Greater Treasure Coast Regional Planning Model Version 3.3 (GTCRPM3.3) Additional Enhancements Technical Report Draft

Transcription

1 Greater Treasure Coast Regional Planning Model Version 3.3 (GTCRPM3.3) Additional Enhancements Technical Report Draft Submitted to: Florida Department of Transportation District IV Submitted by: The Corradino Group, Inc. December 2010

2 Table of Contents 1. GTCRPM3.3 Enhancements Overview 1 2. CUBE Analyst Matrix Estimation Why CUBE Analyst CUBE Analyst Matrix Estimation Methodology CUBE Analyst Application to GTCRPM Model Control Files and Input Files Development Phased Implementation of CUBE Analyst Matrix Estimator Further Adjustments to Trip Tables and Production of Future Adjusted Trips Model Application Base Year Calibration and Future Year Scenario Results 7 3. External Trip Forecasting Enhancements External Stations External Trip Forecasting User Guidelines for Intermediate Year Data Additional Reporting Features Input Files Network Attributes Catalog Keys Window Area HEVAL Instructions Select Link Analysis Instructions References 30 i:\projects\3909\two18\gtcrpmiii-v1.03 additional enhancements.docx Page i

3 List of Figures Figure 2-1 Flow Chart of Implementation of CUBE Analyst Matrix Estimation in a Feedback Loop 5 Figure 2-2 Screenlines, Cutlines, and Cordons Locations in Brevard and Indian River Counties 11 Figure 2-3 Screenlines, Cutlines, and Cordons Locations in St. Lucie and Martin Counties 12 Figure 2-4 Screenlines, Cutlines, and Cordons Locations in Palm Beach County 13 Figure 2-5A Maximum Allowable Deviation in Total Screenline and Cutline Volumes GTCRPM3.3 Analyst Version 14 Figure 2-5B Maximum Allowable Deviation in Total Screenline and Cutline Volumes GTCRPM3.3 Original Version 14 Figure 2-6A Scatterplot of Assigned Volumes vs. Observed Counts GTCRPM3.3 Analyst Version 15 Figure 2-6B Scatterplot of Assigned Volumes vs. Observed Counts GTCRPM3.3 Original Version 15 Figure 3-1 External Stations in GTCRPM Page ii

4 List of Tables Table 2-1 GTCRPM3.3 CUBE Analyst Version Percent Root Mean Square Summary by Volume Groups 8 Table 2-2 GTCRPM3.3 CUBE Analyst Version Volume to Count Ratios by Facility Type Groups 9 Table 2-3 GTCRPM3.3 CUBE Analyst Version Volume to Count Ratios by Area Type Groups 10 Table 2-4A Screenline/Cutline Results GTCRPM3.3 Analyst Version 16 Table 2-4B Screenline/Cutline Results GTCRPM3.3 Original Version 17 Table 2-5 Trip Length Distribution Comparisons between GTCRPM3.3 Original and GTCRPM3.3 Analyst Versions 18 Table 2-6 Base and Future Year Results I-95 Corridor in Treasure Coast Area 19 Table 2-7 Base and Future Year Results System Wide Model Parameters 20 Table 3-1 Internal and External Station Numbers in GTCRPM Table 3-2 External Station Counts Forecasting for Table 3-3 External Seed Percentage Matrix for GTCRPM3.3 (Input file EE_SEED_NEW.dbf) 24 Table 3-4 EE and IE/EI Targets in GTCRPM Page iii

5 Page 1 1. GTCRPM3.3 Enhancements Overview This document details the additional enhancements to the Greater Treasure Coast Regional Planning Model Version 3.3 (GTCRPM3.3). This effort supplements the GTCRPM3.3 model enhancements document dated June Primary enhancements made to the model are model post processing adjustments using CUBE Analyst matrix estimation, external trip forecasting enhancements, and new Cube Voyager based reporting features. Post processing adjustments using CUBE Analyst software included the development of an adjusted base year Origin-Destination matrix. The most recent Greater Treasure Coast Regional Planning Model Version 3.3 (GTCRPM 3.3) was developed in such a way that the user does not need to rerun or use CUBE Analyst software. In the model catalog, the user was given choice of whether or not to use the post processing adjustments prior to running the model. The catalog file was designed in such a way that the user can run regular model and/or the CUBE Analyst adjustments using one single model catalog. The model catalog file contained 4 scenarios: 2 for the base year and 2 for future year, instead of the original 2 scenarios in the earlier release. External trip forecasting enhancements included a thorough review of external-external (EE) trip patterns in the GTCRPM area. Earlier versions used EE patterns directly extracted from the Florida Statewide Model (SWM). The review found that the EE trip table needed further adjustments. The seed matrix was modified after a review of trip patterns with FDOT staff. The original FRATAR model was replaced with the new seed matrix and new CUBE applications were written to handle this process. In addition, corresponding changes to the External-Internal input files were made, so that the total of EE and EI matched traffic counts at external stations. The process was carried over to the future year, and future year inputs also were modified. Finally, this report describes the Cube Voyager based HEVAL reporting features, incorporated into GTCRPM model. Several new reporting features were added based on posted speed ranges. Additionally, a few sample reports were setup using the CUBE Reports feature, for interactively generating Scatter Plots and Pie Charts. Select link Analysis and Window Area HEVAL processes were setup to be more user-friendly. The report is organized into 5 sections. Section 1 details overview of enhancements; Section 2 describes CUBE Analyst post processing adjustments development and application details; Section 3 describes the external trip forecasting enhancements; Section 4 documents the enhancements to model reporting features; and finally, section 5 lists the references.

6 Page 2 2. CUBE Analyst Matrix Estimation 2.1 Why CUBE Analyst? Historically, post-processing techniques are implemented to travel demand model results, for using in various model-application studies, such as Corridor Planning Studies (CPS), Project Development and Environmental (PD&E) studies, Interchange Justification Report (IJR) studies, and Development of Regional Impact (DRI) studies. These post-processing techniques consider differences between traffic counts and model assigned volumes in base year as fixed errors from the travel demand model. Adjustments are made to the future year model assigned volumes to counter these errors. These post-processing techniques, sometimes quite cumbersome, consume enormous manpower and resources, to produce reasonable forecasts. While predicting future traffic itself has uncertainty, post-processing techniques sometimes become inconsistent within an alternative, and among different alternatives. In addition, the model s trip distribution errors are not successfully corrected in these post processing techniques. Thus, in regional studies such as long range plan development, the post processing techniques are less than ideal. Matrix estimation using CUBE Analyst, for the same reason, is a handy tool that can be used to build trip tables that minimize the differences between base year volumes and counts. The quality of the base trip table that is input to CUBE Analyst, in conjunction with the quality of the counts is what determines the accuracy of new estimated trip table. Hence trip distribution in traditional 4-step model is still very important. 2.2 CUBE Analyst Matrix Estimation Methodology The reader is referred to the CUBE Analyst Users Guide, supplied by Citilabs, for background details on this process. Important notes on CUBE Analyst calibration are mentioned here. Since matrix estimation is an optimization technique that modifies matrix cell values to match the volumes and counts, caution should be exercised in modifying the control files that are input to CUBE Analyst. The modifications performed to the control files, such as, confidence levels of input counts/screen lines, and prior trip table cell values are adjusted during the CUBE Analyst model calibration stage. The Analyst-produced output matrix is assigned to the highway network to verify the new overall RMSE for improvement. The output matrix is fed back to the CUBE Analyst procedure, to repeat matrix estimation, until model-assigned RMSE is below the targeted value or cannot be reduced further. The feedback loop can also be terminated using a maximum number of iterations. Finally, the output matrix should be reviewed for reasonableness. The feedback loop in the CUBE Analyst matrix estimation process is important because of improvements in prior matrix and path file/intercept file in the current iteration can improve the matrix estimation in next iteration. A thorough testing process is mandated in this calibration process to determine maximum number of iterations for feedback loop process, and how to terminate the feedback loop to achieve optimum results.

7 Page CUBE Analyst Application to GTCRPM Model The CUBE Analyst matrix estimation process was implemented in an application subgroup that is not executed unless the user specifies to do so. In other words, this subgroup has an execution order number of zero. The process is available in the GTCRPM model if the user intends to make further modifications to the calibration process Control Files and Input Files Development The following control files were used as inputs in CUBE Analyst matrix estimation for GTCRPM. The memory to be allocated to the program was set to 900 MB. It should be noted that CUBE Analyst requires large amount of memory when set to be applied on very large regional models like GTCRPM. 1. Prior Trip Table: This is the trip table produced from the GTCRPM3.3 mode choice model, which is also an input to the highway assignment. Each cell value was given a confidence level of one percent. The confidence levels of cell values can be given from percent depending on how much change CUBE analyst is allowed to make to matrix cells. A one percent confidence suited GTCRPM, as this allows the matrix estimator program to change the cell values. 2. Screenline file: Each individual traffic count was chosen to be a separate screenline. In other words, by providing this input, CUBE Analyst attempts to modify the trip table to match individual traffic counts. GTCRPM has a total of 2,985 links with counts. All counts were considered as independent screenlines. Since GTCRPM is a daily model, and directional counts should balance out on a daily basis, both links (AB/BA) on a two-way street were given the same screenline number. 3. Trip-ends file: The trip end file represents the production/attraction file. It was obtained simply as the row sum and column sum of the prior input matrix. A confidence value of 100 percent was given to the trip ends. Using this method, the size of the original trip table was preserved. If this is not used, it was found that CUBE Analyst significantly changes the matrix size to match volumes and counts. By using a trip end input file, original total origins and destinations were not modified. Instead, only the distribution of O-D pairs was changed. 4. Intercept file: This file is a CUBE Analyst program-specific file, similar to a path file. This file represents all the screenlines that are intercepted by an O-D pair in the prior matrix assignment to the model network. It should be noted that the GTCRPM trip table has multiple classes of tables, including DA, SR2, SR3+ and Trucks. However the model does not have counts corresponding to those classes to correct each individual trip tables. Hence, in order to prepare input to CUBE Analyst, trips for all classes were summed to one single trip table. Hence, the raw corrected matrix obtained from CUBE Analyst was an aggregated total trip matrix. This corrected matrix was again disaggregated into DA, SR2, SR3+ and Truck classes, using the original trip table s (individual/total) ratios Phased Implementation of CUBE Analyst Matrix Estimator The entire CUBE analyst application was setup in a feedback loop. The loop had a maximum of ten iterations. CUBE Analyst estimated an output matrix within each of the individual iterations. The

8 Page 4 output matrix was again input to the GTCRPM highway assignment to evaluate the model RMSE. An RMSE evaluation step at the regional level was added, and the resulting RMSE value was compared to the cut-off RMSE of 15 percent. The cut-off RMSE, in this case, is defined as the RMSE at which the feedback loop will be terminated. The loop will be terminated if either the RMSE criteria or the maximum iterations are met. Figure 2-1 presents the flow chart of CUBE Analyst implementation in GTCRPM model. Due to the memory issues faced in running CUBE Analyst when all traffic counts were used in one single step, phased input of screenlines was performed. This process was achieved by randomly selecting a set of screenlines for each iteration. To select random counts, random numbers with values between 0 and 1 were assigned to all links with counts in each of the iterations. All freeway and toll road links with counts were selected because of the importance of these links. Arterials with counts were selected if the random numbers were less than or equal to 0.3. This reduced the number of counts selected for screenline numbering to less than a thousand, and helped to reduce the memory required by Analyst. In each of the individual iterations, the Analyst-produced output matrix was fed back as an input matrix to the next iteration s Analyst step. Using this method, the Analyst output matrix was improved in the subsequent iteration. The iterative process was continued until either the RMSE of the assignment was less than 15 percent, or a maximum of 10 iterations were performed, whichever occurs first. Confidence levels of screenlines were set to 100 percent for Indian River, St. Lucie and Martin Counties. Confidence levels for screenlines on the external cordon, and for those counts at Brevard/Indian River and Palm Beach/Martin County lines, were also set to 100 percent. Confidence levels of screenlines in Palm Beach and Brevard counties were set to 50 percent. The reduced confidence levels for counts outside the Treasure Coast counties were used in order to provide higher importance to Treasure Coast counts, since the Treasure Coast Counties are the focus of the model Further Adjustments to Trip Tables and Production of Future Adjusted Trips CUBE Analyst reduced model RMSE values significantly and improved model statistics. The Analyst produced trip table was compared with the original trip table to ensure there are no large deviations of the cell values between the two. If there are large deviations observed, certain low and high cut off factors were used to control the analyst produced matrix. The question then arises is how to develop these cut off factors. These factors were developed just based on common sense to control the Analyst matrix, and the RMSE of assignment using this final trip table is verified. A few experiments were conducted in order to finally calibrate these empirical factors. The following rules were used to limit Analyst corrections. Let us define Analyst matrix cell values as Analyst, and original matrix cell values as Original, and adjusted cell values as Adjusted. Further adjustments to the cells are made only if: If Analyst < 0.1* Original, Adjusted=0.1*Original If Analyst > Minimum of (10*original, 50+original), Adjusted= minimum of (10*original, 50+original)

9 Page 5 Figure 2-1 Flow Chart of Implementation of CUBE Analysis Matrix Estimation in a Feedback Loop

10 Page 6 Finally, the adjusted future trip tables are produced using the following formulas: Growth = (Unadjusted future) (Unadjusted base) Adjusted future trips = (Analyst adjusted base trips) + Growth Model Application This section describes about how the adjustments to the base trip tables estimated by Analyst were transferred to the Analysis Years. The user does not need to run CUBE Analyst to make use of trip table adjustments. This was undertaken by the model developer, and the resulting modified trip table was input into Applications\Analyst folder. If the user intends to make use of CUBE Analyst trip table adjustment process, the catalog key Analyst should be set to 1. By default, this catalog key has a value of zero, signifying only traditional 4-step process will be implemented without any user input to this catalog key. Certain adjustments to model Flow Charts were made in this process. Step number 25 was added under the Mode application group, which produces the final adjusted trip tables for assignment. As can be seen in the figure below, Adjusted Future, and Adjusted Base are the newly created trip tables. These tables are produced in the general model stream, but will be used in the assignment only if user sets the Analyst catalog key to 1. If the Analyst catalog key is zero, the original trip table output from Step 24 will be used in the assignment. The GTCRPM3.3 model package comes with 4 scenarios as shown in the figure below. Scenarios that do not have Analyst in their names use trip tables produced in general model stream, and are not subjected to any further adjustments. Scenarios titled BASEAN, Y2030AN use Adjusted Future trip tables produced in Step 25.

11 Page Base Year Calibration and Future Year Scenario Results Tables 2-1 through 2-3 demonstrate the RMSE and Volume/Count ratios using the final calibrated Analyst Matrix. Significant reductions in RMSE were achieved using the Analyst produced trip table. At the same time, the Volume/Count ratios generally decreased for all counties, although remained close to 1.0. Figures 2-2 through 2-4 illustrate the screenlines in the GTCRPM model. Note that these screenlines are the original screenlines from the 4-step model, and are not the screenlines (actually counts) used in the Analyst matrix estimation. Figure 2-5A and Table 2-4A present the screenline results from Analyst version and Figure 2-5B and Table 2-4B show the screenline results from the original GTCRPM model. Figure 2-5A compares screenline-specific percent deviation from the counts to the maximum allowable percent deviation as indicated by NCHRP 255. This figure shows that all screenlines except one fall under the maximum allowable deviation curve. The results indicate that at the screenline level, the total volume and counts match closely. The improved screenline volume-to-count ratios also show the improvements in trip distribution. Similarly, Figure 2-6A presents the scatter plot of volumes and counts from the Analyst version of GTCRPM3.3. The figure shows improvements when compared to the original GTCRPM3.3 results shown in Figure 2-6B. Many data points in Figure 2-6A are closer to the 45-degree line, indicating that volumes closely match counts. Trip Length Frequency Distribution and Average trip length for GTCRPM Analyst version were compared against the original GTCRPM3.3 version. This check was performed to assess whether significant changes were made by CUBE Analyst to match volumes to counts. The average trip length of GTCRPM3.3 Analyst version is minutes, where as this statistic is minutes for GTCRPM3.3 original version. The trip length frequency comparisons are presented in Table 2-5. These statistics indicate that the new trip table produced by CUBE Analyst did not deviate much from the original trip table, and the trip making behavior of the region was not greatly distorted to match volumes and counts. Table 2-6 presents an illustration of link level volumes of I-95 segments in Treasure Coast Counties. The results of all four GTCRPM scenarios are presented for comparison purposes. Generally, the results appear to be reasonable. Table 2-7 illustrates the systemwide parameters as output by the four model scenarios.

12 Table 2-1 GTCRPM3.3 CUBE Analyst Version Percent Root Mean Square Summary by Volume Groups GTCRPM3.3 Year 2005 Palm Northern Maximum Volume Group Indian St. Brevard N* N* N* Martin N* Beach N* Palm Beach N* Region N* Allowable River Lucie (PB) N of SR 80 RMSE 1-5,000; 91.2% % % % % % % % 5,000-10,000; 24.9% % % % % % % % 10,000-20,000; 18.2% % % % % % % 1, % 20,000-30,000; 11.2% % 2 4.0% % % % % % 30,000-40,000; 10.5% 6 4.6% 2 3.7% % % % % 40,000-50,000; 4.7% % 8 4.7% % 50,000-60,000; 2.2% 2 2.2% 2 2.2% % 60,000-70, % % % 70,000-80,000; 14.1% % % % 80,000-90,000; 9.3% % 4 9.3% % 1-500,000; 18.3% % % % % 1, % % 2, % GTCRPM Version 3.3 (GTCRPM3.3) Additional Enhancements Technical Report Draft Page 8

13 Table 2-2 GTCRPM3.3 CUBE Analyst Version GTCRPM3.3 Volume to Count Ratios by Facility Type Groups GTCRPM3.3 Year 2005 Facility Palm Northern Description Indian Type (FT) Brevard N* N* St. Lucie N Martin N Beach N Palm Beach River (PB) N of SR 80 N Region N 1 Freeways Uninterrupted Roadway Higher Speed Interrupted Facility , ,947 6 Lower Speed Facility & Collector Ramps HOV facilities Toll Facilities All Facilities , ,985 N*: Total Number of counts GTCRPM Version 3.3 (GTCRPM3.3) Additional Enhancements Technical Report Draft Page 9

14 Table 2-3 GTCRPM3.3 CUBE Analyst Version Volume to Count Ratios by Area Type Groups GTCRPM3.3 Year 2005 Area Palm Northern Type Description Indian Brevard N N St. Lucie N Martin N Beach N Palm Beach N Region N (AT) River (PB) N of SR 80 1 CBD Fringe Residential OBD , ,038 5 Rural All Area Types , ,985 N*: Total Number of counts GTCRPM Version 3.3 (GTCRPM3.3) Additional Enhancements Technical Report Draft Page 10

15 Figure 2-2 Screenlines, Cutlines, and Cordons Locations in Brevard and Indian River Counties GTCRPM3.3 Page 11

16 Figure 2-3 Screenlines, Cutlines, and Cordons Locations in St. Lucie and Martin Counties GTCRPM3.3 Page 12

17 Figure 2-4 Screenlines, Cutlines, and Cordons Locations in Palm Beach County GTCRPM3.3 Page 13

18 Figure 2-5A Maximum Allowable Deviation in Total Screenline and Cutline Volumes GTCRPM3.3 Analyst Version Maximum Allowable Deviation of Screenlines and Cutlines 60 Percent Deviation of Total Scrrenline Volumes Count Total in Thousands NCHRP 255 Maximum Allowable Deviation GTCRPM3.3 Percent Deviation Figure 2-5B Maximum Allowable Deviation in Total Screenline and Cutline Volumes GTCRPM3.3 Original Version Maximum Allowable Deviation of Screenlines and Cutlines Page Percent Deviation of Total Scrrenline Volumes Count Total in Thousands NCHRP 255 Maximum Allowable Deviation GTCRPM3.3 Percent Deviation

19 Figure 2-6A Scatterplot of Assigned Volumes vs. Observed Counts GTCRPM3.3 Analyst Version Figure 2-6B Scatterplot of Assigned Volumes vs. Observed Counts GTCRPM3.3 Original Version Page 15

20 Table 2-4A Screenline/Cutline Results GTCRPM3.3 Analyst Version County/Region SL, CL& Corridor Number GTCRPM Enhanced Model Total Volume Total Count V/C % Diff in Total Volume SL,CL Description 2 365, , N-S N. of Yamato Rd 3 332, , N-S N. of CR , , N-S Crossing CR , , N-S N. of CR , , N-S S. of Indian town Rd 8 74,662 70, E-W Crossing Turnpike 9 150, , E-W Crossing I-95 PB County , , E-W Crossing Turnpike , , E-W Crossing CR , , E-W Crossing I , , E-W Crossing I ,153 95, E-W Crossing Turnpike , , E-W Crossing I , , E-W Crossing CR , , E-W Crossing , , N-S Turnpike Martin County , , N-S Crossing CR 708 Martin County , , E-W SR 76 St. Lucie County 22 33,582 36, E-W Crossing US 1 St. Lucie County , , N-S Crossing Midway Rd IR County 19 17,603 17, E-W Crossing I-95 IR County , , N-S Crossing Oslo Rd IR County 46 87,345 85, N-S Crossing SR 60 Brevard County , , E-W Crossing I-95 Brevard County , , N-S N. of SR 192 Treasure Coast Area , , E-W Causeways 21 74,228 80, E-W Causeways External Cordon , , Cordon - The Study Area 95 2,761,499 2,988, I-95 Palm Beach All 99 26,446,183 27,525, All others Page 16

21 Table 2-4B Screenline/Cutline Results GTCRPM3.3 Original Version County/Region SL, CL& Corridor Number GTCRPM Enhanced Model Total Volume Total Count V/C % Diff in Total Volume SL,CL Description 2 445, , N-S N. of Yamato Rd 3 371, , N-S N. of CR , , N-S Crossing CR , , N-S N. of CR , , N-S S. of Indian town Rd 8 93,037 70, E-W Crossing Turnpike 9 143, , E-W Crossing I-95 PB County , , E-W Crossing Turnpike , , E-W Crossing CR , , E-W Crossing I , , E-W Crossing I ,531 95, E-W Crossing Turnpike , , E-W Crossing I , , E-W Crossing CR , , E-W E-W Crossing , , N-S Turnpike Martin County , , N-S Crossing CR 708 Martin County , , E-W SR 76 St. Lucie County 22 44,542 36, E-W Crossing US 1 St. Lucie County , , N-S Crossing Midway Rd IR County 19 22,368 17, E-W Crossing I-95 IR County , , N-S Crossing Oslo Rd IR County 46 82,986 85, N-S Crossing SR 60 Brevard County , , E-W Crossing I-95 Brevard County , , N-S N. of SR 192 Treasure Coast Area , , E-W Causeways 21 94,336 80, E-W Causeways External Cordon , , Cordon - The Study Area 95 3,252,426 2,988, I-95 Palm Beach All 99 28,137,614 27,541, All others Page 17

22 Table 2-5 Trip Length Distribution Comparisons between GTCRPM3.3 Original & GTCRPM3.3 Analyst Versions GTCRPM3.3 Original Version GTCRPM 3.3 Analyst Version Trip Length %Trips Trip Length %Trips ============== ============== =========================== =============================== ===================== ====================== ============== ============ ======== ======= ===== ==== === == = = = = Total Trips: 6,359,690 Average Trip Length GTCRPM3.3 Original: minutes Average Trip Length GTCRPM3.3 Analyst: minutes GTCRPM Version 3.3 (GTCRPM3.3) Additional Enhancements Technical Report Draft Page 18

23 Table 2-6 Base and Future Year Results I-95 Corridor in Treasure Coast Area Seg Description CO 2005 Count GTCRPM3.3 Base GTCRPM3.3 VOL/CNT GTCRPM3.3A Base GTCRPM3.3A VOL/CNT GTCRPM GTCRPM3.3A MA/PB CL MA , , ,460 2 N of SW Bridge Rd MA 64,500 73, , , ,354 3 N of Kanner HWY MA , , , ,907 4 NW of High Meadow Ave MA , ,188 91,553 5 N of Martin HWY MA 54,622 39, , , ,984 6 N of MA/SL CL SL 54,622 39, , , ,036 7 N of Gatlin Blvd. SL 46,500 48, , , ,130 8 N of SW Juliet AVE SL 46,500 48, , , ,039 9 N of St. Lucie Blvd. SL 55,500 58, , , , N of Midway Rd. SL 59,900 58, , , , N of Sr 70/Okeechobee SL 52,084 52, , , , N of CR 68/Orange SL 50,500 46, , , , N of Cr 614/Indrio SL 39,500 33, , ,082 96, N of Oslo IR 39,500 33, , ,398 81, N of St HWY 60 IR 38,000 35, , ,699 66, N of Cr 512/ Fellsmere IR 39,616 31, , ,713 58,552 GTCRPM Version 3.3 (GTCRPM3.3) Additional Enhancements Technical Report Draft Page 19

24 Table 2-7 Base and Future Year Results Systemwide Model Parameters Parameter Base Future Base-Analyst Future-Analyst VMT 49,243,379 76,676,041 47,675,052 76,038,178 VHT 1,245,621 2,047,334 1,171,593 2,004,855 Original Speed (mph) Congested Speed (mph) Change in Speed (mph) Percent Change in Speed -5.76% -9.52% -5.03% -9.19% Number of Directional Links 16,899 18,081 16,899 18,081 Total Lane Miles 8,706 10,934 8,706 10,934 Total Directional Miles 5,588 6,285 5,588 6,285 Volume to Count Ratio 1.01 NA 0.96 NA Total Volume/Capacity Ratio LOS E Total Volume/Capacity Ratio LOS D Total Volume/Capacity Ratio LOS C Travel Time Value per vehicle hour (TTV) $9.94 $ $21.13 Vehicle Operating Cost(VOC) per Vehicle Mile $0.50 $ Page 20

25 3. External Trip Forecasting Enhancements 3.1 External Stations GTCRPM has 33 external stations (Figure 3-1). External station node numbers range from County-specific dummy stations are reserved for potential future incorporation of new external stations. The TAZ numbering scheme, presented in Table 3-1, shows the external station numbers in each county. Figure 3-1 External Stations in GTCRPM3.3 Page 21

26 Table 3-1 Internal and External Station Numbers in GTCRPM3.3 County Internal Zones Internal Dummy Zones External Zones Number Range Unused Number Range* Number Range Brevard Indian River St. Lucie Martin Palm Beach * Dummy zones are allocated for possible future applications. 3.2 External Trip Forecasting External trip modeling in GTCRPM3.3 involved following steps: 1. Identifying external station AADT for the analysis year: Identifying the external station AADT is simple for the base year, as it is extracted from the counts database. For future year 2035, a separate analysis was conducted to determine the growth rates at each external station. The growth rates were applied to the 2005 counts, to obtain 2035 forecasts. Various sources of data were used to determine this growth rate. The process is presented in Table Identifying EE trip percentages at each external station: Total EE trip percentages at each external station were obtained from the Statewide Model. These percentages were used in developing the EETARG_{year}{ALT}.dbf input file. 3. Preparing the seed percentage Matrix to perform External-External trip distribution: The initial EE seed percentage matrix was obtained from the Florida Statewide Model. This matrix was modified using local knowledge. Thirteen external stations with a significant number of EE trips were used in the EE seed matrix. Table 3-3 presents the new seed percentage matrix. Page 22

27 Table 3-2 External Station Counts Forecasting for 2035 Annual 2005 Percent Count 2035 Estimate G.R Source Comment Station County Description 3001 Brevard A1A 29,432 34, % CFRPM Brevard Riverside Dr (513) 24,240 27, % CFRPM Brevard Pineapple Ave % CFRPM Brevard Harbor City Blvd (1) 44,000 54, % CFRPM Brevard Stewart Rd % CFRPM Brevard Croton Rd 15,558 25, % CFRPM Brevard Wickham Rd (509) 37,453 42, % CFRPM Brevard John Rodes Blvd (511) 13,622 20, % CFRPM Brevard I-95 77, , % CFRPM Brevard Eau Galie Blvd % CFRPM Brevard SR ,000 15, % CFRPM Indian River SR60 5,022 9, % BEBR, IR/OSC 3022 Indian River Tpke (SR 91) 27,552 52, % BEBR, IR/OSC SWM Model Projections were verified. It was noted many 3031 St Lucie CR68 2,314 4, % BEBR, SL/OKEE arterials did not load well. Hence, 3032 St Lucie CR70 4,946 8, % BEBR, SL/OKEE annual growth rate was applied 3033 St Lucie CR % BEBR, SL/OKEE based on recent BEBR population 3041 Martin SR710 7,684 9, % BEBR, MA/OKEE projections Martin US441 5,352 6, % BEBR, MA/OKEE 3051 Palm Beach Lox Rd (827) % SERPM NEEDS 3052 Palm Beach Nob Hill Rd % SERPM NEEDS 3053 Palm Beach University Drive % SERPM NEEDS 3054 Palm Beach Riverside Dr % SERPM NEEDS 3055 Palm Beach US ,000 72, % University Dr. Study (Alt D) University Drive study considered 3056 Palm Beach Lyons Rd 34,000 61, % University Dr. Study (Alt D) detailed analysis on these roadways between Broward/Palm Beach county lines 3057 Palm Beach Tpke (SR 91) 98, , % University Dr. Study (Alt D) 3058 Palm Beach Powerline Rd (845) 35,500 52, % SERPM NEEDS 3059 Palm Beach Military Trail (809) 27,500 29, % SERPM NEEDS 3060 Palm Beach I , , % SERPM NEEDS 3061 Palm Beach Dixie Hwy 15,800 42, % SERPM NEEDS 3062 Palm Beach US 1/SR 5 30,500 39, % SERPM NEEDS 3063 Palm Beach A1A 10,500 12, % SERPM NEEDS 3064 Palm Beach US 27 15,654 19, % BEBR, PB/HENDRY Model Under loads in base year 3065 Palm Beach/Broward US 27 9,600 16, % SERPM NEEDS GTCRPM Version 3.3 (GTCRPM3.3) Additional Enhancements Technical Report Draft Page 23

28 Table 3-3 External Seed Percentage Matrix for GTCRPM3.3 (Input file EE_SEED_NEW.dbf) TOTAL % 0% 90% 1% 4% 5% 0% 0% 0% 0% 0% 0% 0% 100% % 0% 90% 2% 3% 5% 0% 0% 0% 0% 0% 0% 0% 100% % 42% 0% 0% 0% 41% 0% 0% 0% 0% 0% 0% 0% 100% % 8% 0% 0% 0% 90% 0% 0% 0% 0% 0% 0% 0% 100% % 7% 5% 1% 1% 81% 1% 0% 0% 1% 1% 0% 0% 100% % 1% 21% 11% 17% 0% 11% 0% 0% 24% 14% 0% 0% 100% % 1% 6% 0% 0% 92% 0% 0% 0% 0% 0% 0% 0% 100% % 0% 0% 0% 0% 0% 0% 0% 0% 70% 30% 0% 0% 100% % 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 100% 100% % 0% 0% 0% 0% 24% 0% 76% 0% 0% 0% 0% 0% 100% % 0% 0% 0% 0% 32% 0% 68% 0% 0% 0% 0% 0% 100% % 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 100% 100% % 0% 0% 0% 0% 0% 0% 0% 17% 0% 0% 83% 0% 100% GTCRPM Version 3.3 (GTCRPM3.3) Additional Enhancements Technical Report Draft Page 24

29 4. Modifying CUBE Voyager (CV) application to develop balanced EE matrix: The CV application was modified to develop reasonable EE distribution matrix. The FRATAR model was modified to achieve percent distribution shown in the EE seed matrix (Shown in Table 3-3) and the target EE trips (shown in Table 3-4). 5. Extracting total EE trips forecasted from the model to identify the total EI/IE trips: This was achieved by subtracting the EE trips from the station AADT, outside the model application. 6. Developing Zdata4: The EI/IE trips from step 5 are in vehicle trip format. These are converted to person trips by an average vehicle trip to person trip converting factor of 3.4. This conversion factor is an initial factor to obtain EI/IE person trips. The EI/IE values were adjusted to match external model volumes and counts using iterative model runs. Few model runs were performed iteratively by adjusting the IE/EI trips to achieve station-specific model volumes and AADT matching within 10 percent error. 3.3 User Guidelines for Developing Intermediate Year Data The input files needed for EE trip forecasting are: EE_SEED_New.dbf EETARG_{year}{ALT}.dbf, where {year} is the catalog key with 2 digit code for year of analysis, {ALT} is catalog key for alternative code Zdata4.{Year}{ALT}.dbf, where {year} is the catalog key with 2 digit code for year of analysis, {ALT} is catalog key for alternative code EE_SEED_NEW.dbf should not be modified unless the user intends to modify the seed matrix percentages. The other two input files can be developed by straight interpolation between 2005 and Page 25

30 Table 3-4 EE and IE/EI Targets in GTCRPM3.3 IE-P IE-A IE-P_35 IE-A_35 Fraction EE_05_ IE/EI_05 EE_35_ IEEI_35_ Stn AADT (Approx (Approx AADT_35 (Approx (Approx EE Model_2way Veh Trips Model_2way Veh trips 0.7*F) 0.7*F) 0.7*L) 0.7*L) , ,723 2,047 32,677 23,654 23, , ,259 4,784 22,475 16,316 16, , ,301 12,146 42,156 26,900 26, , ,669 25, ,728 19,296 19, , ,149 3,238 38,911 27,585 27, , ,488 5,893 14,595 10,261 10, , ,228 31, ,814 90,611 90, , ,012 4,404 10,608 7,558 7, , , ,482 8,772 8, , ,018 28,858 23,160 22,188 22, , , ,008 2,733 2, , , ,566 5,560 5, , , ,643 7,233 7, , ,717 1,910 4,806 4,137 4, , , ,280 51,629 51, , , ,720 36,306 36, , ,000 29, , , , , , ,238 35,316 35, , , ,487 21,445 21, , ,854 12, , , , , , ,730 27,045 27, , , ,699 27,333 27, , , ,472 9,139 9, , ,928 9,384 10,544 8,579 8, , ,500 11,294 5,206 6,374 6,374 Page 26

31 4. Additional Reporting Features Several additional reporting features were incorporated into GTCRPM. The standard FSUTMS HEVAL FORTRAN program was removed from the model stream and CUBE Voyager based HEVAL reporting features were added. This will facilitate easy modification of reports and convenient debugging of any errors encountered due to changes performed to model networks. The primary reference of CV scripts and reporting features is the SERPM6.5 new reporting features (Reference 1). Users normally do not need to modify the lookup tables or input files to run the model. All of these input files are largely self explanatory. If a detailed description of the lookup tables and format of new input files is needed, the user is referred to the Reference 1. Three different output files are produced in the Voyager based HEVAL process. Overall Summary (HEVAL-AD-OverallSmry-GTC.PRN), Detailed Summary (HEVAL-AD-Detailed-GTC.PRN), and Screenline Summary (HEVAL-AD-ScreenSmry-GTC.PRN) are the three outputs of HEVAL process. The reports are produced for each County, and for the GTCRPM region. In addition, Windowed Area Reporting is also made available in the model, for the predefined subarea network. In addition to standard HEVAL reports, Travel Time Values (TTV) and Vehicle Operating Costs (VOC), reports by posted speed bins are also produced in the new HEVAL reports. Furthermore, a few sample reports using the CUBE reports features were also created. These reports will not be created as part of the model run stream. The user should double-click these.rep files in order to produce and display these reports. The included reports are: - a scatter plot of Volumes to counts, and pie charts of VMT and VHT by County. Users should note that additional reports can be also be setup interactively, using these CUBE reports feature. However, it should be noted that the attributes on which reports to be generated must be present in the loaded networks used in the CUBE reports generation. Note that new capacity values, reported by the 2009 Florida LOS handbook, were used in reporting LOS C, LOS D and LOS E capacities. The model validation was based on the 2002 LOS handbook, and was not altered during this work. In order to perform reporting by using the 2009 LOS handbook, a capacity adjustment step was added subsequent to the model s highway assignment step. Several new input files and catalog keys and network attributes were added while developing Voyager based HEVAL scripts. 4.1 Input Files Following new input files are added to model s input folder. FTRptGrp_TTVLkup.dbf Facility type specific travel time values ATRptLabels.dbf Area type labels. Page 27

32 VOCSpdLkup.dbf- Vehicle operating cost values lookup table. FRWYPCE.DBF- Capacity (LOS-C, D, E) lookup table for Freeways. HWYPCE.DBF- Capacity (LOS-C, D, E) lookup table for Uninterrupted Roadways. LOFFPCE.DBF- Capacity (LOS-C,D,E) lookup table for Loop Off Ramps. LONPCE.DBF -Capacity (LOS-C, D, E) lookup table for Loop On Ramps. LOWPCE.DBF- Capacity (LOS-C, D, E) lookup table for Low Speed Roadways & Collectors (posted speed <35mph). OFFPCE.DBF -Capacity (LOS-C, D, E) lookup table for Off Ramps. 21 ONPCE.DBF- Capacity (LOS-C, D, E) lookup table for On Ramps. TOFFPCE.DBF- Capacity (LOS-C, D, E) lookup table for Toll Off Ramps. TONPCE.DBF -Capacity (LOS-C, D, E) lookup table for Toll Off Ramps. 4.2 Network Attributes CNT_FLG: This is an optional attribute for running the model. It is needed only if the model developer wants to rerun the Analyst matrix estimation. A value of 1 indicates external cordon counts; a value of 2 indicates Palm Beach/Martin County line counts; and value of 3 indicates Brevard/Indian River County line counts. DETAILNET: A value of 1 indicates links included in the SUBAREA evaluation. 4.3 Catalog Keys Conf_p: Defaults to 100 indicates 100 percent confidence in production trip end values, input to CUBE Analyst. Conf_p: Defaults to 100 indicates 100 percent confidence in attraction trip end values, input to CUBE Analyst Analyst: Defaults to zero in non-analyst scenarios. If a value of 1 is input in this catalog key, Analyst corrected trip table process will be used. Inflation_TTV: Inflation value for estimating future year travel time values AnalysisYr_TTV_VOC: Indicates the analysis year for estimating TTV and VOC BaseYr_TTV_VOC: Indicates the base year for estimating TTV and VOC Inflation_VOC: Inflation value for estimating future year travel time values CountYear: Indicates the year of traffic count collection Userdefinedsubarea: Defaults to zero, indicating no subarea outputs will be produced in general model stream. A value of 1 triggers the SUBEVAL process to run, producing the Window Area HEVAL reports 4.4 Window Area HEVAL Instructions The following instructions should be used to produce Window Area Evaluation statistics. These evaluation statistics can be useful in determining the model performance measures within a specified window area, as opposed to the entire GTCRPM region or the County-specific evaluations produced by the default model run. The user must identify the subarea within which HEVAL evaluations should be performed. The subarea code is given to the input network attribute Page 28

33 DETAILNET. All links within the subarea for which special evaluation should be performed must be coded with a value of 1. This can be achieved by the below instructions: 1. Edit Input network (Red colored box "Starting Network - Step 3 of main application) 2. Define/Save or Restore your Polygon of Subarea [Polygon - New/Save or - Restore] 3. Void "DetailNet" link attribute of any pre-defined subarea (DetailNet=0; Use Set1: Void DetailNet) [Link-Compute-Apply/ok-Close] 4. Populate/compute "DetailNet" link attribute with "1" (DetailNet=1; Use Set2: Populate DetailNet) [Link-Compute-Apply/ok-Close] (use Applies to as "All items inside and crossing polygon NOW") 5. Save the network Note: Make sure that the 'UserDefinedSubarea' catalog key is set to 1 for windowed evaluation. The user can also identify the subarea in the output network that is input to the SUBEVAL process. 4.5 Select Link Analysis Instructions 1. Code {RunSELLINK} Catalog Key with 1. This is 0 by default in a regular model run 2. Code the links on which Select Link Analysis should be run in {SELLINK} catalog key. 3. Examples: {SELLINK} = This means, Select link will be run on directional link If direction is also needed in the analysis, this can be coded as , OR, simply * also represents non-directional analysis. 4. This process handles a logical OR condition of select link. That is, if the user specifies links , in the {SELLINK} catalog key, all trips that use any or both of the two links will be selected. 5. If an AND condition is required, the Highway Assignment script should be modified to reflect (L= &&L= ). Page 29

34 5. References Southeast Florida Regional Planning Model 6.5 New Time-of-Day and 24 Hour Subarea Models TECHNICAL REPORT Model Adjustment and Application Guideline, FDOT-District 4, prepared by The Corradino Group, June 2010 Assessment of Southeast Florida Road User Costs, Final Report, FDOT District 4, Prepared by Cambridge Systematics, February Southeast Florida Road and Transit User Cost Study, Draft Report, FDOT District 4, Prepared by Cambridge Systematics, March A Revised RUC Calculator Spreadsheet Program (RUC_TUC_Calculators_final.xls; see {VOC_Lookup_Table} tab), FDOT District 4, Updated program by The Corradino Group, March Page 30