage and Lee, 2009 ) describes the content of these reports in detail and provides examples of specific report selection. ETL Operations The ETL process must accept the raw archive data, co mbine it with the facility data that describes the properties of each detector in the system and load the combined data into the CDW database. All of these operations are accomplished by a specially developed utility program called Sun ETL Utility. This program is described in STEWARD Final Report Appendix 4 (Courage and Lee, 2009 ) Figure 5 8 illustrates the flow of data involved in the ETL process. The elements of Figure 5 8 focus on the ETL Utility p rogram, which uses two types of data input:

PAGE 64

64 The SunGuide archive data files, which are received as raw data input by the ETL Utility. These files are eventually discarded from the ETL process. They are kept as raw data on separate media to be furnished t o researchers who require the raw data. The facility data files, which are developed as a part of the configuration process for each facility. The facility configuration process was described previously. The ETL Utility program produces three output files : Daily reports, which may be used by the facility operators to assess problems with the detector system. A sample of a daily report is presented in Figure 5 9 The following terminology applies to each system element (lane, sta tion or ramp): construction reported no volumes during the entire day. Null minutes indicate int ervals in which no report was received from any system element, suggesting that the system was down during that interval. identified in the configuration file. Orphan elements are usual ly the result of new additions or misspellings in the configuration file. Elimination of orphans is an important maintenance task. Detector data, summarized by 5, 15 and 60 minute periods for each detected lane in the facility; Station data, accumulated f rom the detectors assigned to each station on the facility. A station consists of one or more detected lanes that carry traffic in the same direction on the same roadway. The station data for each day of operation is loaded into the STEWARD database. A c ombination of the station and lane data is used to produce the QA reports described elsewhere in this document. Current Status of the System STEWARD receives TSS archive data from District 2, 4, 5, 6, and 7 daily from 1,200 stations. Most detectors are rad ar/video detector types, which cover up to 8 lanes at one location. Approximately 4200 lanes are covered by STEW A RD. Table 5 1 shows

PAGE 65

65 the information on the facilities and detector stations as of 9 30 2009. Table 5 2 shows the data available from STEWARD for each district as of 9 30 2009. STEWARD provides traffic data and reports through its web site. Table 5 3 shows visitor statistics for three months. Average visitors were more tha n 60 per day and more than 240 different IPs per month accessed the STEWARD web pages. The principal users are located in Gainesville, Miami and Tallahassee, FL.

PAGE 66

66 Table 5 1 Status of STEWARD facilities and stations Dist rict Facility Number of Stations 2 I 95, I 295 192 4 I 4, I 75, I 95, I 595, SR869 334 5 I 4, I 95 452 6 I 75, I 95, I 195, SR 826, US 1 233 7 I 4, I 275 150 Table 5 2 TSS d ata a vailability in STEWARD District Data avail able on the STEWARD web site 2 6 28 07 to current 4 5 1 08 to current 5 4 7 09 to current 6 5 26 08 to current 7 1 8 09 to current Table 5 3 STEWARD w eb s tatistics for June July and August 2009 Jun. 2009 Jul. 2009 A ug. 2009 Visitors Total Visitors 2,176 2,572 1,452 Average Visitors per Day 70 80 64 Total Unique IPs 243 280 293 IP location of top downloader Gainesville, FL (17GB) Miami, FL (23GB) Tallahassee, FL (72GB) Page Views Total Page Views 10,784 12,566 11,525 Average Page Views per Visitor 4.96 4.89 5.57 Bandwidth Total Bandwidth 41.30 GB 119.17 GB 172.90 GB Average Bandwidth per Day 1.33 GB 3.72 GB 5.40 GB Average Bandwidth per Hit 1.19 MB 3.66 MB 5.42 MB Average Bandwidth per Visitor: 19.44 MB 47.45 MB 85.13 MB Hits Total Hits 35,570 33,337 32,651 Average Hits per Day 1,147 1,041 1,020 Average Hits per Visitor 16.35 12.96 15.77

PAGE 67

67 Figure 5 1 STEWARD system configuration Fig ure 5 2 Automated data flow diagram for the SunGuide archive data

PAGE 68

68 Figure 5 3 Example screen capture from the installation process

PAGE 69

69 Figure 5 4 STEWARD web architecture

PAGE 70

70 Figure 5 5 STEWARD o verview Page

PAGE 71

71 Figure 5 6 Example of an interactive satellite photo map

PAGE 72

72 Figure 5 7 GIS map example

PAGE 73

73 Figure 5 8 ETL u tility data flow

PAGE 74

74 Figure 5 9 Sample daily report from the ETL Process

PAGE 75

75 CHAPTER 6 DEVELOPM ENT OF QUALITY ASSURANCE PROCEDURE The literature review presented in Chapter 3 includes a description of the quality assurance (QA) procedures that are commonly applied in US cities. This section will apply these procedures to the data in the STEWARD arc hives. It will also explore the potential for expanding the procedures to include additional QA tests and rules that are made possible by the comprehensive nature of the STEWARD data. The current QA procedures are applied to each detected lane. Because the STEWARD database is organized geographically, it is possible to create additional tests that examine the consistency of data among the individual lanes at a station and among contiguous stations along a facility. These are four levels of quality assura nce procedures for STEWARD: Level 1: completeness tests Level 2: data validity tests Le vel 3: station level tests Level 4: system level tests Each category focuses on different aspects of the traffic data. The first two levels apply the current QA proce dures to individual lanes. The Level 1 data completeness test checks the detector malfunctions, communication failures, archive errors, etc. The Level 2 validity test checks that the traffic data are within the operational data range such as the maximum o r minimum allowable values. The last two levels examine consistency among groups of lanes. The Level 3 station level data validation examines the variation between traffic conditions in the lanes that comprise a station. The Level 4 system level data val idation examines the variation between traffic conditions between adjacent stations. For these two additional

PAGE 76

76 levels, the measures that should be expected to show consistency are identified and examples of consistent and inconsistent data are presented. L evel 1 Completeness Test The Level 1 completeness test verifies that the traffic data collection, transfer and archiving system is functioning properly. Therefore, it will identify the system hardware or data communication issues observed in the archived d ata. For example, the traffic data are produced from the freeway detectors every 20 or 30 seconds and delivered into the district SunGuide systems. Within the SunGuide systems, these data are processed and archived into the database. The daily traffic dat a items are retrieved, formatted and transferred into STEWARD. During these operations, several problems could arise: Detector malfunctions Communication errors betwee n detectors and SunGuide system Data processing errors, such as duplicate traffic data o n the same ti mestamp at the same location Communication erro rs between SunGuide and STEWARD To identify these problems, following items will be tested and verified: Av ailability of the district data Missing detector dat a All zero or stuck detector data Dup licate or negative scan data Availability of the District Data The availability check for the district data is simple but critical to the overall system performance. STEWARD receives the traffic data everyday from each district. This test will verify tha t STEWARD receives valid data files from each d istrict on time. If the data

PAGE 77

77 files are not available, the entire traffic data on that day would be missing for that district. The district data availability for STEWARD is shown in Table 6 1 : It is observed that three, ten and fourteen days of traffic data are missing from district 2, 4 and 6, respectively, during the second half of 2008. Note also that the data file for 12/2/08 from District 4 has a mechanical error in the archived file format and is therefore not valid. This case would be another example of missing dates from the district. Missing Detector Data The availability test for the detector data checks all of the detector data in the traffic data file. If there are any p roblems in detector or communication errors during the data transfer, all of the detector data would not be available. This test covers all of the traffic data that are collected and archived. Completeness is defined as the degree to which data values ar e present in the attributes that require them This is a percentage value calculated from the available number of data values as a percent of the number of total expected data values. (6 1) Where n avai lable values = the number of records or rows with available values present n total expected = the total number of records or rows expected In this calculation, completeness is defined to verify the availability rather than the validity. Complet eness of traffic data from District 2, District 4 and District 6 were examined for the month of Oct. 2008. The results are as follows:

PAGE 78

78 Completeness District 2 = (total number of traffic data values during the period) / (total number of detected lanes 31 days (24 hours / 20 sec)) = 44 696 974 / (681 31* (24 60 60) /20) = 0.49 Completeness District 4 = 70 357 757/75 129 120 = 0.94 Completeness District 6 = 56 819 105/ 92,404,800 = 0.61 If a station produces no data (including all zero volume data) for a day, it is defined as a null station for that day. In District 2, an average of 74 out of 190 stations are null stations in October 2008. Null stations (74/19 0 = 38.9%) impact the completeness of District 2 data. In District 6, 54 out of 233 stations are null stations. This null station ratio (54/233 = 23.2%) impacts the most of completeness of District 6 data. Null stations in these districts are the result o f the system implementation schedule. If a lane produces no data (including all zero volume data) for a day, it is defined as a null lane for that day. The occurrence of null lanes at non null stations was minimal in all districts. For the period examin ed, it is clear that there were problems in the District 2 data collection and archiving systems. These problems have been resolved since the period of the analysis. All Z ero or S tuck D etectors e data values for the traffic detector for a time period. These data could be all zero or one fixed value over a time period. The time periods for the all zero or stuck data test are suggested as 8 consecutive identical values from an FHWA report on monito ring urban freeways ( Turner et al., 2004 )

PAGE 79

79 A threshold of five consecutive minutes has been used in this study to check if the detector is in the all zero or stuck condition. The sampling rates for District 2, 4, and 6 are 20sec and approximately 15 data s amples (5min/20sec) are tested to have all zero or stuck detectors. If the sample sizes are too small (equal or less than 1), these data are not applied. These rules were applied to detector data from District 2, 4, and 6 for the month of October 2008. T he results were as Table 6 2 Most of the detectors in District 2,4 and 6 are RTMS detectors. Given the limited scope of the data collection and the fact that the systems were in various stages of implementation, it is difficult to draw general conclusions on the relative significance of these results. District detectors mostly exhibit the all zero problems rather than stuck data problems. Duplicate or N egative s can D ata Duplicate (zero scan) or negative scan data problems are specific issues in the SunGuide system. Duplicate data records are defined by multiple records from the same detector with the same time stamp. Therefore, two or more detector records are archived in the traffic data file with the same detector/lane ID wit h no time intervals (zero scan intervals). The SunGuide data archive system is designed to log the traffic data into the file in chronological order. Negative scan intervals are defined by records in which a time stamp for a given record indicates an ear lier time than the preceding record. These rules were applied to detector data from District 2, 4, and 6 for the month of October 2008. The results were as Table 6 3 District 2 and District 6 had these problems at one point an d more than 10% of the daily traffic data records were reported with zero or negative scan intervals. It is our

PAGE 80

80 understanding that these problems have been resolved by the SunGuide contractor. The suspicious data were not archived into the STEWARD databas e. Level 1 Test Summary The level 1 test focuses on the system operation related traffic data. As STEWARD is not involved in the data generation, communication, archive and transfer, it needs to verify that the delivered traffic data files have all the exp ected data with predefined formats. The causes and results at this level are summarized as Table 6 4 Most of problems in this level could be resolved with the support of each d istrict. Level 2 Data Validity Test Data validity t ests in this level mainly check that traffic data are in an acceptable operational range. Most of quality control methods offered in the FHWA report on monitoring urban freeways were used for the test. Eight validation rules were set up and applied to th e District 2, 4 and 6 data. The rules are: Maximum volume Maximum occupancy Maximum speed Multivariate consistency (zero speed with non zero volume) Multivariate consistency (zero volume with non zero speed) Multivariate consistency (zero volume, zero spee d with non zero occupancy) Truncated occupancy values of zero Maximum estimated density Maximum V olume T est The maximum volume for each lane should be less than: 17 vehicles for 20 second polling intervals 25 vehicles for 30 second polling intervals 250 vehicles for 5 minute data aggregation

PAGE 81

81 750 vehicles for 15 minute data aggregation 3000 vehicles for 1 hour data aggregation This rule was applied to 20 second data from District 2, 4 and 6 during October 2008. The results are presented in Table 6 5 Note that less than 1% of the observations failed this test. Maximum Occupancy Test Maximum allowable occupancy for each lane data should be less than 95% for 20 or 30 second data 80% for 1 to 5 minute data This rule was applied to 20second data from District 2, 4 and 6 during October 2008. The results are presented in Table 6 6 Note that less than 1% of the observations failed this test. It is also observed that District 2 had a zero failure rate. Min imum and M aximum S peed T est s The minimum allowable speed for each lane should be higher than 5 mph and the maximum allowable speed for each lane data should be less than 100 mph for 20 or 30 second data 80 mph for 1 to 5 minute data These rules were appl ied to 20second data from District 2, 4 and 6 during October 2008. The results are presented in Table 6 7 In most cases, the failure rates were below 0.05%. The only exception is the minimum speed test in District 6 which show ed a failure rate of 3.32%. This suggests that some attention to the calibration of certain detectors in District 6 might be desirable. M ultivariate Consistency T est There are three cases that are related with multiple variables: Zero speed and non zero v olume case

PAGE 82

82 Zero volume and non zero speed case Zero speed an d volume and non zero occupancy case These rules were applied to 20 second data from District 2, 4 and 6 during Oct. 2008. The results are presented in Table 6 8 It w as observed that District 4 detector s tended to generate non zero speed values with zero volume when the traffic volumes are very low in early morning. The reason for this is not known. Truncated O ccupancy V alues Test Older detectors on the roadway have a lower resolution in occupancy data. With very low volumes, the occupancy might fall below one percent and be truncated to a zero value. This test was applied to 20 second data from District 2, 4 and 6 during Oct. 2008. The results are presented in Table 6 9 Failure rates for this test were very low. Maximum Estimated D ensity For this test, the maximum allowable density for each lane should be less than the jam density for the facility. The jam density is assumed to be 220 ve h/lane/mile; corresponding to a spacing of 24 ft between the front bumpers of successive vehicles. The density may be estimated by dividing the flow rate by the average speed for the interval. This test was applied to the 20 second data from District 2, 4 and 6 during Oct. 2008. The results are presented in Table 6 10 Failure rates for this test were very low. Summary of Level 2 Tests These eight criteria were implemented in the STEWARD ETL processing. Error codes are defined for these criteria as shown in Table 6 11 If one of the data items fails two or more criteria, the sum of the error codes will be recorded. Data items that fail

PAGE 83

83 any of these criteria are marked with the appropriate error code a nd are not used in the aggregated STEWARD reports. The results of applying the eight criteria to the data from District 2, 4 and 6 during Oct. 2008 are summarized as Table 6 12 The values in the following tables indicate the pe rcentage of records that failed one or more of the eight quality check criteria. Level 3 Station D ata V alidation The Level 1 and Level 2 tests are applied to the data from individual lanes to identify problems with the detectors. The Level 3 station da ta validation is applied to the aggregated data for all lanes at the station. This procedure uses traffic flow principles to identify inconsistencies among the lane specific data. The relationship between volume, speed and occupancy and other performance measures from the following STEWARD reports are used for this purpose. Maximum flow rates for the station Effective vehicle length (EVL) Lane balance Daily volume variation Annual volume variation To set up the criteria, a set of stations was chosen with location and time limits that were known to be free of problems associated with system malfunctions, construction, etc. that would generally be detected by the lower level tests. The test sample had the following characteristics: Facility: District 2 I 9 5 northbound, south of I 10 Date: Oct., 2008, weekdays (23 days) Time: Morning peak (7:00 AM~10:00 AM) rush hour traffic Among twenty three active stations, four stations (210471, 210511, 210531, 210671 ) are selected

PAGE 84

84 The location and configurations of th ese four stations are described in Figure 6 1 and Table 6 13 Maximum Flow R ates Maximum flow rates can be used to identify the stations that produce excessive traffic volumes and therefo re might need calibration or other maintenance attention. STEWARD provides the maximum flow rate per day from all stations. These rates are calculated from 5min, 15min or 1hr traffic volume and excessive maximum flow rates could be the result of detector o ver counting. Also, if one of the detected lanes has excessive flow rate problems and the detector covers several lanes of the station at the same time, all the lanes in that detector might need to be checked for over counting problems. In the Level 2 te sts the threshold for maximum flow rate from an individual lane was set at 3000 veh/ln /h For purposes of this test, a station level threshold value of 2400 veh/ln/h which is more in line with the HCM capacity estimates, will be used. For data profiling, flow rate histograms for the four selected station s were created as shown in Figure 6 2 from 15min traffic volume data. From the cumulative percentages, 99.3% of flow rates are less than 2400 veh/ln/h which was selected as the t hreshold criterion. This criterion was applied to the STEWARD maximum flow rate report with the following conditions: District 2 I 95 Northbound and southbound between I 10 and I 295 Weekdays of Oct. 2008 Table 6 14 shows the frequency table of the max flow rates from this data report. It is observed that 95.2% of the observations have less than 2400 veh/ln/h The other 5% are producing volumes that exceed the upper limits of capacity indicated by the HCM

PAGE 85

85 proce dures for analysis of basic freeway segments. While it is not impossible that higher flow rates could occur, there might be a need to examine the operation of these stations in more detail. The flow rate observations in excess of 2400 veh/ln/h were exami ned in more detail to determine whether the threshold value for this test should be raised, keeping in reported in the HCM. It could be argued that observations that exceed a typical value would not necessarily indicate a data quality problem. As indicated in the cumulative distribution plot of Figure 6 3 the observations that exceeded 2400 veh/ln/h were all below 3400 veh/ln/h The cumulative dis tribution reached the 99% level at a flow rate of approximately 2900 veh/ln/h Raising the will increase the usefulness of the maximum flow rate test without an unreasonabl e departure from the HCM results. Therefore a value of 2900 veh/ln/h will be applied. EVL Characteristics The EVL is defined as the length of the vehicle plus the length of the detection zone because a vehicle will be detected as long as it any part of i t remains within the detection zone. While the EVL can be calculated at the individual lane level (i.e., Level 2), it has been considered as a Level 3 characteristic for purposes of this discussion because it has not been included in the Level 2 tests des cribed in the literature. The EVL can be cal culated from the measured values of volume, speed and occupancy using the basic traffic flow equations: q = k x U (6 2)

PAGE 86

86 (6 3) Occupancy = k x (EVL) (6 4 ) Where q =Flow rate ( v eh /hr ) k =Density ( veh/hr ) U = Space mean speed ( mi/hr ) V = Tim e mean speed ( mi/hr ) Then ( EVL ) = V x Occupancy / q (6 5) The first equation is a simple flow density speed relationship. The second equation uses the assumption that the speeds are constant over the link, which makes the space mean speed the same as the measured time mean speed. The third equation comes from the density definit ion and previous two equations ( FHWA, 1999 ) From this relationship, the EVLs are estimated using the four selected detector s. Figure 6 4 shows the EVLs for station 210471. The EVL shows relatively constant values at medium to high flow rates. But as the flow rate decreases below this level, high EVLs begin to occur. Some of the higher values are ass ociated with congested operation, with low flow rates and high density (occupancy). Under this condition, successive vehicles can be counted as a single long vehicle. The average EVLs for the four selected station s are calculated as follows. During the c alculation, high occupancy data (occupancy >18%) are excluded to avoid the oversaturated region.

PAGE 87

87 Station 210471: 21.4ft Station 210511: 21.5ft Station 210513: 20.1ft Station 210513: 24.2ft The average EVL should be in the order of 21ft, which is the sum o f average passenger vehicle length (15ft) and average detector length (6ft). Note that determination of an EVL requires a reasonable sample size. EVLs computed for individual polling intervals tend to be very erratic. Five minutes should be considered as the minimum aggregation period. For data profiling, the flow rate histograms for the four selected station s were created as shown in Figure 6 5 from 5 minute traffic volume data. From the cumulative percentages, 96.7% of the EVL s are less than 30ft, which was therefore selected as the threshold criterion for purposes of this test. This criterion was applied to STEWARD EVL report with the following characteristics: District 2 I 95 Northbound and southbound between I 10 and I 295 Weekdays of October. 2008 Figure 6 6 shows the histogram of the EVLs of all stations from this report. It is observed that 6.4% of station observations have EVLs larger than 30ft. It is also noted that 9.6% of the observations have lengths of 10ft or less. Figure 6 7 show flow occupancy relationship from one of stations with low EVL problems. Note that the average occupancy in the low EVL cases is much smaller than the normal case (>10ft) and the slop e of the flow occupancy graph for the small EVL case is much steeper than that for the normal case. Since EVL is proportional to (Occupancy / Flow), it would be anticipated that low EVLs would be associated with low occupancy values. This

PAGE 88

88 example illustr ates how the EVLs may be used to provide additional insight into the operation of a SunGuide detector station. Lane Volume Balance Ratio As indicated in Chapter 3, t he lane volume balance ratio is expressed as the ratio of the highest to lowes t lane volume at each station. STEWARD provides lane volume balance ratio lane by minute and 1 hour aggregation levels. From this relationship, the lane volume balance ratio was estimated using the four selected detectors. Figure 6 8 shows the lane volume balance ratio for station 210 51 1. It is observed that the value converges to a level near 1. 0 at high flow rates. This would be expected because, as the flow rates increase, the lane volume for all lanes should be similar. The maximum number is less than 2.0. Figure 6 9 shows the histogram of the lane volume balance ra tio from all four stations. It shows that 96.1% of lane volume data are less than 2. During the estimation, the observations with the large occupancy values (>18%) were excluded to avoid downstream situations that might affect the natural lane volume balan ce ratio For example, when the lanes are partially closed during an incident, lane volume balance ratio could increase substantially. This criterion was applied to the STEWARD lane volume balance ratio report with the following characteristics: District 2 I 95 Northbound and southbound between I 10 and I 295 Weekdays of Oct. 2008

PAGE 89

89 Table 6 15 shows the histogram of the lane volume balance ratio from all the stations in the STEWARD report. The maximum value of lane volume balance ratio is limited to 99 during the ETL process to avoid meaningless results. Therefore, the highest possible value is 99 times of the lower lane volume. Around 22.4% of lane volume balance ratio s are larger than 2. There are several scenarios that might ex plain these cases: Detector configuration problems Lane configuration problems Installation and calibration issues Incidents Downstream origin/destination issues Detector configuration problems are the most systematic of all lane volume balance ratio pro blems and therefore tend to produce substantially higher values. As an example of a possible detector configuration problem, a ll of the stations with lane volume balance ratios of 10 or more are shown in Table 6 16 It is observ es that station 210032 has 761 occurrences of excessive lane volume balance ratio accounting for 11.5% of the entire time period. Station 210032 appears to have a detector configuration problem. Average flow rates for the lanes from Station 210032 are pre sented in T able 6 17 Note that Lane 2 has relatively high flow rates for the sampling period and Lane 3 has minimal flow rates. This station is located on the F uller W arren B ridge and covers I 95 SB as shows in Figure 6 11 Lane 3 is a diverging lane for I 95 and the exit ramp and is wider than other lanes. This detector might need calibration to cover Lane 3 more precisely. Lane configuration and utilization problems are also systemic and therefore shou ld be quite evident in the lane volume balance ratio report. For example: Station 200201

PAGE 90

90 would appear to have an unusual lane configuration and usage. The average flow rates for three lanes are presented in Table 6 18 This table shows that the Lane 1 volume is very small compared to the other lanes. As Figure 6 10 shows, this station has only two apparent through lanes plus what appears to be an auxiliary lane. If the auxiliary lane (La ne 3) had very low volumes, the situation would have been more credible. The problem is that Lane 1, which is the inside lane, is the one with the low volume. In the absence of a reason for a much lower lane 1 volume, this suggests a possible configurat ion error. Detector calibration issues are more subtle. For example, Figure 6 11 shows the flow rates for Lanes 1 and 3 at station 210451. The lane volume balance ratio for this station varies from 1.37 to 3.46 and the traffic vo lume of lane 1 is 2.4 times larger than that of the lane 3. Unless a physical or operational reason can be found for this disparity, recalibration of the detectors might be warranted. Incidents can also cause large values of lane volume balance ratio beca use they tend to constrain the flow in specific lanes of the freeway. An example of the effect of an incident on lane volume balance ratio and other performance measures is presented in Chapter 9. Daily V olume V ariation Daily volume variation can be used t o verify the station data quality. Figure 6 13 shows the hourly volume variation at station 210471 over the year of 2008. The station three th rough lanes. The x axis shows the time of day and the y axis shows the station flow rates (v eh/ h r ). The average flow rates are shown on this figure along with upper

PAGE 91

91 and lower 95% confidence bounds based on the computed standard deviation. The highest flow rates are observed during the morning peak around 7:30 AM. Figure 6 14 shows an example of a suspiciously low daily volume from station 210231 for the same time period. The station is located at n orth of Spring Glen R oa d with th ree northbound through lanes. There are no ramps near this station but relatively small rush hour flow rates (3500 veh/hr or 1160 veh/ln/h ) are observed in the morning peak. While the hourly variation appears to have a reasonable shape, all of the flow ra tes are much lower than the surrounding stations. This is the only one of twenty nine northbound stations on I 95 that exhibits this problem. Unless a reason for the discrepancy can be found, this station might require some attention. Annual Volume V ari ation quality. Figure 6 15 shows the total 24 hour volume at station 210471 for year of 2008. The x axis shows the day of the year and the y axis shows the total 24 hour volume. This graph excludes all the weekend data but not holidays, which are shown as days with low volumes in November and December. Low day volume problems are also found in August. Note that some days show unreasonably low volumes in dicating possible system problems. Most of the stations have similar issues and some have no traffic data for several days per year; probably the result of detector and communication problems, etc. As an extreme example, Figure 6 16 shows a gap in daily volume that lasts for nearly four months. Most of District 2 detectors exhibit periods of missing data from several days to months. Many of the problems were associated with construction activities. More recent data suggest tha t these problems have been resolved.

PAGE 92

92 Level 4: System Level T ests Level 4 system level data validation uses the continuity of traffic flow to check the quality of the traffic data. The following measures can be expected to be continuous within the system. EVL Volume To set up the criteria, a set of stations were chosen within spatial and temporal limits that were free of incidents Facility: District2 I 95 northbound, south of I 10 Date: Oct. 6th, 2008, Monday Time: Morning peak (7:00 AM~10:00 AM) rush h our traffic Eighteen active stations The same section of the facility was examined over the same peak period for five consecutive days from Oct 20 to Oct 24, 2008 Continuity of EVL EVLs for each station are expected to be continues within the system. Figure 6 17 shows the EVL over the system. As described in the previous sections, EVLs are expected to have an average of 21 ft and a maximum value of less than 30 ft. In Figure 6 17 there are two issues: consecutive large EVLs at station 210341 and one large EVL estimate from station 210391. Station 210341 shows more than 30 ft EVLs successively. The problem comes from low volume counts and low occupancy values that could have been detected and recalibrate d in a Level 3 test. Station 210391 shows one value above 50ft. The traffic data suggests that there might be an incident at this point that caused the condition of low volume with high occupancy. In this oversaturated situation, the EVL could be relativel y large.

PAGE 93

93 To detect the abrupt increase in EVLs, the following procedures are proposed. The ratio of EVLs between the upstream and downstream stations should be relatively constant. Within the sample period (Oct. 6th, 2008), the changes are less than 10% for over 79% of the time periods. Changes over 50% would be typical of the incident case at station 210391, which shows a 68% increase from 8:00 AM to 9:00 AM. The maximum value of all other cases is 22%. This rule was applied to one weekday (10/20/08~10 /24/09, 7:00 AM~10:00 AM) and four data points were found that have more than 50% ratio changes. These points are shown as (a) in Figure 6 18 If the EVLs of two or more consecutive stations increase together, the first station th at has the abrupt changes would be identified. For example, the EVL of Station (b) in increases with the upstream station and both increases will be detected as one incident at the station upstream of Station (b). Continuity of Volume Traffic volumes shoul d also be continuous over time within the system, taking into account the entrance and exit ramp volumes. Unexplained discontinuities in traffic volumes over reasonably long periods such as 1 hour could be taken as a sign of a potential detector problem. An example of volume continuity analysis will be presented here. The freeway section for this example is 11.26 miles long, with 18 stations in the northbound direction. There are 9 interchanges with 15 on/off ramps. Nine ramps are covered to count the exit/entry volumes. The number of lanes starts from four lanes and decreases to three from milepost 340mi to 349.4mi. Figure 6 19 shows the hourly traffic volume by freeway milepost.

PAGE 94

94 Several discontinuities are observed in this figure, which shows three stations that are consistently undercounting at mileposts 347.054, 345.177, and 343.216. These stations have 30% or more hourly volume differences compared to their upstream stations. The first two stations do not have exit or en try ramps between them and their upstream stations. Therefore, these detectors might need to be recalibrated. On the other hand, the third station has neighboring exit ramps, so the difference in volumes cannot be attributed entirely to counting accuracy Without these two undercounting stations the volume differences between consecutive detectors are relatively minimal. System level analyses should be able to identify data quality problems that would be missed by lower level tests. Reliability could be improved with segment specific tests involving some knowledge of the facility configuration. Maximum reliability could be obtained by establishing historical benchmark values for comparison with the daily measures. The task of creating benchmark data wi ll be possible when the system has been in operation for a few years, but that task is clearly beyond the scope of this dissertation.

PAGE 95

95 Table 6 1 Availability of the D istrict 2,4, and 6 data (7/1/08~12/31/08) Number of missing dates Availability District 2 3 1 3/184= 98.4 % District 4 10 1 10/184= 94.6 % District 6 14 1 14/184 = 92.4 % Table 6 2 All zero or stuck data for the D istrict 2,4, and 6 data (7/1/08~12/31/08) Number of all zero data (V/S/O) for 5min (1) Number of stuck data (V/S/O) for 5min (2) Total expected number of traffic for 5min (3) Percent ratio (1+2)/3 District 2 45,254 230 6,079,968 0.8% District 4 24,478 250 5,008,608 0.5% District 6 144,222 2241 6,160,320 2.4% Table 6 3 D uplicate or negative scan records in D istricts 2, 4, and 6 (10/1/08~10/30/08) Duplicate or negative scan data Total traffic data Percent ratio District 2 571 5,681,9105 0.001% District 4 4,091 70,357,757 0.006% District 6 48 9,779 56,819,105 0.9% Table 6 4 Summary of Level 1 test Causes Results Detector malfunctions Detector turn off/lane closure Detector SunGuide communication problems No detector data All zero or Stuck data values SunGuide dat a archiving and retrieving problems Duplicate or negative time scan CRC error on the archive files Delivery problems Missing the whole district data Table 6 5 Maximum volume test for the D istrict 2,4, and 6 data (7/1/08~12/3 1/08) District 2 District 4 District 6 Maximum volume 0.13% 0.28% 0.66% Table 6 6 Maximum occupancy test for the D istrict 2,4, and 6 data (7/1/08~12/31/08) District 2 District 4 District 6 Maximum occupancy 0.00% 0.01% 0 .25% Table 6 7 Speed test for the D istrict 2,4, and 6 data (7/1/08~12/31/08) District 2 District 4 District 6 Minimum speed 0.03% 0.02% 3.32% Maximum speed 0.00% 0.00% 0.05%

PAGE 96

96 Table 6 8 Multi variate consistency test for the D istrict 2,4, and 6 data (7/1/08~12/31/08) District 2 District 4 District 6 Zero speed and non zero volume 0.01% 0.02% 3.22% Zero volume and non zero speed case 0.01% 10.24% 3.43% Zero speed an d volume and non zero occ upancy case 0.01% 0.00% 0.44% Table 6 9 Truncated occupancy values test for the D istrict 2,4, and 6 data (7/1/08~12/31/08) District 2 District 4 District 6 Zero occupancy and non zero volume 0.00% 0.01% 0.03% Table 6 10 Maximum estimated density for the D istrict 2,4, and 6 data (7/1/08~12/31/08) District 2 District 4 District 6 Maximum estimated density 0.05% 0.00% 0.18% Table 6 11 Error code for level 2 Tests Er ror type Error code Maximum volume 1 Maximum occupancy 2 Minimum speed 4 Maximum speed 8 Zero speed and non zero volume 16 Zero volume and non zero speed case 32 Zero speed an d volume and non zero occupancy case 64 Zero occupancy and non zero volum e 128 Maximum estimated density 256 Table 6 12 Summary of L evel 2 test for the D istrict 2,4, and 6 data (7/1/08~12/31/08) District 2 District 4 District 6 Eight QC rules 0.24% 10.58% 11.58%

PAGE 97

97 Table 6 13 Configuration of four selected station s Station ID Description State milepost Number of lanes 210471 I 95 NB South of Butler Blvd 342.905 3 210511 I 95 NB North of Baymeadows Rd 341.499 3 210531 I 95 NB Entrance from Baymeadows Rd 340.7 53 3 210671 I 95 NB Entrance from Philips Hwy 338.768 4 + 1(on ramp)

PAGE 98

98 Table 6 14 Maximum flow rate frequencies Maximum flow rates Frequency Cumulative % 100 18 0.36% 200 6 0.48% 300 38 1.23% 400 53 2.29% 500 65 3.58% 60 0 125 6.07% 700 93 7.92% 800 46 8.83% 900 67 10.17% 1000 94 12.04% 1100 118 14.39% 1200 128 16.93% 1300 129 19.50% 1400 229 24.05% 1500 247 28.97% 1600 267 34.28% 1700 346 41.17% 1800 378 48.69% 1900 414 56.92% 2000 521 67.29% 2100 515 77.54 % 2200 378 85.06% 2300 287 90.77% 2400 225 95.24% 2500 98 97.19% 2600 41 98.01% 2700 26 98.53% 2800 20 98.93% 2900 21 99.34% 3000 16 99.66% 3100 8 99.82% 3200 3 99.88% 3300 4 99.96% 3400 2 100.00% More 0 100.00%

PAGE 99

99 Table 6 15 Histogram o f lane volume balance ratio Lane volume balance Ratio Frequency Cumulative % 1 3348 6.17% 2 32226 65.57% 3 8917 82.01% 4 3086 87.70% 5 1557 90.57% 10 2583 95.33% 15 1061 97.29% 20 548 98.30% 25 292 98.84% 30 183 99.17% 35 110 99.38% 40 65 99.49% 45 47 99.58% 50 51 99.68% 55 44 99.76% 60 13 99.78% 65 8 99.80% 70 10 99.81% 75 14 99.84% 80 25 99.89% 85 15 99.91% 90 5 99.92% 95 6 99.93% 100 36 100.00% Table 6 16 Frequency of lane volu me balance ratios > 10.0 Station ID Frequency 210032 761 200201 698 210711 280 200082 269 210412 230 200112 118 210122 51 210562 49 200192 45 200281 23 210681 11 210422 7 210041 6 210192 5 210642 4 200031 4 200141 2 210702 1 210162 1

PAGE 100

100 T able 6 17 Detector configuration problem Lane1 Lane2 Lane3 Average flow rate ( veh/ln/h ) 1452 2277 140 Table 6 18 Lane configuration problem Lane1 Lane2 Lane3 Average flow rate (veh/ln/h ) 78 872 1166

PAGE 101

101 Figure 6 1 Location of four selected station s

PAGE 102

102 Figure 6 2 Flow rate histogram from four selected station s Figure 6 3 Cumulative percentage over 2400 ve h/ln/h from I 95 stations

PAGE 103

103 Figure 6 4 EVL with flow rates at Station 210471 Figure 6 5 EVL histogram from four selected station s

PAGE 104

104 Figure 6 6 EVL histogram from all stations Figure 6 7 Flow occupancy diagram for low EVL s (Station 210192)

PAGE 105

105 Figure 6 8 Changes of lane volume balance ratio with flow rates at station 210511 Figure 6 9 Lane volume balance ratio histogram from four selected station s

PAGE 106

106 Figure 6 10 Location of Station 210032 Figure 6 11 Flow rates of lane1 and lane 3 from station 210451

PAGE 107

107 F igure 6 12 Location of Station 200201

PAGE 108

108 Figure 6 13 Average flow rates of station 210471 for year of 2008 data Figure 6 14 Average flow rates of station 210 231 for ye ar of 2008 data

PAGE 109

109 Figure 6 15 Total day volume for year of 2008 at station 210471 Figure 6 16 Total day volume for year of 2008 at station 210531

PAGE 110

110 Figure 6 17 EVLs over mileposts during 10/3/09 morning peak Figure 6 18 EVL s over mileposts over five weekday data Station 210341 Station 21039 1 (a) (b)

PAGE 111

111 Figure 6 19 Hourly station volume over system milepost

PAGE 112

112 CHAPTER 7 OPERATIONAL AP PLICATIONS FOR ARCHIVED DATA The data available from STEWARD can be used for both operational and research purposes. The two purposes are discussed in separate chapter s of this document because they tend to have different requirements. This chapter deals with operational applications. It begins with a discussion of the detailed reports that are available from STEWARD at various levels. It concludes with a discussion of current and potential real world applications for these reports Summary of Available Reports The following reports are available from STEWARD: Diagnostic Procedures problems encountered, with a summary at the end. A record is also added to the conversion His tory file to summarize the results for that day. The following diagnostic items are reported: FileName: The date is embedded in the file name From: The time at which the first record was received To: The time at which the last record was received Elapsed Minutes: Should be 1440 if the system ran for the whole day Null Minutes: Number of minutes in which no report was received from any detector: This should be zero unless the system was off line for a portion of the delay Total Records: The number of recor ds processed: Should be consistent from day to day Total Count: The sum of all of the volumes reported: Some variation is expected, especially by day of week.

PAGE 113

113 Missed Scans Number of times a report was not received within the specified polling interval: Ne gative/Zero Scans: Negative scans indicate that the time stamp for a report was before the time stamp for the previous report. Zero scans indicate duplicate reports with the same time stamp. There should be no negative or zero scans. Orphan Stations, La nes and Ramps: Stations, Lanes and Ramps that are in the daily archive file but not in the configuration file Null Stations, Lanes and Ramps: Stations, Lanes and Ramps that are in the configuration file but reported no data for the day Offline Stations, La nes and Ramps: Stations, Lanes and Ramps that are flagged as offline to avoid showing up as nulls in the conversion report. Station Level Reports Four reports are provided in the station level reports: on level traffic data (speed, volume and occupancy) and its statistics. The details of each column are described in Table 7 1 Th e details of each column are described in Table 7 2 statistics. The details of each column are described in Table 7 3 effective vehicle length data and its statistics. The details of each column are described in Table 7 4 Section Level Reports Two new section level repo rts have been developed: measures for each segment on the section and provides totals for the section as a whole. Table 7 5 and Table 7 6 describe the details of the columns for each segment and describe the baseline. See Table 7 7 for the sample report. he measures that are used nationally in travel time reliability assessment. It also presents two measures (Percent on time arrivals and on time delay) that are specific to Florida and frequency table as a separate table. Table 7 8 describes the details of the columns

PAGE 114

114 for each segment. Table 7 9 describes the baseline and Table 7 10 describes frequency table for the travel time. See Table 7 11 for the samp le report. Facility Level Reports Three reports are presents the facility level traffic data (speed, volume and occupancy) and its statistics in the facility. The details of each column are desc ribed in Table 7 12 statistics. The details for each column are described in Tabl e 7 13 Th facility. The details of each column are described in Table 7 14 Traffic Volume Data for Traffic Counting Programs The FDOT Statistics Office maintains several continuous telemetered traffic count stations on Florida highways. Three permanent traffic counters are located on I 95 within the District 2 SunGuide system. With the cooperation of the Statistics Office, the research team was able to compare th e data from one count station to the archived counts generated by SunGuide and stored in the STEWARD database. The permanent count station was located in the southbound lanes of Interstate 95 between Emerson Street and University Blvd. The two adjacent S unGuide detector stations were located approximately 1000 ft north and 700 ft south of the permanent count station. Figure 7 1 shows an example comparison between the hourly counts from the permanent count station and the two Sun Guide detectors. Note that a near perfect agreement is apparent here. This will not always be the case and comparison of data from the two sources could potentially improve the accuracy of both sources. This possibility will be explored in the next phas e of the project. There is clearly a potential benefit that could be derived from a mutual exchange of traffic count data between the ITS centers and the Statistics Office. The Statistics Office data could provide an important reference for

PAGE 115

115 calibrating t he ITS detectors, most of which are microwave based. The ITS data could provide a useful supplement to the statewide traffic count coverage now in place. Converting ITS Data to FDOT Counts The traffic counts in the SunGuide archives have essentially the same content as the FDOT Statistics Office and District Planning Office traffic count files. A desktop utility program has been developed to convert the count data in the SunGuide data archive to either of the FDOT count formats. This program, called IT SCounts, is summarized in STEWARD Final Report Appendix 4b (Courage and Lee, 2009 ) An overview of the data flow is shown in Figure 7 2 The STEWARD database is accessed via the Internet to download traffic count data in comma d elimited (CSV) format. The ITSCounts utility program accepts the CSV formatted files as input and converts these file to either the FDOT Central or District Office count formats. A separately formatted spreadsheet file is also produced to facilitate plot ting of the results. Reporting of Traffic Volume Trends Another desktop utility program called SunVol was developed to analyze traffic volume trends over a full year to examine the variability of data from day to day and to identify questionable days. Sa mple outputs from this utility program are illustrated in Figure 7 3 The full program documentation is included in STEWARD Final Report Appendix 4c (Courage and Lee, 2009 ) Plots similar to Figure 7 3 have been prepared for all SunGuide sensor stations. They may be accessed from the Resources Page of the STEWARD website. Integration with Statewide Crash Data Records The Florida Department of Transportation maintains a crash database known as CARS, whic h is implemented in the ir mainframe computer. The CARS database includes crash

PAGE 116

116 information dating back to 2001. The project team was given remote access to the CARS system to retrieve the detailed output as a comma delimited text file. Each crash record has the following 38 fields: Crash Report Number Crash Date Time of Crash DOT County Number Section Number Subsection Number Located Mile point Nearest Node Number Located Route Id (lowest numbered "SR" route) DOT Site Location Side of Road ( for 1st harmful event) Lane of Accident (for 1st harmful event) Road Surface Condition (crash report form) Lighting Condition (crash report form) Weather Condition (crash report form) Traffic Control (1st value from crash report form) Road Conditions at Time of Crash (1st value from crash report form) Crash Rate Class Category (CAR code) Average Daily Traffic (RCI) Crash Level Alcohol Involved Code (crash report form) 1st Harmful Event for At Fault Vehicle (crash report form)

PAGE 117

117 Vehicle Type for At Fault Vehicle (crash report form) Vehicle Use Code for At Fault Vehicle (crash report form) First Point of Impact for At Fault Vehicle (crash report form) Vehicle Movement Code for At Fault Vehicle (crash report form) Direction of Travel for At Fault V ehicle (crash report form) 1st Contributing Cause Driver/ Pedestrian for At Fault Section (crash report form) Driver/ Pedestrian Age for At Fault Section (based on driver/ped birth date from crash report form) Vehicle Type for Next Vehicle (crash report form) Vehicle Use Code for Next Vehicle (crash report form) 1st Point of Impact for Next Vehicle (crash report form) Vehicle Movement Code for Next Vehicle (crash report form) Direction of Travel for Next Vehicle (crash report form) Contributing Caus e Driver/ Pedestrian for Next Section (crash report form) Driver/ Pedestrian Age for Next Section (based on driver/ped birth date from crash report form) Total Number of Vehicles in Crash Total Number of Traffic Fatalities in Crash (Traffic Fatality is person with Injury Severity value of "5") Total Number of Injuries in Crash (Injury is person with Injury Severity value of "2", "3" or "4") The archived data offer s an excellent potential for integration with the crash records An example of the analysi s of a selected crash will be presented in Chapter 9. Integration with the Roadway Characteristics Inventory FDOT maintains a comprehensive roadway characteristics inventory (RCI) database containing several descriptive fields for each roadway segment. For example,

PAGE 118

118 the RCI data for I 95 in Jacksonville indicates that there are 106 segments, including 3 mainline segments, 9 one way segments and, 94 ramps. The integration of the RCI data with STEWARD is limited at this point to the provision of a field in th e STEWARD facility data to indicate the RCI roadway segment for each TSS station. This will facilitate access within FDOT to the RCI Data for any SunGuide detector station. There are some possibilities for the creation of more automated links to this inf ormation. This question should be explored with STEWARD users when the user base has expanded sufficiently. General Support for Periodic Reporting Requirements As indicated earlier in this chapter several performance measure reports are generated from bo th the TSS and TVT data. When the STEWARD user base is expanded, these reports will serve to facilitate the periodic reporting requirements for the districts. Changes to the performance report content and format to meet the district expectations will be made as necessary. Instead of adhering to a rigid format, all reports are now generated as comma separated value (CSV) files that may be directly imported into office productivity programs such as Microsoft Excel, etc. This will allow the districts to m odify the actual presentation formats to meet their individual preferences. Diagnostic Support for TMC Detector Operation and Maintenance As indicated earlier in this c ha p ter several diagnostic reports are generated from both the TSS and TVT data. In the current phase of development, the data were obtained from District 2 on a more or less monthly basis. When the acquisition of the archived data is streamlined as proposed in the next phase, the diagnostic reports will be able to be generated on a schedu le that will give more timely feedback to the

PAGE 119

119 personnel at the SunGuide TMCs. This feedback should provide useful support for the maintenance of their detector systems and communications facilities Other Applications for the STEWARD Reports The balance of this chapter will describe some current and potential real world applications for the archive data reports. The projects covered in this section were carried out by others and are not a product of this dissertation. They are mentioned here as a demonstr ation of the usefulness of the STEWARD data, keeping in mind that demonstrating the value of STEWARD was one of the stated objectives of the dissertation. Work Zone Crash Analysis The primary focus of this research, which is sponsored by the Southeastern Transportation Center, is to determine the impact of reduced capacity on crashes in work zone queues in Jacksonville. The study is being performed by the University of Florida, in cooperation with the Florida Department of Transportation, Jacksonville Tra ffic Management Center and the Florida Highway Patrol. The study focuses on the I 95 Trout River Bridge reconstruction project and the Interstate 10/Interstate 95 interchange project in Jacksonville. Crash data were obtained from the Florida Highway Patr ol and those crashes occurring in the vicinity of the identified work zones were isolated from the larger crash data set. STEWARD data were used to confirm the traffic impacts that are caused by incidents near the work area. The dates used for this proje ct were from June 2007 to December 2007. The STEWARD data included 15 minute aggregations of traffic volumes and speeds from the stations closest to the work zone.

PAGE 120

120 Support for Identification of Recurring Congestion The consulting firm of RS&H is currentl y conducting a Bus in Shoulder s tudy for the Jacksonville Transportation Authority. One of their tasks is to identify recurring congestion on I 95 in Jacksonville speeds below 35 mph. T heir initial request was for monthly station level and lane level, volume and speed data on I 95 in the Jacksonville area. The STEWARD web site and documentation were provided to access and retrieve the traffic data via the STEWARD web pages. This is an ongoing activity. RS &H has made some constructive suggestions regarding possible improvement of the data report formats to facilitate their use. We will pursue the need for changes with RS&H and other users during the next phase of the project. We have been advised that sim ilar studies will be conducted in other districts. Travel Time Reliability Reporting As part of Strategic Intermodal System (SIS) management, two research projects on travel time reliability models were developed for freeways travel time reliability. The f irst project used data from Philadelphia, PA and the second project is evaluating the feasibility of using truck travel time data collected by the Federal Highway Administration (FHWA) and the American Transportation Research Institute (ATRI) to estimate t ravel times and determine the travel time reliability for freeways in Florida. The UF research team for that project obtained data for the I 95 freeway in Florida from the STEWARD website and are using it for model development. In addition to supporting t he specific study, we anticipate a continuing involvement with the supply of data reporting requirements

PAGE 121

121 Table 7 1 Column description for the all data fields report Measure Unit Description DAY N/A Date TIME N/A Time STATION_ID N/A A statewide unique station identifier in the format DFnnnS, where: D is the district number F is the facility number within the district nnn is the sequence n umber of the station within the facility S is direction (1=increasing mileposts, 2=decreasing mileposts) FWY_SPD Mi/Hr Volume weighted average thru speed FWY_VOL Veh Sum of thru volume FWY_OCC Percentage Average thru occupancy SPD_CV Veh Coefficient o f variation for speed VOL_RATIO Ratio Ratio of max volume lane to min volume lane ENTRY_VOL Veh Sum of entry ramp volume EXIT_VOL Veh Sum of exit ramp volume FWY_QA Ratio Percentage of freeway volume observation hit rate 100 (Thru Volume Observed)/( Thru Volume observation expected) ENTRY_QA Ratio Percentage of on ramp volume observation hit rate 100*(on ramp volume observed)/(on ramp volume observation expected) EXIT_QA Ratio Percentage of off ramp volume observation hit rate 100*(off ramp volume o bserved)/(off ramp volume observation expected) HOV_VOL Veh Sum of HOV lane volume HOV_SPD Mi/Hr Volume weighted average HOV speed HOV_OCC Percentage Average HOV occupancy HOV_QA N/A Percentage of HOV volume observation hit rate 100*(HOV volume observe d)/(HOV volume observation expected)

PAGE 122

122 Table 7 2 Column descriptions for the station traffic counts report Measure Unit Description DATE N/A Date TIME N/A Time FACILITY Facility ID : District 2 0: North of I 95, 1: S outh of I 95, 2: I 295 District 4 1: I 95, 2: I 595 District 6 0: I 95, 1: SR 826, 3: US 1, 4: I 195, 5: 1 75 STATION_ID N/A A statewide unique station identifier in the format DFnnnS, where: D is the district number F is the facility number within the district nnn is the sequence number of the station within the facility S is direction (1=increasing mileposts, 2=decreasing mileposts) NUM_OF_LANES N/A Number of lanes DIRECTION N/A Direction (N/S/E/W) STATION_DESC Mi le Station description STATION_MP N/A Station milepost COUNT_STATION N/A Count station ID provided by FDOT statistical office TOTAL Veh Total thru volume LANE1_VOL + LANE2_VOL+ LANE3_VOL+ LANE4_VOL + LANE5_VOL + LANE6_VOL LANE1_VOL Veh Lane1 volume LA NE2_VOL Veh Lane2 volume LANE3_VOL Veh Lane3 volume LANE4_VOL Veh Lane4 volume LANE5_VOL Veh Lane5 volume LANE6_VOL Veh Lane6 volume BALANCE Ratio Ratio of max volume lane to min volume lane FWY_QA Percentage Percentage of freeway volume observation hit rate 100 (Thru Volume Observed)/(Thru Volume observation expected) ON_RAMP1 Veh On ramp1 volume ON_RAMP2 Veh On ramp2 volume ON_RAMP3 Veh On ramp3 volume ON_RAMP_QA Percentage Percentage of on ramp volume observation hit rate 100*(on ramp volume observed)/(on ramp volume observation expected) OFF_RAMP1 Veh Off ramp1 volume OFF_RAMP2 Veh Off ramp2 volume OFF_RAMP3 Veh Off ramp3 volume OFF_RAMP_QA Percentage Percentage of off ramp volume observation hit rate 100*(off ramp volume observed)/(off r amp volume observation expected) COUNTY N/A County ID provided by FDOT statistical office

PAGE 123

123 Table 7 3 Column description for the maximum flow report Measure Unit Description DATE N/A Date FACILITY N/A Facility ID District 2 0: North of I 95, 1: South of I 95, 2: I 295 District 4 1: I 95, 2: I 595 District 6 0: I 95, 1: SR 826, 3: US 1, 4: I 195, 5: 1 75 STATION_ID N/A A statewide unique station identifier in the format DFnnnS, where: D is the district number F is the facility number within the district nnn is the sequence number of the station within the facility S is direction (1=increasing mileposts, 2=decreasing mileposts) DIRECTION N/A Direction 1: NB or EB (increasing mile posts,) 2: SB or WB (decreasing mileposts) STATION_DESC N/A Station description STATION_MP Mi Station milepost LANE_NUM N/A Number of lanes MAX_FLOW veh/ln/h Max hourly flow rate at the selected stations MAX_TIME N/A Timestamp when Max_Flow occurred

PAGE 124

124 Table 7 4 Column description for the effective vehicle length report Measure Unit Description DATE N/A Date TIME FACILITY N/A Facility ID for each d istrict District 2 0: North of I 95, 1: South of I 95, 2: I 295 District 4 1: I 95, 2: I 595 District 6 0: I 95, 1: SR 826, 3: US 1, 4: I 195, 5: 1 75 STATION_ID N/A A statewide unique station identifier in the format DFnnnS, where: D is the district number F is the facility number wit hin the district nnn is the sequence number of the station within the facility S is direction (1=increasing mileposts, 2=decreasing mileposts) DIRECTION N/A Direction 1: NB or EB (increasing mileposts,) 2: SB or WB (decreasing mileposts) STATION_DESC N/ A Station description STATION_MP Mi Station milepost LANE1_VOL Veh Lane1 volume LANE1_SPD mi/hr Lane1 volume weighted average speed LANE1_OCC percentage Lane1 average occupancy LANE1_EFF_DET_LENGTH Ft Lane1 effective vehicle length E V L = Speed Occup ancy/ Flow = (LANE1_SPD 5280ft/mi) (LANE1_OCC /100) / (LANE1_VOL 12) LANE2_VOL Veh Lane2 volume LANE2_SPD mi/hr Lane2 volume weighted average speed LANE2_OCC percentage Lane2 average occupancy LANE2_EFF_DET_LENGTH Ft Lane2 effective vehicle length LANE3_VOL Veh Lane3 volume LANE3_SPD mi/hr Lane3 volume weighted average speed LANE3_OCC percentage Lane3 average occupancy LANE3_EFF_DET_LENGTH Ft Lane3 effective vehicle length LANE4_VOL Veh Lane4 volume LANE4_SPD mi/hr Lane4 volume weighted average speed LANE4_OCC percentage Lane4 average occupancy LANE4_EFF_DET_LENGTH Ft Lane4 effective vehicle length LANE5_VOL Veh Lane5 volume LANE5_SPD mi/hr Lane5 volume weighted average speed LANE5_OCC percentage Lane5 av erage occupancy

PAGE 125

125 Table 7 4 Continued Measure Unit Description LANE5_EFF_DET_LENGTH Ft Lane5 effective vehicle length LANE6_VOL Veh Lane6 volume LANE6_SPD mi/hr Lane6 volume weighted average speed LANE6_OCC percentage Lane6 average occupancy LANE6_EFF _DET_LENGTH Ft Lane6 effective vehicle length

PAGE 126

126 Table 7 5 Column descriptions for the performance measures report Measure Unit Description SEGMENT N/A Segment is defined as the link between the current station (downnode) and the upnode station in the upper row MP Mile Milepost of current station LENGTH Mile Segment length (upnode MP downnode MP) AVERAGE VOLUME Veh Daily average of total link volume Sum(((Upnode thru volume + upnode entry volume) + (downnode thru volu me + upnode exit volume)) /2) /(Number of days) LANES N/A Number of thru lanes in downnode VOL per LANE Veh/ln/h Average per lane hourly flow rate Average of (AVERAGE_VOLUME/LANES/hours) VEH MILES Veh Mi Daily average of Veh Miles AVERAGE_VOLUME LE NGTH VEH HOURS Veh Hr Daily average of Veh Hours Link volume LENGTH / Link Speed = Sum (((Upnode thru volume + upnode entry volume) + (downnode thru volume + upnode exit volume)) /2 LENGTH / downnode speed ) / (Number of days) SPEED Mi/Hr Volume weighted downnode thru speed Sum(downnode speed ((Upnode thru volume + upnode entry volume) + (downnode thru volume + upnode exit volume)) /2) /Sum(((Upnode thru volume + upnode entry volume) + (downnode thru volume + upnode exit volume)) /2) D ELAY Veh Hr Congestion delay: When downnode thru speed is less than the reference speed, delay is calculated as sum of the differences between the link travel time measured and the reference link travel time. Reference speed is defined as 2/3 of the speed limit. Sum(((Upnode thru volume + upnode entry volume) + (downnode thru volume + upnode exit volume)) /2 LENGTH (1/downnode speed 1.5/ downnode speed limit))) KINETIC ENERGY (10 6 ) Veh Mi /Hr Daily average of Kinetic energy Sum (((Upnode thru volu me + upnode entry volume) + (downnode thru volume + upnode exit volume)) /2 downnode speed ) / (Number of days) PERCENT OBSERVA TIONS Percentage Percentage of downnode data observation hit rate 100*(downnode volume observed)/((distinct numbers of d ays that downnode volume observed)*(number of observations expected per day)) DENSITY Veh/Mi/Ln Max density using 15min data. Max ( Hourly flow rate / downnode speed/ Number of lanes) = Max (4*Sum((Upnode thru volume+ upnode entry volume) + (downnode thru volume + upnode exit volume) /2) during 15min /downnode speed /Number of lanes)

PAGE 127

127 Table 7 5. Continued Measure Unit Description v/c RATIO Percent age Max volume/capacity ratio using 15min data Max (Hourly flow rate / Lane capacity / Number of lanes) = Max ( Sum((Upnode thru volume + upnode entry volume) + (downnode thru volume + upnode exit volume) /2) during 15min /Lane Capacity(2200 veh/ln/h ) / Number of lanes) LOS N/A Level of service If v/c ratio > 100% LOS = F Else if Density > 35 LOS = E Else if Density > 26 LOS = D Else if Density > 18 LOS = C Else if Density > 11 LOS = B Else LOS = A

PAGE 128

128 Table 7 6 Column descriptions for the performance me asures report baseline Measure Unit Description SEGMENT N/A N/A MP N/A N/A LENGTH Mi Total segment length AVERAGE VOLUME N/A N/A LANES N/A N/A VOL per LANE N/A N/A VEH MILES Veh Mi Daily Veh Miles for the total segments Sum of (VEH MILES) VEH HO URS Veh Hr Daily Veh Hours for the total segments Sum of (VEH HOURS) SPEED Mi/Hr Segment average speed VEH MILES / VEH HOURS DELAY (Veh Hr) Veh Hr Sum of Congestion delay Sum of (Delay) KINETIC ENERGY Veh Mi/Hr Daily average of total kinetic energy Sum of (Kinetic Energy) PERCENT OBSERVATIONS N/A N/A DENSITY N/A N/A v/c RATIO N/A N/A LOS N/A N/A

PAGE 129

129 Table 7 7 Sample performance measures report SEGMENT MP LENGTH AVER AGE LANES VOL per VEH VEH SPEED DELAY KINETIC PERCE NT DEN SITY V/C LOS VOLUME LANE MILES HOURS (Veh Hr) ENERGY OBSER VATIONS RATIO I 95 SB Entrance from Bowden Rd I 95 SB South of Bowden Rd 344.56 0.29 290 3 386 84 1 58.6 0 0.02 1 3.6 0.1 A I 95 SB North of Butler Blvd 344.21 0.35 62393 3 885 21650 351 62.6 0 3.91 97.9 36.1 0.9 E I 95 SB Exit to Butler Blvd 343.85 0.36 60068 3 852 21684 320 67.9 0 4.08 97.9 27.6 0.9 D I 95 SB Entrance from Butler Blvd WB 343.67 0.18 38534 3 547 6975 1 05 66.7 0 2.57 97.9 18 0.6 B I 95 SB Entrance from Butler Blvd EB 343.26 0.41 46300 3 676 19122 295 67.9 12 3.14 95.1 39.4 0.8 E I 95 SB South of Butler Blvd 342.9 0.35 49024 3 716 17306 257 67.3 0 3.3 95.1 26.2 0.8 D I 95 SB Between Butler and Bayme adows 342.48 0.42 52768 3 748 22163 313 70.9 0 3.74 97.9 27.2 0.9 D I 95 SB North of Baymeadows Rd 341.94 0.55 53128 3 754 29168 435 67.2 0 3.57 97.9 27.8 0.9 D I 95 SB Entrance from Baymeadows Rd 341.11 0.38 142 3 190 54 1 63.8 0 0.01 1 1.6 0 A Total s: 3.3 138206 2078 66.5 12 24.33

PAGE 130

130 Table 7 8 Colum n description for the travel time reliability report Measure Unit Description SEGMENT N/A Segment is defined as the link between the current station (downnode) and the upnode station in the upper row MP Mi Milepost of current station LENGTH Mi Total segment length AVERAGE VOLUME N/A Compensated downnode daily thru volume Sum (Downnode thru volume / ((thru volume observed)/(thru volume observation expected))) / (Number of days) LANES N/A Number of thru lanes in downnode SPEED Mi/hr Volume weighted Downnode thru speed Sum(Downnode thru seed Compensated volume) / Sum(Compensated volume) = Sum (Downnode thru seed (Downnode thru volume / ((thru volume observed)/ (thru volume observation expected))) / sum (Downnode thru volume / ((thru volume observed)/(thru volume observation expected)) AV_TT Min/Veh Average travel time in minutes Average (60* LENGTH/SPEED) TT INDEX Ratio Average TT / Reference TT Average(D ownnode speed limit / Downnode speed) PCNT ONTIME Percent Percent of 5 min intervals in which AvTT =< (TT with 10mi below the speed limit) 95% TT Min 95 %ile value of travel time in this segment BUFFER INDEX Ratio Buffer index (95% TT AV_TT) / AV_TT ONTIME DELAY Min/Veh On time delay per vehicle: When downnode thru speed is less than the reference speed, delay is calculated as sum of the differences between the link travel time measured and the reference link travel time divided by the number of vehicles. Reference speed is defined as 10 MPH below the speed limit. 60* LENGTH Compensated volume *(1/(downnode speed ) 1/(downnode reference speed))/ Compensated volume = 60* LENGTH (1/(downnode speed ) 1/(downnode reference speed)) = 60* LENGTH (1/(downnode speed ) 1/((downnode speed limit) 10)) CONGESTION DELAY Min/Veh Congestion delay per vehicle: When downnode thru speed is less than the reference speed, delay is calculated as sum of the differences between the link travel time measured and the reference link travel time. Reference speed is defined as 2/3 of the speed limit. 60 LENGTH Compensated volume *(1/(downnode speed ) 1/(downnode reference speed))/ Compensated volume = 60 LENGTH *(1/(downnode speed ) 1/(downnode reference speed)) = 60* LENGTH (1/downnode speed 1.5/ downnode speed limit)))

PAGE 131

131 Table 7 9 Column descriptions for the travel time reliability report baseline Measure Unit Description SEGMENT N/A N/A MP N/A N/A LENGTH Mi Segment length (d istance between upnode and downnode) AVERAGE VOLUME Veh N/A LANES N/A N/A SPEED Mi/hr Average SPEED LENGTH / ( AV _TT/60) AV_TT Min/Veh Sum of AV_TT in the segment TT INDEX Ratio Average TT / Reference TT Average(Downnode speed limit / (Length wei ghted average speed)) PCNT ONTIME Percent Percent of 5 min intervals in which (section AvTT) =< (section TT with 10mi below the speed limit) 95% TT Min 95 %ile value of travel time in this section BUFFER INDEX Ratio Section buffer index (95% TT AV_TT ) / AV_TT ONTIME DELAY Min/Veh Sum of on time delay per vehicle CONGESTION DELAY Min/Veh Sum of congestion delay per vehicle

PAGE 132

132 Table 7 10 Colum n descriptions for the frequency table in the travel time reliability report M easure Unit Description INDEX N/A Row index LOWER LIMIT(min) Min Lower limit of the histogram is the minimum value of section travel time. LOWER LIMIT is calculated as follows: LOWER LIMIT = Min (Travel Time) +(Max (Travel Time) Min (Travel Time)) /NUMBER_OF_BINS (INDEX 1) UPPER LIMIT(min) Min UPPER LIMIT = Min (Travel Time) +( Max (Travel Time) Min (Travel Time)) /NUMBER_OF_BINS INDEX COUNT N/A Number of section travel time in this bin CUMULATIVE PERCENT Percentage Cumulativ e percentage of travel time

PAGE 133

133 Table 7 11 Sample performance measures report SEGMENT MP LENGTH AVERAGE LANES SPEED AV_TT TT PCNT 95% BUFFER ONTIME CONGESTION VOLUME INDEX ONTIME TT INDEX DELAY DELAY (Min/Veh) (Min/Veh) I 95 SB Entrance from Bowden Rd 344.85 I 95 SB South of Bowden Rd 344.56 0.29 64450 3 54.91 0.31 1.17 87.94 0.47 0.4 8 0.03 0.02 I 95 SB North of Butler Blvd 344.21 0.35 62881 3 62.71 0.33 1.02 89.01 0.43 0.3 0.01 0 I 95 SB Exit to Butler Blvd 343.85 0.36 41343 3 67.81 0.32 0.97 100 0.35 0.08 0 0 I 95 SB Entrance from Butler Blvd WB 343.67 0.18 37859 3 66.48 0.17 1 .02 90.78 0.21 0.23 0 0 I 95 SB Entrance from Butler Blvd EB 343.26 0.41 49212 3 68.51 0.37 0.98 97.06 0.4 0.08 0 0 I 95 SB South of Butler Blvd 342.9 0.35 53041 3 67.3 0.32 0.97 100 0.34 0.06 0 0 I 95 SB Between Butler and Baymeadows 342.48 0.42 547 47 3 70.81 0.36 0.93 100 0.39 0.08 0 0 I 95 SB North of Baymeadows Rd 341.94 0.55 53657 3 67.15 0.5 0.99 99.65 0.55 0.1 0 0 I 95 SB Entrance from Baymeadows Rd 341.11 0.38 39461 3 63.23 0.37 1.04 95.74 0.41 0.11 0 0 Totals: 3.3 64.68 3.47 1.14 99.2 9 3.78 0.09 0.05 0.02 Frequency Table INDEX LOWER UPPER COUNT CUMULATIVE LIMIT(min) LIMIT(min) PERCENT 1 6.23 6.33 19 6.74 2 6.33 6.44 44 22.34 3 6.44 6.55 64 45.04 4 6.55 6.66 42 59.93 5 6.66 6.77 21 67.38 6 6.77 6.88 27 76.95 7 6.88 6.99 19 83.69 8 6.99 7.1 17 89.72 9 7.1 7.21 12 93.97 10 7.21 7.32 6 96.1 11 7.32 7.43 7 98.58 12 7.43 7.54 2 99.29 16 7.87 7.98 1 99.65 20 8.31 8.42 1 100

PAGE 134

134 Table 7 12 Column descriptions for the all data fields rep ort Measure Unit Description DATE N/A Date TIME N/A Time STATION_ID N/A A statewide unique station identifier in the format DFnnnS, where: D is the district number F is the facility number within the district nnn is the sequence number of the station wi thin the facility S is direction (1=increasing mileposts, 2=decreasing mileposts) STATION_DESC N/A Station description STATION_MP Mile Station milepost LANE1_VOL Veh Lane1 volume LANE1_SPD mi/hr Lane1 volume weighted average speed LANE1_OCC percentage Lane1 average occupancy LANE2_VOL Veh Lane2 volume LANE2_SPD mi/hr Lane2 volume weighted average speed LANE2_OCC percentage Lane2 average occupancy LANE3_VOL Veh Lane3 volume LANE3_SPD mi/hr Lane3 volume weighted average speed LANE3_OCC percentage L ane3 average occupancy LANE4_VOL Veh Lane4 volume LANE4_SPD mi/hr Lane4 volume weighted average speed LANE4_OCC percentage Lane4 average occupancy LANE5_VOL Veh Lane5 volume LANE5_SPD mi/hr Lane5 volume weighted average speed LANE5_OCC percentage Lan e5 average occupancy LANE6_VOL Veh Lane6 volume LANE6_SPD mi/hr Lane6 volume weighted average speed LANE6_OCC percentage Lane6 average occupancy ONRAMP1_VOL Veh On ramp1 volume ONRAMP1_SPD mi/hr On ramp1 volume weighted average speed ONRAMP1_OCC perc entage On ramp1 average occupancy ONRAMP2_VOL Veh On ramp2 volume ONRAMP2_SPD mi/hr On ramp2 volume weighted average speed ONRAMP2_OCC percentage On ramp2 average occupancy ONRAMP3_VOL Veh On ramp3 volume ONRAMP3_SPD mi/hr On ramp3 volume weighted ave rage speed ONRAMP3_OCC percentage On ramp3 average occupancy OFFRAMP1_VOL Veh Off ramp1 volume OFFRAMP1_SPD mi/hr Off ramp1 volume weighted average speed OFFRAMP1_OCC percentage Off ramp1 average occupancy OFFRAMP2_VOL Veh Off ramp2 volume OFFRAMP2_S PD mi/hr Off ramp2 volume weighted average speed OFFRAMP2_OCC percentage Off ramp2 average occupancy OFFRAMP3_VOL Veh Off ramp3 volume OFFRAMP3_SPD mi/hr Off ramp3 volume weighted average speed OFFRAMP3_OCC percentage Off ramp3 average occupancy

PAGE 135

135 Tabl e 7 13 Column descriptions for the volume map and i/o balance report Measure Unit Description DATE N/A Date TIME N/A Time STATION_ID N/A A statewide unique station identifier in the format DFnnnS, where: D is the district num ber F is the facility number within the district nnn is the sequence number of the station within the facility S is direction (1=increasing mileposts, 2=decreasing mileposts) STATION_MP Mile Station milepost UPNODE_ID N/A Upnode ID. Has the same format a s Station_ID. Upnode is defined as the first upward station from the current station with STATION_ID ENTRY_VOLUME Veh Station entry ramp volume FWY_VOLUME Veh Station thru volume EXIT_VOLUME Veh Station exit ramp volume LINK_INPUT Veh Link input volume Link is defined between the upnode and the current node. Upnode entry ramp volume + upnode thru volume LINK_OUTPUT Veh Link output volume. Link is defined between the upnode and the current node. Downnode thru volume + downnode exit ramp volume DIFFERE NCE Veh Difference between the link input and output volumes Link_input Link_output PCNT_DIFF Percentage Percentage difference of link input and output volumes 100 Difference / ((Link_input + Link_output)/2)

PAGE 136

136 Table 7 14 Column descriptions for the facility traffic counts report Measure Unit Description DATE N/A Date TIME N/A Time STATION_ID N/A A statewide unique station identifier in the format DFnnnS, where: D is the district number F is the facility number within th e district nnn is the sequence number of the station within the facility S is direction (1=increasing mileposts, 2=decreasing mileposts) STATION_DESC N/A Station description STATION_MP Mile Station milepost COUNT_STATION N/A Count station ID from Florid a DOT statistical office TOTAL Veh Total thru volume LANE1_VOL + LANE2_VOL+ LANE3_VOL+ LANE4_VOL + LANE5_VOL + LANE6_VOL LANE1_VOL Veh Lane1 volume LANE2_VOL Veh Lane2 volume LANE3_VOL Veh Lane3 volume LANE4_VOL Veh Lane4 volume LANE5_VOL Veh Lane5 v olume LANE6_VOL Veh Lane6 volume FWY_QA Percentage Percentage of freeway volume observation hit rate 100 (Thru Volume Observed)/(Thru Volume observation expected) ON_RAMP1 Veh On ramp1 volume ON_RAMP2 Veh On ramp2 volume ON_RAMP3 Veh On ramp3 volume ON_RAMP_QA Percentage Percentage of on ramp volume observation hit rate 100*(on ramp volume observed)/(on ramp volume observation expected) OFF_RAMP1 Veh Off ramp1 volume OFF_RAMP2 Veh Off ramp2 volume OFF_RAMP3 Veh Off ramp3 volume OFF_RAMP_QA Perce ntage Percentage of off ramp volume observation hit rate 100*(off ramp volume observed)/(off ramp volume observation expected) COUNTY N/A County ID

PAGE 137

137 Figure 7 1 Example SunGuide TSS an d statistics office count comparison

PAGE 138

138 Figure 7 2 Overview of the ITSCounts data flow

PAGE 139

139 Figure 7 3 Sampl e page from the facility count analysis report f or District 6

PAGE 140

140 CHAPTER 8 RESEARCH APPLICATIONS FOR ARCHIVED DATA The previous chapter of this dissertation dealt with operational applications for the STEWARD data. Research applications will be covered in this chapter Research applications differ from operational applications in two respects: They tend to require data at a finer level of granularity than operational applications. Therefore they often require custom data with 1 minute aggregations. Data at this aggregation level are created as a part of the ETL process but they are not stored in the STEWARD database because of s torage space requirements. They frequently need to refer to data from other sources to develop relationships with external factors such as roadway characteristics, incidents, etc. This chapter considers some current examples and potential research applic ations for archived freeway data. The activities described in this chapter have been conducted in connection with other projects and do not reflect the accomplishments described in this dissertation. They are included here primarily as a demonstration of the ability of STEWARD to perform useful functions. Developing this ability was mentioned as one of the principal challenges to be addressed by the project. The previous chapter introduced the various reports and materialized views available as STEWARD resources. Those resources apply equally to this chapter Analysis of Breakdown at a Freeway Ramp The objective of National Cooperative Highway Research Program (NCHRP) Project 3 87, which started in October 2006, is to develop procedures for selecting ramp management strategies for a freeway section under the threat of flow breakdown. These procedures will be evaluated using simulation in conjunction with field data. One of the current sites in the data collection plan will be within the District 2 S unGuide facility on Interstate 95. The archived volume, speed and occupancy data is well suited

PAGE 141

141 application that will use short interval aggregations to model the breakdown of traffic flow on a freeway in the vicinity of an entrance ramp. Simulation Support for SunGuide A SunGuide simulation support project is being carried out in District 6 by a team from Florida International University under FDOT Research Project BDK80 Task Order No. 977 3 The goal of this project is to explore the development of micro simulation methods and tools to support the SunGuide system implementation, operation, testing and evaluation. STEWARD is associated with this project in two ways: One of the objectives of that project is to provide support for the future development and testing of STEWARD by producing data in the SunGuide archive file format based on simulation outputs One of the project tasks involves the d evelopment of s imulation m od eling a pplications for SunGuide using data downloaded from the STEWARD web site. The research team has already made extensive use of the STEWARD data for that purpose. The STEWARD SunGuide interface components include: A data quality check Daily pattern i dentification Period segmentation Spatial conciliation a nd missing volume estimation Free flow speed estimation. The project team has enumerated the benefits of the STEWARD interface in the following terms: Significant improvement in mo deling and analysis of traffic Lower the cost o f simulation and other analysis Much mor e details in time and space Provide an important and a new source of data for planning and traffic analyses They have suggested that the interface modules can be used by a TMC to:

PAGE 142

142 Assess Segment ti me of the day into intervals Estimate system demands The team has already developed p rocedures to fine tune simulation model parameters to produce throughputs/capacities, volumes, speeds, an d occupancy close to those of detector data in STEWARD. Current Managed Lane Applications University of Florida TRC researchers are now using the STEWARD data for two projects: NCHRP 3 96 Analysis of Managed Lanes on Freeway Facilities. T hree different m ethods are applied to the data from I 95 HOV lanes to estimate and compare the capacity of HOV and general purpose lanes. Managed Lane Operations Adjusted Time of Day Pricing vs. Near Real Time Dynamic Pricing (FDOT BDK 77 977 02) One of the tasks uses b efore/after data from the 95 Express project to systematically examine whether and how reduced lane, shoulder widths and designs of ingress/egress points affect the capacity of the managed lanes. It is anticipated that HOTTER will be useful (perhaps with s ome modifications) for these projects and others to be undertaken in the future. Other TRC Research Applications Current, past and future projects related to STEWARD and carried out by University of Florida researchers include: Modeling the Location of Cra shes within Work Zones Dr. Siva Srinivasan, PI The objective of this study is to model the location of crashes within work zones as a function of the lengths of the different work zone segments, traffic volume, weather,

PAGE 143

143 and other exogenous factors. Data f rom crash reports were augmented with spatial attributes by using geographic information systems. The results from a multinomial logit model were used to construct the crash probabilities per lane mile for the different work zone segments. A Case Study in Spatial Misclassification of Work Zone Crashes Dr. Siva Srinivasan, PI Studies associated with work zone crashes are often based on law enforcement traffic crash reports. Work zone crashes are typically segregated from larger, statewide, crash data sets b zone finds that such an assumption may be flawed. CDW information was used to match traffic information with the work zone crash data. Analyzing the Effectiveness of Enhanced Penalty Zones and Police Enforcement as Freeway Speed Control Measures Dr. Siva Srinivasan PI The objective of our study is to examine the simultaneous impacts of police enforcement and increased penalties on freeway speeds and crash characteristics. The project will analyze crash, traffic enforcement and roadway traffic data from the CDW reac h the objective. (Anticipated start date January 2010). C apacity of Florida F reeways FDOT Project BDK 75 977 08 Dr. Scott Washburn, PI CDW data are being used in an FDOT funded project to assess the capacity of Florida freeways. In this project, speed and flow data from Jacksonville, Miami, Ft.

PAGE 144

144 Lauderdale, Orlando, and Tampa are being used to develop speed flow relationships and capacity distributions for a variety of basic freeway segments. Travel Time Reliability Dr. Lily Elefteriadou, PI CDW data wer e used in several FDOT funded projects on travel time reliability (BD 545 70, BD 545 75, and BDK 77 977 02). The data were used to identify areas with congestion around the Jacksonville area, and to extract speed, flow, and travel time information from tho se locations. Freeway Work Zone Capacity Dr Lily Elefteriadou, PI CDW data were also used in an FDOT funded project to assess the capacity of freeway work zones (BD 545 82). In this project, speed and flow data were obtained for a work zone along I 95 in Jacksonville. NCHRP 3 Breakdown Dr Lily Elefteriadou, PI CDW data were used during the initial stages of this project for site selection. Researchers were examining various sites around the co untry for obtaining speed and flow data to develop probability of breakdown models. Doctoral Dissertation Project: Dr. Alexandra Kondyli data collection locations around Jacks onville.

PAGE 145

145 Other STEWARD Users Other current and potential users that have communicated with the STEWARD project team include: I dentification of Recurring Congestion (RS&H) Mobility Monitoring Program (Cambridge Systematics) Evaluation of DMS effectiveness f or diversion (HNTB) Central and District Office Traffic Counts (FDOT)

PAGE 146

146 CHAPTER 9 DATA ANALYSIS EXAMPLES The use of the STEWARD data for the development of a variety of reports presenting useful performance measures was described in Chapter 7. This chapte r presents examples of the use of the STEWARD data for more specific investigations. Four examples will be presented. The first compares the speed flow density relationships obtained from a selected station with the relationships found in the literature The second examines the effect of a selected incident on the performance of the facility. The third deals with the extraction of measures that could be used to evaluate the performance of a managed lane on the freeway. The fourth deals with travel tim e reliability reporting. Speed Flow Density Relationships The literature on the relationships between speed, flow rate and density as the macroscopic descriptors of traffic flow was summarized in Chapter 3. The purpose of this example is to illustrate h ow the archived data from STEWARD may be used to estimate these descriptors and to evaluate how well the results matched those in the literature. Archived data from a selected station over a period of one month at a five minute aggregation level will be u sed for this example. The raw data from the detectors is obtained at 20 second polling intervals. The data items include a count of the number of vehicles that passed the detector during the interval, the average speed of all vehicles during the interva l and the proportion of time that the detector was occupied by a vehicle. The speed, flow rate and density can be estimated from these data items as follows:

PAGE 147

147 Speed A distinction between time mean speed and space mean speed must be made at the outset. The time mean speed is represented by a simple arithmetic mean of the individual vehicle speeds. The space mean speed is calculated using the harmonic mean from the following equation ( FHWA, 1992 ) : (9 1) Where U s is the space mean speed U i is the speed of vehicle i N is the number of vehicles observed The space mean speed is the appropriate choice for all computations involving the speed flow density relati onships. The speed values from the radar detectors at this location are produced by a proprietary algorithm and the exact definition is not clear. Furthermore, the individual vehicle speeds required for computation of the space mean speed are not availab le. Therefore, the speed values included in the raw data must be used as the best available estimator of the space mean speed. Density The density estimation is more complex. From the definition, occupancy is the fraction of time that vehicles are over th e detector and could be described as follows: (9 2) Where L i and U i are vehicle length and speed d : detector length

PAGE 148

148 T : time interval Using the basic macrosco pic relationship (9 3) Where q is flow rate k : density U s : space mean speed The occupancy equation can be simplified as follows: Where L is average vehicle length d : detector length k : density C k : sum of the average vehicle length and detector length C k is the effective vehicle length (EVL) described in the previous section. Therefore, the d ensity could be estimated from the measured occupancy with the following equations. (9 4) To illustrate this point with a numerical example, let us assume that the EVL is 17.3 feet and the occupancy is 0.1. The density would then be computed as (5280 x 0.1)/17.3 = 30.5 veh/lane/mi. Since density is expressed in units of vehicles per mile per lane, this is clearly a spatial measure that applies to an entire segment, whereas the fiel d data represent the conditions at a single point. The point measurement is the only information available to

PAGE 149

149 support density estimates. This measure uses the same units as density and is c surveillance systems to use concentration as a direct estimate of density and to refer to the computed results as Flow Rate A distinction must be made between volume and flow rate at this point. Both measures are expressed in units of vehicl es per unit of time, with a period of one hour generally used to represent the unit of time. Volume refers to the actual number of less than one hour and represen ts a normalized value that would be obtained if the conditions persisted for an hour. So the flow rate over a five minute period would be determined by multiplying the accumulated five minute vehicle count by 12. The result may be applied to an individua l lane or to all lanes at a station. Since the density is expressed on a per lane basis it is necessary to express the flow rate in the same manner to preserve the speed flow density relationships. Examples of R elationships The selected detector station ( ID 210511) is located on Interstate 95 North of Baymeadows Rd in Jacksonville. The freeway at this station carried northbound traffic in 3 lanes. The analysis period included all of October 2008 during the morning peak (7:00 AM~10:00 AM). The figures pr esented in this section show the speed flow density relationships estimated from the STEWARD data at the selected location. The relationship between the speed and flow rate is shown in Figure 9 1 This figure shows a uniform fre e flow speed around 68 MPH in the undersaturated area. In the oversaturated region, the flow rates drop with speed. The capacity is reached at a

PAGE 150

150 flow rate of 2236 veh/ln/h and a speed of 65.6mi/h. The density at capacity estimated by dividing the flow ra te by the speed is computed as 2236 / 65.6 = 34.1 veh/mi/lane. Graphically, this figure compares very well with the corresponding figure from the HCM, presented previously in Figure 3 4. The more quantitative HCM figure presented previously in Figure 3 5 shows that the capacity is reached at a somewhat lower speed and density. The HCM procedure for basic freeway segment analysis suggests that the capacity of a segment corresponds to a density of 45 veh/mi/lane. The density was calculated from the occu pancy using an assumed effective vehicle length (EVL) of 17.3ft. The relationship between flow rate and density is shown Figure 9 2 The trend line in undersaturated area shows that the flow rates increase with density below 43 v eh/mi/lane. The slope of the trend line represents the speed of a backward wave at a given density. The slope at a density of zero indicates the free flow speed. Since the flow density relationship shown in the figure appears to be linear, the shock wave speed and free flow speed should be equal at an estimated value of 53.5 mi/h. The maximum flow rate under these conditions is 2260 veh/ln/h at 43 veh/mi/lane. The projection of the trend line in the oversaturated area to the horizontal axis suggests a j am density of 227 veh/mi/lane. This suggests a spacing of 5280/227 = 23.3 ft between the front bumpers of successive vehicles at jam density. The corresponding figure from the HCM, as shown previously in Figure 3 6 is 190 veh/mi/lane, representing a spac ing of 27.8 feet. Since the highest recorded density on this figure was 140 veh/mi/lane, the linear projection of the trend line to the horizontal axis might not be appropriate.

PAGE 151

151 The slope of the trend line in this area represents the backward wave speed which is estimated at approximately 11 mph. This is slightly lower than the 16.5 mph estimate from the HCM. In comparing these values, it must be pointed out that the HCM figures represent broad national averages obtained from many locations, whereas t he data presented here represent a single location. The HCM does not present any data on the variability of this value. VMT or VHT can be easily obtained from the STEWARD performance measures report. From these reports, the VMT and VHT are 6,272 veh mi an d 108 veh hr, respectively. The corresponding space mean speed is 58.1mi/hr. The difference between the detector data and STEWARD measures comes mainly from the definition of the link. STEWARD defines the link from the upstream station to the current stat ion and the measures are calculated from the average volume and speed of these two stations. Figure 9 3 shows the speed density relationship at this location. The density is calculated from the occupancy using a constant effectiv e vehicle length of 17.3ft and the speed is the measured time mean speed. This figure shows that, in the low density area (uncongested region), the speed approaches the free flow speed. As the density increases, the relationship fits better with Greenberg 2 =0.9076) than the Greenshields linear relationship (R 2 =0.864) as follows. The jam density is calculated as 150.3 veh/ lane/ mi. The speed density equation thus becomes (9 5) The jam density computed here suggests a spacing of approximately 35 ft between the front bumpers of successive vehicles. This is somewhat higher than the

PAGE 152

152 values previously mentioned. Jam density is a somewhat abstract concept because it n ever actually occurs. It is simply a mathematical property of the speed flow density model. From the preceding discussion on speed, flow rate and density, it is clear that density may be computed either from the flow rate and speed or from the occupancy a nd EVL. If the detectors are producing credible data, then these two methods should produce comparable results. The results of the two methods for the same example are shown in Figure 9 3 The density computed from the speed an d flow rate is plotted against the occupancy values obtained directly from the detectors. Note that a tight linear relationship is demonstrated in this figure, suggesting that the detectors are producing data with a credible speed flow density relationshi p. Crash and Incident Analysis Applications Crashes and other incidents cause perturbations that show up in the archived data. The occurrence of incidents might be expected to be evident in each of the basic archive data items, including the flow rate, sp eed and occupancy. Flow rates and speeds are likely to decrease during the period of an incident, and occupancy is likely to increase because of the higher traffic densities. This example will investigate the effect of a selected crash on the basic archi ve data and the measures that are derived from these data. The FDOT CARS system described in Chapter 7 was the source of the crash data Overall Crash Characteristics The sample data retrieved from CARS covered all of 2008. A total of 196 crashes on I 95 between Station Milepost: 338.0 and 348.7 were included. The locations with respect to the roadway were:

PAGE 153

153 North bound: 74 M iddle/median : 47 South bound: 75 Figure 9 5 shows a distribution of crashes by day of week. Note that the n umber of crashes (37) on Friday is almost twice of that on Tuesday (18). Figure 9 6 shows the number of crashes by month of the year. The maximum number of crashes (27) occurred in August, followed by April with 21. Figure 9 7 shows the number of crashes side of the road as northbound, southbound and median. The frequency of crashes was higher at specific locations, such as southbound side of st ate milepost 350 where I 95 merges with Acosta expressway and the northbound side of state milepost 351 where I 95 merges with I 10. Figure 9 8 shows the number of crashes by time of the day. The crash distribution is typical with peaks during the rush hour. The southbound direction had the maximum frequency at 6:00 PM and the northbound at 8:00 AM. Sample Crash Analysis Incident data from CARS system can be associated with STEWARD archive data to support a more detailed investigat ion of the effect on the operation of the facility. The selected crash had the following characteristics. CARS crash number: 769954660 Location: 343.947 mi Date: 10 3 08 12:56PM L ane of accident: 3 Weather condition: Clear Total number of vehicles: 2 Tota l number of injuries: 2

PAGE 154

154 This incident occurred on a Friday at the south end of the Jacksonville downtown area. It took place on the southbound roadway. Figure 9 9 shows the location of incident on a satellite photo map. Hourly flow rates, speed variation occupancy have been suggested as the most common precursors for incident analysis ( Kuchangi, 2006 ) In this example, hourly flow rates, occupancy, speed and speed variance will be verified for this incident. Also the overall de lay caused by this incident will be calculated by a number of methods. Hourly flow rates Figure 9 10 shows the hourly flow rate per lane at milepost 344.56mi on 10 3 08. CARS reported that the incident occurred at 12:56PM. The e ffect is evident on the flow rate graph, which shows a decrease near 12:40 PM. The red circle in the figure shows the two flow rate drops during the incident, which occurred around 1:00 PM and 1:45 PM and the values were below the 1000 veh/ln/h To establ Figure 9 11 shows the hourly flow rate per lane at the same location on three Fridays without incidents during the same time frame This figure presents the average flow rates for three days (10 10 08 10 17 08, and 10 24 08). It shows a similar trend of the flow rates except the flow does not drop at the time of the incident. Figure 9 12 shows five minute volume counts for the incident and non incident cases. The differenc es are plotted in Figure 9 13 as the cumulative volume difference between the non incident case and the incident case from 12:00. From this figure, it is observed that the incident starts at about 12:55 and the capacity decreases. Also, the queue starts to build up until the cumulative difference between the incident volume count and non incident volume count reaches 1,058 vehicles at 14:10. After that, the

PAGE 155

155 queue starts discharge until 15:15 and the cumulative differences become st abilized around 900veh. From this information, the delay can be estimated using queuing analysis. Mean arrival rate ( ) = 9610veh / 145min = 3977veh/hr Reduced service rate ( ) = 4498veh/ 80min = 3374 veh/hr Service rate ( ) = (9610veh 4498veh)/ 65min = 4719 veh/hr Time duration of queue ( t Q ) = = 80 x (4719 3374) / (4719 3977) = 145min Total delay = = 80min x 145min x (3977 3374) /2 = 971.5 veh hr Average vehicle delay = (971.5 veh hr/ 9610veh) = 6.1 veh min/veh Figure 9 14 shows the traditional queuing diagram for this incident. Occupancy Figure 9 15 shows a n abrupt occupancy increase at the same location, milepost 344.56 mi. Between 12:45 PM to 2:30 PM, the occupancy increases over 45%.

PAGE 156

156 Figure 9 16 presents a contour graph to show the changes in the occupancy over the time and space. The x axis shows the milepost and the y axis shows time of day. The incident started near the red circle (milepost 344 at 12:40) and the occupancy increases are evident on this figure. Speed Figure 9 17 shows the speed changes at the same location. Between 12:45 to 14:30, there were speed drops up to 3 mph that lasted until 14:00. As demonstrated previously, density can be estimated from the occupancy data. Using the density, flow, and speed data, total delay and average individual delay could be estimated with shock wave analysis. Figure 9 18 shows the speed variation over the time and space and shock waves can be observed. A backward forming shock wave is shown as w AB and a forwar d recovery shock wave is shown as w B D This incident shows a temporary capacity increase around 13:00, which creates a temporary backward shock wave (w B D1 ) and forward recovery shock wave (w AB1 ) in the middle of the incident. By definition, the area of a t ime space domain of congestion multiplied by the density of the traffic flows under congestion is the total vehicle hours of travel in congestion ( Nam,1998 ) Total delay can be calculated from the difference between the travel time without incident and the travel time with incident. For the calculation of the travel time without the incident, averages of the three day density for each station are used. Total travel time = ( Density x ( distance x time )) = 642 veh hr Total delay = ( travel time non incident Total travel time incident )

PAGE 157

157 = 510 veh hr Average vehicle delay = ( delay/flow ) = 9.03veh min/veh Est imation of d elay from the t ravel t ime r eliability r eport The total delay can be also calculated from the Travel Time Reliability Report. Congestion delay is referenced for purposes of that report to a travel time index of 1.5. The travel time index is def ined as the ratio of the actual travel time to the travel time at the free flow speed. The speed limit will be used to represent the free flow speed. The From the definition, congestion delay is ca lculated as follows: Total congested delay = distance volume ( 1/speed 1.5 / ( speed limit)) (when speed is less than (speed limit)/1.5) = 159.8 (Veh Hr) But the total delay consists of the delays from the incident (non recurring) and everyday congestions (recurring). N on recurring delay can be estimated from the average delay from the same location without incidents. The same section does not have an incident from 10/10 /08 to 1/24/08 on every Friday. The daily average congestion delay is 10.4 (Veh Hr) and it can be assumed as the recurring delay. The delay from the incident (non recurring delay) would be calculated from the difference of the total delay and the recurrin g delay: Non recurring delay = total congested delay recurring delay = 159.8 10.4 = 149.4 (Veh Hr)

PAGE 158

158 Average vehicle delay = (971.5 veh hr/ 9610 veh) = 1.6 veh min/veh Compariso n of d elay e stimation m ethods Three different methods were applied to estimate the total delay and the average vehicle delay. Each method has a sound theoretical basis but substantial differences in their results were observ ed, largely due to differences in assumptions and definitions. It is clear that the travel time reliability method estimates the smallest delay of the three methods but it bases its computation on a different definition of delay because it ignores any del ay that occurs when the speeds are less than the free flow speed, but greater than 2/3 of the free flow speed. It also covers a time period that extends beyond the incident on both sides because of the need to start and end the analysis at the beginning o f an hour for purposes of this report. The speed/shock wave method estimates the highest delay because it focuses exclusively on the time period in which delays were observed. The purpose of this exercise was to identify the various approaches to computi ng incident delay from the archived data. A substantial effort beyond the scope of this dissertation would be required to provide useful guidance on their relative merits. Lane volume balance ratio Figure 9 19 sho ws the lane volume balance ratio (LVB R ) during the incident. The x axis shows the milepost and the y axis shows the time. Several peak values of LVBR are found during the incident near milepost 344 mi. At some points the LVBR approaches 8.0, indicating a s evere unbalance in the lane utilization. These larger LVBR

PAGE 159

159 vehicles were moving through some of the lanes. This condition could be verified from Figure 9 20 Three contour graphs represent the occupancy in each of the three lanes. The x axis shows the milepost and the y axis shows the time. The vehicles are moving from right to left (decreasing milepost). As the figure shows, the incident occurs in lane 3 (rightmost lane) and therefore, the occupancy values from Lane 1 and Lane 2 are less affected than Lane 3. Speed variance The speed variance is also a potential indicator of the incident. Figure 9 21 shows the speed CV at the i ncident location. It shows peak values around 8:30, 12:50, 16:00 and 18:00, but there is only one the incident record in CARS in that area on the day in question. Inspection of the Florida Highway Patrol incident log indicated four incidents on that day, but none of them could be expected to affect this location. Speed variance is a measure that is unique to STEWARD and is not used for operational analysis. It was created mainly to support future research into congestion modeling and incident analysis. It certainly provides an indication of some type of perturbation in the traffic stream. No specific conclusions on its value can be drawn from this single example. The speed CV certainly peaked during the incident but similar peaks were observed at othe r times of the day that were incident free. It is interesting to note that some of the other peaks occurred during the AM and PM peak periods. There is a chance that peaks in the speed CV are simply a natural phenomenon that is associated with the onset and resolution of congestion. It appears, however, that a substantial investigation well beyond the scope of this dissertation would be required to support definitive conclusions.

PAGE 160

160 Kinetic e nergy The concept of kinetic energy as the product of speed and f low rate was introduced in Chapter 4. Like speed variance, this measure is not used by SunGuide for operational analysis, but was incorporated into STEWARD to support future research. Because it reflects changes in both flow rate and speed, it offers gre ater sensitivity to traffic stream perturbations. The sensitivity of kinetic energy is illustrated in Figure 9 22 which shows the variation in volume and kinetic energy throughout the period of the incident. It is clear from t his figure that the kinetic energy dropped much more dramatically than the volume, reaching a level near zero during the incident. Other less significant variations were observed outside of the incident time range. No quantitative conclusions can be dra wn from this one example, however, it can be said that kinetic energy has a potential application for screening the archived data over long periods of time to isolate periods of perturbation for further analysis. Analysis of Managed Lanes The concept of managed lanes is gaining popularity in congested urban areas. Managed lanes fall into two categories: High occupancy Vehicle (HOV) lanes in which the use is restricted to vehicles with a specified minimum occupancy. High Occupancy Toll (HOT) lanes in w hich a toll fee is charged to all vehicles using the lane. HOV lanes have been in use for several years. HOT lanes are a more recent concept, largely because they require an ITS infrastructure to support their use. Both types of managed lanes are in use in SunGuide systems in Florida. The archive data available from STEWARD offers an excellent potential for studying the effectiveness of

PAGE 161

161 managed lanes. This section explores the measures that can be computed for this purpose and presents an example of a managed lane analysis. The procedures for evaluating the performance of managed lanes were speci fied section of roadway to provide performance information for the managed lanes and the general use lanes. The operation of this utility program is described in the ST EWARD Final Report Appendix 4d ( Courage and Lee, 2009 ) The archived data for analys is are downloaded from the STEWARD web site in the manner described previously in this dissertation. The required user specified parameters include: The lane numbers for the managed lanes The type of lane (HOT or HOV) The passenger occupancy for HOV and g eneral use lanes The cost to the motorist per vehicle mile for HOT lane use The spatial and temporal limits of the analysis Managed Lane Performance Measures The basic performance measures obtained directly from the data include: Managed Lane Volume, V m Managed Lane Speed, S m General Lane Volume, V g General Lane Speed, S g Vehicle Speed Difference, D s = S m S g Vehicle Speed Ratio, R s = S m / S g With information on the facility length the following performance measures can be derived: Vehicle miles travel ed in the managed lanes, VMT m = L* V m Vehicle miles traveled in the general lanes, VMT g = L V g Total vehicle miles traveled, VMT = VMT m + VMT g

PAGE 162

162 Travel time per vehicle in the managed lanes, TTV m = L/ S m Travel time per vehicle in the general lanes, TTV g = L S g Vehicle hours spent in the managed lanes VH m = TTV m V m Vehicle hours spent in the general lanes VH g = TTV g V g Total vehicle hours spent, VH = VH m + VH g Average vehicle speed for all lanes, VS = (VMT m +VMT g ) / ( VH m +VH g ) The to determine changes in facility productivity resulting from the managed lane operation. The operational effectiveness of a managed lane may also be assessed in an absolute sens e (i.e., without a before and after study) by comparing the average vehicle speeds and travel times in the managed lanes and the general lanes. The following measures may be obtained: Travel time difference, D tt = TTV g TTV m Speed difference, D s = S m S p S peed Ratio, R s = S m / S p Negative values in the differences or ratios less than 1.0 would indicate that the operation in the HOV lane was worse than the general lanes. Additional performance measures that could be computed for HOT lanes, based on th e price per vehicle mile, PVM, include: Cost per vehicle hour saved, CVH = PVM L /D tt Revenue, R = PVM VMT m Additional performance measures that could be computed for HOV lanes, based on the passenger occupancy in the managed lanes and the general l anes, PPV m and PPV g include: Passenger miles traveled in the managed lanes, PMT m = VMT m PP V m Passenger miles traveled in the general lanes, PMT g = VMT g PP V g Total passenger miles traveled, PMT = PMT m + PMT g Travel time per vehicle in the managed la nes, TTV m = L/ S m Travel time per vehicle in the general lanes, TTV g = L S g

PAGE 163

163 Passenger hours spent in the managed lanes PH m = VH m PPV m Passenger hours spent in the general lanes PH g = VH g PPV g Total Passenger hours spent, PH = PH m + PH g Th e a verage passenger speed PS for the facility may be computed as a passenger occupancy weighted average of the vehicle speeds in the managed lanes and the general lanes. An increase in speed for high occupancy vehicles, coupled with generally higher veh icle occupancy should increase the average passenger speed to a level greater than the average vehicle speed VS. The relationship between vehicle speeds in the HOV lanes and the general lanes provides an indication of the advantage given to the HOV lanes at the expense of the general lanes. It does not necessarily reflect the overall value of the HOV lane to the transportation system. For example, a n HOV lane that accommodates little or no traffic would provide a great advantage to its occupants but wou ld be of limited value to the transportation system. The relationship between the average passenger speed and the average vehicle speed on the facility offers a better measure of the value of the HOV lane operation because it also reflects the degree of utilization of the HOV lane, in terms of both the traffic volumes and the passenger occupancy levels. For purposes of this discussion, The followin g measures may be comput ed: HOV performance difference, PS VS, expressed in mph HOV performance ratio, PS/VS Both measures reflect the degree to which the average passenger is moving faster than the average vehicle. If there is no difference in the two speeds, then it is difficu lt to argue that the HOV lane provides any value to the transportation system.

PAGE 164

164 Example Results The performance measures have been incorporated into an experimental version of HOTTER to demonstrate their potential. There is very limited experience with th eir application at this point. A data set from a section of I 95 which now includes and HOV lane in District 4 was selected for demonstration. The common performance measures are shown below: Vehicle miles traveled in the managed lanes: 2514 5 Vehi cle hours spent in the managed lanes: 372 Average s peed in the managed lanes (mph): 67.5 Vehicle miles traveled in the general lanes: 143069 Vehicle hours spent in the general lanes: 2374 Average s peed in the general lanes (mph): 60 .3 Vehicle miles traveled in all lanes: 168214 Vehicle hours spent in all lanes: 2746 Average vehicle s peed in all lanes (mph): 61.2 Travel time per vehicle in the managed lanes (min): 22.4 Travel time per vehicle in the general lanes (min): 25.1 Travel time difference (min); 2.7 Vehicle speed difference (mph): 7.28 Vehicle speed Ratio: 1.12 It was assumed that the HOV lane had an average occupancy of 2.1 PPV and that the general lanes had an average occupancy of 1 .2 PPV. The HOV operational analysis results were as follows:

PAGE 165

165 Passenger miles traveled in the managed lanes: 52803 Passenger miles traveled in the general lanes: 171682 Total passenger miles traveled: 224486 Passenger hours spent in the man aged lanes: 781 Passenger hours spent in the general lanes: 2848 Total p assenger hours spent: 3630 Average passenger speed for the facility (mph): 64.89 HOV performance difference (mph): 3.65 HOV performance ratio: 1.06 There ar e currently no HOT lane facilities providing data to STEWARD. Therefore, t o demonstrate the HOT lane analysis capabilities of HOTTER, it was hypothetically assumed that the HOV lane was instead a HOT lane with a pricing of $ 1.00 per trip. The results ind icated a c ost of $ 2 2.19 per hour of travel time saved in comparison with the general lanes HOT lanes would normally be expected to offer a substantially greater travel time difference to attract participation by the motorist. Since this example is hypot hetical the only conclusion that can be drawn is that the speed difference associated with the HOV operation would not be worth $1.00 to many drivers. The main purpose for including the example was to illustrate the potential to evaluate a real HOT lane f rom the STEWARD data at some point in the future. These examples demonstrate the ability to produce potentially useful results; however, more experience with this application in addition to stakeholder feedback will be required before meaningful applicatio n guidelines can be developed.

PAGE 166

166 Travel Time Reliability Travel time reliability and the measures by which it can be assessed have been mentioned throughout this dissertation. The need for reporting of travel time related measures was introduced in Chapter 4. The STEWARD report that presents these measures was described in detail in Chapter 7. A number of research projects using STEWARD data for dealing with travel time reliability were summarized in Chapter 8. The use of the travel time report for assess ing incident delay was discussed earlier in this chapter. Because of the importance of travel time reliability assessment, an example will be presented here using data from STEWARD. A southbound section of Interstate 95 in Jacksonville between the entra nce from Emerson Street and the entrance from WB Butler Blvd. will be used to demonstrate the travel time reporting features. The data sample covered the period 4:00 PM to 6 :00 PM for all weekdays in 2008 A total of 253 days are represented in this exam ple Not all stations reported valid data for the entire period. The number of valid days per station ranged from 59 to 249. The average number of days of valid data per station was 211 This relatively short section (3.58.mi) was chosen for demonstrat ion to simplify the discussion. The travel time reliability table produced by STEWARD for this example is presented in Table 9 1 Most of the segments in this section were relatively congestion free, but some congestion and delay s may be observed in the segments near the center of the section. The following measures are presented in the table for each segment:: Average Speed : The segment speeds ranged from 41.36 to 73.47 mph

PAGE 167

167 Average Travel Time : The segment travel times ranged fr om 0.15 to 0.44 min/veh. The average travel time for the section was 3. 58 minutes. 95 Percentile Travel Time: This value was based on the travel time distribution. The values ranged from 0.16 to 1.12 min/veh. The 95 percentile travel time for the sectio n was 5 .36 minutes. Travel Tim e Index: This value represents the ratio of the average travel time to the travel time at the speed limit. The maximum value for any segment was 1.93. Note that some values fell slightly below 1.0, indicating that the actual speeds exceeded the speed limit by a small amount. The overall travel time index for the section was 1.29, indicating a moderate level of congestion. Percent on Time: This value represents the percent of vehicles that were able to make their trip within 10 mph of the speed limit. This value varied by segment from 30.65% to nearly 100% with the higher values at the south end of the section. The average value was 66.1%, indicating that about 1/3 of the vehicles were not able to complete their trips within 10 mph of the speed limit. Buffer Index: The buffer index is defined in Table 7 8 as (95% Travel Time Average Travel time) / Average Travel time It is intended to convey the amount of extra time a person would have to allow t o be 95% sure of being able to make the trip within the allotted time. The buffer index for the section is 0.53. Therefore, based on an average

PAGE 168

168 travel time of 4.5 minutes, a person would have to allow 4.5 (1+0.53), or 6.88 minutes for this portion of t he trip. On Time Delay: This value represents the extra time spent in the section over and above the time that would be spent at 10 mph below the speed limit. The total for the section is 0.76 minutes. Congestion Delay: This value represents the extra t ime spent in the section over and above the time that would be spent at a travel time index of 1.5. The total for the section is 0.47 minutes. The congestion delay is lower than the on time delay because it is referenced to a lower speed. Congestion del ay can generally be taken as an indication of a capacity deficiency, whereas on time delay is considered to be more related to driver satisfaction.

PAGE 169

169 Table 9 1 Res ults for travel time reliability example Av Travel Time TT % On Buffer Ontime Congestion Segment Speed Av 95% Index Time Index Delay Delay Units mph (Min/Veh) (Min/Veh) Entrance from Emerson North of Spring Glen Rd 58.07 0.44 0.68 1.19 69.76 0.56 0.05 0.02 South of Spring Glen Rd 57.34 0.4 0.74 1.28 65.4 0.86 0.07 0.04 South of University Blvd 56.46 0.39 0.65 1.26 64.03 0.69 0.05 0.03 Exit to University Blvd EB 60.81 0.37 0.54 1.12 75.5 0.49 0.03 0.01 Exit to University Blvd WB 53.68 0.34 0.63 1.39 55 0.85 0.07 0.04 B e tw ee n University and Bowden 50.63 0.26 0.48 1.49 52.03 0.86 0.07 0.05 Entrance from Bowden 44.54 0.58 1.12 1.93 39.45 0.92 0.23 0.17 South of Bowden 41.35 0.47 0.75 1.76 30.65 0.58 0.16 0.11 North of Butler Blvd 54.86 0.39 0.5 1.22 46.18 0.27 0.03 0 Exit to Butler Blvd 66.55 0.3 3 0.36 0.98 98.67 0.09 0 0 Entrance from Butler WB 73.47 0.15 0.16 0.89 99.44 0.09 0 0 Entrance from Butler EB 66.49 0.38 0.41 0.99 96.91 0.1 0 0 Totals: N/A 4.5 7.02 1.29 66.1 0.5 3 0.76 0.47

PAGE 170

170 Figure 9 1 Traffic flow ra te speed graph at Station 210511 (Oct., 2008 weekdays, morning peak: 7:00 AM~10:00 AM)

PAGE 171

171 Figure 9 2 Density fl ow rates graph at s tation 210511 (Oct., 2008 weekdays, morning peak: 7:00 AM~10:00 AM)

PAGE 172

172 Figure 9 3 Density s peed graph at s tation 210511 (Oct., 2008 weekdays, morning peak: 7:00 AM~10:00 AM)

PAGE 173

173 Figure 9 4 Comparison of density computation methods

PAGE 174

174 Figure 9 5 Number of crashes by d ay of the week Figure 9 6 Number of crashes by m onth of the year

PAGE 175

175 Figure 9 7 Number of crashes by c ounty milepost Figure 9 8 Number of crashes by t ime of day

PAGE 176

176 Figure 9 9 Incident location in aerial map

PAGE 177

177 Figure 9 10 Hourly flow rates changes with the incident

PAGE 178

178 Figure 9 11 Hourly flow rates changes without incident

PAGE 179

179 Figure 9 12 Comparison of five minute volume counts with incident and non incident case

PAGE 180

180 Figure 9 13 C umulative differences between incident and non incident volume counts

PAGE 181

181 Figure 9 14 Queuing diagram for the incident

PAGE 182

182 Figure 9 15 Occupancy changes on the incident day

PAGE 183

183 Figure 9 16 Occupancy contour graph with milepost

PAGE 184

184 Figure 9 17 Speed changes at incident

PAGE 185

185 Figure 9 18 Speed contour graph during the incident Figure 9 19 T i me space diagram for lane volume ba la nce during the incident w AB w B D (A) (B) (D) w AB 1 w B D1

PAGE 186

186 Figure 9 20 Time space diagram for lane occupancy during the incident

PAGE 187

187 Figure 9 21 Changes of speed c oefficient of v ariance (CV) during the incident

PAGE 188

188 Figure 9 22 Flow rate and kinetic energy at the point of the incident by time of day

PAGE 189

189 CHAPTER 10 CONCLUSIONS AND RECOMMENDATIONS Conclusions The goals of the project described in this dissertation were 1) to design a data archiving system capable of producing a s et of useful reports and 2) to demonstrate the value of the system to researchers and practitioners. Both of these goals have been met. This project has created an important resource for a wide variety of traffic data users in Florida, including both pr actitioners and researchers. The web site developed as a part of the project provides the capability to download several reports summarized over a range of temporal and spatial limits. The data can serve a variety of purposes including: Identif ication of detector malfunctions C alibration guidance for detectors Q uality assessment tests on data Development of d aily performance measures Fulfillment of periodic reporting requirements Evaluation of special projects, such as managed lanes. Provi sion of data for research and special studies There are several projects and activities that have already benefited from the available data. As noted previously, University of Florida and Florida International University researchers have already made good use of the data The web site has shown a continued high level of activity. It is anticipated that activity levels will increase as more data become available and awareness of the STEWARD capabilities increases. While the system implementation schedules in the district s created some delays in the provision of archive data, cooperation at the district level was excellent, and was a

PAGE 190

190 strong factor in the success of the project. A fully functional scheme is now in place for automated transmittal and processing of archive d ata. The feedback from the project team to the districts and to the SunGuide contractor was helpful in resolving some technical issues with SunGuide. The diagnostic reports furnished to the districts should be valuable in the maintenance of their detector systems. These reports indicate that, in general, the detectors are functioning well. The quality assurance procedures indicate that the completeness and validity of the data is on a par with, and sometimes exceeds, the corresponding measures in other s ystems throughout this country. The traffic volume data produced by the SunGuide archive should be useful to the district and statewide traffic counting programs. The capability to examine data from all detector stations and to create traffic count files in both the d istrict and Central Office formats should facilitate the extraction of counts. Preliminary experience indicates that the accuracy of the count data varies among stations for reasons enumerated previously. It appears, however, that with caref ul selection of stations, the FDOT traffic counting programs will benefit from the availability of the data. A number of specific observations can be made from the insight developed during the course of the project: It has been demonstrated that the arch ive data characteristics are consistent with the principles of traffic flow theory. Relationships between the macroscopic descriptors of traffic flow demonstrated good agreement with those found in the literature, and with the empirical data presented in the Highway Capacity Manual. The speed, flow rate and occupancy values produced by the RTMS detectors are not measured independently but are derived based on proprietary algorithms. A comparison of the relationship between the density estimated from the flow rate and speed values and the density estimated from the vehicle occupancy values suggests that the measures are at least internally consistent.

PAGE 191

191 Several measures such as lane volume balance ratio, speed variation, kinetic energy and effective vehicl e length were incorporated into the reports. These measures are not widely used for operational purposes but they were demonstrated through the use of examples to offer some potential for future applications. With the exception of known problem areas, usu ally resulting from construction, the traffic sensor subsystems in all districts appear to be functioning reliably and producing credible data. One specific detector problem area involves the failure of the detectors to report data during periods of extrem ely low volume, typically from, 1AM to 5AM. The cause of this problem is unknown The quantity of data that must be transmitted daily from each TMC can be accommodated by the ETL procedures that have been developed for this purpose. The current facilities for data processing are adequate to accommodate the prototype system operation, but additional speed, storage and bandwidth will be required if the system utilization expands significantly beyond its present level. The quality assurance procedures describ ed in the literature can be improved by incorporating additional QA tests that consider the relationships between the data from all of the lanes at a detector station as well as the consistency of the data between adjacent detector stations. The lane volu me balance ratio at a given station is a good example of a characteristic that is not generally considered in current procedures. The consistency of the effective vehicle length between adjacent stations provides additional useful information. Detectors t hat produce unreasonable traffic volumes generally require adjustment and calibration. Threshold levels are required for determining when volumes are too high. Investigation of the distribution of maximum flow rates suggests that a threshold level of 290 0 v eh/ln/h which is approximately 20% greater than the typical capacity suggested by the HCM is appropriate for screening detectors. Flow rates in excess of this threshold occur in less than 1 percent of the observations. The traffic counts produced by th e detectors can be extracted in a practical manner to augment the FDOT traffic counting programs. Some care needs to be exercised in choosing the appropriate detector stations and days for extraction. The desktop processing utilities have proven to be ve ry helpful for this purpose. Effective vehicle length and lane volume balance ratios offer useful information at medium to high volumes. These measures can be misleading under very light traffic. They should only be applied during the 7AM to 7PM time fra me.

PAGE 192

192 Delay resulting from incidents may be estimated by a variety of methods. Those that focus directly on density at the point of the incident are likely to produce more credible results. Speed and kinetic energy fluctuations were shown to be associated w ith the sample incident that was studied; however no quantitative conclusions could be formulated. A substantially larger study of many incidents would be required to support definitive observations. Recommendations Recommendations from this study fall in to two categories, including recommendations about future system operation and future research. Recommendations on system operation include: STEWARD should continue to operate as long as resources permit. The current usage, as evidenced by web site activ ity, justifies continued operation. It is reasonable to expect that the usage will increase. The University of Florida should continue to apply resources from the Center for Multimodal Solutions for Congestion Mitigation to the extent that they are avail able. Funding should be sought to establish a permanent home for STEWARD, probably as a part of the FDOT Intelligent Transportation Systems establishment. The STEWARD web site should be maintained and a library page should be created for reports from proj ects that utilized STEWARD data. Data from other SunGuide archive systems should be brought on board as they become operational. The step by step facility configuration process was documented in detail with that in mind. The links between STEWARD and the SunSim Project being carried out by Florida International University should be strengthened. The workshop material developed under this project should continue to be used to promote the use of STEWARD. The material was developed in accordance with the pr oject scope for live instructor delivery. This material could be expanded for interactive web delivery. In that format, it would reach a much wider audience of intended users. Districts should consider adding detectors to lanes upstream of exit ramps whe re they are currently omitted. Such detectors would greatly improve the accuracy of traffic counts extracted from the data. The additional detectors should be configured to provide archive data but not to be used in travel time estimation.

PAGE 193

193 Districts shou ld also consider placing detectors on entrance and exit ramps to form input/output analysis is very important to identifying inaccuracies in traffic volumes from detectors. STE WARD will continue to be a resource for research projects that need freeway operational data. The University of Florida and other Universities within Florida should continue to develop research proposals to further the body of knowledge in freeway operati ons and congestion management. The potential use of archived data for this purpose was covered in this dissertation. Some additional possibilities include: Extension of the procedure to examine the speed flow density relationships: These relationships are the basis of the computational methodology of Chapter 11 of the 2010 Highway Capacity Manual. More detailed knowledge of the relationships could support an improved quality control methodology. Integration of Steward system and other transportation datab ases, such as FDOT CARS, Florida highway patrol incidents, lane closure, or weather database. It would require the systematic methodology to analyze the impacts of one event in one database with other databases. For example, the relationship between the la ne closure and crash increases during the bad weather condition. Data visualization is one of the new topics in traffic data warehouses in US. This would provide the Steward stakeholders the easy access to the existing traffic databases and their applicat ions, and the new data analysis tools between the different types of databases. Investigation of the basic continuity relationship described in Chapter 3 of this dissertation. Continued research into managed lane operations using the HOTTER desktop utility program described in this dissertation. Continued research into improving the validity of traffic count data for FDOT traffic counting programs Continued research into travel time variability reporting with a view to using actual data from STEWARD instead of a surrogate modeling process. Investigation of the validity of data from various types of detectors (radar, loop video, etc.). More insight is needed into the cost/performance tradeoffs for these devices.

PAGE 194

194 Investigation of turbulence as a predictor of incidents. This would require a significant study effort involving a substantial amount of archive data at aggregation levels below the 5 minute level available from the STEWARD web site. It would also require a substantial amount of incident data. The objectives of a project of this nature would be to 1) Develop means of quantifying turbulence using traffic flow principles, 2) Develop a statistical model that describes the relationships between the turbulence measures and the incidents and 3) Verify the model with a new set of traffic data. Development of the real time identification of breakdown events : Automated procedure to identify the breakdown events on the roadways would be developed using the archived traffic data and incident logs. The real tim e analysis could be emulated and verified within the Steward system.

PAGE 195

195 LIST OF REFERENCES Bertini, R L., Hansen, S Matthews, S Rodriguez, A ., and Delcambre, A ( 2005 ) PORTAL: Implementing a New Generation Archived Data User Service in Portland Oregon. In 12th World Congress on ITS San Francisco Bertini, R L. and Makler J ( 2008 ) Beyond Archiving: Developing and Attracting Users of an Archived Data User Service In TRB 87th An n ual Meeting Washington, D.C Bertini, R L. ( 2009 ) Portland Transport ation Archive Listing (Portal). . Chen, Chao. (2003) Freeway Performance Measurement System (PeMS), California PATH research report UCB ITS PRR 2003 22 Computer Science Corporation (2005) CHART II System Architecture Mary land DOT Courage, K G., and Lee S ( 200 8) Development of a Central Data Warehouse for Statewide ITS and Transportation Data in Florida Phase II: Proof of Concept Final Report University of Florida. Courage, K G., and Lee, S. ( 200 9) Statewide Transpor tation Engineering Warehouse for Archived Regional Data (STEWARD) Final Report University of Florida. Dellenback, S and Duncan, T. ( 2008 ) SunGuide: Software Users Manual. FDOT, SwRI Duncan, T and Halbert, J. (2008) SunGuide Database Design Documen t 3.1.0. FDOT, SwRI Hallenbeck, M E., Ishimaru, J M., and Nee, J ( 2003 ) Measurement Of Recurring Versus Non Recurring Congestion, Washingt on State Transportation Center Transportation Res ea rch Board ( 2010 ) Highway capacity Manual. Transportation Re s ea rch Board Washington, D.C. Hranac, R. ( 2009 ) Session: Transforming Archived ITS data into information. In TRB 88th An n ual Meeting Washington, D.C. Kuchangi, S. ( 2006 ) A Categorical Model f or Traffic Incident Likelihood Estimation, exas A&M University Maryland DOT. (2009) CHART June 9th, 2009 < http://www.chart.state.md.us/ > May, A D. ( 1990 ) Traffic Flow Theory. Prentice Hall. Print.

PAGE 196

196 Pack, M L. ( 2009 ) Regional Integrated Transpor tation Information System. 2009 June 9 th < http://www.cattlab.umd.edu/index.php?page=research&a=00023 > PATH. (2005) Pe MS System Overview. June 8th 2009. . Tufte, K A., Ahn, S Bertini, R L., Auffray, B and Rucker, J ( 2007 ) Toward the systematic Improvement of Data Quality in the Portland, Or egon Reg ional Transportation Archive Listing (PORTAL). TRB 86th An n ua l Meeting, Washington, D. C. Turner, S Margiotta, R and Lomax T ( 2004 ) Monitoring Urban Freeways in 2003: Current Conditions and Trends from Archived Operations Data. Publication FHWA HOP 05 018, Federal Highway Administration. Turner, S. (2007) Qua lity Control Procedures for Archived Operations Traffic Data: Synthesis of Practice and Recommendations. Battelle Work Order Number 03 007 Federal Highway Administration. Washin gton, D.C. Wongsuphasawat, K Pack, M L., Filippova, D VanDaniker, M a nd Olea, A. (2008) Visual Analytics for Transportation Incident Datasets. TRB 88th Annual Meeting Was hington, D.C.

PAGE 197

197 BIOGRAPHICAL SKETCH Seokjoo Lee was born in Seoul, Republic of Korea. He ho lds a master s degree in civil engineering and in electrical a nd computer engineering from the University of Florida and a B achelor of S cience degree in electrical engineering from Korea Advanced Institute of Science and Technology.