Technical Report # 3 Model Validation & Calibration. February 2010

Transcription

1 Technical Report # 3 Model Validation & Calibration February 2010

2 Technical Report # 3 Model Validation & Calibration Envision 2035 Long Range Plan Update Prepared for North Florida Transportation Planning Organization 1022 Prudential Drive Jacksonville, Florida Prepared by 7406 Fullerton Street, Suite 350 Jacksonville, Florida (904) In Association with Renaissance Planning Group The Corradino Group, Inc. M. Victoria Pennington, Marketing & Public Affairs February 2010

3 Commonly Used Abbreviations 4TK A/ALT AADT AM/PK AP ASE AT/ATYPE BPR BRT CBD COM/CmTk CTPP CV DA DBF DRI EC EE EI FCCJ FDOT FF FSUTMS FT/FTYPE FTA FTI GIS Four-wheeled Truck Alternative Annual Average Daily Traffic Peak Period (for reference to Transit model) Attraction-to-Production Automated Skyway Express Area Type Bureau of Public Roads Bus Rapid Transit Central Business District Combination truck-trailer Census Transportation Planning Package Cube-Voyager Drive Alone Database Format Development of Regional Impact Existing-plus-Committed External-External External-Internal Florida Community College-Jacksonville Florida Department of Transportation Friction Factor Florida Standard Urban Transportation Model Structure Facility Type Federal Transit Administration Florida Traffic Information Geographic Information System HBNW/HBO Home-Based-Non-Work/Home-Based-Other HBShop/HBSH Home-Based-Shopping HBSocRec/HBSR/ HBSCR Home-Based-Social-Recreation HBW Home-Based Work HCM Highway Capacity Manual HD/HDTK/ HTRK Heavy Duty truck HEVAL Highway Evaluation HOV/HO High (Multiple) Occupancy Vehicle IE IVT/IVTT JTA JUATS KNR LD/LDTK/ LTRK LOS LRT LRTP Internal-External In-Vehicle (Travel) Time Jacksonville Transportation Authority Jacksonville Urbanized Area Transportation Study Kiss-Ride/Drop-Off Light Duty truck Level of Service Light Rail Transit Long Range Transportation Plan MC Mode Choice MD/MIDDAY/OP Off-peak period (for reference to Transit model) MOCF Model Output Conversion Factor MPO Metropolitan Planning Organization

4 Commonly Used Abbreviations (continued) NCHRP NERPM NFHTS NFETS NHB NL OBD OD OVT PA PCWALK PSCF PSWADT PT PNR QRFM RMSE RTS National Cooperative Highway Research Program Northeast Regional Planning Model Northeast Florida Household Travel Survey Northeast Florida External Travel Survey Non-Home-Based Number of Lanes Outer Business District Origin-Destination Out-of-Vehicle Time Production-to-Attraction Percent Walk Peak Season Conversion Factor Peak Season Weekday Daily Traffic Public Transport Park-Ride Quick Response Freight Manual Root Mean Square Error Regional Transportation System SERPM SOV/SO SPGEN SR SU/SUTK/ MTRK TAC TAZ TD TG TMIP TPO UB V/C V/G VMT VHT YY ZDATA Southeast Regional Planning Model Single-occupancy vehicle Special Generator Shared Ride Single Unit (medium Duty) Truck Technical Advisory Committee Traffic Analysis Zone Trip Distribution Trip Generation Travel Model Improvement Program Transportation Planning Organization User Benefit Volume-over-Capacity Ratio Volume-over-Count Ratio Vehicle Miles of Travel Vehicle Hours of Travel Year Zonal Data

5 Table of Contents List of Figures... iv List of Tables... vi 1. Introduction Model Background Model Enhancement Summary Model Modules Report Organization Highway Network Network Background Network and TAZ Update and Review Highway Network Update TAZ Update Review of Highway Network Transit Network Elements Coded onto the Highway Network Review of Transit Network Review of Transit Ridership Data Traffic Count and Screenline Data Sources Count Coding Review of Traffic Count Data Review of Screenlines Updates of Speeds and Capacities and Validation External Trips Model Description Data Development and Validation Adjustments Results and Comparisons Trip Generation Trip Generation Process Zonal Socioeconomic Data Summary Trip Generation Validation Adjustments Trip Generation Validation Results Highway Paths and Skims Model Process Model Validation Trip Distribution Trip Distribution Model Process Peak-Period Highway Assignment Trip Table for Pre-Mode Choice Assignment Subarea Balancing Model Validation Comparison of Journey-To-Work and Model HBW Trips HBW Travel Patterns Results and Comparisons February 2010 Page i

6 7. Transit Network, Path and Skim, and Fare Transit Network Transit Network Elements Coded onto Highway Network Bus Only Links Park-Ride (PNR) Station Coding Station Data Information Micro-Coding Stations and Fixed-Guideway Links Transit Route Walk Coverage Non-Transit Connectors Walk Access Connectors Park-Ride Access Connectors Drop-Off Access Connectors Fringe PNR Connectors Downtown Drop-Off Access Connectors Transfer/Sidewalk Connectors All-Walk Connectors Transit Network Summary and Speed Validation Transit Paths and Skims Transit Paths Transit Skims Transit Paths using Guideway Links (SELECTLINK) Transit Fares Mode Choice Model Structure Model Choice Calibration Model Choice Reports Calibration Results Auto-Occupancy Rates Transit Assignment Model Process Additional Reporting Summary of BRT Trips Model Validation Results and Comparisons Highway Assignment Model Process and Validation Adjustments Modified Volume-Delay Functions UROAD Factors CONFAC Factors Model Validation Results and Comparisons Systemwide Volume-over-Count and RMSE Statistics Screenline, Cutline and Corridor Volume-over-Count Ratios Volume-over-Count Ratios by FT and AT Groups Average Volume and Vehicle-Miles and Vehicle-Hours of Travel Summary and Conclusion February 2010 Page ii

7 12. List of References Appendix A Appendix B Appendix C Selected Validated Data and Parameter Summary Selected Model Data Summary Year 2035 External Trip Estimation I:\projects\3843-Jax\Reports\NERPM4-TR1&2\NERPM4-tr1&2.doc February 2010 Page iii

8 List of Figures Figure 1-1 Study Area Boundaries and Valid Number of TAZ Figure 1-2 Model Macro Flow Chart Figure 2-1 Year 2005 Highway Network Highlighting Facility Types Figure 2-2 Year 2005 Highway Network Highlighting Area Types Figure 2-3 Year 2005 Highway Network Highlighting Number of Directional Lanes Figure 2-4 FDOT Traffic Count Locations Figure 2-5 Locally Collected Traffic Count Locations Figure 2-6 Screenline, Cutline and Corridor Locations Figure 3-1 External Station Locations Figure 3-2 Bandwidth Plot of External Assigned Volumes Figure 3-3 Highlighted Facilities Prohibiting External Trips Figure 4-1 District Boundaries Figure 4-2 Trip Attraction Districts Figure 5-1 Comparison of Highway Paths from I95 North External Station to Four Other Major External Stations Figure 5-2 Comparison of Isochromes (10 minutes increment) from a Downtown Jacksonville TAZ Figure 6-1 CTPP Districts Figure 6-2 Duval County CTPP Districts Figure 6-3 Time Trip Length Frequency Distributions Figure 6-4 Distance Trip Length Frequency Distributions Figure 7-1 Year 2005 Transit Route Coverage Figure (Base Year) Transit Stations Figure (Base Year) Zoomed Downtown Area Transit Stations Figure 7-4 Peak Period 2005 Transit Walk Coverage Figure 7-5 Off-peak Period 2005 Transit Walk Coverage Figure 8-1 Mode Choice Nesting Structure Figure 9-1 Skyway Volume Plots using ONLINKREC Output Figure 9-2 Station Activity (Transit On/Off) Plot ASE Route Southbound Direction Figure 9-3 Line Volumes at Stop (Transit Line Profile) Plot ASE Route Southbound Direction Figure 9-4 Scatterplot and Accuracy Statistics of Transit Route Boardings NERPM Validation Figure 9-5 Scatterplot and Accuracy Statistics of Transit Route Boardings JTA/RTS 2005 Validation Figure 10-1 Modified BPR Volume-Delay Functions Figure 10-2 Scattergram of the Assigned Volumes versus the Counts of NERPM4 Model Figure 10-2 Scattergram of the Assigned Volumes versus the Counts of NERPM2000 Model Figure 10-4 Total Screenline Volumes and Maximum Desirable Deviation Figure A-1 Highway Network and Speed-Capacity Scripting Changes in JTA/RTS Transit Model Updates... A-11 Figure A-2 Snippet of SPDCAP Table Showing JTA/RTS Transit Model Speed Modification... A-14 Figure A-3 List of NERPM4 Speed and Capacity Modifiers of SPDCAP File... A-16 February 2010 Page iv

9 List of Tables Table 1-1 Summary of Year 2005 Based NERPM (NERPM4) TAZs Table 2-1 Definitions of Area Types and Terminal Time Values Table 2-2 Definitions of Facility Types Table 2-3 Link Traffic Count Summary by Facility and Area Types Table 2-4 Link Traffic Count Summary by Facility Type and County Table HEVAL Speed Summary by Facility and Area Type Combinations of NERPM Table (Trend) HEVAL Speed Summary by Facility and Area Type Combinations of NERPM Table 2-7 Summary of 2005 Lane-Mile and Capacity by Facility and Area Type Combinations Table 2-8 Comparison of Model Input and Congested Speeds by Facility and Model Table External Station Traffic Count Summary Table External Vehicle Trip Table Summary Table External Station Traffic Projection Summary Table External Vehicle Trip Table Summary Table 3-5 Comparison of 2005 External Station Traffic Counts and Volumes Table 4-1 Dwelling Unit Weights Table 4-2 Summary of External Station Trip Factors by Mode of Travel Table 4-3 Validated Trip Production Rates Table 4-4 Validated Trip Attraction Rates Table 4-5 Summary and Comparison of Trip Generation Outputs Table 6-1 Pre-assignment Peak Period Factors Table 6-2 Comparison of CTPP and Model Estimated HBW Trips by Duval County CTPP Districts Table 6-3 Comparison of CTPP and Model Estimated HBW Trips by County Table 6-4 Comparison of 2035 and 2005 Model Estimated HBW Trips by Duval County CTPP Districts Table 6-5 Comparison of 2035 and 2005 Model Estimated HBW Trips by County Table 6-6 Summary and Comparison of 2005 Trip Length and Intrazonal Trips by Purpose and Vehicle Trips by Mode Table 6-7 Summary of 2035 (Trend Scenario) Trip Length and Intrazonal Trips by Purpose and Vehicle Trips by Mode Table 7-1 New Fields for Transit-only Links in the Highway Network Table 7-2 PNR Link Facility Types Table 7-3 Fields Required for Transit-Only Links Table 7-4 Node Fields for Station Data Table 7-5 Transit Mode Definitions Table 7-6 Transit Operator Definitions Table 7-7 Non-Transit Modes Built Details Table 7-8 Comparison of Transit Network Summary Statistics by Mode and Operators Table 7-9 Transit Paths Settings Table 7-10 Tables in the Transit Skim Matrices Table 7-11 Tables in the Selectlink Matrix Table 7-12 Transit Boarding Fares (2007 fares) Table 7-13 Transit Transfer Fares February 2010 Page v

10 List of Tables (continued) Table 8-1 Summary of Mode Choice Transit Coefficients and Calibrated CBD Constants Table 8-2 Summary of Calibrated Mode Choice Constants for Various Sub-modes Table 8-3 Summary of Target Trip Shares for Mode Choice Calibration Table 8-4 Comparison of 2005 Home-Based-Work Model Estimated and Target Trips by Mode and Household Market Table 8-5 Comparison of 2005 Home-Based-Non-Work Model Estimated and Table 8-6 Target Trips by Mode and Household Market Comparison of 2005 Non-Home-Based Model Estimated and Target Trips by Mode Table 8-7 Comparison of Transit Trip Transfer Rates by Path Table 8-8 Mode Choice Calibration Results for CBD Trips Table 8-9 NCHRP 365 Auto Occupancy Rates by Urbanized Population, Income and Purpose Table 9-1 Comparison of Systemwide Transit Boarding by Operators Table 9-2 Comparison of 2005 Model Estimated and Observed Transit Ridership by Route Table 10-1 Comparison of Systemwide Highway Model Validation Statistics Table 10-2 Comparison of Systemwide Root-Mean-Square-Error Statistics Table RMSE and Volume-over-Count Statistics by Count Range Group and County Table 10-4 Comparison of Volume-over-Count Ratios by Facility Type Group Table Volume-over-Count Ratios of Screenlines, Cutlines and Cordons Table 10-6 Volume-over-Count Ratio by Facility Types, Area Types and Counties Table 10-7 Volume-over-Count Ratio by Facility and Area Type Combinations Table 10-8 Comparison of 2005 and 2035 Average Link Volume, VHT, VMT and Percent VMT by Facility and Area Types Table A-1 List of TRANSPD.DBF File Highway-to-Transit Speed Conversion Parameters... A-1 Table A-2 Validated Special Generators Trips... A-3 Table A-3 Validated Friction Factors... A-4 Table A-4 List of Validated Turning Penalties... A-7 Table A-5 List of VFACTORS File... A-9 Table B-1 Year 2005 (Base) Key Socioeconomic Data Totals by District and County... B-1 Table B-2 Year 2035 (Trend) Key Socioeconomic Data Totals by District and County... B-2 Table B-3 Change in Key Socioeconomic Datasets between 2005 (Base) and 2035 (Trend) by District and County... B-3 Table B-4 Percent Growth in Key Socioeconomic Datasets between 2005 (Base) and 2035 (Trend) by District and County... B-4 Table B-5 Year 2005 Station Data Information... B-5 Table B HEVAL Pre-assignment Speed Summary of NERPM4... B-6 Table B HEVAL Pre-assignment Speed Summary of JTA/RTS Model... B-7 Table B HEVAL Assignment Speed Summary of JTA/RTS Model... B-8 Table B HEVAL Assignment Speed Summary of NERPM2000 Model... B-9 Table B-10 Summary of Year 2000 Based NERPM (NERPM2000) TAZs...B-10 Table B-11 Summary of Year 2005 Based JTA/RTS (JTA/RTS-2005) Model TAZs...B-11 Table B-12 External Station Traffic Information...B-12 Table B Census Journey-To-Work (JTW) Trip Flow Summary of Duval County CTPP Districts...B-14 Table B Census Journey-To-Work (JTW) Trip Flow Summary of Six Counties of NERPM4...B-15 Table B-15 Table B-16 List of Zones of Sub-area Balancing Attraction Districts...B-16 Summary of Base (2005) Year Transit Route Characteristics and Observed Ridership...B-17 February 2010 Page vi

11 1. Introduction Version 4 of Northeast Regional Planning Model (NERPM4) is a multimodal travel demand model covering the six urban counties (Baker, Clay, Duval, Nassau, Putnam and St. Johns) of Northeast Florida. Figure 1-1 shows the coverage of the study area along with the number of active internal TAZs in each of the six constituent counties of NERPM4. NERPM4 is the latest version of NERPM and uses Cube- Voyager (CV) and Public Transport (PT) as the new FSUTMS modeling platform for highway and transit travel estimation. NERPM4 includes many improvements that were implemented in earlier versions of the Jacksonville area regional travel models [JTA/RTS 2005 (see References 1-3) and NERPM2000 (see References 4-12)]. The NERPM4 and JTA/RTS 2005 transit models are essentially same. However, NERPM4 uses more consistent speeds in both highway and transit modeling processes and expands the study area. Similar to the JTA/RTS 2005 model, NERPM4 has been structured to utilize Cube s parallel-processing capability, Cube Cluster, and runs optimally on a computer with a quad-core processor. Although there are total of 1,862 active TAZs, dummy zones were added to make 2,494 internal TAZs, allowing for future expansion. The 29 external stations are numbered 2,550-2,578. More details on the zone structure are provided in section and Tables 1-1, B-10 and B-11 of this report. Table 1-1 lists the zones in NERPM4 by regional district. It separates the TAZs that existed in the 2000 based models and the extra and split TAZs in NERPM4. Dummy zones are also listed in this table. Model development and validation of NERPM4 is part of overall 2035 long range transportation plan update study [Reference 13]. The work effort included the following tasks: Data Collection, Review and Update (Task 2 of Reference 13): Collect Previous Studies Collect, Review and Develop Zonal Data Review and Update External Trip Data Review and Update 2005 Highway Network Review and Designate Screenlines, Cutlines and Cordon Lines Collect and Review 2005 Traffic Count Data Collect 2005 Transit System Operations and Ridership Data Review and Update 2005 Transit Network and Transit Services Review Trip Generation Rate Review Trip Length Frequency Distribution Review and Update Auto Occupancy Rates Review and Update Model Parameters and Rates Refine and Update existing Cube Voyager model(s) Mapping Documentation on data collection, review and update Model Validation and Calibration (Tasks 3 & 10 of Reference 13): Expansion of the Regional Model to include Baker and Putnam counties Validate External Trips Validate the Trip Generation Model Review Validation of the Path Building Model Validate the Trip Distribution Model Validate the Mode Choice Model February 2010 Page 1-1

12 Validate the Transit Assignment Model Validate the Highway Assignment Model Final Validation of the six county regional model Documentation on validation and user application guide February 2010 Page 1-2

13 Figure 1-1: Study Area Boundaries and Valid Number of TAZ Nassau (108) Baker (29) Duval (1281) Clay (184) St. John s (216) Putnam (44) Note: County Name with valid number of TAZ in parenthesis Total number of Valid TAZ in six Counties = 1,862 February 2010 Page 1-3

14 Table 1-1: Summary of Year 2005 Based NERPM (NERPM4) TAZs February 2010 Page 1-4

15 Table 1-1 (contd.): Summary of Year 2005 Based NERPM (NERPM4) TAZs The process by which the travel demand model is refined until it closely replicates observed travel patterns (both speeds and counts/ridership) is called validation. This report describes the 2005 validation efforts and results. The validated model parameters were then applied and tested with the 2035 NERPM4 model. February 2010 Page 1-5

16 1.1 Model Background NERPM was originally prepared for a base year of 1998 and a future year of 2025 with the assistance of the District s general planning consultant. The original software platform for NERPM was TRANPLAN and other programs developed as part of the Florida Standard Urban Transportation Model Structure (FSUTMS). Many of the parameters and assumptions used in NERPM were derived from the North Florida Household Travel Survey (NFHTS) and the Northeast Florida External Travel Survey (NFETS), both of which were conducted in the year In the fall of 2002, the NERPM 2000 model validation study [see References 4-6] was initiated with a consultant team led by Cambridge Systematics, Inc. and was later used in 2030 long-range transportation plan (LRTP) update. The NERPM 2000 model used Cube-Voyager, which is the new platform for FSUTMS. Before NERPM (1998 based) and NERPM2000, an earlier version of the model known as Jacksonville Urbanized Area Transportation Study (JUATS) was in place. It covered a smaller study area than the four county study area adopted in NERPM (1998 based) and NERPM2000. In 2007, transit components of NERPM2000 were completely modeled in Public Transport (PT) in a study initiated by the Jacksonville Transit Authority (JTA). This is known as the JTA/RTS 2005 model [see References 1-3]. The model allowed JTA to study a possible expansion of the transit system with introduction of premium transit modes including BRT, LRT and commuter rail. In early 2008, the North Florida TPO selected a team of consultants lead by PBS&J to conduct the 2035 LRTP update [Reference 13]. The Corradino Group updated and validated the model (NERPM4) as a subcontractor to PBS&J. The NERPM4 model enhanced the 2005 JTA/RTS model and expanded the study area to include Baker and Putnam Counties. There were inconsistencies in the 2005 JTA/RTS model in the speeds between the final highway assignment and the speeds used in the transit model. This was primarily due to the fact that the goal of the JTA/RTS model was to enhance transit model and had less concern for the highway model. The validation of NERPM4 made the highway and transit components consistent, and was based on 2005 traffic counts and transit data collected and assembled for NERPM4. Traffic analysis zones were updated and expanded in all versions of the NERPM and JTA/RTS models with the latest update in NERPM4. More information on NERPM4 is presented in this report. 1.2 Model Enhancement Summary The development of 2005 and 2035 NERPM4 represents a new generation of modeling techniques applied to the six county region of Northeast Florida. It is an outgrowth of the two other recent regional models (JTA/RTS 2005 and NERPM2000), which covered a four county region (Clay, Duval, Nassau and St. Johns). Both NERPM4 and JTA/RTS 2005 models are an update to the Northeast Regional Planning Model validated for 2000 (NERPM2000). The purposes for developing the JTA/RTS model were to apply the new FSUTMS transit modeling standards using PT and to ensure that it met standards for FTA New Starts Analyses. Both NERPM4 and transit component of the JTA/RTS models are calibrated to the year NERPM4 highway and transit components were validated for the six county region. The JTA/RTS and NERPM4 transit models are practically the same. The following major revisions were embodied in NERPM4. February 2010 Page 1-6

17 Zones/Zonal Data The total number of active zones increased from 1,312 (NERPM2000) or 1,619 (NERPM2000) to 1,862 (NERPM4). All the files in the input folder and the parameters folder were accordingly modified to reflect the new and expanded zone system. Dummy zones (total of 481) were added for possible expansion of the model region in future The 2005 and 2035 ZDATA were updated by PBS&J in association with the North Florida TPO and other county and FDOT planning staffs. Networks and External Trips The highway network was initially adopted from JTA/RTS model and was then updated by Consultant team to include revisions suggested by FDOT District 2 to correct numerous access and minor coding issues identified as part of model development and validation. The networks of Baker and Putnam counties were added and external trips of the model were adjusted accordingly. The turning penalty file was revised to prohibit illegal ramp movements. Two separate turning penalty files for pre-assignment and final highway assignment steps of JTA/RTS model are consolidated into one file for use in NERPM4. The fixed-guideway links and stations were micro-coded. The station data information was added to the node layer in the highway network. The NERPM2000 special generator file for external trip validation was removed in NERPM4. The special code to prohibit through traffic movement implemented in NERPM2000 was continued in NERPM4. Trip Generation Starting with NERPM2000, the trip generation process was scripted in CV. The trip generation process includes 12 expanded trip purposes used in NERPM2000. The dwelling unit variable used in NERPM2000 model s HBW trip attraction equation was removed in JTA/RTS and NERPM4 models. The subarea balancing process adopted in NERPM2000 and continued in JTA/RTS 2005 was deactivated in NERPM4. Starting with NERPM2000, a new truck model with a structure similar to the one recommended in the Quick Response Freight Manual (QRFM) was added. The truck model includes three truck purposes (four-tired, Single Unit and Combination), treating trucks as a separate mode from generation through assignment. Trip Distribution The MCSEED FORTRAN program (used in NERPM2000) was scripted in Cube Voyager and its logic was revised to better handle the newly subdivided zones in JTA/RTS and NERPM4 models. The estimated work travel pattern was validated. The scripted speed and capacity adjustments for JTA/RTS transit model network developments were omitted and those adjustments are carried to the speed-capacity table used in highway network processing steps in NERPM4. February 2010 Page 1-7

18 The NERPM4 model removed the 24-hour pre-distribution highway assignment steps that were used in HBW trip distribution. It now uses the 2-hour peak period preassignment for both final HBW trip distributions as well as for peak period transit network speeds. Transit Network, Path and Mode Choice The transit model in PUBLIC TRANSPORT (PT) was implemented to conform to the new FSUTMS transit modeling standards. The transit network represents a single transit route file for peak and off-peak periods in PT. An AM peak period highway assignment was incorporated in the JTA/RTS model and was adopted in NERPM4 to estimate reasonable auto and transit speeds for the mode choice model as well as the final HBW trip distribution. The transit model was calibrated and validated to year The results of the auto-transit speed relationship study were incorporated in JTA/RTS model and carried to NERPM4. The mode choice structure of JTA/RTS model was modified to reflect the transit travel market as reflected in the 2006 bus-rider survey (on-board survey) and was unchanged in NERPM4. The mode choice model of the JTA/RTS model was scripted in Voyager, including an auto calibration and routines to generate Summit input files. The mode choice model was calibrated based on the bus-rider survey. A SELECTLINK methodology was added to compute in-vehicle travel time on the transit guideway for future BRT analyses. Logic to compute additional benefits from mode specific constants was added for use in New Starts studies. Results of NERPM4 transit model were compared to JTA/RTS 2005 model as well as targets that were assembled from survey and observed ridership. Transit Assignment Transit assignment reports, including station activity, visual representation of transit loadings on links and BRT related trip summaries were developed. Highway Assignment and Evaluation The NERPM4 evaluation application includes numerous updates that generate key model validation result summaries. The NERPM4 highway evaluation application generates model evaluation outputs for the region as a whole as well as for the six constituent counties. The NERPM4 highway assignment step uses modified volume-delay function parameters. The NERPM4 model validation efforts included systematic speed and capacity adjustments to produce reasonable speeds by facility type as well as validation of regional traffic counts. Numerous tables and figures compare the NERPM4 model validation results to those from the earlier NERPM2000 model as well as FDOT and national standards. February 2010 Page 1-8

19 1.3 Model Modules The overall structure of the NERPM4 model is shown in the form of a flowchart in Figure 1-2. It has 12 component modules. The macro flowchart identifies all the user-supplied input and parameter files that are used by each of the modules. It also shows the standard FSUTMS routines (RMSE, HEVAL, AUTOCON and TAReport ). The output files of greatest interest to users are also shown in this macro flow chart. Users should consult Technical Report 3 - Model Application Guidelines for a detailed description of these modules, their input/output files, as well as inputs and outputs of the standard routines. The transit modeling modules follow the FSUTMS transit modeling standards adopted by the FDOT Systems Planning Office. This flow chart typically follows the standard 4-step process (trip generation, trip distribution, mode split and assignment) to estimate travel demand. Trip Generation determines the total number of trips produced and attracted each day for each trip purpose. Trip Distribution finds the number of person trips that travel between all pairs of zones. The Mode Split step finds the number of trips using each available mode between a production/attraction zone pair. The Trip Assignment step determines which route highway and transit trips will follow. The end results include traffic volumes, transit boardings, line volumes and mode-of-access data. However, the standard steps now contain many enhancements. For example, the distribution step now has a peak period pre-assignment for generation of congested skims in final work trip distribution and mode choice. Also, the steps within the transit model are different to reflect the travel market, networks and travel patterns of the Jacksonville area. These components of the model shown in Figure 1-2 are processed in a serial fashion to complete the travel demand simulation. February 2010 Page 1-9

20 Figure 1-2: Model Macro Flow Chart February 2010 Page 1-10

21 1.4 Report Organization This report (TR1&2) describes model data, calibration and validation. It presents the model validation efforts and results of both 2005 and 2035 NERPM4 models. A companion to this report is the Model Application Guidelines (TR3), which describes the model features and operation and then guides the users for its application. These reports assume that the reader is familiar with Cube Voyager and standard modeling techniques used in Florida. The model was run with CV Versions and (current official release of CV). In this report, the term calibration and validation are used interchangeably. In fact, calibration and validation are separate tasks, although many transportation planners/modelers try to do both at the same time. Calibration applies to each step in the modeling process, while validation applies to the model as a whole. In calibration, each model step has one or more parameters that can be adjusted to assure that the step is replicating known travel behavior. Very often calibration is performed by statistical methods. Validation primarily involves comparing a base-year forecast to known traffic levels (counts and ridership). A poor quality validation would indicate the need for additional calibration. This Model Data, Calibration and Validation report (TR 1&2) is divided into twelve chapters and three appendices. Chapter 1, Introduction, describes the model enhancements, model process and report organization. Chapter 2, Highway Network, describes the new network, review of key network attributes, review and updates of key network data, and updates of speeds and capacities and validation. Chapter 3, External Trip, contains a description of the external model and its validation. Chapter 4, Trip Generation, summarizes the key aspects of trip generation model, zonal data and rates used in the model and the results. Chapter 5, Highway Paths and Skims, describes the paths and skims used in model validation. Chapter 6, Trip Distribution, provides the description of the trip distribution model process. It then summarizes and compares the key results. Chapter 7, Transit Network, Path and Skim, and Fare, describes the transit network, path and skim, and fare. Numerous tables are used to summarize the model results. Chapter 8, Mode Choice, describes the mode choice model. It uses a nested logit structure for mode choice analysis. Numerous tables and figures are used to summarize the model results. Chapter 9, Transit Assignment, summarizes and compares the results of the transit assignment process. Chapter 10, Highway Assignment, describes parameters and results of the assignment process and compares the results against established criteria. Chapter 11, Summary and Conclusion, presents the highlights of the NERPM4 model validation process and offers suggestions for future model enhancements. Chapter 13, List of References, provides a list of references on NERPM and JTA-RTS, and other Florida and national resources referenced in NERPM4 technical reports. Appendix A, Selected Validated Data Summary, presents several validated model parameters, which were referenced in this report. Appendix B, Selected Model Data Summary, presents sixteen summary tables of model data that are referred to in this report. Appendix C, Year 2035 External Trip Estimation, presents a technical memo on processing of historical traffic counts for estimation of 2035 external trip data for model validation. February 2010 Page 1-11

22 The following chapters describe the changes to the various modeling steps and the calibration and the validation results. February 2010 Page 1-12

23 2. Highway Network Highway networks represent the transportation system in the study area covered by the travel demand model. The highway network includes a series of interconnected links each containing a set of attributes relevant to simulating highway conditions. Highway network processing is the second module in the CV application (see Figure 1-1). Users should consult model application guidelines for the details of all user input and parameter files of this module. The most critical of these attributes in FSUTMS models are those pertaining to area type (AT), facility type (FT), and number of lanes (NL). With these attributes, the highway network module of the model includes calculation of speeds and capacities that the model uses later in the model stream for trip distribution and trip assignment. The NERPM4 transportation network includes both highway and relevant micro-coded transit network data that were prepared based on the network used for 2005 JTA/RTS model as a starting point and then modified based on year 2005 data made available by FDOT, North Florida TPO, JTA, and the county governments of Nassau, Clay, St. Johns, Baker and Putnam Counties. Once the development of initial highway and transit networks was completed, efforts commenced on reviewing and refining the data. Initial stages of network review concentrated on ensuring appropriate network characteristics coding and the correct configuration of the network. This was accomplished primarily through the review of plots generated by Cube. Later stages of review occurred throughout the validation process with an eye toward improving model performance. The extensiveness of the NERPM4 networks demanded frequent oversight and review. This chapter of the report describes the review and refinement of both the highway and transit networks during the NERPM model validation. It includes a discussion of the review of traffic counts and transit ridership data. This chapter also describes the update of the highway network and presents summaries of network speeds and traffic counts. 2.1 Network Background Staring with the 2000 NERPM model, a master network was used for all alternatives. The master network deals primarily with management of model networks and relates to the organization of the networks for all of the scenarios and alternatives in a model. A master network is one large network database that contains all of the links and nodes for all alternatives and scenarios for a given model. In the initial stage of model validation of NERPM4, efforts were made so that master highway network contains all of the links for each of the scenarios. However, the consultant team faced difficulties to maintain this master network for needs alternatives coding and runs for the Long-Range Transportation Plan (LRTP) 2035 Update. The consultant team then decided to use a master network that is scenario specific. It is important to remember that the scenario network is isolated from the master network during the highway network process and that the model will only run on the network consistent with the scenario that the modeler has selected. The FSUTMS transit modeling standards require some transit network information to be coded onto the highway network. Hence, the network coding process is different from what it has been used in the previous versions of NERPM; however, if the transit layer is open at the same time as the highway layer, edits to both can be made simultaneously. The transit modeling process in NERPM4 implemented the same approach as the 2005-based JTA/RTA model, using Voyager s Public Transport (PT) program. The transit station areas in the network (MicrocodedHnet4_YYA.NET) were developed from the JTA/RTS model s master network file. This network was micro-coded transit station information as well as all fixed February 2010 Page 2-1

24 guideway facilities and optional transit links. As noted earlier, scenario specific networks were used in the long range plan development process. The fields or attributes of highway network whose values vary by alternative are coded with field names ending in _YYA. If the links do not exist in the alternative, the field FTYPE_YYA is coded as 0. In the 2005 JTA/RTS model, the pre-assignment estimated speeds were compared to observed data to make sure that both auto and the transit speeds were reasonably represented in the model. The observed data for auto speeds were obtained from the 2006 Florida Highway Data CD and the Jacksonville auto speeds data. The transit travel time data were taken from the public time table provided by the JTA. After making adjustments to the networks to reflect the observed characteristics, a separate two-hour peak period assignment was performed in order to get reasonable speeds for the transit model. The primary focus during the auto speed calibration process was the major roadways in potential BRT corridors. In the JTA/RTS model, refinements were made to the base year highway network using a script that runs within the model to make it more reflective of the known capacity or speed characteristics. Some of those refinements include: Correction to the laneage on I-95 between the I-95/I-10 interchange and the I-95/Atlantic Blvd interchange, Modification to the turning penalty (TCARDS) files to prohibit illogical ramp movements along I-95, Adjustment to the free-flow speeds for freeways and arterials to reflect the posted speeds from 2006 Florida Highway Data, Adjustment to the free flow speeds on the St. Johns bridges to reflect the posted speeds on these bridges, However, the 2005 based JTA/RTS model did not carry the changes in the speeds to the initial unloaded network and thus caused inconsistencies in speed that used in final highway assignment process. The changes were made using a script that runs within the model. These changes were not applied to networks used in trip distribution and highway assignment. The 2005 JTA/RTS procedures noted above were eliminated from NERPM4 because they caused inconsistencies between the travel times used in the highway and transit models. 2.2 Network and TAZ Update and Review NERPM4 includes six counties. The counties of Baker and Putnam were added to the original four counties of Clay, Duval, Nassau, and St. John s. This section of the report describes updating of the NERPM networks to base year 2005 conditions. NERPM4 has a base year of 2005 like the JTA/RTS model, and a new future year of The highway network contains many updates and corrections. Updates started with the 2000 NERPM and 2005 JTA/RTS base-year networks. The 2005 highway network included updated traffic counts, two additional counties, new external stations and a review of the highway network in the four existing counties. TAZs and zonal data were added for Baker and Putnam Counties. February 2010 Page 2-2

25 2.2.1 Highway Network Update The existing JTA/RTS highway network is a four county and multiple-year network. Several updates were required. These updates include adding 2005 traffic counts, adding new highway links in the two new counties, updating turn prohibitions and penalties, updating the external stations to account for the two new counties, and review of the network in the existing four counties for changes through The traffic counts update is discussed in next section. New highway links and centroid connectors were added to the existing network for the two new counties of Baker and Putnam. The roadways and their attributes were added manually with a PDF document used for reference. The existing networks highway links and centroid connectors were also reviewed. Link attributes were reviewed and some new attributes were added. The attributes that added included county numbers and district numbers for use in the model. With the addition of the two new counties, some of the external stations were no longer valid. Some external stations had to be moved and others were created for the two new counties. Figure 3-1 shows the new external station locations. The new external stations did not have any information available from the previous model so research and professional judgment were applied to develop the data. Every effort was made to have counts for all external stations. The traffic counts on external stations were updated to 2005 AADT unless there was no information available in the 2005 FDOT Traffic Information CD. Year 2008 counts were acquired for the missing counts and adjusted to 2005 based on historic growth rates. Table 3-1 shows the estimated Peak Season Weekday Average Daily Traffic (PSWADT) derived from AADT of the external station information acquired during this process. The review of the existing network required some changes to highway links and centroid connectors to update them to The addition of new TAZs required some changes to the centroid connectors. This task was labor intensive due to the manual checking required for all TAZs in the two added counties. Other tasks required were: Ramps had to be corrected with the proper facility type codes One way links were checked and corrected, some from two way to one way links Freeway links were checked for incorrect alignments and interchanges and were updated as necessary Links that did not have a connection to the network were removed (stub links) After the highway network was reviewed and updated, the turn penalty and prohibitions were reviewed and updated to accommodate the two new counties and any changes to the highway network. The changes were made and descriptions were added to all prohibitions and penalties. Comments were also added to all turning penalty and prohibitor records. Next, the highway network was reviewed by the TPO, FDOT and Consultants. Consultants reviewed the network to add projects that have been completed up to 2005 and the FDOT review was done on the updated network. Resulting changes included: Branan Field Road, alignment from Blanding Road to 103 RD Street and the number of lanes is 2. Racetrack Road, number of lanes varies from Durbin Creek Boulevard to US 1. Wonderwood Expressway, number of lanes varies from I-295 to SR A1A. Atlantic Boulevard does not have access to I-95 and the nodes should be moved to display a separation. Geometry issues with many links were updated. February 2010 Page 2-3

26 I-95 NB access to J. Turner Butler Boulevard WB is no longer available after Southside Connector interchanges were not added in the 2005 network from Arlington to Merrill. Not available until after Network nodes that were questionable in the packet were moved for better visibility. Towne Center Drive did not exist in 2005 but is built now. It was added in the future network. After an initial model run was completed, the highway network was checked for problem areas. This check is for zero volume roadways and out of range (or extreme) values of volume over count ratios. Micro-coded transit stations were highway network were reviewed and adjusted as needed. During initial part of NERPM4 model validation, the consultant maintained a master network file for coding alternative years. Table 4-2 of TR3-Model application Guidelines describes the selected network attributes in the unloaded alternative specific network. Users should check and update (if necessary) the following node attributes for each of the network years (YY) and alternatives (A): TSRANGE_YYA Maximum roadway distance for auto access connectors (miles); TSPARKSPACE_YYA Number of parking spaces (PNR station only); TSCOSTAM_YYA Station parking cost (cents) in peak period; TSCOSTMD_YYA Station parking cost (cents) in off-peak period; and TSTYPE_YYA Types of access available at station. The following link attributes must be checked and updated (if necessary) for each of the network year (YY) and alternative (A): FTYPE_YYA Two-digit facility type; ATYPE_YYA Two-digit area type; LANES_YYA Number of lanes per direction; and TWOWAY_YYA Whether link is two-way or not. There are also network characteristics that are not distinguished by network year, nor are they brought into the model. These attributes are still important as in the highway evaluation and mapping. The network link attributes not distinguished by network year, are: COUNTY Geographic location code (1=Nassau, 2=Duval, 3=St. Johns, 4=Clay, 5=Baker and 6=Putnam); SCREENLINE Whether link belongs to a screenline, cutline or corridor (a code of 99 or BLANK is used for links not located on a screenline); DISTANCE Link distance; and NERPM Unique ID number used to convert GIS links to highway network link-node structure for the NERPM2000 model. Other transit related link attributes (for a description, see Table 4-2 of TR3 Model Application Guidelines) that users should check are PNRTERMTIME, KNRTERMTIME, TBDIST, TBSTIME, TFGDIST, TFGTIME and TFGMODE. The NERPM4 master network is stored as a CV network and contains data for most significant roadways in the study area. Also included are roadway extensions and new roads representing future alternatives, projects in the North Florida TPO LRTP. It should be mentioned that some links are not part of certain highway networks. For those roadway segments that are to be included in a specific model highway network, this status is switched on or off on the basis of the FTYPE_YYA attribute field (where February 2010 Page 2-4

27 YYA equals the two digit network year and alternative A). Any link which contains a two-digit FSUTMS facility code in this attribute will be included in the model network for the alternative. Any link with a value of zero for this attribute will not be brought into the model. The base year 2005 network as it existed was compared to FDOT and the North Florida TPO data sources in order to identify and rectify any inconsistencies that may have existed. Where necessary, links were added to or removed from the network. Most of the links that were added were necessary in order to accommodate the new TAZ structure arising from zone splits. Links were removed from areas where network detail was considered to be too fine to support a reasonable TAZ structure TAZ Update Prior to NERPM4 the model consisted of only Clay, Duval, Nassau, and St. Johns Counties. In this update, the existing areas were reviewed and two new counties were added, Baker and Putnam, as explained in the highway network section. Some existing TAZs were split for the NERPM4 model. Once the updates to the existing areas were finished, the new county TAZs were added. Figure 1-1 shows the new TAZ area. The next task was to add the socioeconomic data and the attributes of the TAZs to the new file. The existing four counties and two new counties were combined into one new file and then added to the shape file. The attributes of the TAZs were updated and some new attributes were added to complete the TAZ area update. Table 1-1 presents a summary of TAZs by regional districts. It also lists the dummy zones that could be added for a DRI study and/or expansion of TAZs for any future model update study. CTPP districts, model districts, counties, attraction districts, and old TAZ numbers were added to the TAZ shape file. The CTPP districts from the previous model were tagged to the new TAZ area for use in validation. Model districts were created for subarea use in the model. The TAZ file was also tagged with original and NERPM TAZ numbers. The file was found to have errors in the shapes when the new counties were joined to the existing counties. The errors were corrected before model runs and summaries could be started. The NERPM4_TAZ_w05ZonalData.shp file in model s media folder was reviewed by FDOT and one change was requested. The change is regarding the exclusion of water from the two new counties. The water areas were removed from the TAZ file in the counties of Nassau, Duval, Clay and St. Johns Review of Highway Network Similar to earlier updates (NERPM 1998 & 2000, JTA/RTS 2005), the 2005 NERPM4 update incorporated the Florida Department of Transportation Model Task Force (MTF) endorsed 2-digit coding. FSUTMS includes a set of standard area type and facility type definitions used to describe key roadway characteristics. Area types are used to define the land use adjacent to each roadway link. Facility types classify each roadway link according to its function and/or design characteristics. While MTF s HNET Procedural Enhancements Study provides fairly concise definitions for FSUTMS area types and facility types, there is still room for subjective judgment in the coding of highway networks. Such subjectivity is beneficial when it reflects an individual s personal knowledge and experience of the study area. Initial highway network characteristics were recoded periodically in order to enhance model performance. The two-digit area type (AT) classification is shown in Table 2-1, while the two-digit facility type (FT) classification is shown in Table 2-2. Area types and facility types conform to the standard 2-digit definitions approved by the Model Task Force (MTF). A few new area types (AT 35) and facility types (FT 29, 40, 52, 59 and 69) were used in the NERPM4 network. Table 2-1 also presents the terminal time for each of the area types. February 2010 Page 2-5

28 After establishing a preliminary highway network, efforts were turned towards examining the network for consistency and accuracy. A series of plots displaying facility types, area types, and number of lanes were generated using Cube software. These three network characteristics form the basis for the FSUTMS speed and capacity lookup table. These plots were reviewed by subarea, corridor, and link, based on local area knowledge, variable consistency, and logic. Network characteristics were modified as needed in order to better reflect a more correct representation of real world conditions using available satellite photography. The NERPM4 highway networks were reviewed and edited for the following link characteristics: Facility Type Area Type Number of Lanes Centroid Connections and Locations Added Network Detail Turn Prohibitors Network Geometry As part of model validation efforts, consistency of the TAZ structure and that of highway network was checked by overlaying the two layers in Cube software. The TIGER street network was used to check the centroid connectors. In the regional model, the interfaces of six counties were examined in detail. Numerous plots were made to display key network attributes (facility and number of lanes) along with model volumes and counts and their ratios of NERPM4 networks. Problems with facility types and number of lanes were investigated through using CUBE and the color-coded plots. Numerous changes were made to the networks based on the review of these plots. February 2010 Page 2-6

29 Table 2-1: Definitions of Area Types and Terminal Time Values Terminal Area Types Area Type Descriptions Times (minutes) 1x CBD Areas (AT 10 is default) 5 11 Urbanized Area (over 500,000) Primary City Central Business District 5 12 Urbanized Area (under 500,000) Primary City Central Business District 4 13 Other Urbanized Area Central Business District and Small City Downtown 4 14 Non-Urbanized Area Small City Downtown 4 2x CBD Fringe Areas (AT 20 is default) 3 21 Central Business District Fringe Areas 3 22 Industrial Fringe 3 23 Strip Commercial 3 3x Residential Areas (AT 30 is default) 1 31 Residential Area of Urbanized Areas 1 32 Undeveloped Portions of Urbanized Areas 1 33 Transitioning Areas/Urban Areas over 5,000 Population 1 34 Beach Residential 2 35 St. Johns River Bridges 1 4x OBD Areas (AT 40 is default) 2 41 High Density Outlying Business District 2 42 Other Outlying Business District 2 43 Beach Outlying Business District x Rural Areas (AT 50 is default) 1 51 Developed Rural Areas/Small Cities under 5,000 Population 1 52 Undeveloped Rural Areas Centroid locations and connectors were reviewed and some changes were made to better reflect access to the roadway network. Numerous plots were made at the early stages of this study to allow for the review and update of facility type, area type and number of lanes coding. Plots showing the highway network facility and area type designations are shown in Figures 2-1 and 2-2, respectively. Figure 2-3 presents the number of lanes. Periodically, throughout the model validation, the network was reviewed to ensure that no inconsistencies were overlooked Transit Network Elements Coded onto the Highway Network The FSUTMS transit modeling standards eliminate the need of manually-coded ASCII files by requiring that certain transit network components be coded onto the highway network. These include bus-only February 2010 Page 2-7

30 links, fixed-guideway transit links, and transit stations. The section below provides a summary of the transit elements that are coded in the highway network. A more detailed description of the coding procedure is provided in the Application Guidelines document (TR3). February 2010 Page 2-8

31 Table 2-2: Definitions of Facility Types 1x 2x 3x 4x 5x 6x FT Facility Type Descriptions Freeways and Expressways (FT 10 is default) 11 Urban Freeway Group 1 (cities of 500,000 or more) 12 Other Freeway (not in Group 1) 15 Collector/Distributor Lane 16 Controlled Access Expressway 17 Controlled Access Parkway Divided Arterials (FT 20 is default) 21 Divided Arterial Unsignalized (55 mph) 22 Divided Arterial Unsignalized (45 mph) 23 Divided Arterial Class 1a 24 Divided Arterial Class 1b 25 Divided Arterial Class II/III 29 Mayport Ferry Undivivided Arterials (FT 30 is default) 31 Undivided Arterial Unsignalized with Turn Bays 32 Undivided Arterial Class 1a with Turn Bays 33 Undivided Arterial Class 1b with Turn Bays 34 Undivided Arterial Class II/III with Turn Bays 35 Undivided Arterial Unsignalized without Turn Bays 36 Undivided Arterial Class 1a without Turn Bays 37 Undivided Arterial Class 1b without Turn Bays 38 Undivided Arterial Class II/III without Turn Bays Collectors (FT 40 is default) 41 Major Local Divided Roadway 42 Major Local Undivided Roadway with Turn Bays 43 Major Local Undivided Roadway without Turn Bays 44 Other Local Divided Roadway 45 Other Local Undivided Roadway with Turn Bays 46 Other Local Undivided Roadway without Turn Bays 47 Low-Speed Local Connector 48 Very Low-Speed Local Connector 49 Transit Only Driveway Link Centroid Connectors (FT 50 is default) 51 Basic Centroid Connector 52 External Station Centroid Connector 59 Transit Only Platform/Escalator and Walk Access Link One-Way Facilities (FT 60 is default) 61 One-Way Facility Unsignalized 62 One-Way Facility Class 1a 63 One-Way Facility Class 1b 64 One-Way Facility Class II/III 65 Frontage Road Unsignalized 66 Frontage Road Class 1a 67 Frontage Road Class 1b 68 Frontage Road Class II/III 69 Transit Only Skyway Link February 2010 Page 2-9

32 Table 2-2 (contd.): Definitions of Facility Types 7x 8x 9x FT Facility Type Descriptions Ramps (FT 70 is default) 71 Freeway On-Ramp 72 Freeway Loop On-Ramp 73 Other On-Ramp 74 Other Loop On-Ramp 75 Freeway Off-Ramp 76 Freeway Loop Off-Ramp 77 Other Off-Ramp 78 Other Loop Off-Ramp 79 Freeway-Freeway High-Speed Ramp HOV Facilities (FT 80 is default) 81 Freeway Group 1 HOV Lane (Barrier Separated) 82 Other Freeway HOV Lane (Barrier Separated) 83 Freeway Group 1 HOV Lane (Non-Separated) 84 Other Freeway HOV Lane (Non-Separated) 85 Non-Freeway HOV Lane 86 AM&PM Peak HOV Ramp 87 AM Peak Only HOV Ramp 88 PM Peak Only HOV Ramp 89 All-Day HOV Ramp Toll Facilities (FT 90 is default) 91 Freeway Group 1 Toll Facility 92 Other Freeway Toll Facility 93 Expressway/Parkway Toll Facility 94 Divided Arterial Toll Facility 95 Undivided Arterial Toll Facility 97 Toll On-Ramp 98 Toll Off-Ramp 99 Toll Plaza Bus-only links are streets that are critical for the correct coding of transit routes but are not represented or loaded in the modeled highway network. They are coded with a facility type 49. The time and the distance on these links can be manually set using link variables TBSTIME and TBSDIST. The connector links from the street to the BRT station are also coded with facility type 49. The default transit time on these links is one minute but can vary based on the characteristics of individual stations. Transit links that connect fixed-guideway stations are coded with a facility type 69. Examples of such links are the Skyway links in the base year model. The station nodes for the Skyway and BRT services are coded separately from the bus stop nodes to more accurately reflect the different boarding platforms. The bus stop and the fixed-guideway stations are connected by a transfer link coded with a facility type of 59. The default walk time on these links is one minute but can vary based on the characteristics of individual stations. Links with facility types 49, 59 and 69 are excluded from highway skimming and assignment. The new fields created in the highway network to represent the distance and the speed correctly on such links are shown in Table 6-1 of TR3-Model Application Guidelines. February 2010 Page 2-10

33 Figure 2-1: Year 2005 Highway Network Highlighting Facility Types February 2010 Page 2-11

34 Figure 2-2: Year 2005 Highway Network Highlighting Area Types February 2010 Page 2-12

35 Figure 2-3: Year 2005 Highway Network Highlighting Number of Directional Lanes February 2010 Page 2-13

36 Transit station data is coded on the node layer of the highway network. Transit stations include formal park-ride or drop-off (kiss-ride) locations, major transit hubs, and BRT or fixed-guideway stations. The variables shown in Table 6-4 of TR3-Model Application Guidelines are included in the node layer. For formal JTA park-ride stations, the terminal times can be set using PNRTERMTIME (for park-ride) and KNRTERMTIME (for drop-off/kiss-ride) fields on the highway link layer for links which connect the park-ride (PNR) node to the station/stop node(s). The TSTYPE node field describes the nature of the parking station. The AUTOCON program builds park-ride connectors to stations with TSTYPE=1, 3, 5 or 6. The new PNR station coding as designed for FDOT Systems Planning provides a more detailed representation of the individual facilities within the station. Separate nodes are used to represent the parking lot, the rail platform/fixed-guideway station and the bus stop. The station information is coded only on the PNR node to minimize coding. Drive-access connectors generated by the AUTOCON program, which recognizes these coding procedures, will be built to the station and/or stop nodes and not to the PNR node. Driveway links (with facility type 49) are added to access PNR lot from the street. The PNR node is connected to the bus stop and the station with links coded as facility type 59. The terminal times at the parking lot are coded on these links in the fields PNRTERMTIME (for park-ride) and KNRTERMTIME (for drop-off/kiss-ride). It is important for the station information to be correctly coded on the PNR node since the station data file used by the AUTOCON program is generated from the PNR nodes during the model run. Coding information on all transit link fields is not necessary. Only certain fields are required; the fields that must be coded depend on the type of transit link being used. Table 6-3 of TR3-Model Application Guidelines lists the variables which are required by the model Review of Transit Network NERPM4 uses the 2005 based JTA/RTS model PT based transit network. It was derived from a series of GIS shape files containing stop and route data. These shape files were developed based on route scheduling data and transit stop coordinate data provided by the Jacksonville Transportation Authority (JTA). The transit networks in NERPM reflect transit service operated by JTA during the year Chapter 7 of this report has more description of the transit network Review of Transit Ridership Data Accurate transit ridership data is vital to the correct assessment of transit assignment models. Data were provided by the Jacksonville Transportation Authority concerning the average daily boardings and alightings of transit vehicles that was used in JTA/RTS 2005 model validation. These ridership data, along with other transit route characteristics data assembled for NERPM4 model validation, are summarized in Table B-16 of Appendix B. Transit target numbers used in NERPM4 were gathered from the RTA/JTS model along with Census 2000 Journey-to-Work statistics. 2.3 Traffic Count and Screenline The validation of any travel demand model relies upon the existence of extensive base year traffic count data. Volume-to-Count (V/G) ratios generated by the model are used to measure the ability of a travel demand highway assignment model to simulate known traffic conditions. Traffic counts are needed for a variety of different roadway categories distributed throughout the study area in order to validate highway assignment performance along screenlines, facility types, area types, and lane configurations. February 2010 Page 2-14

37 An updated and accurate traffic count database is a critical tool in the travel model calibration and validation step. In some cases, the task of adding traffic counts to the model for use has to be done manually, which is labor intensive and time consuming. In the NERPM4 model, the traffic counts for state roadways were added by some creative GIS techniques. However, this process did not eliminate the need for manual coding. The directional links, freeways and expressways had to be coded as one way pairs and ramps had to be checked due to the offset between the database and the NERPM4 network. Local counts were not available in GIS format, so they also had to be manually coded. Like most FSUTMS models, NERPM assigns trips to the highway network in terms of peak-season weekday average daily traffic (PSWADT). Typically, traffic count data collected from various reliable sources are reported in average annual daily traffic (AADT). These AADT figures are then converted to PSWADT based on the model output conversion factor (MOCF) provided by FDOT and made available on the CD entitled 2005 Florida Traffic Information. Along with MOCFs, this CD contains geographically related data on traffic count location and AADTs. Traffic count data were stored into a separate GIS shape file that contained not only the AADT information, but also data source, count station number, and a unique reference number that allowed the count station shape file to be associated with the highway network shape file. The counts were then added to the appropriate links in the highway network shape file. If particular locations existed without counts for the year 2005 but were necessary in order to preserve screenlines, count volumes were estimated by using past count data at that location or other locations nearby to establish a rate of growth. If no count data existed for a given location for either the base year or any other given year, then no count was entered for that location. A separate count study was not undertaken as part of the NERPM4 validation effort, other than the referenced external counts Data Sources For the base year model validation, the NERPM4 regional model requires 2005 Annual Average Daily Traffic (AADT) values coded to the network links directionally. In this process, there were two different sources used for input. FDOT s AADT point data was used to begin the process. To supplement this database, local AADT volumes were obtained from the four existing counties. All counts were already adjusted for seasonal and truck factors AADT volumes were not available on some local roadways, so a growth factor was applied for adjustment. The two main data sources used in count coding were: FDOT traffic counts: FDOT AADT shape file for NERPM4 region was created from the 2005 Florida Traffic Count DVD. Clay, Duval, Nassau and St. John s County local counts: A spreadsheet of Annual Average Daily Traffic (AADT). In addition, Corradino staff obtained some local counts of Baker and Putnam counties from the PBS&J staff. They were coded manually on the network. Traffic count data for the study area came from multiple sources. First, the 2005 Florida Traffic Information CD from FDOT represents traffic count data that were given the highest level of confidence. Whenever possible, traffic count data from FDOT were used. Second, when data were not available from FDOT, locally assembled count data were used. Counts located on state roads tended to have data February 2010 Page 2-15

38 corresponding to the FDOT CD. Third, if neither the FDOT CD nor the locally obtained counts for a particular location existed then data were sought from the counties. This was typically necessary only for the more rural portions of Nassau, Clay, St. Johns, Baker and Putnam Counties. Finally, year 2000 model external station traffic counts were reviewed when the locations were the same as in NERPM2000 and JTA/RTS Count Coding The first task in the count coding process was to match traffic count point files with the directional links in the network by doing an extensive GIS analysis on the FDOT data. The analysis was completed using a GIS overlay process. Overlay is a procedure that estimates the attributes of one or more features by superimposing them over other features and deciding the extent to which they overlap. The overlay process is similar to tagging but controlled by the bandwidth sizes. Several overlays were computed, varying the bandwidths from one foot to 200 feet. All tagged points using a one-foot bandwidth were reviewed for accuracy and the next overlay was computed using a twofoot bandwidth. In this process, all the points that were tagged accurately in the one-foot bandwidth were excluded from the overlay computation using two-foot bandwidth. This controlled tagging process was repeated up to a maximum bandwidth of 200 feet. In addition, all freeway and ramp counts were checked with extensive manual reviews for accuracy and all the points that were not tagged within a 200 feet bandwidth were manually reviewed and coded. Even though the overlay process facilitates automatic tagging of points to the lines, careful review for accuracy is needed at freeways, ramps, intersections and model links that have higher offsets from the street centerline files. The key used to identify this set in the network is CNTSRC_YR05 equal to FDOT. Local traffic count data was manually coded to the line layer using the description of traffic count station as the identifier. The key used to identify this set in the network is CNTSRC_YR05 equal to Local. Growth factors were applied to all the local counts that were not taken in the year Separate growth factors were developed based on the historical data of the roadway or similar roadways for estimating 2005 counts. Local traffic count source data ranged from 2001 to Review of Traffic Count Data The assessment of highway assignment model performance depends on the accuracy of traffic counts along highway network links. Highway evaluation statistics, such as volume-to-count ratios (V/G), vehicle-miles traveled (VMT), vehicle-hours traveled (VHT), and root mean squared error (RMSE), are calculated on the basis of traffic count data. Errors in the traffic count data can distort the accuracy of the assignment model so as to make the model appear to be more or less accurate than it actually is. The 2005 counts (Average Annual Daily Traffic AADT_YR05) were reviewed for reasonableness and edits were made as needed. Count data are used by the HEVAL routine in validation mode to compare the model generated traffic volumes against the traffic counts. Care was taken to ensure that count data were available for model validation. Network plots were produced to facilitate quality control and oversight on traffic count data. Figures 2-4 and 2-5 show FDOT and locally collected traffic count locations in the NERPM4 area for the 2005 base year model validation. As a continual part of model validation, traffic count data were reviewed, reevaluated, and if appropriate, corrected to improve model accuracy. Occasionally, poor validation statistics could be the result of an error with the count data. If the data were entered incorrectly, such as by placing the count on the wrong link or entering the wrong count value, the result of the volume-to-count (V/G) ratio for the link could be February 2010 Page 2-16

39 in error. Many such occurrences were evident as drastic spikes or dips in the V/G ratios inconsistent with other V/G statistics occurring along a given corridor. Such counts were corrected when they were discovered. In some cases, the exact location of a traffic count station was unclear. This could have mainly been due to slight inconsistencies between the traffic count station location shape file and the NERPM highway network shape file. These inconsistencies, while slight, may have located traffic counts on inappropriate sides of centroid connectors. Count station descriptors in the data provided from the North Florida TPO, the counties, and FDOT make it unlikely that traffic counts were located on inappropriate sides of network link (non centroid connector) intersections. If it was possible to move a count slightly, then the count was moved if doing so improved model accuracy without compromising the true count location. February 2010 Page 2-17

40 Figure 2-4: FDOT Traffic Count Locations February 2010 Page 2-18

41 Figure 2-5: Locally Collected Traffic Count Locations February 2010 Page 2-19

42 These counts were carefully reviewed for reasonableness and edits were made where necessary. The count data are used by the HEVAL routine in validation mode to compare the model generated traffic volumes against the traffic counts. Care was taken to ensure that enough count data were available for model validation. Table 2-3 presents a summary of the links by main facility and area types with traffic counts. For the whole NERPM4 region, 6.5% the links have traffic counts. Table 2-3 also presents the number of links with traffic counts as well as the total number of links. The information in Table 2-3 is valuable for judging the model statistics by facility and area type for each county by the variation of the percentage of links with traffic counts. Unlike NERPM2000, NERPM4 has many ramp counts. For other facility types, the percentage of links having traffic counts varies from 3.22% (collector) to 9.84% (freeways). By area type, the percentage of links having traffic counts varies from 5.69% (rural) to 7.11% (CBD). There are significant differences in the percentages between the counties (varies from 5.21% in Clay county to 11.96% in Baker county) of links with traffic counts. Table 2-4 also presents the counts availability by the facility and area types for each six counties Review of Screenlines Screenlines and cutlines define groups of roadways that travel in the same direction, and carry traffic considered significant within the study area. The following characteristics are considered during the review (addition/deletion) of screenline and cutline locations: Availability of traffic counts Representative of travel patterns Minimized duplication of travel patterns. Comparisons between screenline locations and traffic count locations ensured that screenlines were as complete as possible. Whenever possible, screenline locations with missing counts were relocated to nearby links with traffic count data. Occasionally, when it was not possible to relocate a screenline due to a lack of counts, the screenline would be shortened so as to minimize the number of screenline locations without counts. Screenlines, cutlines and cordons are drawn across the model network throughout various parts of the study area for summary of traffic volumes in subareas and along major corridors. These screenlines are used to report an aggregate volume-to-count ratio for all of the links that comprise any given screenline. This allows for measurement of travel flows between various parts of the study area. The starting point for developing screenlines for NERPM4 was to review the screenlines that were already present in NERPM 2000 and JTA/RTS models. The scrreenline and cutline codes established for the 2000 NERPM network were also preserved in the 2005 network to the extent they were applicable. These screenlines were checked to ensure that their orientation coincided with traffic count locations. February 2010 Page 2-20

43 Table 2-3: Link Traffic Count Summary by Facility and Area Types February 2010 Page 2-21

44 Table 2-4: Link Traffic Count Summary by Facility Type and County February 2010 Page 2-22

45 Screenlines typically follow natural features, major transportation facilities, or political boundaries. Also, screenlines can be used to cordon off certain portions of the study area in order to measure the flows into and out of those areas (such as measuring the flow of travel demand into and out of smaller urban areas CBDs or the external model boundary). The screenlines, cutlines and cordons, as shown in Figure 2-6, cover all major regional and local travel movements in sufficient detail. Every effort was made to maintain consistency between screenline locations and traffic count locations. When a count was missing, either the count would be produced from an exhaustive review of count data sources or the screenline was moved to a nearby count location that was a reasonable substitute for the missing count. After securing the orientation of current screenlines, it was necessary to determine where new screenlines were needed and where old screenlines were obsolete or redundant. Of these, only three significant changes took place. First, two new screenlines (no. 23 & 40) were added as a cordon around the towns of Macclenny of Baker County and Palatka of Putnam County. Second, the external station cordon line was updated to cover the entire study area. This was vital for validation of the external model. 2.4 Updates of Speeds and Capacities and Validation Speeds, capacities and volume/delay functions play an important role in nearly all facets of the travel demand model. Initially, a speed-capacity table was developed in the NERPM4 model that was consistent with the changes made in 2005 based JTA/RTS transit model input speed. Figure A-1 (see Appendix A) presents a list of changes in network and speed capacity modifications of the CV script of JTA/RTS model. All changes were directly incorporated in the NRPM 4 SPDCAP table and network. A snippet of the SPDCAP table with proper comments for JTA/RTS changes is shown in Figure A-2 of Appendix A. The initial speed and capacity table that was used in both 2000 based NERPM and 2005 JTA/RTS models is consistent with the 2002 Quality/ Level of Service Handbook from FDOT. The speeds and capacities used in the final validation for NERPM 2000 were incorporated directly from the NERPM 1998 model validation. The 1998 NERPM speed and capacity table was also result of an auto-calibrated SPDCAP file. The NERPM 1998 SPDCAP file was originally based on the JUATS 1998 speed and capacity table but was ultimately modified through the use of the Automatic Model Calibration System developed for the 2000 Treasure Coast Regional Planning Model update. This process automatically conducts an iterative series of runs and updates to fine tune input files to given levels of tolerances. In NERPM4, after implementation of the JTA/RTS transit model related speed changes, an iterative process of manual adjustments was conducted in order to improve model validation while maintaining a logical hierarchy of speeds. The initial speed is one of the key model parameters adjusted during the validation process. This adjustment can make specific transportation facilities more or less attractive, thereby causing the model to produce estimates that are closer in magnitude to observed conditions. Several changes were made to the method for estimating initial speeds during the course of 2005 model validation process. The adjustments to the initial speeds were an iterative process designed to yield better estimates of traffic volumes that reflect observed traffic flows as well as to replicate observed speeds. Primarily, adjustments were made to be specific to certain area and facility type combinations so as to avoid unintended impacts. Figure A-3 (see Appendix A) presents all speed and capacity modifiers of validated speed capacity table. No systematic patterns of changes were noticed in these modifiers. February 2010 Page 2-23

46 Figure 2-6: Screenline, Cutline and Corridor Locations See Duval Insert (Next page) February 2010 Page 2-24

47 Figure 2-6 (contd.): Screenline, Cutline and Corridor Locations Duval Inset DUVAL SCREENLINES February 2010 Page 2-25

48 Data from the HEVAL summary were used to check the hierarchy of the speeds and capacities. The original speeds were also compared to the model generated congested speeds. Tables 2-5 and 2-6 present summaries of 2005 base and 2035 E-plus-C trend model runs. The logical hierarchies of speeds are evident in these tables. Statistics on original and congested speed, change in speed and percent change in speeds are reported for each cell by main facility type. For the 2005 model, an overall decrease in 5.74 mph (13.7%) is shown between the original and congested speeds. The percent change in speeds among the facility types in the 2005 validation run ranges from -2.5% (one-way) to -13.5% (ramps). Once again, the trend of speed decrease due to congestion by facility type is reasonable. For the 2035 model, the overall decrease between the original and congested speeds is mph (30.6%). The percent change in speeds among the facility type in 2035 run ranges from -4.4% (one-way) to -24% (ramps). The change is reasonable because of the increase in the number of trips in the 2035 model. Logical hierarchies in speed are exhibited in Tables 2-5 and 2-6. By facility type, higher volume facilities are more congested, and rural areas are less congested. None of these results are contrary to the observed travel characteristics in the region. Speeds for each of counties were examined by both facility and area types and their combinations from corresponding period HEVAL outputs. They provided more insight into speeds for each county and assisted in model validation efforts. A summary of the systemwide capacities (in vehicle-per-lane-per-hour) was made from the CV generated cross tab reports. This summary was made for each of the main facility and area type combinations. Table 2-7 is a summary in vehicles per hour per lane. This table also shows the number of lane miles for each combination of facility and area types. The NERPM4 speeds and capacities in Tables 2-5, 2-6 and 2-7 conform to the expected hierarchies among the facility and area types and the reported values are very reasonable. The NERPM4 model enhanced the 2005 JTA/RTS model and expanded the study area to include Baker and Putnam counties. There were inconsistencies in the 2005 JTA/RTS model in the speeds between the final highway assignment and the speeds used in the transit model. This was primarily due to the fact that the goal of the JTA/RTS model was to enhance transit model and was not concerned with the highway model. The validation of NERPM4 resulted in consistency between the highway and transit components based on 2005 traffic counts and transit data. Speeds statistics from the JTA/RTS and NERPM 2000 models were generated by HEVAL and compared with NERPM4 model speeds. Table 2-8 presents these comparisons by the main facility type groups. Further summaries of the initial and congested speeds of JTA/RTS and NERPM4 models final highway assignments by their FT and AT group combinations are shown in Tables B-8 and B-9 (see Appendix B), respectively. Both NERPM4 and JTA/RTS models includes a pre-assignment step for distribution of work trips with congested skims. The pre-assignment step performs a two hour AM peak period highway assignment. The congested skims of the pre-assignment step also used in peak period transit modeling. Tables B-6 and B-7 of Appendix B presents summaries of initial and congested speeds of NERPM4 and JTA/RTS 2005 models. These summaries were made by each main facility and area type combinations. The following conclusions are based on the summaries of Table 2-8: JTA/RTS model s initial speeds used in final highway assignment and in pre-assignment for transit model are not consistent. NERPM4 model uses same initial speeds throughout the model chains. February 2010 Page 2-26

49 NERPM4 model s initial speeds are very similar to those used in JTA/RTS pre-assignment for transit model validation. NERPM4 model s congested speeds are also very similar to those resulted from JTA/RTS preassignment for transit model validation. NERPM4 model s initial and congested speeds are significantly higher than those used in 2000 based NERPM model. Those differences are more evident for higher facility groups (for example, freeways and ramps) February 2010 Page 2-27

50 Table 2-5: 2005 HEVAL Speed Summary by Facility and Area Type Combinations of NERPM4 February 2010 Page 2-28

51 Table 2-6: 2035 (Trend) HEVAL Speed Summary by Facility and Area Type Combinations of NERPM4 February 2010 Page 2-29

52 Table 2-7: Summary of 2005 Lane-Mile and Capacity by Facility and Area Type Combinations February 2010 Page 2-30

53 Table 2-8: Comparison of Model Input and Congested Speeds by Facility and Model February 2010 Page 2-31

54 3. External Trips This chapter presents the validation of external trips. Highway external trips are divided into externalinternal (IE and EI) and through (EE) vehicle trip ends. Modeling EE trips is the first module in CV application (see Figure 1-1). The external trip module requires an EE trip table that contains EE vehicle trip between external stations. External stations are intersections between the network and the study area boundary. These stations serve as ports of entry and exit from/to the study area. The NERPM4 study area consists of all of six counties (Nassau, Duval, Clay, St. Johns, Baker and Putnam) of northeast Florida (see Figure 1-1). All recent earlier regional models [NERPM2000 & JTA/RTS, see References 1-6] study area included four counties (Nassau, Duval, Clay and St. Johns). As a result, some new external stations were added in NERPM4 and a few external stations were moved to new locations. There are 29 external zones in NERPM4 as opposed to 23 stations in earlier 4-county models. These zones are numbered 2,550 through 2,578 and are shown in Figure Model Description Most FSUTMS models use a set of simplified assumptions to determine which external trips are eligible for travel in high-occupancy vehicle (HOV) lanes. The typical FSUTMS process assumes that all external-external (EE) trips are HOV eligible and internal-external (IE) trips are not HOV eligible. For the NERPM2000 model [Reference 4-6], it was recommended that survey-based estimates of HOVs and commercial vehicles be added to the external model so that the future year model could test policies and projects that treat these travel modes differently. The current NERPM4 model continued these assumptions for modeling external trips. The resulting external model described in this section adds the capability of testing lanes and facilities that are restricted to trucks, HOVs, and/or through trips. The external-internal data in NERPM is different from the standard FSUTMS ZDATA4 file. The NERPM4 external-internal file is named as EITRIPS and is in DBF format. In addition to the number of trips by external zone, the NERPM external-internal data file includes the percent trips for the following trip purposes or modes: Single-Occupancy Vehicle (SOV); High-Occupancy Vehicle (HOV); Light-Duty Truck (LDTK); and Heavy-Duty Truck (HDTK). SOV trips are defined as trips with one person per vehicle, whereas HOV trips are defined as trips with two or more persons per vehicle. Definitions of light- and heavy-duty truck were based on official FDOT truck classification data. The EETRIPS input file is generally the residual left after estimating IE trips in EITRIPS. The percentage of EE trips was applied to the PSWADT per external zone and then distributed from each origin zone to each destination zone. A revised distribution pattern was estimated for NERPM4 using current traffic data and the distribution pattern from earlier models. The EETRIPS file includes the total EE trips per external origin and destination zone interchange, whereas the EXTAOFAC input file includes the percent of trips by external purpose or mode from each February 2010 Page 3-1

55 external origin zone to each destination zone. EXTAOFAC is a new file not found in other FSUTMS models. Users should consult Technical Report 3 - Model Application Guidelines for a description of model inputs, parameters and output files. Five trip tables were generated in the final output external trip matrix, one for each mode of travel, and the total. February 2010 Page 3-2

56 Figure 3-1: External Station Locations Note: See Table 3-1 for description of Station February 2010 Page 3-3

57 3.2 Data Development and Validation Adjustments The external model was revised to use survey-based percent truck estimates from the 2000 NERPM model. In order to test the impact of truck exclusive lanes and high-occupancy vehicle lanes within the NERPM, the external model was divided into the following four modes or purposes: 1. Low-Occupancy Vehicle (LOV); 2. High-Occupancy Vehicle (HOV); 3. Light-Duty Truck (LDTK); and 4. Heavy-Duty Truck (HDTK). LOV trips are defined as trips with one person per vehicle, whereas HOV trips are defined as trips with two or more persons per vehicle. The percentage of trips per external zone allocated for the LDTK trip purpose was derived from the six-tire vehicle class percentages used in NERPM2000, which used information from the 2000 North Florida External Travel Survey. Table B-12 in Appendix B summarizes the external model data that were used in NERPM 2000 and JTA/RTS models. The percentage of trips per external zone allocated for the HDTK trip purpose was based on either the three-axle/semis vehicle class percentage in the 2000 NERPM External O/D Survey or FDOT vehicle classes 6 through 13, whichever was more appropriate for the subject external station. Classification data from official FDOT traffic count stations, as reported on the 2005 Florida Traffic Information CD, were used wherever these coincided with external stations. For the local road external stations, reasonable assumptions were made on truck trip percentages. Three input files were revised to reflect the external model refinements: ZDATA4 (Internal-External Productions); EETRIPS (External-to-External Trip Table); and EXTAOFAC (External Auto Occupancy Factors). Although the ZDATA4 input file is part of the trip generation process, it is integral to generating the EETRIPS file. The percent IE/EE split for each external zone was primarily derived from the NERPM200 model. Adjustments were made to account for the expanded study area and the new external stations. Table 3-1 summarizes the traffic counts and their distribution among internal and internal-external trips. It also presents the percentages of trips by mode of travel. In the ZDATA4 file, the percent IE trips was applied to the 2005 PSWADT to calculate the total number of IE trips for each zone. The percentages allocated for each external trip purpose were then applied to the total PSWADT per external zone. The EETRIPS input file is generally the residual left after estimating IE trips in ZDATA4. Validation of the EETRIPS file was based on extrapolation and professional judgment. The EETRIP file validation should generally rely upon recently collected roadside or cordon line surveys to determine the proportion of the vehicle traffic that passes through the study area. Since recent data are not available, this study builds on the 2000 EETRIP file based on the percent distribution of external trips from earlier models and then adjusts them slightly after comparing the 2000 and 2005 traffic counts at the external stations. The percentage of EE trips was applied to the PSWADT per external zone and then distributed from each origin zone to each destination zone using distribution patterns from the 2000 EETRIPS file. Professional judgment was applied to account for the expanded NERPM4 study area as well as addition of several new external stations. The initial through trip table was assigned to the model and the bandwidth plots were examined to assess the external-external model flows for reasonability. Adjustments were made where appropriate. Figure 3-2 presents a bandwidth plot of 2005 external trips. The FDOT, MPO and February 2010 Page 3-4

58 Consultant staffs reviewed the resultant through trip table to affirm the reasonableness of the data for model validation. This was necessary particularly because of the expansion of the study area. Table 3-1: 2005 External Station Traffic Count Summary February 2010 Page 3-5

59 The EE movements between minor facilities that were not near to each other were either removed or adjusted downward to better match volume-over-count ratios along screenlines significantly impacted by external trips. Adjustments also were made to EE movements that simulated u-turn trips crossing over to the same general origin as the destination (e.g., trips from one external zone to Bradford County to another external zone bordering Bradford County). These adjustments proved successful in eliminating over-assignments in some of the rural areas near the model boundary. The validated 2005 external through trip table is summarized in Table 3-2. February 2010 Page 3-6

60 Figure 3-2: Bandwidth Plot of External Assigned Volumes February 2010 Page 3-7

61 Table 3-2: 2005 External Vehicle Trip Table Summary February 2010 Page 3-8

62 A similar process was used to develop the 2035 EETRIPS file. It used the growth factors that were derived by trend analysis of historical traffic counts for each external station. Appendix C summarizes the results of the trend analysis. The growth factors of this analysis were then applied to the 2005 PSWADT. Table 3-3 summaries 2035 external station traffic projections for use in 2035 LRTP model runs. Overall, there is a 39% growth in through trips from 2005 to For the 2035 external model, the distribution of trips between external and inter-external and those among the modes of travel assumed the same percentages as in the model. The same station specific growth factors were used to extrapolate the through trip flows through a Fratar model. The CV scripts to perform this extrapolation shown below: ; Do not change filenames or add or remove FILEI/FILEO statements using an editor. Use Cube/Application Manager. RUN PGM=FRATAR PRNFILE="D:\NERPM4\APPLICATIONS\GNFRA00B.PRN" MSG='Frartar/Develop 2035 EETRIPS with target trips' FILEI MATI[1] = "D:\Jacksonville\SksWorkOfInputFiles\Externals\EETRIPS05.MAT" FILEO MATO[1] = "D:\Jacksonville\SksWorkOfInputFiles\Externals\EETRIPS35Tem.MAT", MO=1, NAME=XX FILEI ZDATI[1] = "D:\Jacksonville\SksWorkOfInputFiles\Externals\EE2035Target.dbf" ; The FRATAR module does not have any explicit phases. The module does run within an implied ILOOP ; where I is the origin zones. The module implements a FRATAR distribution process by modifying an ; existing matrix based on a set of user supplied production and attraction factors. PARAMETERS ZONES={ZONESA} MAXRMSE=0.01 MAXITERS=50 SETPA P[1]=zi.1.EETRGT, A[1]=zi.1.EETRGT, MW[1]=mi.1.5 CONTROL=PA INCLUDE={ExtZnStrt}-{ZONESA} ACOMP=1,PCOMP=1 MARGINS=1 MARGINREC=y LIST=J,R1(6.0),C1(6.0) Through trip tables used in the model are balanced external trip tables. The 2035 external OD trip table uses in 2035 LRTP model runs is summarized in Table 3-4. Staring with 2000 NERPM, external trips were prohibited to use certain facilities during assignment phase. Without such adjustments there were illogical trip assignments where EE trips diverted from major roadways onto the local network to bypass congestion. It is reasonable to assume that the vast majority of EE travelers passing through the study area would not be familiar enough with the local street network to feel confident in making such trip decisions. To prevent this, special network code (EECODE) was introduced in the NERPM network to prohibit EE trips from using links with an EECODE of 1. This process was continued in NERPM4. Figure 3-3 highlights the facilities that prohibit the external trips. The prohibition includes CR 121 and all facilities in and outs of downtown Jacksonville and St. Augustine core areas. As seen in figure, some major ramps of I-95, I-10 and I-295 and bridges were also included in this external trip prohibition. February 2010 Page 3-9

63 3.3 Results and Comparisons The external trips consist of both IE passenger trips and EE vehicle trips. Total assigned model volumes were compared to the traffic counts and the results are presented in Table 3-5. Results are summarized from the HEVAL validation mode output. The volume/count ratios for all external stations are In particular, a review of the external cordon line and other screenlines close to the model boundary indicate a reasonable match of external travel movements. February 2010 Page 3-10

64 Table 3-3: 2035 External Station Traffic Projection Summary February 2010 Page 3-11

65 Table 3-4: 2035 External Vehicle Trip Table Summary February 2010 Page 3-12

66 Figure 3-3: Highlighted Facilities Prohibiting External Trips February 2010 Page 3-13

67 Table 3-5: Comparison of 2005 External Station Traffic Counts and Volumes External TAZ Roadway Name 2550 I-95 North Est 2005 PSWADT Model Volume Volume/C ount 63,808 63, US 17 North 3,656 3, US 1 North 11,290 11, CR 2 2,280 2, SR 121 2,746 2, SR 2 East SR 2 West Florida Grade CR I-10 West 23,260 23, US 90 4,388 4, CR CR SR 121 3,878 3, US ,106 22, CR 225 3,020 3, SR 16 6,562 6, CR 230 3,230 3, SR 100 North 10,416 10, SR 26 9,264 9, SR 20 15,684 15, CR CR 315 2,020 2, SR 19 3,368 3, US 17 5,894 5, SR 100 4,632 4, US 1 South 2577 I-95 South 9,896 9, ,022 42, SR A1A 5,416 5, , , February 2010 Page 3-14

68 4. Trip Generation Trip generation determines the number of person or vehicle trips produced by or attracted to each zone. The standard FSUTMS GEN model was slightly updated in 2000 NERPM model to separately account for trip generation by 3+ auto households, whereas FSUTMS GEN does not differentiate between 2 and 3+ auto households. In the interest of simulating truck-exclusive and high-occupancy vehicle (HOV) lanes, the NERPM trip generation program was updated to simulate multiple truck trip purposes and to allocate internal-external trips by auto occupancy and truck categories. Trip generation process is the first module in CV application (see Figure 1-1). The process of trip generation was completely scripted in the model chain. Users should consult model application guidelines for details of all user input and parameter files of this module. This chapter summarizes production and attraction data used in 2005 and 2035 model runs and lists few elements of the trip generation model. It then summarizes the overall model process and the validated rates and results. 4.1 Trip Generation Process As described in section 3.3 of Technical Report 3 (Model Application Guidelines), socioeconomic formats for the trip generation model were modified from standard FSUTMS practices. Both production (ZDATA1) and attraction (ZDATA2) zonal data were converted into a single zonal data file (ZDATA) in DBF format. Separate trip rates were developed from the North Florida Household Travel Survey (NFHTS) for 3+ auto households as part of NERPM 2000, whereas earlier version of NERPM did not differentiate between households with 2 or 3+ autos. Production data were modified to identify the percent of 3+ auto households for single-and multifamily dwelling unit types. In order to maintain consistency, new trip production rates were calculated for all cells within the trip generation rate matrix as part of the NERPM 2000 trip generation enhancements. Cross-classification and regression-type models are used in the NERPM4 trip generation model. Crossclassification analysis is used to group households with common socioeconomic characteristics (dwelling unit types single, multi- family or hotel/motel, household size and number of vehicles) together to create relatively homogenous groups for four home based trip purposes. Dwelling unit weights were maintained from earlier versions of NERPM and are depicted in Table 4-1. Regression-type models were used for all other trip purposes. NERPM 2000 also modified the calculation of commercial vehicle trips through introduction of trip rates from the Quick Response Freight Manual (QRFM). Earlier versions of NERPM, like most FSUTMS models, included a single truck-taxi purpose (trips not generally recorded in household surveys). The QRFM includes three commercial vehicle categories: Four-tire vehicles; Single-unit trucks (six+ tires); and Combinations. The implementation of the QRFM trip rates requires four employment types whereas most FSUTMS models only include three employment types. The primary difference between FSUTMS GEN and QRFM is the separation of industrial employment into two categories. Therefore, the attraction data February 2010 Page 4-1

69 starting with NERPM 2000 includes two categories of industrial employment (manufacturing NAICS and other industrial NAICS 01-19). The final modification to trip generation as part of NERPM 2000 was the allocation of internal-external trips into four trip purposes based on auto occupancy and truck categories. This classification is consistent with that used in the external model, as described in chapter 3 of this report. This required a modification to the external-internal (EITRIPS_YYA.DBF) format to include the percent of trips by each auto occupancy and truck classification. Formats used for all use input and parameter files are provided in Technical Report 3 (Model Application Guidelines). A summary of external station trips factors is shown in Table 4-2. February 2010 Page 4-2

70 Table 4-1: Dwelling Unit Weights Percent of Households Per PPDU* Categorg Bin No Average Persons Per Dwelling Unit One-Person Households Two-Person Households Three-Person Households Four-Person Households Five-Person Households * Persons Per Dwelling Unit The result is an expansion from the seven standard FSUTMS trip purposes used in NERPM 1998 to the following 12 trip purposes in NERPM 2000, JTA/RTS 2005 and NERPM4: 1. Home-based work; 2. Home-based shop; 3. Home-based social/recreation; 4. Home-based other; 5. Nonhome-based; 6. Light-duty truck (four-tire commercial); 7. Medium-duty truck (single-unit six-tire trucks); 8. Heavy-duty truck (combinations); 9. Low-occupancy vehicle (LOV) internal-external trips; 10. High-occupancy vehicle (HOV) internal-external trips; 11. Light-duty truck (LDTK) internal-external trips; and 12. Heavy-duty truck (HDTK) internal-external trips. The first five purposes listed above are the same as NERPM Purposes 6 through 8 are based on QRFM trip purposes and reflect what used to constitute the truck-taxi purpose in NERPM Purposes 9 through 12 are consistent with the external model described earlier in chapter 3 of this report. LOVs are single-occupant personal vehicle trips (autos) and HOVs are auto trips with more than one February 2010 Page 4-3

71 passenger. For external trips, trucks are only divided into two categories. As described in chapter 3, heavy-duty internal-external trucks comprise both heavy-duty and medium-duty trucks. Table 4-2: Summary of External Station Trip Factors by Mode of Travel February 2010 Page 4-4

72 The special generator file (named as SPGEN, see Table 3-4 of TR3, Model Application Guidelines) has a format that is slightly different than the standard of ZDATA3 file. Differing from the standard FSUTMS GEN model special generator process, the NERPM trip generation process includes a process in which special generator trip attractions are not adjusted to match trip production totals by purpose. The advantage of this approach, based on the Southeast Florida Regional Planning Model (SERPM), is that when users input a specified number of trips for a special generator, major activity center, or development of regional impact, the number of trips entered in special generator file will not require multiple rounds of adjustment to account for trip attraction balancing. The changes that were made to the model s special generator procedure do not require changes to any input variable. However, the trip attraction balancing procedure has been modified. This adjustment methodology should give logical results unless the special generator trips are a very large fraction of the total number of trips or the sum of productions and attractions are grossly out of balance. 4.2 Zonal Socioeconomic Data Summary Highway traffic volumes and transit ridership are functions of the population and the employment within an area, so it is critical to have these socio-economic data correctly represented in the model. The household data file (ZDATA1) is used to estimate the number of trips produced by each TAZ. Attraction data (ZDATA2) is used by the trip generation model to calculate the trips attracted to TAZs. Both production (ZDATA1) and attraction (ZDATA2) zonal data were converted into a single zonal data file (ZDATA) in DBF format (see Table 3-3 of TR3- Model Application Guidelines). The parking cost data in ZDATA is used by the mode choice model. The short-term parking cost, which is used in home-based non-work trip and non-home based mode choice calculations, and long-term parking cost, which is used in home-based work mode choice calculations, represent three and nine hour average parking costs respectively. The zonal data files used in 2005 and 2030 (E-Plus-C trend) validation runs were primarily developed by PBS&J in association with the North Florida TPO and other county and FDOT planning staffs. Corradino reviewed the data files for errors as part of model validation, and provided suggestions to the PBS&J staff when revisions were needed. During model validation, trip generation errors noted in the error files (LUERRORS.PRN & AO_ERRORS.PRN), such as household size and auto-occupancy related inconsistencies, were resolved. A summary of all production and attraction data files that were used in the 2005 model validation as well as the 2035 models are shown in Tables B-1 and B-2 of Appendix B. Summaries of population and dwelling units are presented for single, multi-family and hotel-and-motel categories. Employment by category and school enrollment also was summarized. Data were summarized by regional districts (1-35) as well as by the six counties. Figure 4-1 depicts the district boundaries. Changes and percent changes in 2005 (base) and 2035 (trend) socioeconomic data were computed and are summarized in Tables B-3 and B-4 by the regional districts and counties. The growth rates show a reasonable pattern in all six counties. In general, population and household growth is highest in the outer counties. Overall, the regional population grows by 58% whereas Duval County grows by just 42%. Employment growth rates are very similar among the counties except for St. Johns County. Socioeconomic data summaries were presented to the technical advisory committee at early stages of regional model validation. It was decided that all data are suitable for use in the 2005 and 2035 NERPM4 models. February 2010 Page 4-5

73 To show the reasonableness of the production and attraction data, socioeconomic data indices reported by the trip generation process were summarized for the region for both 2005 and 2035 NERPM4 models runs and compared to those from NERPM 2000 and JTA/RTS Those indices are displayed in part B of Table 4-5. February 2010 Page 4-6

74 Figure 4-1: District Boundaries Note: See Table B-1 for description of Districts February 2010 Page 4-7

75 Household (HH) Size for Permanent Population Household (HH) Size for All Population Employment/Population Ratio Service/Total Employment Ratio Commercial/Total Employment Ratio Manufacturing/Total Employment Ratio Other Industrial/Total Employment Ratio 4.3 Trip Generation Validation Adjustments In the validation of JTA/RTS 2005 it was noticed that the HBW trip attraction equation included dwelling units. This was different from other Florida models and the JUATS model, both of which use employment as the only independent variable. The regional HBW attraction equation was modified to: HBW Trip Attractions = 1.80 * Employment By using this new attraction equation, the model s CBD work trip attractions were comparable to the Census data. In addition county-wide adjustments were made to the HBW attraction rates to better replicate 2000 Census journey to work flows. Several adjustments were made to the NERPM trip generation model during validation. Due to time constraints, model validation was initiated using NERPM 2000 and JTA/RTS 2005 trip production and attraction rates. It should be mentioned that in the validation process that the QRFM commercial vehicle trip rates had to be reduced by 50 percent in NERPM 2000 model since for every trip attraction estimated, a trip production also was estimated consistent with standard FSUTMS non-home-based trip calculation processes. Both production and attraction rates were further modified so that model produces reasonable results both in the trip generation module and in the context of overall NERPM4model chain validation. The rates were modified so that model generated volumes reasonably replicate the observed counts both at the regional and county levels. Comments to document the changes are in the model script. A summary of the validated NERPM4 production and attraction rates are shown in Tables 4-3 and 4-4, respectively. Final trip rates, along with county specific adjustment factors, are summarized in these tables. Some district based adjustments were made to the non-work purposes to distinguish the CBD, rural and other areas. Several rounds of ZDATA adjustments also were incorporated into the validation. PBS&J was primarily responsible for the preparation of base and future year ZDATA files. It also was discovered that auto availability percentages did not always add up to 100 percent and rounding corrections were made to ZDATA. Several special generator adjustments were made during validation. The process of external-internal trips validation through the special generator adjustments was removed in NERPM4. This was used in both of the recent model validations (NERPM2000 and JTA/RTS 2005). The rationale for these adjustments in the NERPM2000 and JTA/RTS 2005 was to account for routine home-based and non-home-based trips generated within the Northeast Florida region that have a trip attraction outside the region of four counties (Nassau, Duval, Clay and St. Johns). A good example of this situation would be residents of southern Clay and St. Johns County who work in Palatka, located in Putnam County outside the model study February 2010 Page 4-8

76 boundary. In the absence of special generator file trip attractors at these external zones, all home-based work trips produced in these areas would be attracted to zones within the NERPM study area, resulting in trip distribution and assignment errors on screenlines and corridors entering the City of Jacksonville. Since the NERPM4 study region was expanded to cover Baker and Putnam counties, it was decided through an early discussion with the TAC to remove this procedure from the NERPM4 model. Moreover, TAC viewed the earlier process to be unnecessarily complicated. February 2010 Page 4-9

77 Table 4-3: Validated Trip Production Rates February 2010 Page 4-10

78 Table 4-4: Validated Trip Attraction Rates February 2010 Page 4-11

79 No special generators were used during the initial validation runs of the model. Later, the set of special generators, without external stations, used in NERPM2000 and JTA/RTS 2005 model validation was used. The values were then slightly adjusted based on model performance (volume/count ratios) near the special generators. Other adjustments included the revision of special generator trips after visually checking the volume/count ratios of links near each of the special generators. A complete listing of special generators used in the NERPM4 model validation is provided in Table A-2. It is similar to those used in NERPM 2000 and JTA/RTS 2005 model, but without special generators for external zones. A notable difference in the NERPM4 and two other recently validated regional models (NERPM 2000 and JTA/RTS 2005) is that the final NERPM4 model did not use sub-area attraction balancing. As part of NERPM2000 and 2005 based JTA/RTS model validation efforts, subarea balancing was introduced to adjust trip distribution and improve overall validation. The purpose of subarea balancing is to adjust trip attractions to match trip productions locally within defined districts as opposed to adjusting attractions universally throughout the entire model area. It was assumed that that the process allowed for a more accurate distribution of trips. Districts are composed of TAZs that are assumed to have higher levels of interaction with each other than with other TAZs. A key component in this assumption is geographic proximity and similar land uses. NERPM2000 and JTA/RTS have five districts that are used for subarea balancing. These districts were developed during model validation by trial and error. By default, the model is not set up to balance attractions during trip generation. The file defining the subarea balancing districts is ATTRDIST.SYN. In NERPM4, two new districts of Baker and Putnam Counties were added. This file identifies districts and the zones that comprise each district. Figure 4-2 shows the boundaries of these subarea districts. Table B-15 presents the list of zones for these subarea districts. In the earlier part of NERPM4 validation efforts, the sub-area attraction balancing process was tested and later dropped because it was viewed as an application of K-factors by the TAC. In the later and final stage of NERPM4 validation, all TAZs were grouped into one district, effectively eliminating the sub-area attraction balancing process. In addition no K-factors were used in trip distribution. 4.4 Trip Generation Validation Results Throughout the validation process, summaries of trip generation statistics were used to assess model validity. Comparisons between NERPM4, NERPM 2000 and JTA/RTS 2005 were made. Statistical comparisons also were made to other regional models in Florida and other areas in the United States using statistics available in the Model Validation and Reasonableness Checking Manual. Statistics from other models were sometimes aggregated to account for different trip purpose schemes. Part A of Table 4-5 provides a summary of trips by purpose. When compared to NERPM 2000 and JTA/RTS 2005, NERPM4 shows slightly higher percentages of HBW and internal-external trips. This trend is logical for NERPM4 s expanded study area, with the addition of Baker and Putnam Counties. HBW trips were compared to the CTPP targets for the 2005 validation. As shown in part B of Table 4-5 that the overall HBW trip rates are and per households for 2005 and 2035 NERPM4 model runs. The rates for NERPM2000 and JTA/RTS 2005 are and 1.391, respectively. NERPM4 overall attraction balance factors by purpose indicate that its productions and attractions are more aligned with the unbalanced attractions compared to those in the NERPM 2000 and JTA/RTS 2005 models. Part C of Table 4-5 shows comparisons of aggregate trips per household, person, and employee. Overall trip rates from NERPM4 are slightly lower than those in NERPM2000 and JTA/RTS 2005 models. The overall NERPM4 rates are very similar to SERPM65 [9.09 for the 2005 SERPM6.5 validated model, February 2010 Page 4-12

80 Reference 19]. For urbanized areas of more than 1 million persons, the overall weighted daily person rate is 8.5 per NCHRP 365 [Reference 27] and these rates are in the middle of the ranges shown in Table 3-4 of Model Validation and Reasonableness Checking Manual [Reference 24]. These comparisons show that NERPM4 is consistent with other models in terms of aggregate trip rates. In all cases, NERPM4 statistics fall somewhere in the middle of those for other models that were used for comparisons. NHB trips are 22.3% for 2005 NERPM4 and 22.6% for 2035 NERPM4, which is very similar to the percentages shown in other models and reports [23% per Reference 24 and 27, 25 percent per Reference 19]. NERPM4 results generally compare favorably with results from other models in Florida and the US. February 2010 Page 4-13

81 Figure 4-2: Trip Attraction Districts Note: District-based sub-area balancing was not used in NERPM4 validation February 2010 Page 4-14

82 Table 4-5: Summary and Comparison of Trip Generation Outputs February 2010 Page 4-15

83 Table 4-5 (contd.): Summary and Comparison of Trip Generation Outputs February 2010 Page 4-16

84 5. Highway Paths and Skims Minimum travel time paths are calculated using time over the highway and HOV system. In building paths, a turning penalty file is used. Paths are not built through prohibited movements. Initial paths are built using free-flow speeds. This chapter describes the process of building NERPM 4paths and skims and then presents the key modeling data that were used in validation. 5.1 Model Process The NERPM highway path module (see Figure 1-2) uses standard Cube Voyager procedures to build time and distance skim matrices for Single Occupancy Vehicle (SOV) and High Occupancy Vehicle (HOV) paths. The SOV paths are defined as the shortest time path through the portion of the highway network available to single occupant vehicles. SOV paths do not include HOV facilities. HOV paths are defined as the shortest time path through the portion of the network available to passenger cars with two or more persons in the vehicle. Such paths consider both HOV and SOV facilities. Truck traffic was assigned to the SOV network as a class of trip in the multi-class equilibrium assignment. Terminal times and intrazonal times are also added to the interzonal skims. Intrazonal times represent the travel times assumed for trips that begin and end in the same TAZ. These times are calculated as one-half the travel time from one zone to the nearest adjacent zone. Terminal times represent the time involved at either end of a trip to travel from an origin to the network or from the network to a final destination. More specifically, this accounts for the time necessary to walk to or from the vehicle used for any given trip, and to park and un-park. Table 4.1 lists the terminal times by area type used in NERPM4. The model also uses congested speeds to develop congested paths and skims. Skims are updated with terminal times, which are a function of area type, and with intrazonal times, which are the average of half the time to the two nearest TAZs. Turn penalties and prohibitors are also added at this stage. Free-flow travel time skims between zone pairs are developed during the HNET-HPATH module (see Figure 1-2). Highway network characteristics are input to this process. Travel time skims are then updated later in the DISTRIB module. The SOV and HOV paths and skims are needed later for mode choice analysis. To permit analysis of HOV lane impacts, the mode choice model reads two sets of highway impedances. One set represents the highway travel times available to travelers in mixed-flow traffic, while the other represents the reduced travel times available to travelers with occupancies that qualify for the HOV lanes. The model assigns the appropriate travel time to each occupancy alternative and then computes mode share that recognizes the impact of HOV time saving. In NERPM, three variables are considered as significant in determining the minimum paths between any given pair of zones. These variables are as follows: 1. In-Vehicle Travel (IVT) time: IVT time is the primary variable, which is a function of distance and input speed. February 2010 Page 5-1

85 2. Prohibited and penalized movements: In addition to the highway network characteristics, two other files are generally used as input to the HNET-HPATH module. The first of these is the TCARDS_YYA.PEN file. The TCARDS_YYA.PEN file contains a listing of all link penalties and prohibitors in the highway network. It also annotates the types of prohibitors and penalties. The TCARDS file allows for the adjustment of travel times on specific links by either inducing a time penalty to pass from one link to another or prohibiting the movement all together. Prohibitors are generally coded to identify turning movements in the highway network that are not permitted. Another use of prohibitors is in the double-line coding of freeway facilities, toll plazas, and interchanges where they are used to route vehicles to the proper entrance and exit ramps, and to prevent U-turn or illogical movements. NERPM includes all such prohibitors used in the earlier models. They are included, for the most part, on freeways to prohibit illegal U-turns, left turns and illogical movements. Time penalties are added to a highway network for several reasons. They can represent movements that are difficult, such as left turns where no signal protection exists. Some localized penalties that were initially adopted in earlier versions of models were adjusted to improve the performances of the model locally. In general, time penalties are minimized during model validation, as they are difficult to assign when developing future year highway network scenarios. Furthermore, movements influenced by psychological factors such as the crossing of rivers and other large bodies of water, are assigned travel time penalties. 3. Toll Impedance: The second file generally used as input to the HNET-HPATH module is the TOLLLINK file. This file describes toll plaza characteristics that convert toll costs into impedance estimates. NERPM4 adheres to the convention established in earlier model versions that assigns the link representing the ferry a unique area type/facility type classification and emulates toll impedances through speed and capacity adjustments. A dummy TOLLLINK file is used only to maintain standard FSUTMS conventions and allow for toll facility testing in the future. There is only one toll link in NERPM, the Mayport Ferry. Following the convention established by earlier versions of NERPM and JTA/RTS models, the TOLLLINK file included in the model is a dummy file. Instead of utilizing the TOLLLINK file for this facility, the Mayport Ferry was given a unique facility type that appears in the SPDCAP with a speed of 1 mph and a capacity of 100. In NERPM, this facility is facility type 29. In addition, a penalty of one minute was added to the link representing the ferry crossing. In the standard FSUTMS toll facility model, toll related data are specified in the TOLLLINK file. Toll data are required in areas where toll facilities exist or are planned. The purpose of the toll data is to account for the costs and delays (i.e., stopping at a toll plaza to pay the toll) associated with using toll facilities in the computation of travel impedance. These costs and delays impact a potential user s decision of whether or not to travel on the toll facility. The TOLLLINK file contains the following data for each toll plaza link in the highway network: Toll class Toll Type Node numbers at both ends Identification of plaza location Number of lanes Number of plaza lanes February 2010 Page 5-2

86 Toll amount Average service time Deceleration and acceleration code Number of exact change and Automated Vehicle Identification (AVI) lanes Percent of heavy trucks (for reporting only) All nodes referenced in the TOLLLINK file must have a corresponding set of nodes listed in the highway network file. Using these variables, a single composite measure of impedance is used to determine the minimum path between all pairs of zones. The calculations of impedance are based upon the combination of time and distance (on non-toll links) or time and toll (on toll links) are as follows: For non-toll links, IMPED = CTIME * TIME For toll links, IMPED = {CTIME * (SERVT + TIME)} + CTOLL * TOLL Where, CTIME = time coefficient, TIME = travel time on the link, and SERVT = service time on the toll booth Toll costs are converted to travel time and factored by a parameter called a CTOLL. In NERPM4, the value of CTOLL is 0.10, which is a representative regional value. It is entered as a Cube key. Highway path development is a critical component of the model stream. For all pairs of zones, minimum paths are based upon the least impedance. Impedance includes in-vehicle time, prohibited and penalized movements, toll cost and service time. 5.2 Model Validation To check the network for coding errors and to ensure reasonable paths are built through the network, the Cube-Base program was used to check the path building. This program was used to display the path between several selected pairs of centroid in various locations in the network. The routines trace the shortest path using the network impedance of time or distance with the summation of link impedances computed. Numerous paths were drawn on the computer screen to make sure that paths drawn were reasonable. Figure 5-1 presents several representative paths between I-95 north external station to four other selected external stations using both free-flow and congested skims. It should be noted that the model also use external prohibitor codes (see Figure 3-3) for the external station flows used in highway assignment steps. Those prohibitors are not reflected in the paths shown in Figure 5-1. Congestion changes the minimum paths in some cases. Another check on the network was made by computing the travel time from the Jacksonville CBD zones to all other zones in the region. The isochromes of the both free-flow and congested travel time are then displayed graphically as shown in Figure 5-2 to check for errors in path building. Prohibitors include all prohibited movements found in NERPM 2000 and JTA/RTS 2005 models plus a few others that were felt to be needed as a result of network changes or to ensure accurate turning February 2010 Page 5-3

87 movements in questionable areas. However, all unnecessary records were dropped from the list and comments were added for each of the turning penalty records. In the JTA/RTS model, two separate penalty record files are used one during the pre-assignment step and the other during final assignment step. In NERPM4, only one turn penalty record file is used throughout the model. In general, the level of detail involved in the NERPM highway network allows for far fewer prohibitors than would otherwise be necessary. Double line coding allows trips to navigate freeways and interchanges with greater accuracy, precluding the need for many turn prohibitors. Initial testing during the NERPM validation effort began without travel time penalties. Such penalties were applied only after a careful consideration of their necessity. Time penalties were included in the model to better validate travel across the various bridges in the NERPM study area. These penalties were applied in conjunction with other network modifications (including highway network edits) so as to minimize their necessity. It became apparent early on that the use of some penalties would be unavoidable. The consultant took great care to ensure that any given penalty did not surpass earlier validated values in NERPM2000 and JTA/RTS 2005 models and was able to reduce the penalties quite significantly in some cases. New penalties were included in NERPM4 for the expansion of the study areas. Table A-4 of Appendix A lists validated turning penalties used in 2005 NERPM4 model. February 2010 Page 5-4

88 Figure 5-1: Comparison of Highway Paths from I95 North External Station to Four Other Major External Stations Travel time From Station 2550 (I95 North) To Station FF Time (min) Congested Time (min) 2559 (I-10 West) (US 301) (SR 20) (I-95 South) Using Free-Flow Skims Using Congested-Flow Skims Note: Above paths do not reflect external prohibition through EECODE used in assignment Stations: 2550-I95North (Nassau/Georgia State), 2559-I10 West (Baker/Columbia), 2564-US301 (Clay/Bradford) & 2570-SR20 (Putnam/Alachua). February 2010 Page 5-5

89 Figure 5-2: Comparison of Isochromes (10-minute increment) from a Downtown Jacksonville TAZ Using Free-Flow Time Using Congested-Flow Time February 2010 Page 5-6

90 6. Trip Distribution The NERPM4 distribution module distributes trips between zones using a gravity model, and produces a set of congested highway skims that are used for the transit model. Skims are used by the gravity model to link the trip productions and attractions generated by trip generation. These trip interchanges denote person trips traveling specifically from one zone in the model to another. Also, as part of trip distribution, these trips are distributed for each of the 12 internal trip purposes. These person trips are converted to vehicle trips during mode choice and then loaded onto the network during assignment. The results of these functions, in turn, become inputs for transit network development and mode choice estimation. This chapter describes the key model process of trip distribution and then presents the key model data and summaries of model outputs of 2005 and 2035 model runs. 6.1 Trip Distribution Model Process Trip distribution models connect trip productions and attractions between pairs of TAZs. These connections typically are calculated by a Gravity Model. A Gravity Model distributes trips between zones directly proportional to the relative attractiveness of each individual zone and inversely proportional to the friction between them (i.e., time). The result is a matrix of person trips defined in terms of productions and attractions as opposed to origins and destinations. Resulting trip matrixes are processed later in the model chain during mode choice to allocate trips by auto occupancy and transit categories to create the basis for vehicle trips. The general distribution process includes building travel time skims as well as application of the gravity model. Except for through vehicles, NERPM4 uses the Cube Voyager gravity model to distribute trips between production and attraction zones for all trips and purposes. The NERPM4 trip distribution module (see Figure 1-2) performs the following functions: Distribute all non-work trips with free flow skims; Perform a preliminary distribution of work trips using free flow skims; Perform a 2-hour peak period pre-assignment; Develop congested skims (HOV and LOV) and redistribution of work trips with congested skims; and Summarize model distributed flows to compare with CTPP flows. The HNET-HPATH and DISTRIB modules (see Figure 1-2) include updating the travel time skims with intrazonal and terminal times, distributing trips between zones using a Gravity Model, and producing a set of congested highway skims. The primary input data used for DISTRIB is the friction factor (FF) file. This file is used by the Gravity Model to measure the effects of spatial separation between zones for trip distribution. It is generally assumed that trips are less likely to be allocated to destinations with greater travel times if alternative destinations with smaller travel times and similar attractiveness are available. These various inputs are used by the Gravity Model to link the trip productions and attractions generated by GEN. These trip interchanges denote person trips traveling from one zone in the model to another. Also, as part of DISTRIB, these trips are distributed for each of the 12 trip purposes. These person trips are converted into vehicle trips during mode choice and then loaded onto the network during highway assignment. February 2010 Page 6-1

91 NERPM4 also derives congested highway skims for the home-based work (HBW) purpose by conducting a preliminary mode choice and highway assignment step during DISTRIB. These skims are then used to distribute the HBW trips by the Gravity Model and for building the transit network. Congested skims are used for HBW trip distribution as it is assumed that the majority of these trips occur during the peak period. Mode choice models can range from simple person-to-auto trip conversion models to more complicated multi-path/multi-period models that estimate modal shares between various categories of auto and transit ridership. NERPM4 utilizes a first pass simplified mode choice process during DISTRIB that uses a seed trip table that estimates mode share on a zone-by-zone basis. NERPM4 utilizes the scripted version of MCSEED program logic that estimates mode share by zone based on the mode shares of a prior person trip matrix. Section provides more on this MCSEED program logic. Beside the interzonal (zone to zone) travel time, the gravity model requires two additional measures of time intrazonal travel time and out-of-vehicle travel (terminal time). Intrazonal travel time is the time needed for a trip between two sites within the same zone. This time is usually smaller than the interzonal time. In CV scripts, intrazonal times are based on the Nearest Neighbor Theory. The theory states that intrazonal travel time is proportional to the amount of time it takes to get to the nearest adjacent zone or zones. The half of the nearest zone in-vehicle time is used as the measure of intrazonal time. In NERPM4, two adjacent zones are used to compute the intrazonal travel time during the trip distribution step. Intrazonal trips are trips that begin and end in the same zone. They are never loaded onto the network and are effectively omitted from total trips during assignment. They play a significant role in estimating the local VMT for air pollution analysis. Calibration of intrazonal trips is not easy unless a large sample of shorter trips exist in the observed database. These trips, in general, are under reported in most household surveys. The percentage of intrazonal trips estimated by the NERPM4 gravity models is in line with other models. Terminal times are the average times required to get into a vehicle and go from the driveway to the street at the origin (production) end of the trip, and the average time required to park the vehicle and reach the final destination point at the destination (attraction) end of the trips. Like many other Florida models, terminal times vary according to the area type of a zone and are input through lookup data file. The values applied for terminal times in the NERPM4 are summarized in Table 2-1. Terminal times are added to the in-vehicle travel time for both ends of a trip, resulting in total travel time between pair of zones. The resulting travel times are ready for input into the gravity model Peak-Period Highway Assignment Congestion on the roadway network has an impact on trip distribution and should be accounted for in the model. This is particularly true if future congestion levels are significantly different than those in 2005 (the model base year). Using the standard approach of distributing trips strictly on free-flow highway travel time, there would be minimal impact on the overall distribution by the addition of capacity to existing facilities. The NERPM4 trip distribution model differs from the conventional FSUTMS distribution models in that it considers the both free-flow and congested time rather than simply the freeflow highway travel time between origin and destination zones to distribute the work trips which usually made during peak hours. The reason for this approach is to properly account for influence of congestion in work trip distribution. NERPM derives congested highway skims for the home-based work (HBW) purpose by conducting a preliminary mode choice and highway assignment step during the distribution application. These skims February 2010 Page 6-2

92 are then used to distribute the HBW trips by the gravity model and for building the transit network. Congested skims are used for HBW trip distribution as it is assumed that the majority of these trips occur during the peak period. The pre-assignment step is a complete highway assignment based on the initial trip distribution using free-flow highway skims for all purposes and a simplified highway only mode choice process. Therefore, it is import to remember that any changes that the modeler may make to the highway assignment model after mode choice may need to be replicated in trip distribution so as to maintain consistency between the pre-assignment and the final assignment. In the NERPM 2000 model, the standard 24-hour pre-mode choice assignment produced very low auto speeds. Consequently, it was decided in 2005 JTA/RTS (and continued in the NERPM4) model to develop a peak period assignment to better match observed speeds. This assignment reflects trips occurring in the AM peak period (i.e., 7:00 am to 9:00 am). The 2000 Household survey was used to compute the time of day factors that convert the daily trip table to a 2-hour AM peak period trip table (see Table 6-1). A CONFAC value of 0.582, also from the household survey, is used to reflect the peak period capacity. Table 6-1: Pre-assignment Peak Period Factors Purpose/direction Value Source HBW P A HBNW P A NHB O D (symmetrical) HBW A P Household Survey HBNW A P NHB D O (symmetrical) Internal/External Time of Day Modeling Procedures for Implementation External/External in FSUTMS Truck Source: Table 12, JTA/RTS Validation Report In the JTA/RTS model, there were inconsistencies in speeds between pre-assignment and final highway assignment steps. In the NERPM4 model validation, these two speed sets were more consistent. In the pre-assignment step, there is no restriction on the external trips to use only the I-95/I-295 bridges to cross St. Johns River. Like JTA/RTS model, NERPM4 uses very low criterion set for gap, RAAD, AAD and RMSE so that the assignment runs for at least 30 iterations to improve closure Trip Table for Pre-Mode Choice Assignment In the NERPM2000 and JTA/RTS 2005 models, the impedances for HBW trip distribution were derived from a 24-hour peak period highway assignment. In the NERPM2000 model, the trip table used for this assignment was created by the MCSEED program, written in FORTRAN. It subdivided the person trip table from the initial trip distribution into drive alone, shared ride-2, shared ride-3+ and transit trips for HBW, HBNW and NHB purposes for pre-mode choice assignment. The program uses a seed matrix (see CALIB98A.MC3 matrix in parameters folder) from old JUATS model for splitting into the four sub modes. However, the zone structure of the JTA/RTS 2005 (see Table B-11) and NERPM4 (see Table 1-1) is different from the NERPM2000 model (see Table B-10). Over 300 (JTA/RTS model) or 500 (NERPM4 model) zones were split compare to that in NERPM2000. However, the input seed matrix was not updated for the zone splits. February 2010 Page 6-3

93 For the JTA/RTS 2005 model, it was decided to script the MCSEED program logic in Cube-Voyager to correct the problem. Since the zones were split mainly in areas where transit service exists, the trips in these zones were subdivided based on the split of trips attraction to the CBD district. The revised logic splits are closer to the split percentages of the sub modes of the initial seed table than compared to the MCSEED program using JUATS seed table. In NERPM4, the impedances for trip distribution of HBW purpose are derived from a 2-hour peak period highway assignment. It uses the scripted MCSEED program logic originally implemented in JTA/RTS model. However, the logic is applied only once instead of two times as in the JTA/RTS 2005 model. The JTA/RTS model uses two pre-assignment steps (24-hour and 2-hour). The 24-hour assignment is used for development of skims for HBW distribution and that followed by another 2-hour assignment for the final congested skims for use in mode choice and transit model. On the other hand, NERPM4 uses only 2-hour AM peak assignment for both HBW distribution and congested skims for mode choice and the transit model Subarea Balancing As part of the NERPM2000 and 2005 based JTA/RTS model validation efforts, subarea balancing was applied as a means of adjusting trip distribution and improving overall validation. The purpose of subarea balancing is to adjust trip attractions to match trip productions locally within defined districts as opposed to adjusting attractions universally throughout the entire model area. It was assumed that that the process allowed for a more accurate distribution of trips. Districts are comprised of TAZs that are assumed to have higher levels of interaction with each other than with other TAZs. A key component in this assumption is geographic proximity and similar land use. NERPM2000 and JTA/RTS models used five districts for subarea balancing. These districts were developed during model validation using a trial and error process. By default, the model is not set up to balance attractions during trip generation. NERPM 2000 includes five districts that are used for subarea balancing. These districts are: 1. Year 2000 First Coast MPO Study Area; 2. Rural SW Clay County; 3. Southern St. Johns County; 4. Eastern Nassau County; and 5. Western Nassau County. The file defining the subarea balancing districts is ATTRDIST.SYN. In the NERPM4 model, two new districts of Baker and Putnam counties were added. This file identifies districts and the zones that comprise each district. Section 4.3 along with Figure 4-2 provides more description of this subarea attraction balancing process. The file is used in trip generation to create a lookup table, which is then used in trip distribution to balance the attractions according to district. In the earlier part of NERPM4 validation efforts, the sub-area attraction balancing process were tested and was later dropped since it was viewed as application of K-factors. In the later and final stage of NERPM4 validation, all TAZs were grouped into one district, effectively eliminating the sub-area attraction balancing process. 6.2 Model Validation Validation of the NERPM4 trip distribution model primarily involved the modification of the highway and transit networks, omission of features (such as subarea balancing in NERPM2000 and JTA/RTS 2005 model) and adjusting friction factors. Evaluation of the trip distribution model was accomplished by comparing statistics for average trip length and percentages of intrazonal trips between NERPM4 and other comparable models, including NERPM 2000 and JTA/RTS 2005 and 2000 travel survey statistics. February 2010 Page 6-4

94 Errors in the trip distribution phase can lead to significant problems in the execution of subsequent steps in the model chain (i.e., mode choice and trip assignment). Consequently, great efforts were taken to maximize the accuracy of the NERPM4 trip distribution module. This effort also included modifications to network speeds and capacities and modifications of travel time penalties. NERPM4 uses the set of friction factors used in earlier models (NERPM 2000 & JTA/RTS 2005). The gravity model can include friction factors (representing travel impedance between zones) and K- factors (often referred as socioeconomic adjustment factors). For NERPM4, validation of the gravity model centered on the use and adjustment of friction factors used in earlier models. For NERPM4, K- factors were not considered because the gravity model with friction factors alone performed well. Using results of the 2000 Travel Survey for North Florida as targets for average travel times, the initial friction factors used for the NERPM 2000 model validation came from the NERPM 1998 model validation. Due to expansion of the model from seven (in NERPM 1998) to 12 purposes (in NERPM 2000), the friction factor file was modified to accommodate the change. The friction factors for Purpose 6 (Truck-Taxi) were replicated for all three commercial vehicle purposes and the friction factors for Purpose 7 (Internal-External) were replicated for all four internal-external purposes. Inconsistencies in results from the survey and from other comparable models in Florida to NERPM 2000 led to experimentation with adopting purpose specific friction factors from other models. Ultimately, the best results came from replacing the friction factors for home-based shop (HBSH) with those from the 1999Tampa Bay Regional Planning Model and from replacing the NERPM friction factors for commercial vehicles (Truck-Taxi) with those from 1998 JUATS. Both of these sets of friction factors were slightly modified to more accurately reflect the friction factor values used in NERPM Table A-3 of Appendix A lists the validated friction factors. The goal of the trip distribution validation was to make the average trip length reasonably match target values (see Section 6.4) and to match the work trip flow pattern to that of JTW (see Section 6.3). 6.3 Comparison of Journey-To-Work and Model HBW Trips Trip distribution is the process of estimating the trip flows from and to each zone. In recent years, modelers have identified trip distribution as one of the sources of unexpected and often incorrect model behavior. The gravity model is typically calibrated to the average trip length and not by travel market. More often than not, the resulting travel markets from the model are not reflective of actual travel patterns and may lead to major issues during post analysis. Work trips are responsible for the majority of user benefits because of their longer trip lengths and frequency. At a minimum, the work trip distribution patterns generated by the gravity model should be checked for reliability. It is necessary to validate the travel patterns produced by the model. The estimated travel patterns were compared to the flow patterns obtained from the Part III of the Census Transportation Planning Package (CTPP3). The Census data was collected in 2000 (although released in 2003). By making this comparison, it is assumed that the travel patterns have not changed drastically between 2000 and 2005 (model calibration year). CTPP3 data was available only for Duval County. Even though the TAZ-level geography is different between the Census and the model, care was taken to match the districts as closely as possible. Figures 6-1 and 6-2 show the extent of the districts used in this analysis. There are ten CTPP districts for Duval County. In addition, five peripheral counties of NERPM4 study area (Nassau, St. Johns, Clay, Baker and Putnam) were added to summarize the model trip table by county against the CTPP flows. Summaries of February 2010 Page 6-5

95 CTPP flows of Duval county CTTP districts and that of six counties are shown in Tables B-13 and B-14 of Appendix B, respectively. The Census long form asks each respondent to describe their daily work trip in terms of its location, travel time, and mode. The results are tabulated and released as part of the CTPP. It represents the largest data sample of travel patterns in the country. It should be noted that while the model estimates the typical home-to-work trip pattern, the CTPP data identifies journey-to-work (JTW) patterns. The home-to-work flow assumes no stops between the production and attraction end. The journey-to-work flow can have intermediate stops. The comparison is still regarded to be valid considering the considerable sample size of the Census long form. For NERPM4, 2005 home-based work (HBW) trip patterns were examined to identify potential problems. Other trip purposes are not reviewed because of the lack of adequate and reliable data on travel patterns. The HBW trip table obtained from the model is compressed to the same districts and the flow table is shown in Part A of Table 6-2. Part B of Table 6-2 shows the 2000 census flows scaled to the total trips from the model and Part C of Table 6-2 shows the ratio of the model trips over the census trips. The county-to-county HBW trip flows were also summarized and then compared (see Table 6-3). The analysis shows a fairly good representation of the HBW trip patterns. Notable results include: The Duval intra-county work trip flow ratio of model and JTW compares very well and are within a 6% tolerance. The Duval intra-county flows account for the majority of work trips (about 65%). Next to Duval, three other counties (St. Johns, Clay and Putnam) have major intra-county flows and accounts about 15% of overall work trip flows. The flows of these three counties inter-county flows are underestimated by about 27-29%. The inter-county flows are generally overestimated in the model. However they account only 17% of the overall work trip flows. As shown in Part C of Table 6-2, that overall flows among the Duval County CTPP districts match the CTPP flows. The production and attraction ends are also match well with significant number of trips. It should be noted that no K-factors were used in the NERPM4 model. Total trip ends (production or attraction) of the model HBW and JTW trips by districts and counties compared very well. It was concluded that model HBW trips reasonably compare the JTW flow. February 2010 Page 6-6

96 Figure 6-1: CTPP Districts Note: See Figure 6-2 for Duval County CTPP Districts February 2010 Page 6-7

97 Figure 6-2: Duval County CTPP Districts Outer North Northeast West North CBD Southeast CBD East Southwest South Far Southeast February 2010 Page 6-8

98 Table 6-2: Comparison of CTPP and Model Estimated HBW Trips by Duval County CTPP Districts February 2010 Page 6-9

99 Table 6-2 (contd.): Comparison of CTPP and Model Estimated HBW Trips by Duval County CTPP Districts February 2010 Page 6-10

100 Table 6-3: Comparison of CTPP and Model Estimated HBW Trips by County HBW Travel Patterns Part A of Table 6-4 shows the district-to-district flow obtained from the model for the year Part B of Table 6-4 shows the ratio of 2035 trips over 2005 trips, representing the growth in the number of trips in each cell. The travel patterns for 2035 seem reasonable. Except for few cells of the ten-by-ten matrix of districts flows, most of the cells show increase. The growth in trips to the CBD is also not limited. However, there is substantial increase in the trips to and from the few outer districts. February 2010 Page 6-11

101 The growths in 2030 county-to-county flows were summarized in Table 6-5. The intra-county HBW flow that represents majority of flows shows significant growth (31-215%). Duval county intra-county HBW trip has a growth of about 57%. Despite few drops in the inter-county flows, the county s overall HBW production ends of trips show about % increases and those of attraction ends show about 56-96% increase. February 2010 Page 6-12

102 Table 6-4: Comparison of 2035 and 2005 Model Estimated HBW Trips by Duval County CTPP Districts February 2010 Page 6-13

103 Table 6-5: Comparison of 2035 and 2005 Model Estimated HBW Trips by County February 2010 Page 6-14

104 6.4 Results and Comparisons The two fundamental Gravity Model checks discussed in this section are the average trip length by purpose and the percentage of intrazonal trips. An analysis of volume-to-count summaries along screenlines also can be helpful in establishing the accuracy of trip distribution. However, as screenline summaries apply more significantly to the analysis of traffic assignment, these will be discussed later in Chapter 10. One additional check on trip distribution model validity was production-attraction flows by district and/or county summary for the HBW trips between model estimates and CTPP (see Section 6.3). Trip length statistics as well as intrazonal trip percentages are summarized. Table 6-6 presents these summary statistics for the 2005 validation run. Trip length statistics are summarized both in travel time (minutes) and distance (miles). The model generated average trip lengths were compared to the trip lengths from the recent regional models (JTA/RTS 2005 and NERPM 2000) and 2000 Travel Survey statics as well as CTPP 2000 travel time. Notable findings include: The modeled trip length (Table 6-6) closely matches the trip lengths of other models, survey and CTPP. For the all twelve purposes together, the weighted modeled trip length is minutes and the un-weighted trip length is minutes. Un-weighted trip lengths of the JTA/RTS and NERPM 2000 are and minutes, respectively. The differences are mainly due to higher trip lengths in the internal-external trips by having larger study area. The model trip lengths are generally lower than the survey results. In terms of distance, the overall un-weighted and weighted trip lengths of NERPM validated model are and miles, respectively. The differences in average trip lengths of each trip purpose are nearly the same for all models except for HBW purpose. The HBW work trip length (30.38 minutes and miles using congested skims) of NERPM4 is much closer to survey trip length of minutes. Overall CTPP travel length (26.8 minutes) is lower than the model estimated trip length. However, that is much closer to model estimated HBW trip length using free-flow skims (25.67 minutes). Among the first five trip purposes, HBW trips are longer, with a model trip length of minutes and miles (using free flow skims), minutes and miles (using congested skims). Truck trips, in general, showed longer trip lengths of minutes and miles. The overall intrazonal trip percentage is 3.25 percent. By purpose, the intrazonal percentages vary from 0.89% (work trips) to 5.76% (truck-taxi trips). In addition to the sizes of TAZs, intrazonal percentages depend on other factors, including mixed/balanced land uses (homogeneous/heterogeneous nature of the TAZ with respect to dwelling units and employment), extent of local roads, and extent of non-motorized travel. The probability of the shorter trips becoming intrazonal goes up if there is a better balance of households (trip productions) and employment (attractions). Also, large percentages of non-motorized trips are intrazonal trips. No national target values for these percentages are available since urban development patterns and transportation infrastructure are unique to each urban area. However, the values shown in Table 6-6 are very reasonable. For example, the home-based work purpose has the lowest intrazonal percentage of trips, less than 1 percent. Trip length frequency distributions with respect to both time and distance were produced and are shown in Figures 6-3 and 6-4, respectively. It should be noted that reported distributions of these graphics February 2010 Page 6-15

105 include intrazonal and terminal times. The distribution of home-based work trips is most important to the transportation modeling process. It is the purpose that usually determines peak-hour usage, and is often the trip for which facilities are designed. For the work trips, peaking of congested trip times is comparatively flatter than the one using free-flow skims. The reverse is true for the distributions using distance in miles. The distributions of all non-work purposes also show reasonable patterns. The reason of coincidences of distributions of truck and internal-external trip purposes is that they use the same friction factors. February 2010 Page 6-16

106 Table 6-6: Summary and Comparison of 2005 Trip Length and Intrazonal Trips by Purpose and Vehicle Trips by Mode February 2010 Page 6-17

107 Trip Percentages Trip Percentages Trip Percentages Trip Percentages Technical Report # 3 Model Validation & Calibration Figure 6-3: Time Trip Length Frequency Distributions Time in minutes 1x. HBW - Free Flow 1. HBW - Congested Flow HB Shop 3. HB Social Recreation 4. HB Other 5. Non-HB Time in minutes Light-Truck 7. Medium Truck 8. Heavy-Truck Time in minutes Time in minutes 9. IE - SOV 10. IE - HOV 11. IE - Light Truck 12. IE - Heavy Truck 100 February 2010 Page 6-18

108 Trip Percentages Trip Percentages Trip Percentages Trip Percentages Technical Report # 3 Model Validation & Calibration Figure 6-4: Distance Trip Length Frequency Distributions x. HBW - Free Flow 1. HBW - Congested Flow HB Shop 3. HB Social Recreation 4. HB Other 5. Non-HB Distance in miles Distance in miles Light-Truck 7. Medium Truck 8. Heavy-Truck IE - SOV 10. IE - HOV 11. IE - Light Truck 12. IE - Heavy Truck Distance in miles Distance in miles February 2010 Page 6-19

109 Based on the close match between the model trip lengths of NERPM4 and other recent regional models, survey and CTPP as well as reasonable intrazonal trip percentages, friction factors were not adjusted further in the model validation phase. The trip length statistics from the 2035 (trend scenario) model run are summarized in Table trip lengths are similar to those of the 2005 model, but the congestion in future years caused somewhat longer trips. For the HBW purpose, the congested trip lengths are (a) minutes (13.49 miles) and minutes (15.21 miles) for the 2035 and 2005 models, respectively. All other purposes used free-flow skims and the trip length statistics of both 2005 and 2030 models runs are very similar. Nearly the same levels of intrazonal trips are found in both 2005 and 2030 models with slightly higher percentages in the 2030 model. In addition to person trips, the vehicle trip statistics from different period highway assignments are summarized for both 2005 and 2030 model runs (see Tables 6-6 and 6-7). Total trips, distribution by mode, total vehicles as well as intrazonal trips by mode of travel are summarized. Notable findings include: For the 2005 model, there are 62.64, and percent of trips for the drive-alone, shared ride and truck trips, respectively. The results of the 2030 model run are very similar. The overall percentages of intrazonal vehicle trip are 3.31% and 3.96% for the 2005 and 2030 models, respectively. Within the framework of the gravity model trip distribution, the trip length statistics are in close agreement with the recent models. The work trip flow shows the pattern exists in 2000 Census journey-towork flow. The distributions of the vehicular trips by periods and by modes as well as percent intrazonal trips are also very reasonable. February 2010 Page 6-20

110 Table 6-7: Summary of 2035 (Trend Scenario) Trip Length and Intrazonal Trips by Purpose and Vehicle Trips by Mode February 2010 Page 6-21

111 7. Transit Network, Path and Skim, and Fare The transit element uses the Voyager PUBLIC TRANSPORT (PT) module for modeling the transit service in the northeast Florida region. It is similar to the JTA/RTS model. The transit model process in NERPM4 was developed from the 2005 JTA/RTS model. Many of the transit related tables and text in this document are from the JTA/RTS model documentation. The transit network, path and skim and fare process is the fourth module in CV application (see Figure 1-1). Users should consult model application guidelines for details of all user input and parameter files of this module. Transit network modeling is an integral component of NERPM4. Transit networks represent the interconnectivity provided by transit systems. Transit path building involves the generation of zoneto-zone transit paths, transit skims, transit fares and station matrices. These files are built for each of three transit modes (local bus, express bus, and rail) during each of the two periods occurring in NERPM. Transit service in Duval County is provided by the Jacksonville Transportation Authority. Only JTA service, which is concentrated in Duval County with modest service into Clay County, is represented in this model. The average weekday ridership on all JTA buses in 2006 was approximately 36,000. Most of the JTA s service in 2005 operates to and from the downtown transit center (also referred to as the FCCJ - Florida Community College-Jacksonville - station) located in the northern edge of downtown. Figure 7-1 depicts the transit route coverage of 2005 NERPM4 model. This chapter discusses data and validation of the transit network, path and skim, and fares in NERPM Transit Network Transit networks are predominantly based on the unloaded highway network and supplemented with transit data such as route alignments, stop locations and headways. The transit model in NERPM4 has been established to accommodate walk and auto access modes for local bus, express bus, and rail transit service. Beside generation of auto connectors, all elements of transit modeling (network, mode choice and assignment) for NERPM are conducted using Voyager s PT program. Auto connectors are based on a user routine (AUTOCON), which again was developed for the new FSUTMS standard transit model application [References 14 & 15]. Like other FSUTMS models, NERPM4 uses the AM designation for peak transit trips and MD for off-peak transit trips. These are not true time of day descriptions. Rather, the peak period refers to HBW person trips while off-peak refers to all other internal person trip purposes from trip distribution. The JTA/RTS model is calibrated to the transit service that existed in November At that time, JTA offered service through its 39 local and interliner buses almost entirely within Duval County. The interlined buses connected the two opposite ends of Jacksonville through the downtown. Express services were offered to the Beaches and the Orange Park Mall area. The extent of the transit service and the bus routes is shown in Figure 3. The general boarding fare is $1.00, although the use of monthly and weekly passes is encouraged. JTA offers park-ride service at three lots, which have a combined 350 parking spaces. Downtown circulation is provided by the Skyway and the trolley service operating throughout the day. The Skyway runs every five minutes and offers connections from the FCCJ station to the South Bank and the convention center. The one-way fare in October 2007 was $0.50. Skyway offers fringe parking at some stations for a monthly fee; free Skyway passes are included. Three trolleys operate in the downtown area connecting various places of interest. Azalea and Magnolia both start at the FCCJ station and run at 20 minutes frequency during peak hours and 15 minutes frequency during the off-peak hours. Sunflower trolley operates at every 10 minutes. All trolley service is free. February 2010 Page 7-1

112 Figure 7-1: Year 2005 Transit Route Coverage February 2010 Page 7-2

113 7.1.1 Transit Network Elements Coded onto Highway Network The network (MicrocodedHnet4_YYA.NET) was developed from the JTA/RTS model s master network file. This network included micro coded transit station information as well as all fixed guideway facilities and optional transit links. Although a single master network was implemented initially, it was later desired by the long-range plan update consultants to make the network scenario specific. This section describes elements of the transit network that are coded onto highway network. Bus Only Links The transit-only links for the buses and the fixed-guideways are coded on the highway network. Links with facility types 49, 59 and 69 (see Table 4-2 for the list of FT codes) are excluded from highway skimming and assignment. The new fields shown in Table 7-1 are created and reviewed in the highway network to represent the distance and speed correctly on such links. Table 7-1: New Fields for Transit-only Links in the Highway Network Parameter Type Description TBSDIST Numeric Distance for bus transit Recommended Values (Unit) DISTANCE (miles) TBSTIME Numeric Travel time for bus transit (minutes) TFGDIST Numeric Distance for fixed-guideway transit DISTANCE (miles) TFGTIME Numeric Travel time for fixed-guideway transit (minutes) TFGMODE Numeric Applicable fixed-guideway transit mode None Source: Table 3, JTA/RTS Validation Report & JTA/RTS Application Guide Streets that are needed to the represent transit routes (especially in terms of transit speeds) but are not in the highway network are coded as facility type 49. If TBSTIME and TBSDIST fields are not provided, TIME (calculated using the speed obtained using the SPDCAP table) and DISTANCE fields on these links are used to calculate transit speeds. If TBSTIME and TBSDIST fields are provided, the transit speeds on these links are overridden by these values. Park-Ride (PNR) Station Coding The new PNR station coding scheme provides a more realistic representation with separate nodes representing the parking lot, the rail platform/fixed-guideway station and the bus stop. Table 7-2 describes the purpose and facility type of the new links added to the highway network in the new coding scheme: Table 7-2: PNR Link Facility Types Link Purpose Facility Type Driveway link Access PNR lot & Bus stop from street 49 Walk Access link Connector between PNR node and rail/bus stops 59 Escalator link Connector between rail and bus stop 59 Source: Table 4, JTA/RTS Application Guide Report February 2010 Page 7-3

114 Station information is coded on the PNR nodes. Driveway links are added to provide access to PNR lots and bus stops from the street. It is important that the station information be correctly coded only on the PNR node since the station data file is generated directly from the PNR node during the model run. Not all fields created for the transit-only links need to be specified in the highway network. It depends on the purpose for which they are coded. Table 7-3 presents the variables (check mark) which are required by the model. February 2010 Page 7-4

115 Station Data Information The station data (kept in STATDATA.YYA in FSUTMS Tranplan models) are coded on the node layer of the highway network. TSTYPE on the nodes describes the nature of the parking station. The fields shown in Table 7-4 are included in the node layer of the network. The default PNR terminal time to all informal park-ride stations identified during the bus-rider survey analysis is assumed to be two minutes and the KNR terminal time is assumed to be a ½ minute at all stations. However, for formal JTA park-ride stations, these default values can be changed for individual stations using PNRTERMTIME and KNRTERMTIME fields on the highway link layer for links which connect the PNR node to the station/stop node(s). Table 7-3: Fields Required for Transit-only Links Facility Type Variables Purpose Busonly link Connector from the street to the BRT station PNR driveways (connector from the street to the PNR node) Escalator link for walking across platforms (bus stop to station) Walk link from PNR lot node to the station/stop node Fixedguideway links FTYPE_YYA ATYPE_YYA TBSDIST 1 TBSTIME 1 TFGDIST TFGTIME 2 TFGMODE PNRTERMTIME KNRTERMTIME 1 If not coded, speed on the link is calculated using the SPDCAP table 2 If not coded, WALKTIME defaults to one minute, otherwise WALKTIME is equal to TFGTIME Source: Table 5, JTA/RTS Validation & Application Guide Reports Table 7-4: Node Fields for Station Data Parameter Type Description Recommended Values TSNAME Character Station name n/a STATZONE Numeric Nearest zone number n/a TSTYPE_YYA TSRANGE_YYA Numeric Numeric Types of access available at station Maximum roadway distance allowed for auto-access connector (miles) TSPARKSPACE_YYA Numeric Number of parking spaces (parkand-ride stations only) TSCOSTAM_YYA Numeric Parking cost in peak period TSCOSTMD_YYA Numeric Parking cost in off-peak period Source: Table 4, JTA/RTS Validation Report & Table 6, JTA/RTS Application Guide 0 not used 1 PNR and drop-off stations 2 Fringe PNR 3 PNR only (daily) 4 drop-off only 5 PNR-only (peak hours) 6 Informal PNRs (from OD-survey) 2.0 park- or kiss-and-ride 5.0 fixed-guideway 10.0 commuter rail or end of line fixedguideway Spaces Cents February 2010 Page 7-5

116 The AUTOCON program builds park-ride connectors to stations with TSTYPE 1, 3, 5 or 6. It should be noted that TSTYPE 2 indicates is a fringe parking facility. Figures 7-2 and 7-3 depict the transit stations used in 2005 model validation. Table B-5 presents a tabular summary of the 2005 station data information. This summary table is generated by the CV script for the transit station nodes. February 2010 Page 7-6

117 Figure 7-2: 2005 (Base Year) Transit Stations Note: Red: JTA parking lots, Black: informal parking lots, Green: Fringe Parking stations, Pink: KNR-only nodes. See Figure 7-3 for downtown area stations Source: Figure A1, JTA/RTS Application Guide Report. February 2010 Page 7-7

118 Figure 7-3: 2005 (Base Year) Zoomed Downtown Area Transit Stations Note: Red: JTA parking lots, Black: informal parking lots, Green: Fringe Parking stations, Pink: KNR-only nodes. Source: Figure A1, JTA/RTS Application Guide Report February 2010 Page 7-8

119 Micro-Coding Stations and Fixed-Guideway Links Fixed-guideway station nodes for the Skyway and BRT are coded separately from the bus stop nodes in the network. The bus stops and the fixed-guideway stations are connected by a transfer link coded with a facility type of 59. The default walk time on these links is one minute. The walk times should be coded to reflect the best estimate of walking time. The micro-coded stations are joined together by a link with a facility type of 69. Buses running on exclusive lanes, which do not experience congestion, are coded on these links Transit Route Since most existing service operates through downtown Jacksonville, a master stop file was obtained from the JTA and used as a guide to code the downtown stops. Further details can be found in the data development section of the Application Guide ( JTA/RTS Model Documentation Application Guide.pdf ). A single transit line file is maintained in PT format for peak and off-peak periods. HEADWAY[1] corresponds to the frequency of the service during the peak period and HEADWAY[2] corresponds to the off-peak period service. If the service does not exist during a period, the corresponding HEADWAY value is set as zero. The summaries of 2005 JTA transit route characteristics (mode, operators, one-way and circular indicators, peak and off peak frequencies) along with ridership are shown in Table B-16 of Appendix B. Each transit route in the transit line file is assigned a mode number. Modal definitions follow FSUTMS transit modeling standards. The first 20 mode numbers are reserved for the non-transit modes. Transit modes are revised for modes All buses (local bus, express bus and BRT bus) are coded as mode 21. Skyway and trolleys are defined as mode 23 (circulators). Table 7-5 lists the modes used in the model. Table 7-5: Transit Mode Definitions Mode Mode Number Mode Type Walk connectors 1 Non-transit mode PNR connectors 2 Non-transit mode Drop-off (KNR) connectors 3 Non-transit mode All walk connectors 4 Non-transit mode Fringe PNR connectors 5 Non-transit mode Downtown drop-off connectors 6 Non-transit mode Transfer connectors 11 Non-transit mode Sidewalk connectors 12 Non-transit mode Local, express and BRT bus 21 Transit mode Premium service 22 Transit mode Circulators (Skyway & trolley) 23 Transit mode Light/Heavy rail 24 Transit mode Commuter rail 25 Transit mode Other mode 26 Transit mode Project mode 27 Transit mode Source: Table 6, JTA/RTS Validation Report & Table 8, JTA/RTS Application Guide Each transit line is also assigned to its operator. Operators represent the different types of service offered by the system and they are used to define the boarding and transfer fares. The fare attached to the line is mapped to the operators in the factor file. Even though the BRT and the local buses are coded as the same mode, the operator number can be used to distinguish different fares between the two services. Table 7-6 lists the operators used in the model. February 2010 Page 7-9

120 7.1.3 Walk Coverage Integral to the building of a transit network is the availability of access to transit. A critical component to this is the determination of the percentage of each zone that is considered to have walk access to transit. The percent walk of a zone represents the proportion of the zone that is accessible to transit by walking. The values are stored in a file named PCWALK_YYA.DAT and are a key input to the transit model. This file is generated offline using a GIS buffer of a ½ mile around the stops. Table 7-6: Transit Operator Definitions Operator Operator number FareSystem* attached to the Operator JTA local bus 1 1 JTA express bus 2 2 JTA BRT bus 3 3 JTA rail 4 4 JTA commuter rail 5 5 JTA trolley 6 6 JTA Skyway 7 7 * as defined in fares file Source: Table 7, JTA/RTS Validation Report & Table 9, JTA/RTS Application Guide Different percent walk values are generated for peak and off-peak transit services. The percent walk at attraction end of the trip is twice the percent walk at the production end to reflect that attractions are normally oriented near major streets where transit is located. User should consult Appendix D of TR3- Model Application Guidelines for detailed stepwise description to generate this input data file. Figures 7-4 and 7-5 depict walk coverage of peak and off-peak periods of 2005 NERPM4 transit networks Non-Transit Connectors Three main types of access exist within the NERPM model. Walk access is generally provided from centroids along centroid connectors. Auto connectors are generated to connect drivers with park-and-ride lots and transit stations from TAZs. Finally, sidewalk connector links connect TAZs and transit stops to each other in areas of relatively high density or where transit stations are present. The two main purposes of non-transit connectors are to provide access to transit stops (a purpose similar to the centroid connectors in the highway networks) and to provide a transfer connection between two transit services. This section describes the process of creating these non-transit connectors. The 2006 bus-rider survey indicated that there were five access modes in the JTA/RTS system. The access modes and their corresponding mode numbers are: Walk-access (mode 1), Park-ride access (mode 2), Drop-off (or Kiss-Ride) access (mode 3), Fringe PNR access (mode 5), and Downtown drop-off (or Kiss-Ride) access (mode 6). Table 7-7 summarizes all the non-transit modes built details in the model before path-building. February 2010 Page 7-10

121 Walk Access Connectors Walk connectors are built using PT s GENERATE statement using a walking speed of 2.5 mph. Connectors are built from centroids to transit stops up to a maximum distance of 0.6 mile. Centroid connectors longer than 0.4 mile are adjusted to 0.4 mile before building connectors. This provides walkaccess to transit for large zones. Walking is not allowed on the freeways, ramps, HOV lanes and fixedguideway links. The connectors generated by PT are reviewed by the REWALK application for logic consistency with the percent walks. REWALK makes adjustments to the connectors and the percent walk file according to percent walk and connector conditions. February 2010 Page 7-11

122 Figure 7-4: Peak Period 2005 Transit Walk Coverage February 2010 Page 7-12

123 Figure 7-5: Off-peak Period 2005 Transit Walk Coverage February 2010 Page 7-13

124 Table 7-7: Non-Transit Modes Built Details Mode Name FromNodes ToNodes Details 1 Walk access / egress 2 PNR Auto access All zones 3 Drop-off (KNR) Auto access All zones All nodes Max 0.6 miles BRT Stops All nodes Max 3 miles Res/OBD zones All PNR (designated lot and informal*) Res/OBD nodes AUTOCON program Min 0.6 miles; Max 3 miles 4 All Walk All zones All zones Max 7 miles Fringe PNR to Skyway/Trolley Downtown drop-off (CBD-KNR) to Bus All zones Skyway/Trolley PNRs Min 5 miles; Max 20 miles; only Duval County All zones FCCJ Station & Pearl/Bay Min 3 miles; Max 7 miles FG to Bus Stop All nodes All nodes FT59 links BRT Stop Transfer BRT Stops All nodes Max 0.6 mile 12 CBD Transfer CBD nodes All nodes Max 0.6 mile Source: Table 23, JTA/RTS Validation Report Longer walk connectors are built around the BRT stations to reflect the longer distances travelers may walk in order to use the better service available at those locations. Connectors up to a maximum distance of three miles are built from the stations to the zones. These connectors have very long lengths to avoid any disconnections between alternatives during New Starts analysis. Park-Ride Access Connectors The AUTOCON program is used to generate drive connectors around the park-ride stations. The program reads the transit.mas file for reading the parameter values and the necessary input/output filenames. The cost on the PNR connector is a weighted cost that includes parking cost, highway operating cost, travel time and terminal access times. The cost is expressed in equivalent in-vehicle travel time minutes. Several informal parking locations were identified from the 2006 bus-rider survey and subsequently added to the network. The nodes representing these locations have a TSTYPE of 6 so that PNR connectors can be built to these locations. Drop-Off Access Connectors Drop-off connectors are built around all bus stops using PT s GENERATE statement. These connectors are meant for transit riders who are dropped-off by car at bus stops fairly close to their homes. These connecters have a maximum distance of three miles, a figure based on the rider survey data. Peak period connectors are built using congested highway time and the off-peak period are built on free-flow times. Once the connectors are generated using PT, they are post-processed to convert the access time into weighted equivalent in-vehicle travel time minutes and add other costs involved. The weights on the costs are as follows: Drop off time (1.5 weight factor) Terminal time (2.5 minutes with a weight factor of 2.0) Operating cost ($6/hr for peak period and $3/hr for off-peak) February 2010 Page 7-14

125 Fringe PNR Connectors Fringe PNR connectors represent travelers who ride into downtown by car and use transit to circulate through the downtown. The fringe connectors are destined for only the St. Andrews Trolley park-ride lot, the Kings Avenue Skyway Station, the San Marco Skyway Station and the Convention Center Skyway Station. The cost on the connector is time and time equivalent of toll. The connectors are post-processed to include other costs involved during the fringe parking. The connectors are weighted as follows: Drive time (1.5 weight factor) PNR terminal time (2.0 weight factor) Parking cost ($6/hr for peak period and $3/hr for off-peak) Operating cost ($6/hr for peak period and $3/hr for off-peak) Downtown Drop-Off Access Connectors The 2006 bus-rider survey indicated almost one-third of the drop-off trips were being dropped-off in downtown Jacksonville. The nature of this market is very different from traditional drop-off trips, so separate connectors are needed for this unique path. Connectors are built to those locations identified in the survey, namely FCCJ-Downtown and the intersection of Pearl and Bay Streets. Transfer/Sidewalk Connectors Non-transit connectors also allow transfers to occur between transit services. These connectors let people walk from a stop along one bus service to a stop along another service. In the model, transfer connectors (mode 11) and sidewalk connectors (mode 12) are the two modes that serve this purpose. Transfer connectors connect bus stops with fixed-guideway platforms to allow transfers between these two modes. Separate transfer connectors are built around the BRT stations to allow transfers to nearby local service. A maximum distance of 0.6 miles is used to build these transfer connectors. Most of the JTA s transit service in 2005 was oriented around downtown Jacksonville. The main transit center, the FCCJ station, is situated at the northern edge of the downtown. Hence, most of the transfers are likely to occur in the CBD. Moreover, buses run on parallel streets because of one-way streets in the downtown. To facilitate proper transfers, sidewalks are built in the CBD. The sidewalks are built for a maximum distance of 0.6 miles. All-Walk Connectors In addition to the five access modes and transfer connectors, zone-to-zone walk connectors ( all walk ) are also created. These connectors are created using PT s GENERATE statement. The connector time reflects the time it would take to walk from one zone to another. They are created in order to compare them against transit paths developed during the transit path building step. If the all-walk cost is less then the transit cost, the transit path is considered to be unreasonable (since it is faster to walk than to ride transit) and the transit skims for such interchanges are zeroed out. This helps to eliminate very short transit trips. 7.2 Transit Network Summary and Speed Validation The auto-bus speed relationships were derived from speed data collected in have been used as the basis for computing mixed-flow transit speeds for the JTA/RTS model. The relationship is read from a file named, TRANSPD.DBF in the parameters folder. Different curves are used based on the area type February 2010 Page 7-15

126 and the facility type of the links. The transit speed is a ratio of the auto speed for any given link. The ratio varies by facility type and area type and is different for peak and off-peak periods. The speed relationships used for JTS/RTS and NERPM4 models are shown in Table A-1 of Appendix A. It is used as a lookup data file to establish relationships between highway and transit speeds. The transit speeds of the JTA/RTS model were calibrated manually by changing the values of PKSPDRATIO and OPSPDRATIO columns. In the JTA/RTS model validation efforts, the end-to-end route travel times were calibrated at regional-level and corridor-level. The public time table gives the travel times between the major stops along a route. The transit speeds represented in the model are a function of the auto speeds in mixed-flow traffic (i.e., where buses and automobile share lanes). Different auto-transit speed functions are used depending on the street s facility type and area type. The starting auto-transit speed relationship data was derived from the Jacksonville auto-bus speed survey. However, the comparison showed a large difference between the estimated and the observed end-to-end travel times of the buses. The auto-transit speed relationships were then manually modified to produce more reasonable transit speeds. For JTA/RTS model validation, instead of comparing end-to-end run times of individual bus routes, the buses were divided into five categories (EW corridor, NS corridor, Circulator, Others and Express). At the regional level, the calibration works well within the limitation of modeling capabilities. Off-peak period results are better than the peak period results. Reference 2 has more information on the JTA/RTS transit model speed validation. The transit model processes and networks in NERPM4 were the same of those used in the JTA/RTA model. The role of transit model speed validation of NERPM4 is to incorporate the JTA/RTS transit model changes and to make the speeds consistent throughout the model chain. In the JTA/RTS model, refinements were made to the highway network prior to transit model application to make it more reflective of the known capacity and speed characteristics. In NERPM4, transit related changes in speeds and capacities were carried out in the initial highway network processing, thus removing the consistencies in the speeds used in highway and transit model components. Table 2-8 compares the initial and congested speeds of the NERPM4 and JTA/RTS model. The inconsistency in the initial speeds of JTA/RTS model between its highway and transit model is evident in the summaries in Table 2-8. On the other hand, the pre-assigned initial and congested speeds of both of the NERPM4 and JTA/RTS models are very similar. The overall initial speeds that are used in off-peak period transit modeling of the NERPM4 and the JTA/RTS are and mph, respectively. The overall pre-assigned congested speeds that are used in peak period transit modeling of the NERPM4 and the JTA/RTS are and mph, respectively. These pre-assigned highway speeds of the NERPM4 and JTA/RTS models are also very similar for each of the facility and area type groups (see Tables B-6 and B-7 of Appendix B). For example, pre-assigned congested speeds of freeways of the NERPM4 and JTA/RTS models are and mph, respectively. Based on the assumption of 6 peak and 10 off-peak hours, the outputs of the TAREPORT routine were processed to summarize transit network characteristics (Distance in miles, VMT, VHT and speeds) for the peak and off-peak periods. These summaries were made for both 2005 NERPM4 and JTA/RTS models. Table 7-8 presents the summaries from the two validated models. A few notable statistics comparing the 2005 NERPM4 and JTA/RTS transit networks are: There are 1,663 directional route-miles for the 2005 transit network. Of these, 1,449 and 14 miles are for buses and circulators (Skyway & Trolley), respectively. February 2010 Page 7-16

127 For NERPM4, the overall vehicle-miles-of-travel (VMT) in the peak and offpeak hours are 13,867 and 18,730, respectively. For JTA/RTS, the overall vehicle-miles-of-travel (VMT) in the peak and off-peak hours are 13,865 and 18,728, respectively. There are 174 peak period express bus directional route miles. For NERPM4, the peak period VMT for express bus is 1,008, which is the same as for the JTA/RTS model. For NERPM4, the overall vehicle-hours-of-travel (VHT) in the peak and off-peak hours are 777 and 1,026, respectively. For JTA/RTS, the overall vehicle-hours-of-travel (VHT) in the peak and off-peak hours are 825 and 1,134, respectively For NERPM4, the systemwide transit running speeds are and mph during peak and off-peak hours, respectively. For JTA/RTS, the systemwide transit running speeds are and mph during peak and off-peak hours, respectively. The VMT, VHT and speeds by mode, operators and period of the 2005 transit networks of the validated NERPM4 and JTA/RTS models are also very similar. It can be concluded that 2005 NERPM4 transit network and speeds reasonably replicate 2005 JTA/RTS model results. February 2010 Page 7-17

128 Table 7-8: Comparison of Transit Network Summary Statistics by Mode and Operators 7.3 Transit Paths and Skims Transit paths represent the best transit option from an origin zone to a destination zone. The transit paths were built using PT s single-path path-builder logic, accomplished by setting the BESTPATHONLY parameter in the factor file to T. Separate transit paths were developed for each transit market in the FSUTMS transit standards and the 2006 bus-rider survey. There is a corresponding skim matrix that gives the travel component values for each path. February 2010 Page 7-18

129 This section of the report provides information concerning validation of the transit access and path building steps in the model. It also describes SELECTLINK feature in PT used in the model to obtain the time spent on the guideway during a transit trip Transit Paths Nine paths are developed each for the peak and the off-peak periods. 1. Walk to bus: the access mode is walk and the transit modes included during the path building are bus and circulators (modes 21 and 23) 2. Walk to project: the access mode is walk and all transit modes are allowed during path building with at least a portion of the path on the project mode 3. PNR to bus: the access mode is PNR and the transit modes allowed during the path building are bus and circulators (modes 21 and 23) 4. PNR to project: the access mode is PNR and all transit modes are allowed with at least a portion of the path on the project mode 5. Drop-off to bus: the access mode is drop-off and the transit modes included during the pathbuilding are bus and the circulators (modes 21 and 23) 6. Drop-off to project: the access mode is drop-off and all transit modes are allowed during path building with at least a portion of the path on the project mode 7. Downtown drop-off to bus: the access mode is the downtown drop-off connectors and the transit modes included during the path building are bus and circulators (mode 21 and 23) 8. Downtown drop-off to project: the access mode is the downtown drop-off connectors and all transit modes are included during the path building with at least a portion of the path on the project mode 9. Fringe PNR: the access mode is the fringe connectors and the transit modes included during the path building is circulators (mode 23) The path parameters are defined in the individual path s factor files. These files are created in the TRNPREPARE step and can be found in the output folder. All path parameters are consistent with those used in the mode choice model. Two transfers are allowed on all paths, except for the fringe PNR path where no transfer is allowed. A boarding penalty of two minutes and a transfer penalty of 10 minutes are added to the overall transit path cost. The travel time weights are applied in the factor file using RUNFACTOR keyword. Out-of-vehicle times are weighted twice the in-vehicle times. Table 7-9 shows all the paths developed in the model and the settings that vary by paths. Table 7-9: Transit Paths Settings Path Access Modes Egress Modes Factor file Required Modes Conditioning Walk- Bus/BRT Walk Walk WalkBus.FAC Bus/BRT (21) Bus/BRT IVTT > 0; Pass all-walk test Walk- Commuter Rail Commuter Rail IVTT > Walk Walk WalkPrj.FAC Project/Premium (25) 0; Pass all-walk test PNR- Bus/BRT PNR Walk PNRBus.FAC Bus/BRT (21) Bus/BRT IVTT > 0; Pass all-walk test PNR- Commuter Rail Commuter Rail IVTT > PNR Walk PNRPrj.FAC Project/Premium (25) 0; Pass all-walk test Drop-Off- Bus/BRT Drop-Off Walk KNRBus.FAC Bus/BRT (21) Bus/BRT IVTT > 0; Pass all-walk test Drop-Off- Drop-Off Walk` KNRPrj.FAC Commuter Rail Commuter Rail IVTT > February 2010 Page 7-19

130 Project/Premium (25) 0; Pass all-walk test Drop-Off Drop-Off Downtown- Downtown Bus/BRT Walk CBDKNRBus.FAC Bus/BRT (21) Bus/BRT IVTT > 0 Drop-Off Downtown- Project/Premium Drop-Off Downtown Walk CBDKNRPrj.FAC Commuter Rail (25) Commuter Rail IVTT > 0 Auto Fringe parking Walk FringeCir.FAC ASE, Trolley (23) Mode 23 IVTT > 0 fringe Source: Table 24, JTA/RTS Validation Report Paths considered to be unreasonable are removed from consideration by the mode choice model. The allwalk skim created in the connector application is used to check whether using transit is better than just walking to the destination. The transit skims for the interchange are zeroed out if the weighted cost on the all walk path is less than the weighted cost on the transit path. Walk-, PNR- and drop off-transit paths are evaluated in this manner during path conditioning Transit Skims Transit skims are the travel cost components obtained from the transit paths. These are required by the mode choice model in calculating the shares of the different paths. There is a skim matrix for each path and each of these matrices has 15 tables, which are listed in Table Table 7-10: Tables in the Transit Skim Matrices Table # Name Modes Included 1 Walk access time Walk-access, walk-egress, station access 2 Drive-access time to bus Weighted drive-access time plus weighted auto occupancy cost, parking cost & terminal time 3 Sidewalk/transfer time Transfer time 4 Bus in-vehicle time All local, express and BRT bus services 5 Premium bus in-vehicle time Limited stop service 6 Circulator in-vehicle time Circulators, streetcars, trolleys 7 Light/Heavy rail in-vehicle time Light Rail, Heavy Rail 8 Commuter rail in-vehicle time Commuter rail 9 Other mode Typically mode introduced before project mode 10 Project mode in-vehicle time New mode that is object of alternatives analysis 11 Boardings All transit modes 12 Initial wait time 13 Transfer wait time 14 Fare 15 Total transit time All modes Source: Table 25, JTA/RTS Validation Report Transit Paths using Guideway Links (SELECTLINK) The SELECTLINK function is a tool to identify demand using a particular transit link. An application utilizing this feature has been developed to calculate the amount of travel time on the BRT guideways during a trip. This helps in isolating transit trips that use fixed- or BRT-guideway during any portion of their trip. This technique is used to apply additional constants in the mode choice utility program or in the computation of additional user benefits for trips benefiting from the guideway. The final output of the application is a SELECTLINK matrix named SELECTLINK_AYY.MAT. The eight tables in the matrix are shown in Table February 2010 Page 7-20

131 Table 7-11: Tables in the Selectlink Matrix Table # Description 1 Time on guideway for Walk Bus/BRT path (peak period) 2 Time on guideway for PNR Bus/BRT path (peak period) 3 Time on guideway for Drop-off (KNR) Bus/BRT path (peak period) 4 Time on guideway for Downtown drop-off (CBD-KNR) Bus/BRT path (peak period) 5 Time on guideway for Walk Bus/BRT path (off-peak period) 6 Time on guideway for PNR Bus/BRT path (off-peak period) 7 Time on guideway for Drop-off (KNR) Bus/BRT path (off-peak period) 8 Time on guideway for Downtown drop-off (CBD-KNR) Bus/BRT path (off-peak period) Source: Table 25, JTA/RTS Validation Report 7.4 Transit Fares Apart from the free trolley service, all other JTA services use the October 2007 flat-fare system. The fare is deflated to 2005 constant dollars for mode choice evaluation using the {InflTransitFare} catalog key. JTA s fare structure emphasizes the use of weekly or monthly passes (as there is no reduced fare when transferring). In an attempt to properly reflect the typical paid fare, the monthly pass cost is converted to an equivalent one-way fare by dividing by 22 average work-days per month and dividing by two directional trips. The fare on the Skyway includes the parking cost at the stations since most Skyway riders buy monthly parking passes from JTA and use it for Skyway rides. The boarding and the transfer fares are detailed in Tables 7-12 and Table 7-12: Transit Boarding Fares (2007 fares) Operator Boarding Fare Notes Local Buses $0.90 $40 monthly pass Express Buses $1.50 BRT Buses $1.50 Rail $1.50 Commuter Rail $1.50 Trolley Free Skyway $0.45 $20 monthly fare Source: Table 8, JTA/RTS Validation Report & Table 11, JTA/RTS Application Guide Table 7-13: Transit Transfer Fares Faresystem From To Free Free Free Free Free $0.90 Free 2 $0.60 Free Free Free Free $1.50 $ $0.60 Free Free Free Free $1.50 $ $0.60 Free Free Free Free $1.50 $ $0.60 Free Free Free Free $1.50 $ Free Free Free Free Free Free Free 7 Free Free Free Free Free Free Free Source: Table 9, JTA/RTS Validation Report & Table 12, JTA/RTS Application Guide February 2010 Page 7-21

132 8. Mode Choice Mode choice is the process during which person trips are categorized by mode of travel and converted to vehicle trips. Almost all major urban and regional modeling in Florida uses a nested logit model formulation. The nested models subdivide modes into related nestings. This allows the mode choice model to prefer choices within these groups to those outside of the groups. For example, if a trip maker would typically be a driver, when confronted with other options for transportation modes, the driver is more likely to choose some other auto mode, such as carpooling, as opposed to a transit mode, such as local bus. Conversely, if a trip maker typically rides transit, the individual would most likely choose between local bus and express bus options rather than choosing an auto mode. The nested logit model allows for this kind of decision-making. The JTA/RTS and/or NERPM4 mode choice model is set up to accommodate the following modes: Auto - Drive alone auto HOV2 - Carpool, 2 passengers HOV-3+ - Carpool, 3-or-more passengers Fringe Parking Bus/BRT Local and Express buses and BRT Project/Premium Project and Premium (rail) modes The transit modes use following access paths: Walk access Auto PNR access Auto drop-off (KNR) access Auto downtown drop-off access The mode choice model performs the choice analysis of the above modes and access paths for three distinct purposes: HBW Home-based work HBNW Home-based non-work (or other) NHB Non-home-based 8.1 Model Structure The final JTA/RTS mode choice model was completely scripted and to use CV s MATRIX XCHOICE function. NERPM4 uses the same nested logit setup for all transit model components. A nested logit model is a behavioral model that is used to estimate the probability of a decision maker s choice of taking an alternative from a set of alternatives. In a mode choice model, for any given choice, the probability of a mode being chosen is given by: Where: P m = Probability of choosing mode m U m = Utility of mode m M = Number of different modes to be chosen among The nested logit mode choice model structure (see Figure 8-1) works by computing the utility for each of the bottom level choices (for example, drive alone, SR2, SR3+, transit with walk access, transit with PNR, transit with KNR). This utility represents the total economic cost in terms of travel time, cost, and other February 2010 Page 8-1

133 impediments/inducements to travel associated with each mode. It is typically constructed as a linear function of the different components of time and cost as shown below: U m = (C 1 *IVTT m +C 2 *OVTT m +C 3 *WAIT m +C 4 *COST m +C 5 *DATA1 m + +C 0 )*1/NC m Where: U m = Utility for mode m C 1,C 2,C 3,C 4,C 5.. = Statistically estimated coefficients C 0 = Constant IVTT m = In-vehicle travel time for mode m OVTT m = Out-of-vehicle travel time for mode m WAIT m = Wait time for mode m COST m = Travel cost or fare for mode m DATA1 m = Other data elements characterizing the mode m trip K m = Mode-specific constant for mode m NC m = Product of the nest coefficients for all upper nests The utility of a mode is assumed to be a function of attributes that describe the level of service (LOS) provided by the mode (called coefficients), and a mode specific constant. The mode specific constant, also known as mode bias coefficient, is an adjustment parameter that compensates the unknown effects of the variables not included in the utility computation. The choices at the top level are auto, fringe parking, and transit (see Figure 8-1). The auto nest is divided into drive alone and shared ride trips. Shared ride trips are further divided into 2 passenger trips and 3+ passenger trips. Transit nest is divided into various access markets: walk, PNR, drop-off and downtown drop-off trips. The access trips are further sub-divided into bus and project modes. BRT buses are included in the same nest as local buses. The total person trip is divided into zero-car, one-car and two+-car households for the HBW and HBNW purposes. No market segmentation is done for the NHB purpose. The mode choice is run separately for these markets. In addition, the trips are distributed into can walk to transit, must drive to transit and no transit access categories using the data in the percent walk file. February 2010 Page 8-2

134 Figure 8-1: Mode Choice Nesting Structure Top level structure Person Trips Auto Fringe Parking Transit Auto nest Transit nest Source: Figure 3, JTA/RTS Validation Report & Figure 2, JTA/RTS Application Guide Report February 2010 Page 8-3

135 8.2 Mode Choice Calibration The appeal of the nested logit model is its ability to accommodate differential degrees of interdependence between subsets of alternatives. Section 8.1 provides the description of the nested logit structure. The logit parameters (constants and coefficients) are presented in this section. The mode split calculation takes into account the time and cost of travel. Travel time is divided into two general groups: (1) time spent in the vehicle, and (2) time spent outside the vehicle (walking, waiting, transferring, and parking the vehicle). Times are separated in the model because travelers dislike out-of-vehicle travel much more than riding time. The primary validation check of the transit assignment process is a comparison of observed versus modeled boardings. This was checked for the region, by mode and operator. The first step of the validation of a transit assignment occurs during the mode choice model validation. In the mode choice step, the mode-specific constants for the region were derived so that the mode-choice model produces the appropriate share of transit trips for the region and different market segments. The coefficients file is named TRN_COEFFICIENTS.DBF and is located in the parameters folder. The coefficients for each purpose are summarized and compared in Table 8-1. The out-of-vehicle time is weighted twice as much as the in-vehicle travel time. The value of time for HBW is six dollars per hour ($/hr) and 3 $/hr for non-work. The value of time is used to convert the cost components of the utility variables into equivalent in-vehicle travel time. A penalty of 10 in-vehicle time minutes is applied for each transfer. The CBD constants applied to trips that are destined to the CBD are also in this file. The validated NERPM4 values were compared to those in JTA/RTS model. They are very similar in values. Table 8-1 also presents the nesting coefficients. The same nesting coefficients are used for the three purposes. February 2010 Page 8-4

136 Table 8-1: Summary of Mode Choice Transit Coefficients and Calibrated CBD Constants February 2010 Page 8-5

137 The file containing the various utilities and mode choice constants, MC_CONSTANTS.DBF, is located in parameters folder. These constants are based on targets developed using the 2006 bus-rider survey. Utility constants having a value of effectively mandate zero trips for that particular choice. Table 8-2 details the constant values used in the NERPM4 validated model as compare those with the JTA/RTS validated model. The validated constants of NERPM4 and JTA/RTS models are very similar. Constants represent the unknown and the goal of the validation should be to reduce the values of these constants. A higher value of the constants causes the model to be insensitive to changes in the level of service and costs associated with a particular mode. Most of the values of the constants are small (see Table 8-2). Transit network simulation requires a number of input files for each period (peak and off-peak). After reasonableness checks of transit network and path building parameters, the mode specific constants are validated through a series of iterative model runs. An auto-calibration routine (see Section 7.3 of TR3 Model Application Guidelines) was used to adjust the constants based on the shares specified in MC_TARGETS.DBF file in the parameters folder. During calibration, the files MC_CONSTANTS.DBF and TRN_COEFFICIENTS.DBF in the parameters folder are overwritten. MC_TARGETS.DBF file contains both auto and transit trip shares. The transit shares are obtained based on the targets used in 2005 JTA/RTS model validation which based the shares from the 2006 bus-rider survey. The auto trip targets are based on the sub-mode proportions from the initial model runs of NERPM4. The shares used in NERPM4 mode choice model validation are summarized in Table 8-3. Note that for each market segment in a purpose, the trip shares add to Mode Choice Reports A summary mode choice report is created for each purpose. The file is named ModeSum.txt and is located in the output folder of the model. It reports the estimated trips for each available sub-mode and a summary of transfers in the system. A debug is also created for detailed traces of the mode choice computations for an interchange. This file is named ModeDebug.txt and is located in the output folder of the model. To produce this file, the catalog key {DebugMC} must be set to 1 and the desired O/D interchange must be identified using the {SELORIGIN} and {SELDEST} catalog keys. The final step in the mode choice application is adding the auto part of the auto access transit trip in the highway trip table. For each auto-access trip between an interchange, trip equivalent to transit trip divided by a user defined occupancy rate is added from the origin zone to the zone where the parking or drop-off occurs. The mode choice application also produces input files needed to run FTA s Summit program for New Starts Analysis. TPP2UB is a Cube utility that converts specially-formatted Voyager matrices to a binary format that can be read by Summit. A Summit binary file is created for each trip purpose. 8.4 Calibration Results The mode choice model provides estimates of linked trips by mode. The section presents and discusses the mode choice model trip summary and compares the results with the observed data. Tables 8-4, 8-5 and 8-6 compare the target trips and the estimated trips after calibration. The three tables are for HBW, HBNW and NHB purposes respectively. Trip ratio data is the ratio of model estimated trips February 2010 Page 8-6

138 to the target trips. The ratios show that the mode choice is fairly close to the target trips (since it is calibrated to target trips). The observed transfer rates for the JTA system is around 40%. The model estimates an overall transfer rate of 44 for the system. Table 8-7 shows the transfer rate by different paths and compares those to 2005 JTA/RTS model validation results. Transfer rates of both NERPM4 and JTA/RTS are very similar. February 2010 Page 8-7

139 Table 8-2: Summary of Calibrated Mode Choice Constants for Various Sub-modes February 2010 Page 8-8

140 Table 8-3: Summary of Target Trip Shares for Mode Choice Calibration Row # HBW HNBW NHB (*) Market Description Drive alone Shared ride 2 passenger Shared ride 3+ passengers Walk bus 5 Walk project Zero-car PNR bus 7 households PNR project KNR bus 9 KNR project CBD-KNR bus 11 CBD-KNR project Fringe parking Drive alone Shared ride 2 passenger Shared ride 3+ passengers Walk bus 17 Walk project One-car PNR bus 19 households PNR project KNR bus 21 KNR project CBD-KNR bus 23 CBD-KNR project Fringe parking Drive alone Shared ride 2 passenger Shared ride 3+ passengers Walk bus 29 Walk project Two-car PNR bus 31 households PNR project KNR bus 33 KNR project CBD-KNR bus 35 CBD-KNR project Fringe parking Walk bus PNR bus CBD Targets KNR bus CBD-KNR bus (*) Shares for Non-Home-Based trips represent all HH categories February 2010 Page 8-9

141 Table 8-4: Comparison of 2005 Home-Based-Work Model Estimated and Target Trips by Mode and Household Market February 2010 Page 8-10

142 Table 8-5: Comparison of 2005 Home-Based-Non-Work Model Estimated and Target Trips by Mode and Household Market February 2010 Page 8-11

143 Table 8-6: Comparison of 2005 Non-Home-Based Model Estimated and Target Trips by Mode Table 8-7: Comparison of Transit Trip Transfer Rates by Path By Path Linked Trips Boardings NERPM JTA/RTS Transfer Rate Linked Trips Boardings Transfer Rate 1. Walk-Bus 23,056 34, ,935 34, PNR-Bus KNR-Bus 1,595 2, ,586 2, CBDKNR-Bus Fringe Cir 2,171 2, ,183 2, Total 28,147 40, ,021 40, February 2010 Page 8-12

144 The auto person trips are shown by purpose, mode (drive alone, 2 persons shared ride and 2+ persons shared ride) and household type (0, 1 and 2+ cars). Some notable statistics of the 2005 validated model are (see Tables 8-4 to 8-6): All the cells match the corresponding targets of auto occupancies. Driving alone makes up about 72, 35 and 35 percent of the HBW, HBNW and NHB person trips. Of the total HBW person trips, 98.7 percent of trips are made by automobile and 1.3 percent by transit. Of HBNW and NHB trips, the transit shares are 0.4 and 0.6 percent, respectively. Transit trips are shown by purpose and mode and access. Tables 8-4 to 8-6 compare modeled versus target linked transit trips by bus modes for the three trip purposes and transit access modes (walk, park-and-ride, kiss-and-ride, and CBD kiss-and-ride). A few notable observations on the linked transit trips in the 2005 NERPM4 validated model (see Tables 8-4 to 8-6) include: Walk to Bus is the predominant transit access mode serving about 92 percent of the linked transit trips in the region for the HBW purpose. The share of bus for HBW trips is 72 percent. That share of bus is 94% and 76% for the HBNW and NHB trip purposes, respectively. About 43 percent of the total regional transit trips are for the HBW purpose. The ratio of total transit linked trips between the validated model run and the target is The ratios of any cell (market) shown in part C of Tables 8-4 to 8-6 varies from 0.98 to 1.01 by purpose, access and mode. This indicates close agreement. Of all HBW trips (highway and transit), 0-car, 1-car and 2+ cars households make up about 6.9, 33.5 and 59.6 percents of trips, respectively. In the case of all HBW transit trips, 0-car, 1-car and 2+ cars households make up about 52.5, 31.5 and 16.0 percent of trips, respectively. Of all HBNW trips (highway and transit), 0-car, 1-car and 2+ cars households make up about 7.9, 33.4 and 58.7 percent of trips, respectively. In case of all HBNW transit trips, 0-car, 1-car and 2+ cars households make up about 58.9, 25.2 and 15.9 percent of trips, respectively. Overall share of walk and auto access transit trips are 82 and 18 percent, respectively. The auto share is slightly higher for the HBW trips (22 percent). Since CBD is the single largest transit market for the JTA, the mode choice model was also calibrated to make sure the total trip to the CBD is correctly represented in the model. CBD trips were calibrated by purpose ad by access mode. Again, the estimated trips are very close to the observed trips (see Table 8-8). The CBD dummy coefficients are in the coefficient file (see Table 8-1) Auto-Occupancy Rates The auto occupancy rates resulting from the validated model are shown in Tables 8-4, to 8-6. The auto occupancy rate for HBW trips is 1.18 and for HBNW and NHB trips those rates are 1.62 and The model- generated rates match the targets, which are based on 2000 NFHTS. The NERPM4 target rates are generally follow the national rates presented in NCHRP 365. The updated NCHRP rates [Reference 23: Tables 37 & 39, NCHRP 365), which are based on 1990 Nationwide Person Transportation Survey (NPTS), are shown in Table 8-9. The NCHRP 187 auto-occupancy rates for some purposes (for example HBW and HBNW) are quite different from those presented in NCHRP 365. However, auto occupancy rates from 2000 NFHTS are February 2010 Page 8-13

145 used as a gauge of how well the target mode shares are being matched. The NFHTS auto-occupancy rates are very comparable to the 2005 validated model runs. Table 8-8: Mode Choice Calibration Results for CBD Trips A. Observed CBD Transit Trips (*) Access HBW HBNW NHB Total Walk 1,506 1, ,420 PNR KNR CBDKNR Total 1,910 1, ,162 B. Estimated CBD Transit Trips Access HBW HBNW NHB Total Walk 1,498 1, ,406 PNR KNR CBDKNR Total 1,900 1, ,143 C. Ratio (Estimated/Observed) of CBD Transit Trips Access HBW HBNW NHB Total Walk PNR KNR CBDKNR Total (*) Source: Table 33, JTA/RTS Validation Report Table 8-9: NCHRP 365 Auto Occupancy Rates by Urbanized Population, Income and Purpose Trip Purpose (NCHRP Table 37) HBW HBShop HBSocRec HBOther NHB ALL Urban Area Size Updated NCHRP 365 Parameters 50,000 to 199, ,000 to 499, ,000 to 999, ,000, Source: NPTS, 1990 (NCHRP Table 37) HBW HBShop HBSocRec HBOther HBNW NHB ALL Urban Area Size Parameters from NCHRP ,000 to 199, ,000 to 499, ,000 to 999, ,000, February 2010 Page 8-14

146 (NCHRP Table 39) Trip Purpose Urban Area Size HBW HBShop HBSocRec HBOther NHB Low Medium High ALL Source: NPTS, 1990 It was concluded from all these results that the NERPM4 mode choice model was successfully calibrated. In general, the 2005 NERPM4 transit model was well validated based on the guidelines recommended for FSUTMS, and provided a good estimate of trips by mode. The ratios between the estimates and the targets are 1.00 for most of the purposes, car ownerships and drive categories (DA, SR2, SR3+). Transit trip estimates were good, given their market share. This model should prove useful for long range planning purposes, as well as for corridor level analysis, but additional validation may be required for corridor level major transit investment studies. February 2010 Page 8-15

147 9. Transit Assignment The last transit-modeling step assigns the transit trip tables produced by the mode choice model onto the transit paths obtained during the path building step. The transit trips are assigned to the minimum time path by an all-or-nothing method for each combination of mode and access. Unlike trips estimated during the mode choice step, assigned transit trips can be identified on all modes used to get to a destination. In other words, transit trips are measured by route and represent unlinked trips by mode. For the 24-hour transit model, a common modeling practice is to assign all work trips to the peak network and all non-work trips to the off-peak network. For NERPM4, HBW trips are assigned to the peak period transit paths and the non-work trips are assigned to off-peak period paths. Like in other FSUTMS models, the transit trips are assigned in the P-to-A direction. Hence, because of this directionality, the results of P- to-a transit assignment must be interpreted cautiously especially when looking at individual stop- or station-level results. 9.1 Model Process Transit assignment is the process of loading the trips HBW trips are assigned to the peak period transit paths and the non-work trips are assigned to off-peak period paths. The PT assignment process produces an output DBF file which contains a summary of boardings and alightings by route link. The transit trips are allocated independently of highway trips. The resulting loads are reported by line, mode and operator using the TAReport program [References 14 & 15]. The TAReport program summarizes the peak and the off-peak period transit boardings at route-level and at stop-level. The TAReport program uses a DBF file produced by the LINKO statement when the keywords NTLEGS is set to N and ONOFFS is set to T. The TAReport program reads a TAReport.CTL file that is created in the model using a PILOT statement. The program outputs two transit assignment summary files; routelevel summary file is named tasum_ayy.prn and the stop-level summary file is named tasroute_ayy.prn. The route level transit assignment summary report produces the travel time, number of passengers, passenger miles, passenger hours, and maximum load by route, mode, and operator. The transit assignment stop-level report produces the leg travel time, the cumulative travel time, ONS, OFFS, leg distance, cumulative distance and the load for all transit routes. In most cases, the two reports are detailed enough to provide the data needed for planning purposes. A more detailed description of the TAReport program can be found in Reference 14 and/or Section 8.2 of Technical Report 3 (Model Application Guidelines) Additional Reporting In addition to the ASCII report, another assignment is performed using the ONELINKREC and NTBYLINK keywords (setting both to T ) with the LINKO statement. This produces a DBF file with records that reflect the accumulated transit volume (accumulated over all routes using the link) on each highway link. The DBF file is used to create a network with total transit volume on each link. The network displays loaded link transit volumes. An example bandwidth loaded volume is shown in Figure 9-1 for base year skyway volume. For advanced and customized reporting, it may be necessary to modify the existing script in the model. A station activity report is also available, which details the station access and egress volumes by mode of access as well as total boardings and alightings. Activity is reported for each station listed in the February 2010 Page 9-1

148 STATREP_YYA.DAT file. Again, caution should be taken while using this report since the report produces the numbers in P-to-A direction. The loaded transit networks (NETO files) contain the transit route links, access links, and stop nodes of all transit routes included in the scenario. It should be noted that, as with most FSUTMS models, peakperiod transit trips are for the volume-based work purpose and that off-peak transit trips are for homebased non-work and non-home-based trips. The relevant loaded transit network attributes are listed below where # is equals to transit trip tables loaded. February 2010 Page 9-2

149 Figure 9-1: Skyway Volume Plots using ONLINKREC Output ON[#] The number of boardings at this node. OFF[#] The number of alightings at this node. VOL[#] The ridership at this node. The trip tables (#) loaded are: 1. Walk bus and BRT 2. Walk project and premium 3. PNR bus and BRT 4. PNR project and premium 5. KNR bus and BRT 6. KNR project and premium 7. CBD KNR bus and BRT 8. CBD KNR project and premium 9. Fringe parking The modeler can use Cube to visually display and analyze the transit ridership information. To do this, the modeler should select the Show Transit On/Off option in the Transit menu at the top of the screen with a loaded transit network opened in cube. The on volumes are represented in green, the off February 2010 Page 9-3

150 volumes are represented in blue, and the through volumes are represented in teal. Figure 9-2 shows the station activity plot of the Show Transit On/Off display results. Figure 9-2: Station Activity (Transit On/Off) Plot ASE Route Southbound Direction The modeler also can use the Show Transit Line Profile option in the Transit menu to display a graph of the ridership behavior. Figure 9-3 shows what this transit profile (line volumes at stop) looks like. February 2010 Page 9-4

151 Figure 9-3: Line Volumes at Stop (Transit Line Profile) Plot ASE Route Southbound Direction Summary of BRT Trips A separate application, named BRT Reporting, is included in the model to produce assignment reports for only those trips on interchanges that use the BRT guideway. This process helps to isolate the BRT ridership impacts. Specialized reporting for other project modes, including rail, is already available by reviewing the standard reports since those modes are assumed to only use fixed-guideway facilities. 9.2 Model Validation The primary validation check of the transit assignment process is a comparison of observed versus modeled boardings. Boardings were checked for the region, by mode and sub mode and operator. The first step of the validation of a transit assignment occurs during the mode choice model validation. In that step, the mode-specific constants for the region were derived so that the mode-choice model produces the appropriate share of transit trips for the region and different market segments. As a first step in the validation of transit assignment results, an evaluation of the operating data and transit attributes generated by the TAReport program was performed. Speeds along with other statistics (directional distance, peak/off-peak VMT and peak/off-peak VHT -- see Table 7-8) give an indication that the model is replicating the existing transit operating characteristics. As part of the transit model validation effort, year 2005 transit service characteristics and ridership information for all fixed transit services in JTA were assembled by the consultants from transit agencies for use in 2005 model validation. Table B-16 of Appendix B includes a summary of these data. The observed ridership data is the average of the weekday ridership between May 1 and May 11, The bus-rider survey was done in the same year. Ridership information, along with 2000 NFHTS and 2006 February 2010 Page 9-5

152 bus-rider survey data, was used to develop transit targets (see Tables in Chapters 8 and 9). These targets are used mainly to check the reasonableness of key modeling assumptions and model ridership estimates. 9.3 Results and Comparisons The NERPM4 transit model assigns all purposes (HBW, HBNW and NHB) in the P-to-A direction for each time period (peak and off-peak). This is a conventional approach for the transit trip assignment process. This section presents summary results of the transit assignment process for the NERPM4 transit models. Summaries are made from the 2005 validated run and were also compared to both observed data and 2005 JTA/RTS model validated model results (see Table 9-1). This table tabulates the model estimated passenger trips by transit operators (Local Bus, Express Bus, Trolley and Skyway). Overall, systemwide estimated ridership is 4% higher than the observed ridership (40,661 estimated vs. 39,122 observed). Estimates of NERPM4 and JTA/RTS models are very similar. The observed transfer rate for the JTA system is around 40%. The model estimates an overall transfer rate of 46% for the system. Table 8-7 shows the transfer rate by different paths. Table 9-2 presents the observed and estimated ridership by route. It also includes the model estimated rideship from the 2005 JTA/RTS validated model. Even though there are larger differences between the estimated and the observed ridership numbers on the individual routes, it is typical of regional zone-based models. Some of the routes show more variability in the ratios of the estimated trips to the observed trips. The variations are primarily due to the very low number of trips. Graphs (scatter-plots) of the route-level estimated versus observed ridership are presented in Figure 9-4. A similar plot was made by using the 2005 JTA/RTS validated model route level transit riderships and is shown in Figure 9-5. The statistical accuracy statistics, often referred as goodness-of-fit parameters (for example, RMSE and correlation), were also computed and presented in these figures. The systemwide statistics (total and average volume per route and differences) are also shown. The scatter-plots exhibit a good linear trend (a high degree of correlation - 94 percent or higher) without any significant outliers. February 2010 Page 9-6

153 Table 9-1: Comparison of Systemwide Transit Boarding by Operators February 2010 Page 9-7

154 Table 9-2: Comparison of 2005 Model Estimated and Observed Transit Ridership by Route February 2010 Page 9-8

155 Figure 9-4: Scatterplot and Accuracy Statistics of Transit Route Boardings NERPM Validation Estimated Boardings 3000 Jax Riderships (Obs & Est) Predicted Value Linear (Predicted Value) Observed Boardings 3000 RMSE: 47.30% Systemwide Statistics: Observed Model Model-OBS Error(%) Total Volume 39,121 40,661 1, % Ave Volume per Route Regression Statistics Multiple R 94.17% R Square 88.68% Adjusted R Square 86.30% Standard Error Observations 43 ANOVA Statistics: df SS MS F Significance F Regression E E E-21 Residual E E+05 Total E+07 February 2010 Page 9-9

156 Figure 9-5: Scatterplot and Accuracy Statistics of Transit Route Boardings JTA/RTS 2005 Validation Estimated Boardings 3000 Jax Riderships (Obs & Est) Predicted Value Linear (Predicted Value) Observed Boardings 3000 RMSE: 44.53% Systemwide Statistics: Observed Model Model-OBS Error(%) Total Volume 39,121 40,829 1, % Ave Volume per Route Regression Statistics Multiple R 94.82% R Square 89.91% Adjusted R Square 87.53% Standard Error Observations 43 ANOVA Statistics: df SS MS F Significance F Regression E E E-22 Residual E E+05 Total E+07 February 2010 Page 9-10

157 Analyses of regression results are shown in each of these figures. The overall r-squared statistics of the fitted lines are about 89 percent and F-statistics are also very high. These accuracy statistics along with systemwide volumes are presented in Figures 9-4 and 9-5 for NERPM4 and JTA/RTS models, respectively. Once again, similar quality of results is exhibited at the route level by both of these models. Based on the results shown in various tables and figures, it can be concluded that the NERPM4 transit assignment model is validated well. The mode choice model estimated linked trips match the target trips very well (see Tables 8-4 to 8-6). The ratios of the model estimation to the target linked trips are with few percentage points for most of the market segments with trips of significant numbers. The mode choice model accurately estimates mode shares. The transit assignment process results in accurate estimates of weekday travel using transit modes. With the number of trips of significant in numbers, the estimated unlinked trips closely match the observed ridership. However, the estimates of individual modes and routes may vary from the observed ridership. The transit validation results show that NERPM4 does an excellent job of replicating existing transit use in the Jacksonville region. February 2010 Page 9-11

158 10. Highway Assignment The last step of the four-step modeling process is assignment. NERPM4 highway assignment uses an equilibrium assignment process. Truck trips are assigned simultaneously with the drive-alone and shared ride trips in NERPM4. Evaluation of the highway assignment model is based on comparisons between traffic counts and model assigned volumes. Modeled traffic volumes are compared to traffic counts in several ways to determine whether the coded highway network accurately represents the highway system, and to determine whether the various assumptions used in the model chain are reasonable. The highway evaluation program (HEVALDBF) is the primary tool used in comparing simulated volumes with the traffic counts. The assigned 24-hour volumes were compared to the 24-hour traffic counts. Validation also included a 2035 model run to make sure that 2035 results are reasonable. The future year run is done to assess the forecasting ability of the model. Validation of the highway assignment involved the adjustment of the speeds, capacities, penalties and other trip distribution elements as well as modifications to the VFACTORS file. A number of key evaluation statistics are generated during the evaluation phase of the model. Three of these (volume-tocount ratios, vehicle-miles traveled, vehicle hours traveled) are compared by area type, facility type, and laneage. Volume-to-count ratios also were compared by screenline and volume groups. Along with these statistics, the root mean squared error (RMSE) was generated. An RMSE is provided for each of the six counties within the study area as well as for the study area as a whole. This chapter describes validation of the highway assignment model. It includes an overview of the model process, development and adjustment of model parameters, and a review of model results. It provides validation statistics of NERPM4 highway assignments of both 2005 (base) and 2035 (cost-feasible trend) model runs. Key assignments results were summarized in numerous tables and figures Model Process and Validation Adjustments The purpose of highway assignment models is to load auto trips onto the highway network. This process results in traffic estimates on individual links that ultimately attempt to simulate general vehicular travel throughout the study area. In NERPM4, the highway assignment process loads these trips separately by purpose. These purposes are split between internal (trips that have either their origin or destination or both located within the study area) and external-external (trips that have both their origin and destination located outside of the study area). These trips are further subdivided into three vehicle classifications: truck/commercial vehicle, single-occupancy vehicle, and high-occupancy vehicle. In total, there are six purposes that are loaded onto the highway network as follows: 1. Trucks; 2. Single occupancy vehicle; 3. High-occupancy vehicle; 4. Trucks external-external; 5. Single-occupancy vehicle external-external; and 6. High-occupancy vehicle external-external. The NERPM4 model also adds transit related auto access trips to the drive alone (purpose 2 above) and shared ride (purpose 3 above). The highway assignment model uses an equilibrium assignment algorithm. February 2010 Page 10-1

159 In equilibrium, all travelers are assigned to their optimum path; no traveler can have a shorter path available. Each assignment of trips from all zones is considered one assignment iteration. Typically, multiple iterations are required before networks reaches full equilibrium. Link speeds are adjusted after each equilibrium iteration and the next assignment is performed. The 2000 NERPM model produced illogical trip assignments in which EE trips diverted from major roadways onto the local network to bypass congestion. It is reasonable to assume that the vast majority of EE travelers passing through the study area would not be familiar enough with the local street network to make such trip decisions. To prevent this, a special network code (EECODES) was introduced in the NERPM network to prohibit EE trips from using links with an EECODE of 1. That process was continued in NERPM4. The highway assignment process uses turning penalties and prohibitors. Table A-4 lists the validated turning penalties. The assignment model uses revised volume/delay curves developed during calibration and facility-type-specific UROAD factors to convert the input level-of-service E capacities from possible to practical capacities. In NERPM4 model all facility specific UROAD factors are set to 1 and thus avoiding the use practical capacity in the assignment step. This change has required the assignment step to use a different set of capacities in the SPDCAP table. It should be noted that capacities in the SPDCAP table represent LOS E and are based on FDOT LOS handbook. Initially, a speed-capacity table used in the NERPM4 model that was consistent with the changes made in 2005 based JTA/RTS transit model input speed. Section 2.4 of this document has further description on updates of speeds and capacities used in model validation. The assignment step includes an option of generating a path file to perform post analysis (for example, select link analysis). This added capability allows users to perform site specific analysis. Both peak-period pre-assignment in the DISTRIB module and the final highway assignment in HASSIGN module are allowed to run for a maximum of 50 iterations or until the equilibrium process converges according to GAPS (less than equal to ) convergence criterion for three successive iterations. The loaded highway network contains vehicle trip volumes estimated by model. In addition to these volumes, the loaded network also reports congested speed, vehicle-miles traveled (VMT), and congested travel times per link. The loaded network from the 24-hour assignment is directly used in the highway evaluation module to generate evaluation statistics for the 24-hour period. Table 8-4 of Technical Report 3 (Model Application Guidelines) describes the key attributes of the loaded network. The highway evaluation module uses database versions of the Florida HEVAL and RMSE routines. Other summary statistics are generated using CV scripts. Outputs of HEVAL and RMSE routines were used to perform systems evaluation activities and to assist in the model validation process. HEVAL operates in two modes (validation and analysis). The statistical summaries generated by CV scripts are fairly extensive with regard to model validation. However, additional important statistics for assignment validation are generated by the standard FSUTMS routines, HEVALDBF and RMSEDBF. Chapter 9 of TR3 (Model Application Guidelines) has detailed descriptions of the reports generated by the CV scripts as well as standard HEVALDBF and RMSEDBF routines. The validation mode of HEVAL allows the user to print a variety of reports designed to assist in the validation task. The validation mode does not require input data other than the loaded highway network file. The analysis mode requires a series of input parameters to calculate the number of accidents, emissions, fuel consumption, and construction costs in addition to the loaded link record file. February 2010 Page 10-2

160 Modified Volume-Delay Functions An iterative equilibrium technique is used in NERPM4. This is a standard practice in most of the FSUTMS highway models in Florida. In this type of assignment, all of the trips are loaded, the paths are revised, the trips are again loaded, and the procedure is repeated until equilibrium is reached. This technique uses the Bureau of Public Roads (BPR) formulation, in which link travel time is recomputed using the following relationship: T c = T f * {1 + (v/c) } Where, T c = congested link travel time T f = link free-flow travel time v = assigned volume c = link capacity, = BPR parameters Since speed is distance divided travel time, the BPR formulation in terms of speeds is expressed as follows: Where, S c S c = S f / {1 + (v/c) } S f = estimated congested speed = link free-flow speed Like many other recent FSUTMS model, NERPM4 highway assignment process is the incorporation of multiple BPR curves based on the facility type of the roadways. Using different BPR curves for each type of facilities recognizes that each facility type has unique characteristics when responding to congestion. For example, freeways can generally handle a higher level of congestion than surface streets before speeds begin to deteriorate. However, with more congestion, speeds deteriorate to stop-and-go conditions much more quickly on freeways than they do on surface streets. The BPR curve does not accurately estimate speeds for volume/capacity ratios greater than 1.0. Modification of the VFACTORS file involves adjustments to the UROAD factors, CONFAC values, and BPR coefficients. The UROAD factors are used to derive the practical capacity (the point at which vehicles begin to divert from the roadway) as a percentage of the capacities designated in the SPDCAP file. The CONFAC is a peak-to-daily ratio used in converting hourly capacities for comparison with daily volume estimates. The two BPR variables, BPR LOS and BPR EXP, provide for the adjustment of speed and delay curves by specific facility type. The BPR curves determine both the level of congestion (the volume/capacity ratio at which speeds begin to deteriorate) and the rate at which they deteriorate as congestion increases. In NERPM models, modified BPR curves have been used, with different coefficients and exponents for each facility type. Version 4 of NERPM uses multiple BPR curves. The curves are specified in the VFACTORS file. The validated VFACTORS file is depicted Table A-5 of Appendix A. The adjustment to the BPR curves was made by changing and the parameters of BPR functions. In addition, speeds were also adjusted. The facility specific BPR curves, used in the 2005 validated model, are shown in Figure The curves used in the 2005 model validation were also tested in the 2035 model to ensure that the assigned speeds are reasonable. A relatively steeper curve was used for freeways and HOV facilities. The curves for arterials were comparatively less steep. The factors entered in the VFACTORS are usually validated parameters and should not be changed for model application. February 2010 Page 10-3

161 UROAD Factors The volume-delay relationship assumes practical capacity. A UROAD factor of 0.75 has commonly been used since FSUTMS was first developed. The UROAD factors, entered in the VFACTORS file (see Table A-5), convert the possible capacity (LOS E) to the practical/design capacity (LOS C) a condition at which trips generally begin diverting to less congested facilities. Volume-Delay relationships and UROAD factors work together. The capacities calculated in the CV application of highway module are converted to practical capacity for use in the volume-delay relationship. The LOS C capacity is largely subjective and is determined by different methods, depending upon the facility type and traffic control. Thus, there no longer exists a simple method of relating LOS C to LOS E capacity that works across the full range of facilities or traffic controls. For example, LOS C on freeway is determined by traffic density; while LOS on two-lane roads is determined by percent time delay. Because of confusion that model users usually faced to report volume-over-capacity in the post analysis, it was decided to use UROAD factors of 1 for all facilities and use LOS E capacities of the SPDCAP table in the highway assignment. These changes caused further adjustments in the volume-delay parameters. However, it has removed the burden on the users in the interpretation of volume-over-capacity ratios in any post analysis since LOS E based capacity entered in the SPDCAP table was directly used in the assignment step. A tabular summary of capacities for the 2005 validated model by the facility and area type group combination are shown in Table 2-7. February 2010 Page 10-4

162 Congested/Uncongested Speed Ratio Technical Report # 3 Model Validation & Calibration Figure 10-1: Modified BPR Volume-Delay Functions Freeway (0.45,6.75) HOV (0.45,7.00) Divided Arterial (0.49,4.35) Undivided Arterial (0.50,3.75) Collector (0.51,3.15) Centroid Tie (0.10,2.50) 1-Way (0.53,4.5) Frontage (0.475,5.25) Ramps (0.475,4.85) Toll Facility(0.45,6.50) Volume/Capacity Ratio Note: Facility-specific ALPHA and BETA parameters are shown as FACILITY TYPE (ALPHA,BETA) February 2010 Page 10-5

163 CONFAC Factors For the 24-hour model, CONFAC is the ratio between the peak hour traffic and the daily traffic. The FSUTMS programs use the CONFAC parameter to convert hourly capacity to a daily value so that a 24- hour assignment can be made. Historically, the method for obtaining daily capacity restrained traffic assignments has been to divide the hourly capacity by CONFAC (say, 0.10) to reflect the daily highway capacity. The VFACTORS file specifies the value of CONFAC, which is the fraction of the 24-hour trip table that occurs in the peak hour for the purpose of calculating volume/capacity (capacities almost always are stated as hourly volumes). Empirical evidence shows that as overall congestion grows, the value of CONFAC decreases. The theoretical lower limit for CONFAC is (1/24), that is, conditions are equally congested during every hour of the day. The upper limit is 1.00, which would occur when all traffic moves during a single hour (admittedly unlikely). Quick Response values for CONFAC for areas with a population of more than one million are about Generally, Florida s 24-hour travel models use a value between 0.07 and 0.11 among the facilities and region. The NERPM4 model uses smaller values of CONFAC for the limited access facilities (freeways, expressways, HOV facility, and toll facilities a value of 0.095) compared to those used for other facilities (a value of is used for non-freeways), because limited access facilities in general are more congested than other facilities Model Validation In total, sixty-six model runs were executed in order to validate NERPM4. Some of these documented model runs include examination of model statistics from both base and future year runs. Validation was done by minimizing the difference between model simulated volumes and observed counts for the year 2005 distributed throughout the study area. As many count locations were accounted for as possible in order to ensure a wide range of coverage geographically as well as to incorporate as many examples of facilities and land uses located within the study area. Adjustments were made to key elements in the modeling process to achieve this validation. After each run, a summary of the results was compiled and analyzed by the consultant in order to identify problems arising in the model and successful strategies toward validation. Appropriate changes consistent with the discoveries revealed during analysis were then implemented and subsequent runs were executed. This iterative process was continued until validation was achieved. Changes made to the model during highway assignment validation included the changes in BPR volumedelay parameters and UROAD and CONFAC factors, iterative adjustments to speeds and capacities, examine the network and make appropriate changes to the network, adjustments to travel time penalties, and increasing the maximum number of equilibrium iterations. Validation of a traffic assignment involves an examination of several statistics, most of which are related to actual ground counts taken on various links throughout the network. The traffic counts for NERPM4 were obtained mainly from the Florida Traffic Information CDROM. Additional local counts were coded onto the network. Section 2.3 of this report documents the process of traffic count updates and review. Tables 2-3 and 2-4 of Chapter 2 summarize the traffic counts. One key to successful highway model validation is the availability of accurate traffic counts in sufficient quantity. Efforts were made to insure February 2010 Page 10-6

164 that sufficient counts were included in the model for all available area type and facility combinations. The percentages of the links with traffic counts by facility and area types are shown in Tables 2-3 and 2-4 of Chapter 2. These statistics were used to evaluate the validation results presented in this chapter. For example, there will be less confidence in the evaluation results (say volume-over-count ratio) in locations where fewer links have traffic counts. These counts provide the basis for highway assignment evaluation, and are input into the model as link attributes. The highway assignment model was validated by adjusting several model parameters, most notably the parameters of the VFACTORS file and the speeds. Several changes were made to the initial free-flow speeds. The speed modifiers are added at the end of the speed-capacity table and were properly commented. These modifiers are listed in Figure A-3 of Appendix A. The hierarchy of speeds and capacities among the facility and area types were always checked when a change in speed was made. Comparisons between uncongested (original) and congested highway operating speeds provide reliable indicators of congestion and associated delays. Tables 2-5 and 2-6 of Chapter 2 present these speed statistics for the 2005 and 2035 NERPM4 model runs. A comparison of the original and the congested speeds was made for each main facility types. Post-assignment network speeds (often known as congested speeds) reflect a substantial decrease in operating speeds for selected facility and area types. For the 2005 model, there is 5.74 mph (13.7%) decrease in speed from an original speed of mph. Freeway speed decreased by 6.99 mph (11.5%) due to congestion. The percent decrease in speeds is higher in the 2035 model run with an overall decrease in speed of 12.9 mph (30.6%). For 2035 model run, freeway speed decreased by 21.4 mph (35.43%) due to congestion. Section 2.4 provides more discussion of these speed comparisons among NERPM4, JTA/RTS and NERPM 2000 models. Following conclusions drawn in Section 2.4 are repeated here: JTA/RTS model s initial speeds used in final highway assignment and in pre-assignment for transit model are not consistent. NERPM4 model uses same initial speeds throughout the model chains. NERPM4 model s initial speeds are very similar to those used in JTA/RTS pre-assignment for transit model validation. NERPM4 model s congested speeds are also very similar to those resulted from JTA/RTS preassignment for transit model validation. NERPM4 model s initial and congested speeds are significantly higher than those used in 2000 based NERPM model. Those differences are more evident for higher facility groups (for example, freeway and ramps) 10.2 Results and Comparisons Since one of the most common uses of travel demand models is to forecast future traffic volumes in order to identify the impacts of growth over time and better plan to mitigate these impacts. A proper validation of the highway assignment is critical to the meaningful use of travel demand models. The HEVAL and RMSE generated statistics provide the basis on which the ability of the model to simulate observed conditions is judged and include VMT-V/G ratios, VHT-V/G ratios, volume-overcount (V/G) ratios, volume to count comparisons for screenlines and cutlines, and percent root mean square error. The special CV steps also produce summaries that were used in model validation. February 2010 Page 10-7

165 Summaries from HEVAL and RMSE outputs are presented in numerous tables in this chapter for the 24- hour highway loads. Chapters 7-9 of this report provide a detailed discussion on the transit model and validation results. The subsections present the validation results of highway assignment mode largely in tabular form Systemwide Volume-over-Count and RMSE Statistics The ratios of VMT and VHT, as calculated from assigned volumes versus those calculated from ground counts, were available. Further aggregations of these statistics were compared by area type, facility type, and for the total of all links. A ratio of 1.0 indicates exact agreement between the assignment and the traffic counts. The systemwide values (see Tables 10-1) of total VMT-V/G, VHT-V/G and V/G ratios range for the region. Table 10-1: Comparison of Systemwide Highway Model Validation Statistics ITEM NERPM NERPM Trend NERPM 2000 Total Households 571, , ,895 Total Population 1,390,070 2,183,258 1,105,229 Total Number of Links 28,789 28,914 27,212 Total System Miles 3,706 3,775 3,066 Total Lane Miles 8,150 8,560 6,579 Total Directional Miles 6,157 6,302 5,005 Average Non-Centroid Total Volume 11,618 16,945 10,996 Total Non-Centroid VMT(NC-VMT) 43,145,248 67,105,192 33,011,404 NC-VMT per Household NC-VMT per Capita Total Non-Centroid VHT(NC-VHT) 1,267,611 2,620,612 1,565,596 NC-VHT per Household NC-VHT per Capita Total NC INITIAL Speed (mph) Total NC CONGESTED Speed (mph) Total Change in Speed (mph) Total Percent Change in Speed % % % Total VMT-Volume/Count (VMT-V/G) Total VHT-Volume/Count (VHT-V/G) Total Volume/Count (V/G) Total VMT-Volume/Capacity (VMT-V/C) Total VHT-Volume/Capacity (VHT-V/C) Total Volume/Capacity (V/C) Symbol Used: NC = Non-Centroid, VMT = Vehicle-Miles-of -Travel, VHT = Vehicle-Hours-of-Travel, V/G= Volume-over-Count, V/C= Volume-over-Capacity, and RMSE = Root-Mean-Square-Error. February 2010 Page 10-8

166 Table 3 of the FDOT Model Update Task C Report suggests that the systemwide V/C ratios should be within ±5 percent. These overall systemwide V/C ratios indicate that the 2005 model perform extremely well relative to these performance standards. The percent Root Mean Square Error (RMSE) for the total areawide assignment is another aggregate measure to show how well the model chain has replicated ground counts. RMSE is the standard measure of error in system planning model. The smaller percent RMSE in the model indicates higher the level of confidence in the model s ability to replicate existing traffic. The percent root mean squared error (RMSE) indicates whether the simulated network contains an acceptable level of assignment error. This is based on both the areawide and volume group summaries. Table 10-2 and 10-3 summarizes the root mean square error (RMSE) statistics for NERPM4 model by volume group and county. Table 10-2 also compares the systemwide RMSE values by volume groups between NERPM and NERPM 2000 models. Table 10-2: Comparison of Systemwide Root-Mean-Square-Error Statistics NERPM NERPM 2000 Count Range Acceptable RMSE Range All Counties All Counties RMSE N RMSE N <5, , ,000-10, ,000-20, ,000-30, ,000-40, ,000-50, ,000-60, ,000-70, ,000-80, ,000-90, , , , ,000 <14 ALL , ,974 Symbol Used: RMSE = Root-Mean-Square-Error. February 2010 Page 10-9

167 Table 10-3: 2005 RMSE and Volume-over-Count Statistics by Count Range Group and County Table 10-3 (contd.): 2005 RMSE and Volume-over-Count Statistics by Count Range Group and County February 2010 Page 10-10

168 The overall system-wide RMSE values of these two models are very similar despite the higher number of model counts that exist in the expanded six county region of NERPM4. These values fall within the suggested range of percent. Moreover, all regions show a good level of validation. Except in very low volume groups, the RMSE values are either within the range or even below the lower limit of the expected ranges. The RMSE by volume groups for each of the counties are summarized in Table The countywide RMSE values of the six counties range from 31.0 to It also compares those countywide values against the RMSE values of the NERPM 2000 validated model. Percent RMSE provides a comparison of estimated traffic volumes to observed counts by volume groups of different ranges for all links for which traffic counts are available. Accuracy is more stringent for higher volume facilities than for lower volume facilities. The RMSE results for all volume groups greater than 10,000 VPD are either better or within the suggested ranges. In case of the low volume group (<10,000 VPD), the RMSE statistics are slightly higher than the upper limits of suggested range. In case of low volume groups, these slight discrepancies are tolerable because errors introduced in such cases generally do not affect roadway lane requirements. The RMSE statistics by the volume groups indicate that the simulated network contains the correct number and type of facilities and the relative speeds and capacities among facilities have resulted in an accurate assignment of traffic. The range of county-wide V/G ratios is (see Table 10-3) for the six constituent counties. The V/G ratios by the FT group between the NERPM4 and NERPM 2000 model are compared in Table NERPM4 models higher percentages of ramp counts compared to those used in NERPM 2000 model. The volume-over-count ratios by FT group in NERPM4 are as good as those validated ratios for the NERPM 2000 model. Beside volume-over-count ratios, the systemwide results from the 2005 and 2035 NERPM4 validation runs are also summarized in Table 10-1 for the region on the following items: Number of Links System Miles February 2010 Page 10-11

169 Table 10-4: Comparison of Volume-over-Count Ratios by Facility Type Group Facility Type Group Total Volume Toal Count Volume/Count No of Counts (N) 1. Freeway (11-17) 8,717,075 8,536, Divided Arterial (21-28) 10,688,249 11,016, Undivided Arterial (31-38) 4,112,439 3,791, Collector (41-48) 2,944,283 3,318, One-Way & Frontage (61-68) 442, , Ramps (71-79, 97-98) 2,154,332 2,242, HOV (81-85) 9. Toll Facility (91-95) TOTAL 29,058,485 29,317, ,750 Overall RMSE: 34.3 TOTAL (Without Ramps) 26,904,153 27,074, ,462 Overall RMSE (Without Ramps): 33.5 Facility Type Group Total Volume Toal Count Volume/Count No of Counts (N) 1. Freeway (11-17) 6,954,803 6,575, Divided Arterial (21-28) 8,192,880 8,444, Undivided Arterial (31-38) 2,823,107 2,652, Collector (41-48) 2,287,784 2,303, One-Way & Frontage (61-68) 253, , Ramps (71-79, 97-98) 104,275 70, HOV (81-85) 9. Toll Facility (91-95) NERPM NERPM TOTAL 20,616,553 20,229, ,974 Overall RMSE: 34.8 Lane Miles Directional Miles Average Link Volume VMT VMT per household VMT per Capita VHT VHT per household VHT per Capita February 2010 Page 10-12

170 Input (Free-Flow) Speed Model Congested Speed Change and Percent Change in Speed Volume-over-Capacity (V/C) ratios These statistics also were reported for the NERPM 2000 validated model run. NERPM 2000 speeds are very low (especially freeways, see Table 2-8). The volume, VMT, VHT, and speed statistics of NERPM4 are very reasonable. More discussion of these items is provided later. A few comparisons of systemwide model results of 2000 and 2035 runs (see Table 10-1) follows: The total lane-miles are 8,150 and 8,560 in the 2035 and 2005 networks, respectively, which represents about a 5% increase. The average link volumes are 11,618 (2005 model) and 16,945 (2030 model), which represent about a 46% change. The percent changes in uncongested and congested speeds are and in 2005 and 2035 model runs, respectively. The changes in speed are 5.74 mph (2005 model) and mph (2035 model). The 2035 network approaches LOS E (possible) capacity in many cases with volume/capacity (V/C) ratios with systemwide average values of 0.80 compare to 0.57 for the 2000 model run. The VHT statistics per household has changed from 0.91 (55 minutes) in 2005 to 1.2 (72 minutes) in The overall VMT/household has of the NERPM 4 model is about 75 and is very similar to values reported nationally. The growth in 2035 VMT compared to 2005 VMT is approximately 107 percent, which is equal to a 2.09% annual compound growth. The above comparisons suggest that results of both 2005 and 2035 models are very reasonable. Thus, by both systemwide V/G and RMSE measures and other travel measures, the validated models did an excellent job of replicating traffic counts. Scattergrams of assigned volumes versus counts of NERPM4 and NERPM2000 validated models are compared in Figures 10-2 and 10-3, respectively. Data points of both of these plots fall within reasonable boundary of the 45-degreee line. Two measures are usually computes of these data points: (1) the Percent Root Mean Square Error and (2) the Correlation of Coefficients or Coefficient of Determination (or R- Square). The RMSE of the two models were compared earlier in this section. The correlation coefficient (Multiple R) and R-Square statistics are shown in Figures 10-2 and 10-3 for NERPM4 and NERPM2000 validated models. These statistics of both of these validated model are 95% or higher. According Model Validation and Reasonableness Checking Manual [Reference 24], the regionwide R-square statistics should be greater than 88 percent. The NERPM4 as well as NERPM2000 models satisfy this criterion Screenline, Cutline and Corridor Volume-over-Count Ratios Screenlines, cutlines and corridors are groups of roadways oriented in the same direction, and carry traffic considered to be significant within the study area. Analyzing volume-to-count ratios along screenlines allows for examining flows into, out of, and across geographic subareas and corridors. This constitutes a key component of highway assignment as well as assisting in the examination of trip distribution. Screenlines and cutlines of the NERPM4 model are primarily based on those selected for the NERPM2000. The locations of the screenlines and cutlines for the NERPM4 model are depicted in February 2010 Page 10-13

171 Figure 2-6. Among these are the common borders of the county lines shared by the six counties in the study area as well as an external cordon measuring trips coming into and going out of the study area. Two new cordon lines (no. 23 & 40) were added in NERPM4. Also, the external station cordon line (no. 39) was updated to cover the entire six county study area of NERPM4. In addition to aggregate summaries by area type and facility type, screenline summaries were produced by HEVAL. Table 10-5 summarizes volume-to-ground-count (V/G) ratios of all screenlines, cutlines and cordon lines. This table also compares the performances 2005 validated model against the NERPM2000 model. It could be seen that both of these models are performing equally with respect to their V/G ratios. February 2010 Page 10-14

172 Volume Technical Report # 3 Model Validation & Calibration Figure 10-2: Scattergram of the Assigned Volumes versus the Counts of NERPM4 Model Count Regression Statistics - NERPM4 Multiple R 97.44% R Square 94.95% Adjusted R Square 94.91% Standard Error 3,633 Observations 2,750 February 2010 Page 10-15

173 Figure 10-3: Scattergram of the Assigned Volumes versus the Counts of NERPM2000 Model Regression Statistics-NERPM2000 Multiple R 97.42% R Square 94.91% Adjusted R Square 94.86% Standard Error 3,566 Observations 1,974 February 2010 Page 10-16

174 Table 10-5: 2005 Volume-over-Count Ratios of Screenlines, Cutlines and Cordons February 2010 Page 10-17

175 Three out of total six screenline volume-over-count ratios lie within the FDOT-suggested guidelines of ±10 percent. FDOT has established four ranges for measuring accuracy based on total counts comprising each screenline. Screenlines that carry less than 50,000 vehicles per day (VPD) should validate within +/- 20 percent. Screenlines that carry between 50,000 to 75,000 VPD should validate within +/- 15 percent. Screenlines that carry more than 75,000 VPD should validate within +/- 10 percent. External cordons and combined non-screenline links (screenline code 99) should be within more or less five percent. Many of screenlines and cutlines were found to have only a few links. This is particularly true for most of the cutlines. The individual link volume/count ratios were examined in Cube to identify systematic error pattern. None were found. To provide a better understanding of the screenline performance, the deviation of loaded volumes with reference to the ground counts was plotted for each screenline and cutline and compared to the maximum desirable deviation equation per NCHRP 255 [Reference 27]. For the validated 2005 NERPM4 model, the results are presented in Figure A total of 5 out of 40 screenlines, cutlines, cordon lines fall above the maximum desirable deviation line and 35 met their accuracy targets. These x V/G ratios for these five lines were further examined through Cube. It was decided not to use K factors in the model. So, no corrective actions were taken for the improvement of the performance of these lines. This diagram is supposed to display only the screenline s volume. However, the volumes of cutlines and corridors were also displayed to gauge their performances with respect to screenline s desirable deviation. Cutline volumes generally warrant larger deviation than screenline volumes. At lower screenline volumes, the permitted volume deviation is quite large, since such deviations would not result in significant design differences. Conversely, at higher screenline volumes, a lower deviation is desired in order to be confident that any design decisions would be valid. Users should be cautioned to adjust the loaded volumes near the screenline(s) and cutline(s) where the departure from the desirable line is significant enough to alter planning and design decisions Volume-over-Count Ratios by FT and AT Groups Several indicators are available for determining the overall performance of the highway assignment model. The volume-over-count (V/G) statistics are one of the key statistics. The ratios of VMT and VHT, as calculated from assigned volume versus those calculated from ground counts were evaluated for those links where ground counts were available. The simple ratios of assigned volume over count also were recorded. Further aggregations of these statistics were compared by area type, facility type, and for the total of all links. A ratio of 1.0 indicates exact agreement between the assignment and the traffic count. The areawide accuracy of highway assignment is measured by means of various volume-to-count ratios (VMT, VHT, volumes) for area type, facility type, and lanes categories. FDOT standards allow for an accuracy of +/- 15 percent per category and +/- five percent areawide. It is assumed that each combination of area/facility/number of lanes and link group contains a statistically valid number of links. For link groups having less than 100,000 total VMT (or less than 20,000 VHT), only a ±25 percent accuracy level is desired. Although not specified in the Task C report, assigned V/G ratios by their facility and area type were also analyzed. The analysis was based on a ±10 percent accuracy level, as was recommended for screenlines and cutlines. The summaries of daily VMT based volume/count, VHT based volume/count and simple unweighted volume/count statistics by major facility and area type and county are summarized in Tables 10-6 and Table 10-6 also compares these ratios against the ratios of NERPM 2000 model. Following are few notable results: February 2010 Page 10-18

176 Figure 10-4: Total Screenline Volumes and Maximum Desirable Deviation February 2010 Page 10-19

177 Table 10-6: Volume-over-Count Ratio by Facility Types, Area Types and Counties A. VMT, VHT & Volume/Count Ratios by Facility Type NERPM NERPM 2000 Facility VMT VHT Volume/ Average VMT VHT V/G Average Type V/G V/G Count (V/G) V/G Ratios V/G V/G V/G Ratios 1. Freeway (11-17) Divided Arterial (21-28) Undivided Arterial (31-38) Collector (41-48) One-Way & Frontage (61-68) C. Volume/Count Ratios by County 7. Ramps (71-79, 97-98) NERPM NERPM HOV (81-85) County Vol/Count Vol/Count 9. Toll Facility (91-95) 1. Nassau TOTAL: Duval St. Johns B. VMT, VHT & Volume/Count Ratios by Area Type 4. Clay NERPM NERPM Baker 1.01 x Area VMT VHT Volume/ Average VMT VHT V/G Average 6. Putnam 1.10 x Type V/G V/G Count (V/G) V/G Ratios V/G V/G V/G Ratios All Counties CBD (11-14) Fringe (21-23) Residential (31-35) OBD (41-43) Rural (51-52) TOTAL: February 2010 Page 10-20

178 Table 10-7: Volume-over-Count Ratio by Facility and Area Type Combinations 1. VMT-Volume/Count Ratio Area Type Facility CBD Fringe Residential OBD Rural TOTAL Type (11-14) (21-23) (31-35) (41-43) (51-52) 1. Freeway (11-17) Divided Arterial (21-28) Undivided Arterial (31-38) Collector (41-48) One-Way & Frontage (61-68) Ramps (71-79, 97-98) HOV (81-85) 9. Toll Facility (91-95) 2. VHT-Volume/Count Ratio TOTAL Area Type Facility CBD Fringe Residential OBD Rural TOTAL Type (11-14) (21-23) (31-35) (41-43) (51-52) 1. Freeway (11-17) Divided Arterial (21-28) Undivided Arterial (31-38) Collector (41-48) One-Way & Frontage (61-68) Ramps (71-79, 97-98) HOV (81-85) 9. Toll Facility (91-95) TOTAL February 2010 Page 10-21

179 Table 10-7 (contd.): Volume-over-Count Ratio by Facility and Area Type Combinations 3. Volume/Count Ratio Area Type Facility CBD Fringe Residential OBD Rural TOTAL Type (11-14) (21-23) (31-35) (41-43) (51-52) 1. Freeway (11-17) Divided Arterial (21-28) Undivided Arterial (31-38) Collector (41-48) One-Way & Frontage (61-68) Ramps (71-79, 97-98) HOV (81-85) 9. Toll Facility (91-95) TOTAL Averages of 3 Volume/Count Ratios Area Type Facility CBD Fringe Residential OBD Rural TOTAL Type (11-14) (21-23) (31-35) (41-43) (51-52) 1. Freeway (11-17) Divided Arterial (21-28) Undivided Arterial (31-38) Collector (41-48) One-Way & Frontage (61-68) Ramps (71-79, 97-98) HOV (81-85) 9. Toll Facility (91-95) TOTAL February 2010 Page 10-22

180 NERPM4 achieves this areawide accuracy for volume-to-count ratios at 1.02 for VMT, 1.03 for VHT, and 0.99 for volumes. There were only a few occurrences of link group volume-to-count ratios (i.e., primary area types and facility-type categories) that exceeded the standard tolerances by group only a few percentage. Most of these ratios of Table 10-7 are within FDOT acceptable tolerances. For NERPM4, the averages of these three V/G ratios range between for the major facility types for the 6-county region. On the other hand for NERPM 2000, they average range between for the major facility types of the 4-county region. For NERPM4, the area types V/G ratios have a range of for the whole region. On the other hand for NERPM 2000, the averages range between Except for one county, the simple V/G ratios of NERPM4 are better than those attained in NERPM2000 model validation. Table 10-7 demonstrates a detailed record these ratios for each combination of the FT and AT groups. The higher departures occur mainly when the link groups have few links with traffic counts (see Table 2-6). An example of this is the 1-way roadways in the fringe and residential area with only 2 links with traffic counts. Volume/Count ratios by area type and facility type provide measures of trip generation as well as trip distribution characteristics of the model chain. Results of these comparisons suggest that the highway loads replicate the observed vehicular traffic patterns in the six-county NERPM region well Average Volume and Vehicle-Miles and Vehicle-Hours of Travel Assigned volumes multiplied by link distances equals vehicle miles of travel (VMT). The link times in hours multiplied by assigned volumes results in vehicle hours of travel (VHT). These measures of system demand provide insight into other network attributes, such as fuel consumption and emissions. To assess the reasonableness of the loaded volume as well as model performance evaluation, HEVALgenerated average link loads, VMT and VHT by the major facility and area types are summarized in Table 10-8 for both 2005 and 2035 model runs. It also presents the VMT distribution and growths in VMT, VHT and average link volumes. Results are prepared for the whole region by their main FT and AT groups. In 2005, the average directional freeway volume is approximately 36,896 with 36.3% of VMT on freeways. The divided arterial accounts second highest amount of travel (30.1% VMT in 2005). Average loaded volumes by facility type follow the expected trend. Examples are much higher levels of traffic on limited access facilities. The growths in VMT, VHT and average volumes in 2030 compare to 2005 are also very reasonable by facilities. Two important statistics for highway planning, design, and management are VMT and VHT. All national statistics show an increase in these measures every year. For instance, Table 2 of the 1990 Nationwide Personal Transportation Survey reports the following: (a) Daily VMT per household (b) Persons per household (c) Daily VMT per capita [computed as (a)/(b)] February 2010 Page 10-23

181 Area Type Facility Type Technical Report # 3 Model Validation & Calibration Table 10-8: Comparison of 2005 and 2035 Average Link Volume, VHT, VMT and Percent VMT by Facility and Area Types Ave. Volume VHT VMT Percent VMT Ave. Volume VHT VMT Percent VMT Ave. Volume 1. Freeway (11-17) 36, ,471 15,654, % 50, ,753 23,564, % 36.1% 50.4% 50.5% 2. Divided Arterial (21-28) 18, ,722 12,981, % 25, ,679 20,318, % 37.8% 56.0% 56.5% 3. Undivided Arterial (31-38) 11, ,541 7,103, % 17, ,728 9,374, % 46.6% 32.9% 32.0% 4. Collector (41-48) 4, ,289 6,095, % 7, ,996 11,860, % 67.8% 91.1% 94.6% 6. One-Way & Frontage (61-68) 8,683 8, , % 11,071 8, , % 27.5% 8.2% 5.9% 7. Ramps (71-79, 97-98) 6,646 24,300 1,083, % 9,477 37,828 1,747, % 42.6% 55.7% 61.2% 8. HOV (81-85) 9. Toll Facility (91-95) NERPM (Base Scenario) NERPM (Trend Scenario) 1. CBD (11-14) 11,877 19, , % 15,397 25, , % 29.6% 30.5% 29.9% 2. Fringe (21-23) 15, ,805 6,346, % 20, ,667 8,482, % 29.8% 35.0% 33.7% 3. Residential (31-35) 11, ,737 26,623, % 17, ,166 42,965, % 55.7% 62.5% 61.4% 4. OBD (41-43) 14,161 41,107 2,035, % 17,823 56,118 2,767, % 25.9% 36.5% 36.0% 5. Rural (51-52) 5, ,021 7,460, % 9, ,997 12,007, % 59.7% 60.1% 60.9% TOTAL: 11, ,531 43,145, % 16,945 1,360,867 67,105, % 45.9% 55.8% 55.5% VHT VMT February 2010 Page 10-24

182 Daily VMT/HH and VMT/person of the SERPM model [Reference 19] of the 24-hour period are shown in the following table: Daily VMT per Household Palm Beach Broward Miami-Dade All County Daily VMT per Capita Palm Beach Broward Miami-Dade All County The per-capita and per-household VMT and VHT of the 2005 and 2035 NERPM4 runs are calculated and are shown in Table They are: NERPM NERPM Trend VMT per Household VMT per Capita VHT per Household VHT per Capita According to the Model Validation and Reasonableness Checking Manual [Reference 24], reasonable ranges of VMT per household are miles per day for large urban areas and miles per day for small urban areas. The 1990 NPTS reported an average of vehicle miles traveled per household daily. Reasonable ranges of VMT per person are miles per day for large urban areas and miles per day for small urban areas. The FDOT Task C report recommends that the VMT per capita per day be in the range of 10-15, which includes the effects of the mode choice and auto occupancy models. VMT per person for the SERPM and its constituent counties are higher than suggested by FDOT. In general, VMT/HH and VMT/person indices are higher in NERPM4 region compare to national and Southeast Florida Regional Travel Model (SERPM). Table 10-8 presents the distribution of VMT among the facilities for NERPM4 region for the 2005 and 2030 model runs. To gauge the reasonableness of the VMT by functional classification, a table from Reference 24 is reproduced below: February 2010 Page 10-25

183 VMT Distribution National Statistics Functional Small Medium Large Class K 200-1M >1M Freeway/Expressway 18-23% 33-38% 40% Other Principal Arterials 37-43% 27-33% 27% Minor Arterials 25-28% 18-22% 18-22% Collectors 12-15% 8-12% 8-12% Source: Table 7-4, Model Validation and Reasonableness Checking Manual, FHWA, The percent distribution of VMT by the facility for the 2005 and hour period is: NERPM NERPM Freeways & Ramps 38.8% 37.7% Divided & Undivided Arterials 46.6% 44.2% Collectors 14.1% 17.7% One-way Facility 0.5% 0.4% The VMT distribution is highly dependent on the distribution of facility types. The NERPM4 model VMT distribution by facility type follows the national trend very closely and the distribution is among the facilities in both base and future year model runs. The average link volumes of the 2035 NERPM4 24-hour period volumes by facility types were compared in Table This comparison was made for the whole region by the facility. Overall, there is 46 percent growth in link average volumes for the region. The growth in average link volume by facility types is not same. The lower volume facilities (undivided arterial and collectors) show higher percentage increases. Although the overall percent growths in VMT and VHT are higher for collectors, however, this facility groups accounts about 14 percent of overall travel (see VMT distribution). For the region, all the facilities types have shown positive growths. All of the statistics from the NERPM4 model presented in numerous tables and figures in this chapter indicate that the NERPM4 model produces quality results and the model is validated well with respect to FDOT and national standards. The NERPM4 model was validated to 2005 data. In addition, a 2035 NERPM4 model runs was made to see the reasonableness of the model statistics in the future. The data on which the model was based were generally developed from the travel surveys and Census data. The zonal data were developed by the PBS&J staff in consultation with the North Florida TPO and their designated local planning agencies. Traffic count data for 2005 were obtained from FDOT traffic information CD and the counties. Transit supply and ridership data were obtained from the Jacksonville transit authority. The model validation demonstrates that NERPM4 replicates existing travel conditions. Modeling theory suggests that if the model performs well in the validation year, it would provide reasonable travel estimates for other years and travel assumptions. However, occasionally modelers discover that a model that is thought to be well calibrated does not provide reasonable and logical results in future years. February 2010 Page 10-26

184 Because of this, the NERPM4 model validation includes both 2005 and 2035 model runs and compares their results. February 2010 Page 10-27

185 11. Summary and Conclusion The NERPM4 travel demand forecasting model contains many of the same elements as the JTA/RTS 2005 and NERPM2000 models with the addition of modifications to the modeling steps for their consistency. These modifications are described elsewhere in this report. Many of the modeling process were streamlined in the NERPM4 model. NERPM4 uses Cube-Voyager as its modeling platform and was validated using 2005 as base year and a future 2035 model runs. This report describes the validation of NERPM4 model. NERPM4 is an outgrowth of its predecessor models and includes new 2005 base year and more coverage of the study region and has refined zonal boundaries. NERPM4 includes Baker and Putnam counties including 4-county regions modeled in JTA/RTS 2005 and NERPM2000 models. In the JTA/RTS 2005 model updates, the transit modeling steps were completely implemented in PT with validation focused on transit elements. The 2005 and 2035 NERPM4 models provide the MPO, the Department and others with a dependable tool for forecasting travel demand in the six county region of Northeast Florida. A wide range of adjustments was made to the modeling system to produce good validation. Some of the adjustments are global, some are local, and some are combinations of both. The validation statistics demonstrate that NERPM4 does good job of replicating existing travel conditions. The validation of NERPM4 was not limited to the evaluation of the model results to the 2005 traffic counts and transit patronage. The results of 2035 model were compared to the 2005 model results to ensure that the model produces reasonable results. All key model statistics and data were summarized and compared through numerous tables and figures. The NERPM4 transit model does an excellent job of replicating existing transit use and closely matches the results of JTA/RTS 2005 model. This report summarized the model validation efforts for the 2005 and 2035 NERPM4 and compared the results with the predecessor models, surveys and national statistics. It demonstrates the strengths and weakness of the model. It was shown that both highway and transit models do a good job of replicating ground counts and transit use. Model results were also compared to the validation criteria established for FSUTMS and elsewhere in nation. Overall highway evaluation measures indicate a high degree of correlation between observed and estimated traffic volumes as forecasted by the 2005 NERPM4 model. In most cases, the performance of the model meets or exceeds the established criteria. The 2005 model is a reliable tool for system level transportation planning analyses. As with all models, however, the model results should be reviewed and adjusted as needed before using them in planning and design. It is expected NERPM4 model will include enhancements in its future updates. A number of potential model enhancements have been identified for future consideration include the following: Implementation of a lifestyle based trip generation processes that have been implemented in Southeast Florida and/or Tampa regions. Use a dynamic implementation of area types based on density of zonal data. Development of variable trip attraction rates based on area type as well as employment categories. Development of county specific production and attraction rates. Implementation of a new process of free flow speed estimation based on posted speed limits and signalization data. February 2010 Page 11-1

186 Refinement of capacity estimation process that emulates the capacities published in the Florida LOS manual. Implementation of a vehicle availability modeling process and use of separate the trip distribution processes for households with or without autos. The distribution of zero auto household trips should use only transit skims. Implementation of a time-of-day modeling process that includes managed lane modeling, including distribution, mode choice and assignment. Implementation of alternate trip distribution methodologies for school trips using school boundaries and possibly use separate purpose for college and university trips. Addition of special trip purposes or alternative trip distribution techniques for unique activities (for example, airport and seaport trips). Separation of NHB purpose into non-home based work and other purposes. Implementation of feedback loops to iterate between distribution and assignment. Conduct field studies to collect peak and off-peak speeds and validate highway model not only on counts but also on speeds. Assemble transit speeds and validate model not only on riderships but also on transit speeds. Development of a transit model that estimates true peak and off-peak travel. Implementation of a logit model for non-motorized trips. Assemble truck counts and validate truck model based on truck traffic counts. Development of 24-hour highway only and sub-area models. Use of comprehensive travel surveys to calibrate model parameters of updated structures. Many of these enhancements will require new surveys for calibration of model parameters. Coordination among FDOT District 2 Planning, North Florida TPO, and other County staff should facilitate a prioritization of these future model enhancements. While model results have generally improved over those reported in JTA/RTS 2005 and NERPM2000, opportunities still exist for further enhancements to model validity in the future. It is also believed that continued enhancements should be made in the estimation of employment data as this has a tremendous impact on the distribution of trips. The NERPM4 model can estimate the number of vehicles on a future road, passengers on a new local/express bus service, riders on a new rapid transit line, or the response to certain travel demand management polices such as imposing higher parking fees. This information is used in the MPO planning process to aid decision makers in their selection of transportation plan alternatives, polices and programs. In addition, the model results could be used to provide detailed information, such as traffic volumes, rapid transit and bus patronage to state, district and local engineers and planners for use in their design of facilities. February 2010 Page 11-2

187 12. List of References Recent JTA-RTS and/or NERPM Related References: 1. JTA/RTS Model Application Guide, Technical Report, Jacksonville Transportation Authority, Submitted by AECOM Consult, June JTA/RTS Model Documentation Calibration and Validation, Technical Report, Jacksonville Transportation Authority, Submitted by AECOM Consult, June JTA/RTS Model Trip Distribution Documentation, Technical Report, Jacksonville Transportation Authority, Submitted by AECOM Consult, June Users Guide, Northeast Florida Regional Planning Model (NERPM), FDOT- District 2, Prepared by Cambridge Systematics, Inc., April Model Validation, NERPM2000 Technical Report 1 (Draft Final), FDOT- District 2, Prepared by Cambridge Systematics, Inc with Advanced Planning, Inc and Corradino Group., December Data Development and Review, NERPM 2000 Technical Report 1 (Draft Final), FDOT- District 2, Prepared by Cambridge Systematics, Inc with Advanced Planning, Inc and Corradino Group., December Northeast Florida Regional Planning Model Data Projections, First Coast Long-Range Transportation Plan 2030 Update, Final Technical Report 3, First Coast MPO, Prepared by Cambridge Systematics, Inc with Advanced Planning, Inc, June Review of The Northeast Regional Planning Model (NERPM) 2000, First Coast MPO, Prepared by Wade White (Citilabs Inc), January Memorandum NERPM Base Year Model Revalidation for Cube Conversion, To: FDOT/District 2 Planning Staff, From: Cambridge Systematics Staff, 9/30/ Memorandum NERPM Zone Splitting Methodology, To: FDOT/District 2 Planning Staff, From: Cambridge Systematics Staff, 6/26/ Memorandum NERPM Enhancements, To: FDOT/District 2 Planning Staff, From: Cambridge Systematics Staff, 10/16/ Trip Production Rate Analysis for Jacksonville Regional Travel Model, Technical Memorandum, FDOT/District 2, Prepared by Corradino Group with Cambridge Systematics, June Long Range Transportation Plan Update Scope of Services, Issued and Approved by First Coast MPO, December Other Florida References: 14. FSUTMS Transit Model Application Guide, Technical Report, Systems Planning Office, FDOT, Submitted by AECOM Consult, September FSUTMS Transit Model Development Guide, Technical Report, Systems Planning Office, FDOT, Submitted by AECOM Consult, September February 2010 Page 12-1

188 16. A Recommended Approach to Delineating Traffic Analysis Zones in Florida (Draft Report), Systems Planning Office, FDOT, Prepared by Cambridge Systematics, Inc with AECOM Consult, June FSUTMS-Cube Framework Phase II Model Calibration and Validation Standards (Draft Tech Memo 1), Systems Planning Office, FDOT, Prepared by Cambridge Systematics, Inc, April Data Dictionary, Systems Planning Office, FDOT, December Model Data, Calibration and Validation, Technical Reports 1 & 2, SERPM and 2030 Models, FDOT- District 4, Prepared by The Corradino Group, October Model Application Guidelines, Technical Report 3, SERPM and 2030 Models, FDOT District 4, Prepared by The Corradino Group, August Other National References: 21. Users Guide Procedure Guide for the Atlanta Travel Forecasting Model Set, Atlanta Regional Commission (ARC), April ARC Model Documentation Travel Forecasting Model Set for the Atlanta Region 2007 Documentation, Atlanta Regional Commission (ARC), October NCHRP 365 Travel Estimation Techniques for Urban Planning, Transportation Research Board, National Research Council, Model Validation and Reasonableness Checking Manual, Travel Model Improvement Program, Federal Highway Administration, February Quick Response Freight Manual, Travel Model Improvement Program, USDOT, September Calibration and Adjustment of System Planning Models, Federal Highway Administration, USDOT, December NCHRP 255 Highway Traffic Data for Urbanized Area Project Planning and Design, Transportation Research Board, National Research Council, December NCHRP 187 Quick Response Urban Travel Estimation Techniques and Transferable Parameters: User s Guide, Transportation Research Board, National Research Council, Traffic Assignment - methods, application and products, U.S. Department of Transportation (USDOT), August February 2010 Page 12-2

189 Appendix A Selected Validated Data and Parameter Summary Table Page A-1 List of TRANSPD.DBF File Highway-to-Transit Speed Conversion Parameters... A-1 A-2 Validated Special Generators Trips... A-3 A-3 Validated Friction Factors... A-4 A-4 List of Validated Turning Penalties... A-7 A-5 List of VFACTORS File... A-9 Figure Page A-1 Highway Network and Speed-Capacity Scripting Changes in JTA/RTS Transit Model Updates... A-11 A-2 Snippet of SPDCAP table showing JTA/RTS Transit Model Speed Modification... A-14 A-3 List of NERPM4 Speed and Capacity Modifiers of SPDCAP File... A-16 February 2010

190 Table A-1: List of TRANSPD.DBF File Highway-to-Transit Speed Conversion Parameters CURVE_NO LOW_MODE HIGH_MODE LOW_FT HIGH_FT LOW_AT HIGH_AT PKSPDRATIO OPSPDRATIO DESCRIPTION ltd acces DA-CBD DA-CBDFrg DA-IndCom DA-Res DA-OBD UDA-CBD UDA-Fri UDA-IndCom UDA-Res UDA-OBD COL-CBD COL-CBDFr COL-Res COL-OBD Centriod OWY-CBBFr OWY-Res OWY-OBD RamHOVTol All rural ltd acces DA-CBD DA-CBDFrg DA-IndCom DA-Res DA-OBD UDA-CBD UDA-Fri UDA-IndCom UDA-Res UDA-OBD COL-CBD COL-CBDFr COL-Res COL-OBD Centriod OWY-CBBFr OWY-Res OWY-OBD RamHOVTol All rural ltd acces DA-CBDFrg DA-Res DA-OBD UDA-CBD UDA-Fri UDA-Res UDA-OBD COL-CBDFr COL-Res COL-OBD Centriod OWY-CBBFr OWY-Res OWY-OBD RamHOVTol All rural February 2010 Page A-1

191 Table A-1: List of TRANSPD.DBF File Highway-to-Transit Speed Conversion Parameters (contd.) CURVE_NO LOW_MODE HIGH_MODE LOW_FT HIGH_FT LOW_AT HIGH_AT PKSPDRATIO OPSPDRATIO DESCRIPTION ltd acces DA-CBD DA-CBDFrg DA-IndCom DA-Res DA-OBD UDA-CBD UDA-Fri UDA-IndCom UDA-Res UDA-OBD COL-CBD COL-CBDFr COL-Res COL-OBD Centriod OWY-CBBFr OWY-Res OWY-OBD RamHOVTol All rural ltd acces DA-CBD DA-CBDFrg DA-IndCom DA-Res DA-OBD UDA-CBD UDA-Fri UDA-IndCom UDA-Res UDA-OBD COL-CBD COL-CBDFr COL-Res COL-OBD Centriod OWY-CBBFr OWY-Res OWY-OBD RamHOVTol All rural February 2010 Page A-2

192 Table A-2: Validated Special Generators Trips February 2010 Page A-3

193 Table A-3: Validated Friction Factors February 2010 Page A-4

194 Table A-3 (contd.): Validated Friction Factors February 2010 Page A-5

195 Table A-3 (contd.): Validated Friction Factors February 2010 Page A-6

196 Table A-4: List of Validated Turning Penalties February 2010 Page A-7

197 Table A-4 (contd.): List of Validated Turning Penalties February 2010 Page A-8

198 Table A-5: List of VFACTORS File FT = 10, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 11, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 12, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 13, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 14, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 15, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 16, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 17, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 18, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 19, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 20, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 21, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 22, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 23, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 24, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 25, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 26, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 27, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 28, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 29, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 30, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 31, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 32, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 33, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 34, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 35, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 36, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 37, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 38, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 39, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 40, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 41, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 42, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 43, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 44, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 45, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 46, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 47, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 48, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 49, UROADF = , CONFAC = , BPR LOS = , BPR EXP = February 2010 Page A-9

199 Table A-5 (contd.): List of VFACTORS File FT = 50, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 51, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 52, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 53, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 54, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 55, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 56, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 57, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 58, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 59, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 60, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 61, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 62, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 63, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 64, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 65, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 66, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 67, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 68, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 69, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 70, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 71, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 72, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 73, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 74, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 75, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 76, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 77, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 78, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 79, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 80, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 81, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 82, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 83, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 84, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 85, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 86, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 87, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 88, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 89, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 90, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 91, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 92, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 93, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 94, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 95, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 96, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 97, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 98, UROADF = , CONFAC = , BPR LOS = , BPR EXP = FT = 99, UROADF = , CONFAC = , BPR LOS = , BPR EXP = February 2010 Page A-10

200 Figure A-1: Highway Network and Speed-Capacity Scripting Changes in JTA/RTS Transit Model Updates ; Do not change filenames or add or remove FILEI/FILEO statements using an editor. Use Cube/Application Manager. RUN PGM=NETWORK PRNFILE="{SCENARIO_DIR}\output\PANET00A.PRN" MSG='Modify the speeds/capacity for peak period assignment' FILEI LINKI[1] = "{SCENARIO_DIR}\output\UNLOADED.NET" FILEO NETO = "{SCENARIO_DIR}\output\UNLOADED_MOD.NET" PROCESS PHASE=LINKMERGE ;; adjust number of lanes on I-95 (based on Google Earth - AECOM / 09/07/2007) ; SB on I-95 if (A=34374 & B=74008) NUM_LANES=4 if (A=74008 & B=34569) NUM_LANES=4 if (A=34569 & B=34601) NUM_LANES=4 if (A=34601 & B=34658) NUM_LANES=4 if (A=34658 & B=74014) NUM_LANES=4 if (A=74014 & B=34767) NUM_LANES=4 if (A=34767 & B=35094) NUM_LANES=4 if (A=35094 & B=74034) NUM_LANES=5 if (A=74034 & B=35233) NUM_LANES=5 if (A=37006 & B=74146) NUM_LANES=3 if (A=74146 & B=74164) NUM_LANES=3 if (A=74164 & B=74174) NUM_LANES=3 if (A=74174 & B=37766) NUM_LANES=3 if (A=37766 & B=37880) NUM_LANES=3 if (A=37880 & B=74187) NUM_LANES=3 if (A=74187 & B=74197) NUM_LANES=3 if (A=74197 & B=38649) NUM_LANES=3 if (A=38649 & B=74238) NUM_LANES=3 if (A=74238 & B=74242) NUM_LANES=3 if (A=74242 & B=39041) NUM_LANES=3 if (A=39041 & B=39276) NUM_LANES=3 if (A=39276 & B=74256) NUM_LANES=3 if (A=74256 & B=39392) NUM_LANES=3 if (A=39392 & B=39406) NUM_LANES=3 if (A=39547 & B=39633) NUM_LANES=2 ;ramp at I-95/US1/US90 if (A=39633 & B=74303) NUM_LANES=2 ;ramp at I-95/US1/US90 if (A=74303 & B=74305) NUM_LANES=2 ;ramp at I-95/US1/US90 if (A=74305 & B=74841) NUM_LANES=2 ;ramp at I-95/US1/US90 if (A=74841 & B=40752) NUM_LANES=2 ;ramp at I-95/US1/US90 if (A=40752 & B=40781) NUM_LANES=4 if (A=40555 & B=40568) NUM_LANES=4 February 2010 Page A-11

201 Figure A-1 (contd.): Highway Network and Speed-Capacity Scripting Changes in JTA/RTS Transit Model Updates ;NB on I-95 if (A=40777 & B=40627) NUM_LANES=2 if (A=40627 & B=74840) NUM_LANES=2 if (A=74840 & B=39513) NUM_LANES=2 if (A=39172 & B=74244) NUM_LANES=3 if (A=74244 & B=74234) NUM_LANES=3 if (A=74234 & B=38650) NUM_LANES=3 if (A=38650 & B=74189) NUM_LANES=3 if (A=74189 & B=74179) NUM_LANES=3 if (A=74179 & B=37866) NUM_LANES=3 if (A=37866 & B=37765) NUM_LANES=3 if (A=37765 & B=74172) NUM_LANES=3 if (A=74172 & B=74166) NUM_LANES=3 if (A=74166 & B=74148) NUM_LANES=3 if (A=74148 & B=37039) NUM_LANES=3 if (A=37039 & B=36950) NUM_LANES=3 if (A=35102 & B=34775) NUM_LANES=3 ;reduced from 4 lanes to 3 if (A=34775 & B=74016) NUM_LANES=3 if (A=74016 & B=34664) NUM_LANES=3 if (A=34664 & B=34607) NUM_LANES=3 if (A=34607 & B=34576) NUM_LANES=3 if (A=34576 & B=34420) NUM_LANES=4 if (A=34420 & B=34398) NUM_LANES=4 if (A=34398 & B=34370) NUM_LANES=4 ; adjust speeds if (SPEED!=0) SPEED=(SPEED + 5) else SPEED=0 endif if (FACILITY_TYPE=11-12 & AREA_TYPE=22-60 & AREA_TYPE<>35) SPEED=70 ;interstate outside the CBD, <>35 takes care of I- 95 bridge if (FACILITY_TYPE=12 & AREA_TYPE=35) SPEED=65 ;takes care of I-295 bridges on StJohns if (FACILITY_TYPE=16 & AREA_TYPE=35 & NUM_LANES=3) SPEED=30 ;takes care of Acosta Bridge if (FACILITY_TYPE=16 & AREA_TYPE=35 & NUM_LANES<3) SPEED=45 ;takes care of Hart and Mathews Bridge if (FACILITY_TYPE=35 & AREA_TYPE=35) SPEED=55 ;takes care of Shands Bridge February 2010 Page A-12

202 Figure A-1 (contd.): Highway Network and Speed-Capacity Scripting Changes in JTA/RTS Transit Model Updates ; if (FACILITY_TYPE=21-24) SPEED=SPEED+10 ;divided arterials - this increases the speed by 15 overall if (FACILITY_TYPE=21) SPEED=60 ;divided arterials unsignalized if (FACILITY_TYPE=22) SPEED=45 ;divided arterials unsignalized if (FACILITY_TYPE=23 & AREA_TYPE=21-30) SPEED=45 ;divided arterials class 1a if (FACILITY_TYPE=23 & AREA_TYPE=31-40) SPEED=50 ;divided arterials class 1a if (FACILITY_TYPE=23 & AREA_TYPE=41-50) SPEED=40 ;divided arterials class 1a if (FACILITY_TYPE=24 & AREA_TYPE=21-30) SPEED=40 ;divided arterials class 1b if (FACILITY_TYPE=24 & AREA_TYPE=31-40) SPEED=45 ;divided arterials class 1b if (FACILITY_TYPE=24 & AREA_TYPE=41-50) SPEED=40 ;divided arterials class 1b if (FACILITY_TYPE=71-79) SPEED=SPEED+10 ;ramps - this increases the speed by 15 overall if (FACILITY_TYPE=29) SPEED=1 ;Mayport ferry ; adjust capacity if (FACILITY_TYPE=11-19) UROADFACTOR=0.95 ; freeways (changed from 0.9 to 0.95) if (FACILITY_TYPE=71-79) UROADFACTOR=1.00 ; ramps (changed from 0.9 to 1.00) if (FACILITY_TYPE=24 & AREA_TYPE=41-49) UROADFACTOR=1.00 ; Class 1b in OBD (take care of Orange Mall area) if (FACILITY_TYPE>=21 & FACILITY_TYPE<=29 & NUM_LANES>=3) CAPACITY=CAPACITY+200 if (FACILITY_TYPE>=11 & FACILITY_TYPE<=19 & NUM_LANES>=3) CAPACITY=CAPACITY+200 IF (SPEED!=0) TIME=60*(DISTANCEFT/5280)/SPEED TIME2=60*(DISTANCEFT/5280)/SPEED ENDIF TIME=ROUND(TIME*100)/100 TIME2=ROUND(TIME2*100)/100 if (time<0.01) time=0.01 if (time2<0.01) time2=0.01 ENDPROCESS ENDRUN Note: PANET00B.S CV script of JTA/RTS model February 2010 Page A-13

203 Figure A-2: Snippet of SPDCAP Table Showing JTA/RTS Transit Model Speed Modification ;RTS Global +5 mph speed - old speed was 5 mph less ;RTS Global +5 mph speed - old speed was 5 mph less ;RTS Global +5 mph speed - old speed was 5 mph less ;RTS Global +5 mph speed - old speed was 5 mph less ;RTS Global +5 mph speed - old speed was 5 mph less ;RTS Global +5 mph speed - old speed was 5 mph less ;RTS Global +5 mph speed - old speed was 5 mph less ;RTS Global +5 mph speed - old speed was 5 mph less ;RTS Global +5 mph speed - old speed was 5 mph less ;RTS Global +5 mph speed - old speed was 5 mph less ;RTS Global +5 mph speed - old speed was 5 mph less ;RTS Global +5 mph speed - old speed was 5 mph less ;RTS Global +5 mph speed - old speed was 5 mph less ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -ramps overall +15 mph - old speed ;RTS Transit Model Speed modification -ramps overall +15 mph - old speed ;RTS Transit Model Speed modification -ramps overall +15 mph - old speed ;RTS Transit Model Speed modification -ramps overall +15 mph - old speed ;RTS Transit Model Speed modification -ramps overall +15 mph - old speed ;RTS Transit Model Speed modification -ramps overall +15 mph - old speed ;RTS Transit Model Speed modification -ramps overall +15 mph - old speed ;RTS Transit Model Speed modification -ramps overall +15 mph - old speed ;RTS Transit Model Speed modification -ramps overall +15 mph - old speed ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -ramps overall +15 mph - old speed ;RTS Transit Model Speed modification -ramps overall +15 mph - old speed ;RTS Transit Model Speed modification -ramps overall +15 mph - old speed ;RTS Transit Model Speed modification -ramps overall +15 mph - old speed ;RTS Transit Model Speed modification -ramps overall +15 mph - old speed ;RTS Transit Model Speed modification -ramps overall +15 mph - old speed ;RTS Transit Model Speed modification -ramps overall +15 mph - old speed ;RTS Transit Model Speed modification -ramps overall +15 mph - old speed ;RTS Transit Model Speed modification -ramps overall +15 mph - old speed 45.0 February 2010 Page A-14

204 Figure A-2 (contd.): Snippet of SPDCAP table showing JTA/RTS Transit Model Speed Modification ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -divided arterial unsignalized - old speed ;RTS Transit Model Speed modification -divided arterial class 1a - old speed ;RTS Transit Model Speed modification -divided arterial class 1a - old speed ;RTS Transit Model Speed modification -divided arterial class 1a - old speed ;RTS Transit Model Speed modification -divided arterial class 1a - old speed ;RTS Transit Model Speed modification -divided arterial class 1b - old speed ;RTS Transit Model Speed modification -divided arterial class 1b - old speed ;RTS Transit Model Speed modification -divided arterial class 1b - old speed ;RTS Transit Model Speed modification -divided arterial class 1b - old speed ;RTS Transit Model Speed modification -take care of I-295 bridges on St Johns - Old Speed ;RTS Transit Model Speed modification -take care of I-295 bridges on St Johns - Old Speed ;RTS Transit Model Speed modification -take care of I-295 bridges on St Johns - Old Speed ;Mayport Ferry ;Few Dummy Link connection ;Few Dummy Link connection ;Few Dummy Link connection *1.00 ;RTS Transit Model Capacity Modification FT=21-29 & NL>=3 rev cap= old Cap *1.00 ;RTS Transit Model Capacity Modification FT=21-29 & NL>=3 rev cap= old Cap *1.00 ;RTS Transit Model Capacity Modification FT=21-29 & NL>=3 rev cap= old Cap *1.00 ;RTS Transit Model Capacity Modification FT=21-29 & NL>=3 rev cap= old Cap *1.00 ;RTS Transit Model Capacity Modification FT=21-29 & NL>=3 rev cap= old Cap *1.00 ;RTS Transit Model Capacity Modification FT=11-19 & NL>=3 rev cap= old Cap *1.00 ;RTS Transit Model Capacity Modification FT=11-19 & NL>=3 rev cap= old Cap *1.00 ;RTS Transit Model Capacity Modification FT=11-19 & NL>=3 rev cap= old Cap *1.00 ;RTS Transit Model Capacity Modification FT=11-19 & NL>=3 rev cap= old Cap *1.00 ;RTS Transit Model Capacity Modification FT=11-19 & NL>=3 rev cap= old Cap *1.00 ;RTS Transit Model Capacity Modification FT=11-19 & NL>=3 rev cap= old Cap *1.00 ;RTS Transit Model Capacity Modification FT=11-19 & NL>=3 rev cap= old Cap *1.00 ;RTS Transit Model Capacity Modification FT=11-19 & NL>=3 rev cap= old Cap *1.00 ;RTS Transit Model Capacity Modification FT=11-19 & NL>=3 rev cap= old Cap *1.00 ;RTS Transit Model Capacity Modification FT=11-19 & NL>=3 rev cap= old Cap +200 February 2010 Page A-15

205 Figure A-3: List of NERPM4 Speed and Capacity Modifiers of SPDCAP File * ;Start NERPM4 specific adjustments ;CBD-Freeway ;CBD-DividedArt ;CBD-UndivArt ;CBD-Collector ;CBD-1way&Frontage ;CBD-Ramps ;CBD-Ramps ;Fringe-Freeway ;Fringe-DividedArt ;Fringe-UndivArt ;Fringe-Collector * ;Fringe-1way&Frontage ;Fringe-Ramps ;Fringe-Ramps ;Residential-Freeway ;Residential-DividedArt ;Residential-UndivArt ;Residential-Collector ;Residential-1way&Frontage ;Residential-Ramps ;Residential-Ramps ;OBD-Freeway ;OBD-DividedArt ;OBD-UndivArt ;OBD-Collector ;OBD-Ramps ;OBD-Ramps ;Rural-Freeway ;Rural-DividedArt ;Rural-UndivArt ;Rural-Collector ;Rural-Ramps ;Rural-Ramps ;Residential-DividedArt February 2010 Page A-16

206 Appendix B Selected Model Data Summary Table Page B-1 Year 2005 (Base) Key Socioeconomic Data Totals by District and County... B-1 B-2 Year 2035 (Trend) Key Socioeconomic Data Totals by District and County... B-2 B-3 Change in Key Socioeconomic Datasets between 2005 (Base) and 2035 (Trend) by District and County... B-3 B-4 Percent Growth in Key Socioeconomic Datasets between 2005 (Base) and 2035 (Trend) by District and County... B-4 B-5 Year 2005 Station Data Information... B-5 B HEVAL Pre-assignment Speed Summary of NERPM4... B-6 B HEVAL Pre-assignment Speed Summary of JTA/RTS Model... B-7 B HEVAL Assignment Speed Summary of JTA/RTS Model... B-8 B HEVAL Assignment Speed summary of NERPM2000 Model... B-9 B-10 Summary of Year 2000 Based NERPM (NERPM2000) TAZs... B-10 B-11 Summary of Year 2005 Based JTA/RTS (JTA/RTS-2005) Model TAZs... B-11 B-12 External Station Traffic Information... B-12 B Census Journey-To-Work (JTW) Trip Flow Summary of Duval County CTPP Districts... B-14 B Census Journey-To-Work (JTW) Trip Flow Summary of Six Counties of NERPM4... B-15 B-15 List of Zones of Sub-area Balancing Attraction Districts... B-16 B-16 Summary of Base (2005) Year Transit Route Characteristics of Observed Ridership... B-17 February 2010

207 Table B-1: Year 2005 (Base) Key Socioeconomic Data Totals by District and County February 2010 Page B- 1

208 Table B-2: Year 2035 (Trend) Key Socioeconomic Data Totals by District and County February 2010 Page B- 2

209 Table B-3: Change in Key Socioeconomic Datasets between 2005 (Base) and 2035 (Trend) by District and County February 2010 Page B- 3

210 Table B-4: Percent Growth in Key Socioeconomic Datasets between 2005 (Base) and 2035 (Trend) by District and County February 2010 Page B- 4

211 Table B-5: Year 2005 Station Data Information February 2010 Page B- 5

212 Table B-6: 2005 HEVAL Pre-Assignment Speed Summary of NERPM4 February 2010 Page B- 6

213 Table B-7: 2005 HEVAL Pre-Assignment Speed Summary of JTA/RTS Model February 2010 Page B- 7

214 Table B-8: 2005 HEVAL Assignment Speed Summary of JTA/RTS Model February 2010 Page B- 8

215 Table B-9: 2000 HEVAL Assignment Speed Summary of NERPM2000 Model February 2010 Page B- 9

216 Table B-10: Summary of Year 2000 Based NERPM (NERPM2000) TAZs February 2010 Page B- 10

217 Table B-11: Summary of Year 2005 Based JTA/RTS (JTA/RTS-2005) Model TAZs February 2010 Page B- 11

218 Table B-12: External Station Traffic Information February 2010 Page B- 12

219 Table B-12 (contd.): External Station Traffic Information February 2010 Page B- 13

220 Table B-13: 2000 Census Journey-To-Work (JTW) Trip Flow Summary of Duval County CTPP Districts February 2010 Page B- 14

221 Table B-14: 2000 Census Journey-To-Work (JTW) Trip Flow Summary of Six Counties of NERPM4 February 2010 Page B- 15

222 Table B-15: List of Zones of Sub-area Balancing Attraction Districts February 2010 Page B- 16

223 Table B-16: Summary of Base (2005) Year Transit Route Characteristics and Observed Riderships February 2010 Page B- 17

224 Table B-16 (contd.): Summary of Base (2005) Year Transit Route Characteristics and Observed Riderships February 2010 Page B- 18