ABSTRACT Title of Document: EVALUATION OF A SIMULATION PLATFORM FOR ASSESSING PERFORMANCE OF PROPOSED NONBARRIER-SEPARATED MANAGED LANES Xiaohan Chen Master of Science, Civil Engineering 2009 Directed By: Elise Miller-hooks This thesis seeks to ascertain whether or not a chosen simulation software platform, the VISSIM simulation platform, using creative modeling methodologies for prohibiting specified user classes from using the managed lanes and ensuring realistic transitioning and weaving behavior between lanes and at managed lane access points, can provide a suitable framework for modeling and analyzing traffic operations in freeways with nonbarrier separated managed lanes. An additional goal of this thesis is to gain insight into the potential benefits in terms of performance of proposed HOT lane facility designs developed for the State of Maryland. To accomplish this, calibration of model parameters was required. The experimental design associated with the calibration took advantage of findings from results of preliminary sensitivity tests, results of numerical experiments conceived using factorial design with the intention of assessing parameter interactions, review of related literature and advice from PTV, Inc. experts. EVALUATION OF A SIMULATION PLATFORM FOR ASSESSING PERFORMANCE OF PROPOSED NONBARRIER-SEPARATED MANAGED LANES By Xiaohan Chen Thesis submitted to the Faculty of the Graduate School of the Civil and Environmental Engineering University of Maryland, College Park, in partial fulfillment of the requirements for the degree of MS in Civil Engineering March 2009 Advisory Committee: Professor Elise Miller-hooks, Chair Professor Paul Schonfeld Mr. Phillip Tarnoff ? Copyright by Xiaohan Chen March 2009 Acknowledgements I would like to express my gratitude to all those who gave me the possibility to complete this thesis. Without their help and encouragement, this thesis cannot be finished I would like to thank Dr. Elise Miller-hooks, my advisor, for directing me to the correct direction, helping me with all the problems in this research work, and reviewing my thesis in detail. I have furthermore to thank my committee members, Professor Paul Schonfeld and Mr. Philip Tarnoff, for reviewing my thesis. I am deeply indebted to Jason Chou who kindly offered me help with the VISSIM modelling. Also, I would like to thank people from Maryland State Highway Administration who provided data required by VISSIM model. Especially, I would like to give my special thanks to my parents whose patient love enabled me to complete this work. ii Table of Contents ABSTRACT.................................................................................................................. 1 Acknowledgements....................................................................................................... ii Table of Contents.........................................................................................................iii List of Tables ................................................................................................................ v List of Figures.............................................................................................................. vi Chapter 1 Introduction............................................................................................... 1 Chapter 2 Input Data.................................................................................................. 4 2.1 Roadway Geometry ...................................................................................... 4 2.2 Traffic Volume.............................................................................................. 7 2.3 Vehicle Occupancy and Composition......................................................... 14 Chapter 3 Modeling the Existing Facility................................................................ 19 3.1 Modeling the Physical Facility ................................................................... 19 3.2 Modeling Traffic......................................................................................... 20 3.2.1 Vehicle Loading by Classification.......................................................... 20 3.2.2 Origin-Destination Modeling.................................................................. 22 3.2.3 Smooth Transitioning between Lanes and Links.................................... 22 3.3 Additional Modeling Efforts to Perfect the Existing Conditions Model .... 26 Chapter 4 Calibration............................................................................................... 30 4.1 Quality of Simulation Results given Default Parameter Settings............... 31 4.2 Details of Relevant Model Parameters ....................................................... 33 4.2.1 Parameters Impacting Physical Attributes of Vehicles........................... 33 4.2.2 Parameters Affecting Behavior Associated with Movement.................. 34 4.3 Sensitivity Analysis of Key Model Parameters .......................................... 36 4.4 Factorial Design.......................................................................................... 40 4.5 Calibration Results...................................................................................... 41 Chapter 5 Alternatives Models................................................................................ 46 5.1 Additional Data Input ................................................................................. 46 5.2 Alternative Modeling Details and VISSIM?s Suitability............................ 47 5.2.1 Origin-Destination Modeling.................................................................. 48 5.2.2 Access Control........................................................................................ 48 5.2.3 Smooth Transitioning.............................................................................. 49 5.3 Performance of Alternative Managed Lane Designs for 2030 ................... 50 5.3.1 Evaluation of Segment Travel Times, Delays and Densities.................. 50 5.3.2 Evaluation of Network Travel Times and Delays................................... 55 5.4 Conclusions................................................................................................. 58 Chapter 6 Findings................................................................................................... 59 Chapter 7 Conclusion .............................................................................................. 62 Appendix A................................................................................................................. 63 iii iv Appendix B ................................................................................................................. 65 Appendix C ................................................................................................................. 66 References................................................................................................................... 68 List of Tables Table 1 ? Number of Lanes in Existing and Alternatives Designs............................... 6 Table 2 ? The Traffic Volume per Lane of Existing and Alternatives by Segment ... 10 Table 3 ? Average Hourly Vehicle Occupancy during A.M. Peak in 2006 ............... 15 Table 4 ? Fraction within each Vehicle Occupancy Category (2006)........................ 15 Table 5 ? Vehicle Composition 2005 to 2007 ............................................................ 17 Table 6 ? Average Vehicle Composition.................................................................... 18 Table 7 ? Vehicle Class Composition - Existing 2006............................................... 21 Table 8 ? Identification of Bottlenecks....................................................................... 28 Table 9 ? Comparison of Segment Geometry, Maximum Volume and Speed .......... 29 Table 10 ? Travel Times for Existing Conditions given Default Parameter Settings 31 Table 11 ? Statistical Analysis of Existing Condition Simulation Results given Default Parameter Settings ......................................................................................... 33 Table 12 ? Parameters Associated with Lane Changing Behavior............................. 34 Table 13 ? Selected Parameters Associated with Vehicle Following Behavior......... 36 Table 14 ? Extreme Values of Parameters for Preliminary Test ................................ 37 Table 15 ? Suggested to CC-Parameters by Literature............................................... 39 Table 16 ? Look Back Distance Values...................................................................... 40 Table 17 ? 2k Factorial Design Points........................................................................ 41 Table 18 ? Calibrated VISSIM Parameters ................................................................ 42 Table 19 ? Average Travel Time of Calibrated Existing Model from VISSIM Model ..................................................................................................................................... 43 Table 20 ? Statistical Analysis of the Calibrated Existing Condition ........................ 44 Table 21 ?Vehicle Classification of Alternatives for 2030 Demand Estimates ......... 46 Table 22 ? Comparison of the Overall Performance for Entire Study Roadway ....... 56 Table 23 ? Inflow and Outflow Volume..................................................................... 58 v List of Figures Figure 1 ? Study Area: Southbound lanes of I-270 from I-370 to the Spur ................. 3 Figure 2a ? Typical Cross Section ? Existing (Shady Grove Road to Montrose Road)5 Figure 2b ? Typical Cross Section ? Existing (Montrose Road to Spur)?????..5 Figure 3a ? Typical Cross Section ? Alternative 1 (Shady Grove Road to Montrose Road)............................................................................................................................. 5 Figure 3b? Typical Cross Section ? Alternative 1 (Montrose Road to Spur)????5 Figure 4a ? Typical Cross Section ? Alternative 5 (Shady Grove Road to Montrose Road)............................................................................................................................. 6 Figure 4b ? Typical Cross Section ? Alternative 5 (Montrose Road to Spur)???...6 Figure 5 ? Access Point Locations for Alternative Designs......................................... 7 Figure 6 ? Synopsis of 2006 Traffic Volume and Turning Rates................................. 9 Figure 7? Synopsis of 2030 No Build Traffic Volume and Turning Rates.................. 9 Figure 8 ? Synopsis of 2030 Alternative 1 Traffic Volume and Turning Rates......... 10 Figure 9 ? Synopsis of 2030 Alternative 5 Traffic Volume and Turning Rates......... 10 Figure 10 ? I-270 7-mile Roadway Stretch................................................................. 12 Figure 11 ? Traffic Flow Chart for Existing 2006 VISSIM Model............................ 13 Figure 12 ? Traffic Flow Chart for Existing 2030...................................................... 13 Figure 13 ? Traffic Flow Chart for Alternative 1 2030 .............................................. 13 Figure 14 ? Traffic Flow Chart for Alternative 5 2030 .............................................. 14 Figure 15 ? Vehicle Composition Survey Station Locations...................................... 16 Figure 16 ? Vehicle Route Decision at Slip Ramp..................................................... 22 Figure 17 ? Vehicle Abruptly Crossing GP lanes to Exit Mainstream Lanes ............ 23 Figure 18a ? Smooth Transitioning to Enter the Freeway North of Shady Grove Road ..................................................................................................................................... 24 Figure 18b ?Smooth Transitioning to Exit the Freeway North of MD 189????.24 Figure 19 ? Modeling of Continuous Access ............................................................. 25 Figure 20 ? Connecting Acceleration Lane to Freeway at MD 189 On-Ramp .......... 26 Figure 21 ? Average Actual 2004 Segment Speed ..................................................... 27 Figure 22 ? Bottleneck and Hourly Traffic Volume per Lane.................................... 28 vi vii Figure 23 ? Comparison of Survey GP Lane Travel Times with Simulated Travel Times given Default Parameter Settings and Existing Conditions............................. 32 Figure 24 ? Comparison of Survey HOV Lane Travel Times with Simulated Travel Times given Default Parameter Settings and Existing Conditions............................. 32 Figure 25 ? The Sensitivity of Parameters with respect to Travel Time .................... 38 Figure 26 ? The Sensitivity of Parameters with respect to Delay............................... 38 Figure 27 ? Comparison of Calibration and Survey Average Travel Time on GP Lanes........................................................................................................................... 43 Figure 28 ? Comparison of Calibration and Survey Average Travel Time on HOV Lane............................................................................................................................. 44 Figure 29 ? Direction Decision at Slip Ramp for Alternatives................................... 48 Figure 30 ? Access Point 2 Control ............................................................................ 49 Figure 31 ? Unpermitted Weaving at Access Point.................................................... 49 Figure 32 ? Average Travel Time on GP Lanes......................................................... 51 Figure 33 ? Average Travel Time on Managed Lanes ............................................... 51 Figure 34 ? Average Hourly Delay on GP Lanes....................................................... 52 Figure 35 ? Average Delay on Managed Lanes.......................................................... 52 Figure 36 ? Average Density on I-270 Mainstream Lanes by Segment..................... 54 Figure 37 ? Comparison of Overall Performance for Entire Study Roadway............ 56 Chapter 1 Introduction As has been demonstrated in various regions within the United States, the use of Express Toll Lanes (ETLs) or similarly functioning High Occupancy Toll (HOT) lanes can lead to more effective use of existing roadway capacity, improved traffic flow along general purpose lanes and additional revenue to support much needed transportation improvements. This thesis describes outcomes and efforts taken in the second phase of a multi-phase research effort to develop an application of a simulation model for the analysis of managed lanes adjacent to general purpose lanes (concurrent flow lanes). Phase I of this project sought to develop a comprehensive understanding of the current state-of-the-art in modeling and analysis of nonbarrier separated electronic/high occupancy toll (HOT) lane and other concurrent flow lane operations as reported in (Miller-Hooks, Tarnoff, Chen and Chou, 2008). As part of the initial effort, information was gathered through interviews conducted with project managers of existing and proposed HOT lane facilities, modelers and other domain experts and review of related reports and literature. Details of models employed, and analytical tools used, to evaluate the impact of proposed HOT lanes on traffic operations and potential revenue; supplemental analysis tools; lane configurations; tolling strategies; High Occupancy Vehicle (HOV) restrictions; types of separation; how weaving is addressed; and design alternatives for ingress and egress between the HOT and general purpose lanes were provided. Knowledge pertaining to model calibration and validation was gleaned from the interview and literature review processes. Potential data sources for calibrating developed models were also identified. Finally, a proof-of-concept was developed to illustrate how details associated with violation modeling can be handled in the selected modeling framework, the VISSIM simulation platform, which was proposed for use in this and additional subsequent phases of this research effort. The VISSIM micro-simulation platform was chosen over other traffic simulators, because this platform had been successfully employed in modeling the impact of proposed HOT lane facilities on traffic operations in several studies conducted across the country as described in (Miller-Hooks, Tarnoff, Chen and Chou, 2008). While nearly all of these 1 models treated the HOT lane facility as a separate link, effectively modeling a barrier separated facility, preliminary work within the platform indicated that this platform could also successfully be used to model nonbarrier separated facilities. The primary purpose of this second phase of this research effort was to ascertain whether or not the chosen simulation software platform, the VISSIM simulation platform, and modeling methodologies provide a suitable framework for modeling and analyzing traffic operations, including the specific details associated with modeling concurrent flow lanes with designated access points, along significant portions of the Maryland freeways. While the intended use of these lanes is for non-intrusive (barrierless) tolling, the model will also be useful for studying the performance of HOV lane operations. An additional goal of this research phase was to gain initial insight into the performance of proposed HOT lane facility designs. To complete this assessment, simulation models of four managed lane design alternatives associated with a 7-mile stretch of I-270 within the State of Maryland were developed: two existing condition models with mostly continuous access HOV lane operations under 2006 and 2030 traffic demand and two alternative models with limited access HOT lane facilities under 2030 demand. The study area, a segment of Southbound I-270 from I-370 to the Spur, is depicted in Figure 1. The study period is from 6:00 a.m. through 9:00 a.m., i.e. the morning peak hours. Parameters of the existing conditions model with 2006 traffic demand were calibrated based on actual traffic measurements. This thesis describes the developed simulation models, data employed within the modeling and calibration efforts, efforts taken to calibrate the existing conditions model, and results and findings from the assessment of the calibration effort and evaluation of proposed design alternatives. 2 Figure 1 ? Study Area: Southbound lanes of I-270 from I-370 to the Spur Data related to roadway geometry, traffic volume, vehicle composition, and vehicle occupancy are required for the development of the VISSIM model of the 7-mile stretch of I-270. Details associated with the preparation of these required input data are given in Chapter 2. In Chapter 3, the general approach to modeling the study roadway segment and specific implementation details are presented. Difficulties that arose in the modeling effort are described and measures taken to overcome these difficulties are provided. Once created, preferred parameters for use in the VISSIM model of the study roadway segment under existing conditions were identified through extensive calibration efforts as described in Chapter 4. In Chapter 5, proposed alternative designs for the nonbarrier separated HOT lane facility that would replace the HOV lane facility were described. The approach employed within this effort to model access points to the HOT lane facilities is presented and results of analysis of the proposed design alternatives are given. Finally, findings from this research effort, including an assessment of the simulation tool?s adequacy in replicating actual traffic on managed lanes, and evaluation of various concurrent flow lane design alternatives are summarized in Chapter 6. 3 Chapter 2 Input Data Data related to roadway geometry, traffic volume, vehicle composition, and vehicle occupancy are required for the development of the existing conditions and proposed alternative VISSIM models of the 7-mile stretch of I-270. Details associated with the preparation of these required input data are given next. 2.1 Roadway Geometry The geometry of the study roadway segment, including characteristics of the interchanges, and general purpose, HOV and collector-distributor (CD) lanes, were extracted from maps available through GoogleMap. A scale of 1:100 meters was employed for this purpose. The study roadway segment consists of six interchanges connecting I-270 with local roads, including I-370 freeway, Shady Grove Road, Montgomery Avenue (MD 28), Falls Road (MD 189), Montrose Road, and the Spur connection to I-495. The interchanges involve eight on-ramps from local roads to CD lanes, five off-ramps from the CD lanes to the local roads, four slip ramps from CD lanes to general purpose (GP) lanes, and two slip ramps from GP lanes to CD lanes. The I-270 facility hosts a single HOV lane in the southbound direction. This lane spans the entirety of the seven-mile study segment and beyond. The HOV lane splits at the Spur, connecting to I-495 Southbound and Eastbound. Continuous-access to the HOV lane is permitted from the northern-most point of the study roadway segment (at the I-370 interchange) to one mile north of the Spur, at which point access is closed via solid striping. The study roadway can be divided into three segments with constant cross section, the latter two of which are depicted in Figure 2: I-370 to Shady Grove Road, Shady Grove Road to Montrose Road and Montrose Road to the Spur. There are three, rather than two, southbound CD lanes from I-370 to Shady Grove Road. All lanes within the study roadway segment have 12-foot widths. 4 Figure 2a ? Typical Cross Section ? Existing (Shady Grove Road to Montrose Road) Figure 2b ? Typical Cross Section ? Existing (Montrose Road to Spur) Two alternative HOT lane facility designs were considered in this study. The first, known as Alternative 1, employs the existing road layout converting the HOV lane facility to a single, limited access non-barrier separated HOT lane. The second, known as Alternative 5, converts the HOV lane facility to a limited access non-barrier separated HOT lane facility with two HOT lanes. This design accommodates two HOT lanes by converting the inside shoulder (and reducing the shoulder width), as well as the HOV lane, and restriping. Cross sections for these alternatives for portions of the study roadway segment are illustrated in Figures 3 and 4. Figure 3a ? Typical Cross Section ? Alternative 1 (Shady Grove Road to Montrose Road) Figure 3b? Typical Cross Section ? Alternative 1 (Montrose Road to Spur) 5 Figure 4a ? Typical Cross Section ? Alternative 5 (Shady Grove Road to Montrose Road) Figure 4b ? Typical Cross Section ? Alternative 5 (Montrose Road to Spur) Source from: WestSide_Typical_Sections_4-07, Maryland State Highway Administration (SHA) The number of CD, GP and managed lanes for each portion of the study roadway segment are given in Table 1. Table 1 ? Number of Lanes in Existing and Alternatives Designs Segment I-370 to Shady Grove Road Shady Grove Road to Montrose Road Montrose Road to Spur Type of Lane CD Lane GP Lane HOV/ HOT CD Lane GP Lane HOV/ HOT CD Lane GP Lane HOV/ HOT Existing 3 3 1 HOV 2 3 1 HOV 0 5 1 HOV Alternative 1 3 3 1 HOT 2 3 1 HOT 0 5 1 HOT Alternative 5 3 3 2 HOT 2 3 2 HOT 0 5 2 HOT For each alternative HOT lane facility design, three access points to the facility were designated (depicted in Figure 5): Access Point 1: 3,000 feet long; beginning 0.5 mile south of I-370 and ending at Shady Grove Road; allowing access from managed lanes to slip ramp to CD lanes located south of Shady Grove Road. Access Point 2: 3,000 feet long; beginning 2,000 feet north of MD 28 and ending 1,000 feet south of MD 28; allowing vehicles in CD lanes to access managed lanes from slip ramp 6 located south of Shady Grove Road and vehicles from managed lanes to access slip ramp to CD lanes located north of MD 189. Access Point 3: 2,500 feet long; beginning 400 feet north of Montrose Road and ending 2,100 feet south of Montrose Road; allowing vehicles in CD lanes to access managed lanes from slip ramp located north of Montrose Rd. Figure 5 ? Access Point Locations for Alternative Designs 2.2 Traffic Volume The VISSIM simulation software platform permits the input of traffic volume data (i.e. the demand) to be provided in either of two formats: as an origin-destination (O-D) matrix (indicating the number of vehicles that desire to travel between each O-D pair) or using turning percentages at interchanges and between CD lanes, GP lanes and other concurrent flow lanes. With either format, vehicle movements can be simulated between origins and destinations over 7 the simulation period. Additionally, the O-D matrix and turning percentages may be dynamic, i.e. they may vary over the course of the simulation period. For example, O-D matrices may be created for every 15-minute interval. In this study, turning percentages are used to direct traffic between lanes and to potential destinations (specifically, exits from I-270). Traffic demand is assumed to be constant over the study period and, thus, only a single set of turning percentages is employed. Traffic demand and related turning percentages throughout the study roadway segment for existing conditions and the proposed alternatives were set based on two main sources of data: 1. Balanced morning-peak average hourly traffic volumes and turning rates of on- and off- ramps at interchanges employed within the Maryland SHA Western Mobility Study 2006 (Appendix A). 2. Maryland SHA CORSIM model estimates of turning rates on slip ramps between CD and GP lanes. The 2006 existing condition traffic volumes were computed using data collected from the field. Traffic volume predictions were also completed for 2030 for each of three possible roadway geometries: No Build, Alternative 1 and Alternative 5. Traffic volumes and turning percentages for 2006 existing conditions and 2030 predictions obtained from these data sources are synopsized in Table 2 and Figures 6 through 9. These data were collected for each of five segments along the entire study roadway segment, depicted in Figure 10. Note that the middle three segments together constitute the Shady Grove Road to Montrose Road segment of the three-segment study roadway depiction used in Section 2.1 to describe roadway geometry. Turning rates along the study roadway segment for the 2030 forecast year were set to ensure consistency in flow across the segments. 2006 slip ramp usage rates were employed with some modifications that were applied to ensure consistency with 2030 demand estimates, which were given by lane classification (CD, GP or managed lanes). 8 Figure 6 ? Synopsis of 2006 Traffic Volume and Turning Rates Figure 7? Synopsis of 2030 No Build Traffic Volume and Turning Rates 9 Figure 8 ? Synopsis of 2030 Alternative 1 Traffic Volume and Turning Rates Figure 9 ? Synopsis of 2030 Alternative 5 Traffic Volume and Turning Rates Table 2 ? The Traffic Volume per Lane of Existing and Alternatives by Segment 2006 Existing HOV GP CD Segment # of lanes volume % # of lanes volume % # of lanes volume % Total Volume 1 1 1478 15 3 6206 63 3 2167 22 9851 2 1 1261 13 3 5626 58 2 2813 29 9700 3 1 1293 12 3 6034 56 2 3448 32 10775 4 1 1287 12 3 5363 50 2 4076 38 10726 5 1 1404 13 3 9396 87 -- -- -- 10800 10 2030 No Build HOV GP CD Segment # of lanes volume % # of lanes volume % # of lanes volume % Total Volume 1 1 1613 15 3 6773 63 3 2365 22 10751 2 1 1268 13 3 5655 58 2 2828 29 9751 3 1 1197 12 3 5586 56 2 3192 32 9975 4 1 1251 12 3 5213 50 2 3962 38 10426 5 1 1469 13 3 9831 87 -- -- -- 11300 2030 Alternative 1 HOV GP CD Segment # of lanes volume % # of lanes volume % # of lanes volume % Total Volume 1 1 1600 15 3 6771 63 3 2379 22 10750 2 1 1600 16 3 5561 56 2 2739 28 9900 3 1 1600 16 3 4731 48 2 3569 36 9900 4 1 1600 15 3 5775 56 2 2975 29 10350 5 1 1600 14 3 9650 86 -- -- -- 11250 2030 Alternative 5 HOV GP CD Segment # of lanes volume % # of lanes volume % # of lanes volume % Total Volume 1 2 3300 27 3 6586 54 3 2314 19 12200 2 2 3300 29 3 5544 48 2 2731 24 11575 3 2 3325 29 3 4688 41 2 3537 31 11550 4 2 3325 27 3 6023 48 2 3103 25 12451 5 2 3225 24 3 10200 76 -- -- -- 13425 11 Figure 10 ? I-270 7-mile Roadway Stretch Segment 1: I-370 to Shady Grove Road Segment 2: Shady Grove Road to MD 28 Segment 3: MD 28 to MD 189 Segment 4: MD 189 to Montrose Road Segment 5: Montrose Road to Spur The data from Figures 6 through 9 was employed in computing the traffic volume to be loaded into the links of the VISSIM model and turning percentages to be employed between CD and GP lanes and at interchanges. This information is depicted in the traffic flow chart presented in Figure 11. Similar figures are provided in Figures 11 through 14 for the 2030 options. 12 Figure 11 ? Traffic Flow Chart for Existing 2006 VISSIM Model Figure 12 ? Traffic Flow Chart for Existing 2030 Figure 13 ? Traffic Flow Chart for Alternative 1 2030 13 Figure 14 ? Traffic Flow Chart for Alternative 5 2030 2.3 Vehicle Occupancy and Composition Vehicle occupancy, i.e. the number of occupants (including the driver) riding in each vehicle, is a significant characteristic in terms of describing a vehicle?s type in the context of this managed lane study. That is, a vehicle will be permitted to use the HOV lane in the existing conditions model during the study period only if that vehicle contains 2 or more occupants. Likewise, a vehicle will be permitted to use the HOT lane(s) in the alternative roadway configurations considered in this study if that vehicle contains 2 or more occupants or is a suitably equipped HOT lane user. Average morning peak-hour hourly vehicle occupancy data employed within this study was based on data obtained via a survey conducted between 6:00 a.m. and 9:00 a.m. on May 23, 2006. This survey was conducted at one hour intervals at two stations, one located north of Democracy Boulevard and the second located south of Shady Grove Road. Vehicles were categorized as one of several types: personal cars with a single occupant (the driver), personal cars with a driver and one or more passengers, buses (assumed to carry 20 passengers), and trucks. Each lane was counted separately and the average per lane hourly occupancies were computed. The relevant average morning peak-hour number of vehicles per lane per hour by occupancy category is shown in Table 3. 14 Table 3 ? Average Hourly Vehicle Occupancy during A.M. Peak in 2006 Vehicle Type Lane 1* 2+** Buses Trucks Southbound I-270 Spur North of Democracy Boulevard Lane 1 ? GP 659 24 1 13 Lane 2 ? GP 1607 134 3 123 Lane 3 ? GP 1969 76 0 44 Lane 4 ? HOV 161 484 5 10 Southbound I-270 South of Shady Grove Road Lane 1 ? CD 1278 124 8 92 Lane 2 ? CD 1587 98 3 70 Lane 3 ? GP 709 43 2 24 Lane 4 ? GP 1535 227 4 125 Lane 5 ? GP 1693 13 0 39 Lane 6 ? HOV 278 1128 17 9 * Passenger cars or vans with occupancy equals to one. ** Passenger cars or vans with occupancy higher than one. The fraction within each category (i.e. the number of vehicles within each category as a fraction of the total number of vehicles in the roadway segment) is presented in Table 4. Note that it was assumed that this fraction is constant over the entire segment. Table 4 ? Fraction within each Vehicle Occupancy Category (2006) Segment Lane Total 1* 2+* Buses Trucks GP 4235 79.7% 238 4.5% 4 0.1% 181 3.4% I-370 to Montrose Road HOV 5315 161 3.0% 490 9.2% 5 0.1% 10 0.2% CD+GP 6803 74.7% 521 5.7% 16 0.2% 350 3.8% Montrose Road to Spur HOV 9106 278 3.0% 1145 12.6% 17 0.2% 9 0.1% * Passenger cars or vans with occupancy of one. ** Passenger cars or vans with occupancy higher than one. Additional survey data (provided by the Maryland SHA) obtained from six survey 15 stations shown in Table 5 were available for use in this study. The location of survey stations are shown in Figure 15. For the five survey stations located between I-370 and Montrose Road, a 48- hour vehicle composition survey was taken each year during August of 2005 and April, May and August of 2007. For survey stations located between Montrose Road and the Spur, the 48-hour vehicle composition survey was taken in August of 2005 only. Traffic counts by vehicle class were recorded at one hour intervals. The following classes were considered. ? Class 1 ? Motorcycles (MC); ? Class 2 ? Passenger Cars; ? Class 3 ? Light Trucks; ? Class 4 ? Buses; ? Classes 5-9 ? Single-Trailer Trucks; and ? Classes 10-13 ? Multi-Trailer Trucks. Figure 15 ? Vehicle Composition Survey Station Locations The fraction of vehicles falling within each category was obtained by dividing the number of vehicles of a given class by the total number of vehicles counted. For consistency with other sources of input data, all taken from 2006, where available, the average of the 2005 16 and 2007 fractions was computed for each vehicle type. Table 5 shows the average vehicle composition fractions computed from this second data source for each station. Table 5 ? Vehicle Composition 2005 to 2007 Station* Year Truck** Bus Car/MC/Light Truck 2005 6.01% 0.40% 93.59% 2006 6.18% 0.39% 93.43% B2966/S025 2007 6.34% 0.39% 93.27% 2005 5.04% 0.46% 94.50% 2006 4.27% 0.46% 95.27% B2965 2007 3.49% 0.46% 96.05% 2005 5.03% 0.51% 94.47% 2006 5.82% 0.52% 93.65% B2847/S024 2007 6.62% 0.54% 92.84% 2005 5.42% 0.50% 94.08% 2006 5.90% 0.52% 93.59% B2848/S023 2007 6.37% 0.54% 93.10% 2005 5.73% 0.51% 93.76% 2006 5.82% 0.50% 93.68% B2849/S121 2007 5.92% 0.49% 93.59% 2005 6.07% 0.66% 93.27% 2006 6.07% 0.66% 93.27% B2850 2007 N/A N/A N/A * Refer to Figure 15 for station numbering. ** Trucks include Classes 5-13. MC=motorcycle In Table 6, the average vehicle composition over all relevant stations is given for each roadway segment. A number of assumptions were required in finalizing the composition values: 1. Vanpools shown in the raw occupancy data were treated as passenger cars with two or 17 more vehicle occupants. 2. Several different types of trucks are found in the raw vehicle composition data. Light trucks are counted and treated in the model as passenger cars and all other truck types are classified under the truck category, i.e. assuming that they are heavy vehicles. 3. No trucks are allowed in the HOV lane. 4. Motorcycles are modeled as single passenger cars and are not permitted to use the HOV lane. Table 6 ? Average Vehicle Composition Road Segment Truck** Bus Car/MC/Light Truck I-370 to Montrose Road 5.60% 0.48% 93.92% Montrose Road to the Spur 6.07% 0.66% 93.27% ** Trucks include Class 5-13. This second source of data, while studied, was not employed as input to the VISSIM model constructed for this study. The data source described in the previous section was obtained for 2006 directly and provided the additional required occupancy data. It is worth noting that a larger percentage of vehicles fall in the Car/Motorcycle/Light Truck category as obtained from the 2006 occupancy data than from this second data source. In both sources, it is assumed that vehicle composition does not change over time or as a function of volume or other traffic characteristics. No vehicle occupancy and composition data were provided for 2030 traffic volume. Thus, the vehicle classification as input to the VISSIM alternatives models would be computed from the existing occupancy and composition data as shown above. This will be discussed in Section 5.1. 18 Chapter 3 Modeling the Existing Facility A VISSIM simulation model was developed (using version 4.3) to replicate the existing facility along I-270 in the study area, including current traffic patterns, volumes, and driver behavior. While traffic conditions, including traffic volume, vehicle composition, and vehicle occupancy vary over the morning peak (6:00 a.m. to 9:00 a.m.), i.e. the study period, it was assumed that conditions were static over the period. Thus, a static modeling approach was employed. A VISSIM simulation model was developed to replicate the existing facility along I-270 in the study area, including current traffic patterns, volumes, and driver behavior. While traffic conditions, including traffic volume, vehicle composition, and vehicle occupancy vary over the morning peak (6:00 a.m. to 9:00 a.m.), i.e. the study period, it was assumed that conditions were static over the period. Thus, a static modeling approach was employed. In Section 3.1, details of the construction of the model with respect to the facility design are given. This is followed by a description of traffic modeling in Section 3.2. Additional modeling efforts required to perfect the existing conditions model are presented in Section 3.3. Rather than articulate generic techniques employed in creating a VISSIM model, details given in this chapter focus on major decisions taken in the modeling effort and nonstandard modeling techniques employed to better reflect real-world traffic movements. 3.1 Modeling the Physical Facility In constructing the VISSIM model of the existing physical facility, two parallel links, with connecting links were employed. One of the parallel links is used to model the CD lanes and the other is used to model the GP and HOV lanes. Separate links connect the CD lanes with the GP lanes and neighboring roads with the CD lanes. The HOV lane is modeled using the same link as is used to model the GP lane to provide continuous access between the lanes as needed (i.e. between I-370 and Tuckerman Lane, which is one mile north of the Spur). All links employed in the model are of the ?freeway? type as categorized in the simulation platform. This 19 implies that model parameters associated with the ?freeway? category will be set identically for the entire facility. In Chapter 4, these parameters are calibrated. In the calibration effort, if such a parameter is changed, it is changed for all links whose type falls in the same category. 3.2 Modeling Traffic A number of considerations must be taken in modeling traffic within the simulation model. First, vehicles must be loaded into the network model. Vehicles are classified within one of eight categories that represent both vehicle type (e.g. truck, bus, passenger car) and whether or not the vehicle is eligible to and will use the HOV lane. The volume within each vehicle classification must be consistent with the traffic composition and occupancy data obtained from actual traffic conditions as determined from the input data described in Chapter 2. Second, the number of vehicles within each category of classification must be set for each origin and destination. Finally, smooth transitioning of traffic between the CD, GP and HOV lanes must be facilitated. Details of each of these components of modeling the traffic are described in the following subsections. Each run of the VISSIM model entailed 5,400 seconds of simulation time, the first 1,800 seconds of which was considered as the warm-up period. Average results when provided in this thesis, unless otherwise specified, are hourly averages based on the 3,600 seconds of simulation run time. 3.2.1 Vehicle Loading by Classification Vehicles are classified so that vehicles falling within the same class have similar characteristics. For example, vehicles in the same class are assumed to have similar physical features (e.g. same length and weight class), acceleration/deceleration rates (i.e. distributions), occupancies and desired speeds. Additionally, eligibility and desire to use HOV lanes is considered. Only those vehicles falling in classes 2, 4, 6, and 7 use the HOV lane. Eight classes of vehicles were created for use in the existing conditions model: (1) trucks that use only CD and GP lanes (2) trucks that use only CD, GP and HOV lanes (i.e. HOV lane violators) (3) buses that use only CD and GP lanes 20 (4) buses that use CD, GP and HOV lanes (5) single occupancy vehicles (SOVs), i.e. passenger cars or vans with only one passenger onboard, that use only CD and GP lanes (6) single occupancy vehicles (SOVs), i.e. passenger cars or vans with only one passenger onboard, that use CD, GP and HOV lanes (i.e. HOV lane violators) (7) HOVs (passenger cars or vans with more than one person on board) that use CD, GP and HOV lane (8) HOVs (passenger cars or vans with more than one person on board) that use only CD and GP lanes Classes 2 and 6 model trucks and passenger cars that violate the occupancy and vehicle classification restrictions of the HOV lane facility. These violators are assumed to behave similarly to comparable vehicles permitted to legally use the HOV lane in all other respects. Note that there is a short segment of solid striping in the southern most portion of the study roadway segment. A vehicle that crosses the solid striped line would commit an alternative form of violation. Only violations associated with vehicle occupancy and classification restrictions are considered in this study. The composition in terms of these eight classes used in creating the existing conditions model is given in Table 6. The values shown in Table 7 were obtained from the composition and occupancy data described in Chapter 2. Table 7 ? Vehicle Class Composition - Existing 2006 Composition (%) Class Type Occupancy Using HOV? I-370 to Montrose Road Montrose Road to Spur Class 1 truck 1 no 3.4 3.8 Class 2 truck 1 yes 0.2 0.1 Class 3 bus 2+ no 0.1 0.2 Class 4 bus 2+ yes 0.1 0.2 Class 5 passenger car 1 yes 3.0 3.0 Class 6 passenger car 1 no 79.7 74.7 Class 7 passenger car 2+ yes 9.2 12.6 Class 8 passenger car 2+ no 4.5 5.7 21 For trucks and buses, i.e. Classes 1 through 4 employed within the alternatives VISSIM models, the desired speed is set to 43.5 mph, ranging from 42.3 to 48.5 mph. For Classes 5 through 8 employed within the alternatives VISSIM models, the desired speed is set to 50 mph, ranging from 47 to 68 mph. The default linear distribution of speeds was employed and default acceleration and deceleration rates by vehicle type were used. 3.2.2 Origin-Destination Modeling VISSIM permits two methods for controlling vehicle destinations as mentioned previously. In creating the existing conditions model of the study roadway segment, the VISSIM methodology that employs turning rates at all major decision points to achieve required volume exiting at each destination is used. This methodology permits the modeler to set specific decision points from which two or more choices for travel destinations are available. A destination in this context may be an exit from the facility or it may be the decision to travel between CD and GP lanes via slip ramps. The use of turning rates in this context is illustrated in Figure 16 for a single vehicle classification. For this vehicle class, 74% of the vehicles reaching the bar (i.e. the route decision starting point) at the left end of the figure will continue in the mainstream, while 26% will follow the slip ramp as depicted to access the CD lanes. The bars at downstream of the roadway indicate a destination for the decision. Turning rates may vary across vehicles classes. Figure 16 ? Vehicle Route Decision at Slip Ramp 3.2.3 Smooth Transitioning between Lanes and Links Additional modeling effort is required to: prevent vehicles from taking very late decisions that, for example, might call for the vehicle to abruptly cross multiple lanes to exit the facility; prevent vehicles from stopping in a lane while waiting for an appropriate gap to change lanes; prohibit certain vehicle classes from using the HOV lane, while simultaneously allowing other 22 vehicle classes to have continuous access to that lane for the majority of the study roadway segment length; and facilitate smooth transitions between connected links in the model. The first two issues associated with smooth lane changing movements are addressed in Subsection 3.2.3.1. The modeling techniques used to simulate continuous access to the HOV lane for a subset of vehicle classes is described in Subsection 3.2.3.2. This is followed by a description of the methodology for ensuring smooth transitions at network model connections in Subsection 3.2.3.3. 3.2.3.1 Lane Changing Movements Upon first running the created VISSIM model for the study roadway segment, vehicles in the model would often abruptly cross multiple lanes to exit or enter various portions of the segment. This abrupt action involved stopping of vehicles in the middle of a stretch to switch lanes. The stopping behavior occurred as the vehicle waited for a suitable gap for maneuvering to a neighboring lane. That is, when such a gap did not arise upon the vehicle?s decision to switch lanes, the vehicle stops to wait for such a gap. Figure 17 provides an example of such behavior. In the figure, an HOV vehicle (Class 7) attempts to exit the facility from the HOV lane, requiring the vehicle to cross three GP lanes, enter and exit the slip ramp, cross two CD lanes and enter the off-ramp. The vehicle does not take a decision to exit until it reaches a location that is very close to the deceleration lane; thus, it is not possible to cross the GP lanes smoothly without passing the slip ramp and an abrupt crossing action is depicted. Such behavior requires that the vehicle stop to wait for an appropriate gap to change lanes. Other vehicles are interrupted and lanes of the freeway become blocked as a consequence of this behavior. Figure 17 ? Vehicle Abruptly Crossing GP lanes to Exit Mainstream Lanes While some drivers may behave as depicted in Figure 17, most do not. Most vehicles prepare for such decisions through lane changing behavior that facilitates a smoother transition. 23 This desirable and more realistic behavior is depicted in Figure 18. Figure 18 illustrates vehicles maneuvering from an on-ramp to the HOV lane via CD and GP lanes, using a slip ramp, as well as from the HOV lane to an off-ramp via GP and CD lanes, using a slip ramp. The behaviors depicted in this figure show smooth transitions between the HOV lane and the on- and off-ramps. Lanes 1 through 3 are GP lanes, Lane 4 is classified as an HOV lane, and Lanes 5 and 6 are CD lanes in the figure. Figure 18a ? Smooth Transitioning to Enter the Freeway North of Shady Grove Road Figure 18b ?Smooth Transitioning to Exit the Freeway North of MD 189 To achieve the smooth transitioning of vehicles as depicted in Figure 18, two major actions were taken: 1. The length of the Decision Route (defined in Figure 16) is extended or contracted to ensure that vehicles exiting or entering the facility or a portion thereof will have enough time to smoothly change lanes if required. 2. The Look Back Distance associated with any connector (discussed in Section 3.2.3.3) that is used with an exit or entrance is extended such that the vehicles are able to recognize the exit or entrance prior to arriving at the connector. 24 3.2.3.2 HOV Lane Access Control The HOV lane was modeled as a separate lane, as opposed to separate facility (i.e. link). This modeling approach allows HOV users to move freely between the GP and HOV lanes in portions where continuous access is permitted, as is the case between I-370 and Montrose Road under existing conditions. Figure 19 depicts this free movement of HOV users between lanes. Continuous access allows eligible HOV lane users to choose between the HOV and GP lanes as traffic conditions change. While some violators, as depicted, may illegally use the HOV lanes, most single occupant or otherwise non-HOV users will not use the HOV lane. These vehicles are shown in yellow in the figure. Some action was required to prevent these non-HOV users from moving into the HOV lane within the model. Figure 19 ? Modeling of Continuous Access To prevent non-HOV users from using the HOV lane, the lane closure property of the HOV lane is set to ?closed? for the non-HOV users, but to ?open? for the HOV users and violators. GP lanes are open to all the vehicle classes. 3.2.3.3 Transitioning between Model Links via Acceleration and Deceleration Lane Connectors A typical approach to modeling acceleration/deceleration as required at connections between the CD lanes and the local roads or GP lanes (via on-, off- or slip ramps) is to use connectors (so-called tapering). Thus, for example, to model acceleration between an on-ramp and the CD lanes, a connector would connect the on-ramp link with the link representing the CD- lanes. To make this connection, the link representing the CD-lanes would need to be broken at the site of the on-ramp. Thus, a second connector would be required to connect the two portions of the CD lane link. This modeling approach, however, results in a conflict between vehicles approaching from the upstream CD lanes and the on-ramp. Since the two connectors are operated 25 separately, the behavior of the vehicles at this connection location will be haphazard and may even result in two vehicles being present at the same location at the same point in time. Rather than use this more typical modeling methodology for connecting the on-ramp with the CD lanes (or making other similar connections required within the model), only one connector is used. Without additional modeling work, the use of one connector may result in the sudden loss of vehicles from the model, because the vehicles are not told to switch lanes. Thus, route decision points are added to the model at the intersection of the on-ramp and the CD lanes (or other similar connections) as depicted in Figure 20. Figure 20 ? Connecting Acceleration Lane to Freeway at MD 189 On-Ramp 3.3 Additional Modeling Efforts to Perfect the Existing Conditions Model Once traffic is loaded into the network model (employing modeling techniques described in prior subsections), runs were made to assess whether or not traffic is replicated in a way that mimics reality. A major consideration in assessing how well the model did after its creation was its ability to replicate conditions at bottlenecks. To evaluate the model with respect to bottlenecks, two steps were followed. First, bottlenecks in the model were identified using speed data from the input data of actual traffic conditions. Speed differentiation across segments was studied. Second, for each discovered bottleneck, the cause of that bottleneck was surmised based on the roadway geometry. The simulation model was suitably modified to more accurately reflect the bottleneck conditions once identified. Modifications primarily involved changes in Route Decision length. Model parameters, such as the Look-Back Distance parameter mentioned previously, were also adjusted accordingly. Average speed data were taken from a vehicle travel time survey conducted by the Maryland SHA during several morning peak periods (6:00 a.m. and 9:00 a.m.) in April of 2004. 13 samples were provided after removing those data associated with abnormal conditions (e.g. in 26 the event of a traffic incident). The travel time data were given by roadway segment, where speed for each segment (segmentation depicted in Figure 10) is as shown in Figure 21. Segment speeds were calculated from the segment travel times and lengths. The average speed for each segment was taken over all 13 sample speeds (given in Figure 21). Figure 21 ? Average Actual 2004 Segment Speed 44 31 44 63 46 0 10 20 30 40 50 60 70 12345 Roadway Segment Speed (mp h ) Three categories of congestion are defined based on the average speeds: Severely Congested (average speed less than 40 mph), Moderately Congested (average speed between 40 and 50 mph) and Uncongested (average speed no lower than 50 mph). As depicted in Figure 21, Segments 1, 3 and 4 are considered to be moderately congested, Segment 2 is severely congested and Segment 5 is uncongested. Based on the congestion designations, Segments 1 through 4 were considered further. It was assumed that no bottleneck exists within Segment 5. The next step in identifying bottlenecks was to consider the roadway geometry and average hourly traffic volume per lane. To do so, the details as shown in Figure 11 were studied. That is, if the geometry allows for a decision, such as to exit the facility, and simultaneously the traffic volume is found to be high in a nearby location, a bottleneck is suspected. Five such bottlenecks were identified along the study roadway segment, as depicted in Figure 22. Note that the average hourly traffic volume per lane was computed 27 from the traffic volumes given in Figure 11 of Section 2.2 divided by the associated number of lanes. Figure 22 ? Bottleneck and Hourly Traffic Volume per Lane Table 8 ? Identification of Bottlenecks Bottleneck Location A Slip ramp to GP lane from CD lane before Shady Grove Road B Slip ramp to GP lane from CD lane between Shady Grove Road and MD 28 C Slip ramp to GP lane from CD lane between Shady Grove Road and MD 28 D Slip ramp to GP lane from CD lane north of Montrose Road E On CD lane between on- and off-ramps at Montrose Road Bottlenecks cannot necessarily be identified by comparing nominal traffic volumes. For example, the high average hourly traffic volume per lane (2,019 vehicles per lane per hour (vplph)) occurs between Montrose Road and the Spur. Moreover, the average speed for this segment is 63 mph. This segment (Segment 5), thus, is considered to be uncongested. On the contrary, average hour traffic volume per lane in Segment 2, which is considered to be Severely Congested, is not particularly high as compared with similar measurements in other portions of 28 the study roadway segment. The average speed for this segment is 31 mph. Thus, the average hourly traffic volume is lower in value on Segment 2 than Segment 5 as a consequence of the lower speed due to increased congestion. If only volume were considered, one might mistakenly rate Segment 2 as uncongested. To improve the accuracy in determining whether or not a segment of roadway is congested and whether or not a bottleneck exists, speed and roadway geometry must be considered in addition to average hourly traffic volume. Moreover, roadway geometry must be studied to identify the bottleneck cause. Specific geometric considerations, including splits and merges, on- and off-ramps, and slip ramps, and maximum hourly volume per lane and speed are given in Table 9 for each segment. Table 9 ? Comparison of Segment Geometry, Maximum Volume and Speed Road Segment I-370 to Shady Grove Rd (1) Shady Grove Rd to MD 28 (2) MD 28 to MD 189 (3) MD 189 to Montrose Rd (4) Montrose Rd to Spur (5) Splits and Merges 1 0 0 0 1 No. of Ramps 3 3 2 3 2 No. of Slip Ramps 1 3 1 1 0 Total No. of Ramps, Splits and Merges 5 6 3 4 3 Max. Volume per CD Lane 722 1907 1774 2037 1475 Max. Volume per Mainstream (GP and HOV) Lane 1975 1777 1832 1663 2019 Average Segment Speed (mph) 44 31 46 44 63 As shown in Table 9, the average segment speed is directly correlated with the number of slip ramps. Less strong correlation exists between average segment speed and maximum hourly volume per mainstream lane. Thus, it was concluded that the congestion at bottlenecks is due to a combined effect of geometry, especially number of slip ramps, and traffic volume per hour per lane in the GP and HOV lanes. 29 Chapter 4 Calibration A myriad of software products exist for simulating vehicular traffic. These models can replicate many of the characteristics of vehicular behavior. Results from such simulation runs are often used to make decisions pertaining to operational and design changes. In this study, conversion of a HOV lane to a HOT lane facility is considered. Before one takes decisions based on the outcomes from the simulation runs, one must be sure that the simulation adequately replicates traffic conditions and behaviors. In addition to the modeling issues described in Chapter 3, parameters of the VISSIM simulation software can be set so that traffic measures from the simulation best match actual measurements taken from the field. Initial runs were conducted using default parameter settings as described in Section 4.1. Results from these runs show that mean travel times estimated by the simulation model using default parameters were statistically significantly different from observed mean travel times. Thus, calibration of the parameters is essential. In this study, the parameters are calibrated based on average segment travel time. In Section 4.2, relevant model parameters, along with their ranges and default values are presented. In Section 4.3, results of sensitivity analysis in which the parameters were set to their extreme values and simulation runs were conducted are presented. Such experiments provide additional insight into the impact of each parameter on travel time. Results of these preliminary runs, as well as input from PTV America, Inc. (PTV) modelers and the literature were used to identify five parameters as having the greatest impact on model calibration. Even if only a few potential values were chosen for each of these five parameters, the number of runs that would be required to consider all parameter setting combinations would be very large. Thus, effort was taken to design the experiments and conduct a more limited set of runs. In Section 4.4, findings from a limited set of runs chosen based on factorial design are given. Results of these runs provide information about interactions between parameters, as well as the potential impact of specific parameter values. For example, in which direction (i.e. below or above a chosen value) the parameter should be set to obtain a particular behavior (e.g. lower or higher segment travel time) can be observed from the run results. The information concerning 30 parameter interactions aided in choosing a subset of parameter combinations for the final set of runs. Intuition gleaned from results of the runs based on the factorial design is employed in final calibration runs. Final results of the calibration are also provided in Section 4.5. 4.1 Quality of Simulation Results given Default Parameter Settings The VISSIM model of the study area was constructed with existing highway geometry and traffic demand as described in Chapter 2. Initial simulation experiments were conducted using default driving behavior parameter settings to ascertain how well the model does in replicating traffic given that default parameter settings are used as is often done in practice. Mean segment travel times obtained from the simulation results were compared with mean segment actual travel times obtained from the field. A comparison of mean travel times for small sample size (i.e. a t-test) was completed. It was assumed that travel times are normally distributed. The small sample size test was employed, because significantly fewer than 30 travel time samples were obtained through field observations for each roadway segment. Results of the analysis are given in Tables 10 and 11 and Figures 23 and 24. Table 10 ? Travel Times for Existing Conditions given Default Parameter Settings GP HOV Simulated Survey Simulated Survey Segment Ave* SD** Ave* SD** Ave* SD** Ave* SD** 1 63 22 214 67 62 25 90 48 2 125 86 312 59 124 24 259 45 3 60 40 145 81 60 39 89 32 4 83 60 193 39 82 75 129 39 5 82 70 163 62 83 53 91 12 Total 412 -- 668 -- 410 -- 668 -- * Ave = Average ** SD = Standard Deviation 31 Figure 23 ? Comparison of Survey GP Lane Travel Times with Simulated Travel Times given Default Parameter Settings and Existing Conditions 0 50 100 150 200 250 300 350 0123456 Segment Tr ave l Time (s econds) Survey Simulation Figure 24 ? Comparison of Survey HOV Lane Travel Times with Simulated Travel Times given Default Parameter Settings and Existing Conditions 0 50 100 150 200 250 300 0123456 Segment Tr av el Ti me (s ec onds) Survey Simulation 32 Table 11 ? Statistical Analysis of Existing Condition Simulation Results given Default Parameter Settings GP Lane HOV Lane Segment t value v value T (0.025,v) If - T