technical memorandum - stancog
TRANSCRIPT
VISION THAT MOVES YOUR COMMUNITY
PLEASANTON ♦ SAN JOSE ♦ SANTA ROSA ♦ SACRAMENTO ♦ FRESNO Corporate Office: 4305 Hacienda Drive, Suite 550, Pleasanton, CA 94588
Phone: 925.463.0611 Fax: 925.463.3690 www.TJKM.com DBE #40772 ♦ SBE #38780
Technical Memorandum Date: May 3, 2018
To: Ryan Niblock Senior Regional Planner San Joaquin Council of Governments
From:
Lawrence Liao, TJKM
Subject: Revised Draft Technical Memorandum for VMIP2 TCM 2015 Update
INTRODUCTION The objective of the San Joaquin Valley Model Improvement Plan, Phase 2 (“VMIP2”) Three-County Model (TCM) 2015 Update project is to create a 2015 base year model to support the development of 2018 regional transportation plans for San Joaquin Council of Governments (SJCOG), Stanislaus Council of Governments (StanCOG), and Merced County Association of Governments (MCAG). The starting point of the TCM 2015 Update project was the TCM 2008 Model delivered by Fehr and Peers on March 3, 2017. The model development report can be found on StanCOG’s website. This TCM_Base08_20170303 Model was a work-in-progress which was not fully calibrated and validated to 2008 conditions. The assumptions and methodologies of the TCM_Base08_20170303 Model and the validation statistics can be found in the following Appendices.
• Appendix A-- SJV MIP - Highway Validation - TCM_SanJoaquin_20170224.pdf • Appendix B -- SJV MIP - Highway Validation - TCM_Stanislaus_20170224.pdf • Appendix C -- SJV MIP - Highway Validation - TCM_Merced_20170224.pdf
An initial review was conducted to identify any issues with the input data and model scripts of the TCM_Base08_20170303 Model. After identified issues were addressed in 2008 Model, the calibration/validation efforts were focused on the 2015 Model. The identified issues and 2015 validation performance results are provided below.
VISION THAT MOVES YOUR COMMUNITY
ISSUES IDENTIFIED IN TCM_BASE08_20170303 MODEL The input data and model scripts were reviewed extensively before model calibration/validation began. Some of the issues identified in the TCM_Base08_20170303 Model are described below: 1. Future project links with FACTYP=0 due to scripting and project coding Based on the current logic in the master network processing script (“NPNET00A.S”), some future project links will have FACTYP=0, when
• IMP2_PRJYR>0 AND IMP2_FACTYP=0
Facility Type is a key variable in determining the BPR function parameters for each link, and is used in the vehicle-miles traveled (VMT) classification. This issue was causing problems in future year traffic assignment, as well as classification of VMT by facility type for air quality conformity. 2. VMT Reporting was incorrect because no Airbasin defined in TAZ data The VMT reporting script assumes Airbasin is in the second field of the TCM08_Base_TAZData.csv file
VISION THAT MOVES YOUR COMMUNITY
However, the second field in the TAZ data file contains a different data.
3. Terminal Time variables were referenced incorrectly in various scripts The scripts assume PTERM and ATERM are in fields #13 and #14, e.g.,
VISION THAT MOVES YOUR COMMUNITY
But, they are in fields #14 and #15, respectively.
4. Transit skimming scripting error for BUS only mode The BUS only mode has zero In-Vehicle Travel Time due to a scripting error.
All the issues identified in the input data and mode scripts have been addressed in the VMIP2 TCM 2015 update, accordingly.
2015 MODEL CALIBRATION/VALIDATION No assumptions nor methodologies in the original VMIP2 TCM development were modified in this project. The main goal of the 2015 model update effort was to calibrate the model VMT to the 2015 HMPS VMT by resolving the issues found in input data and scripting. The 2015 model has been validated up to the Mode Choice step, currently. The 2008 non-highway validation workbooks were updated for 2015 model calibration and validation. Each non-highway validation workbook includes the following checks:
VISION THAT MOVES YOUR COMMUNITY
• Vehicle availability was validated using Census vehicle ownership cross-classified by household size and income.
• Trip generation was validated for trip productions, attractions, and trip balancing.
o Trip production: A comparison of model total trips by purpose and observed totals from the expanded 2012 CHTS data. A secondary comparison, if needed, can be HBW trips from more aggregate sources such as the CTPP or NHTS. These sources are used with caution since they report ‘usual’ workplace locations and are not directly comparable to model generated workplace locations. Convert person trip rates to ITE rates using Ave Veh Occ by purpose.
o Trip attraction: Compare HBW attractions to total jobs in zone, range of 1.2-1.5 HBW attractions per employee in zone (source TFResource.org)
o Trip balancing: PA totals within +-10% of totals and totals by purpose
• For trip distribution models: The gravity model and any associated friction factors (k-factors) were calibrated iteratively to match average trip lengths by purpose and trip length frequencies by purpose are compared with the household travel survey. As a secondary method, the model volumes are compared to observed traffic volumes or observed survey data of vehicle volumes.
• For mode choice models, observed transit ridership (when available) can be compared against trip tables and the model mode shares for validation. As a secondary method, the mode shares, developed by pooling all SJV households is compared to the local mode shares observed in the CHTS.
The 2015 non-highway validation summaries and highlights can be found in the next section. For Trip Generation, the average person trips per household was compared to the target from 2012 CHTS for each county. The 2015 average person trips per household are consistent with the 2008 model results. For example, the 2015 average person trips per household for SJCOG is 8.9 compared to 9.0 in the TCM_Base08_20170303 Model. The 2015 average person trips per household for each county is shown below:
VISION THAT MOVES YOUR COMMUNITY
VISION THAT MOVES YOUR COMMUNITY
The Trip Distribution validation improved significantly compared to the TCM_Base08_20170303 Model. For example, the following graphic shows a comparison of SJCOG Trip Distribution results between TCM_Base08_20170303 and 2015 models.
The Trip Distribution validation results for StanCOG and MCAG are shown below:
VISION THAT MOVES YOUR COMMUNITY
The mode share validation, especially for Walk, improved significantly compared to the TCM_Base08_20170303 Model. For example, the following graphic shows a comparison of SJCOG mode choice by purpose results between TCM_Base08_20170303 and 2015 models.
The mode share validation results for StanCOG and MCAG are shown below:
VISION THAT MOVES YOUR COMMUNITY
The Caltrans 2015 HPMS VMT used as the validation target is shown below:
Note: StanCOG was referred to as SAAG (Stanislaus Area Association of Governments) in the HPMS table
The 2015 VMT validation results for all three counties are shown below:
VISION THAT MOVES YOUR COMMUNITY
As presented, the static validation for 2015 Trip Generation, Trip Distribution and Mode Choice models have all improved significantly compared to the previous 2008 model validation results. And, the 2015 model VMT for all three counties are within 3% of the 2015 HMPS VMT.
VISION THAT MOVES YOUR COMMUNITY
APPENDICES
• Appendix A-- SJV MIP - Highway Validation - TCM_SanJoaquin_20170224.pdf • Appendix B -- SJV MIP - Highway Validation - TCM_Stanislaus_20170224.pdf • Appendix C -- SJV MIP - Highway Validation - TCM_Merced_20170224.pdf
.
2/24/17 6:29 PM
DAILY Assignment Model/Count by ADT Volume Groups RMSE by ADT Volume Groups
Model/Count Ratio = 1.04 Link Volume M/C Link Volume %RMSE FHWA M/C
Percent Within Caltrans Maximum Deviation = 49% > 75% > 50,000 1.53 > 50,000 53% < 21% 1.54
Percent Root Mean Square Error = 67% < 30% 25,000 - 49,999 #DIV/0! 25,000 - 49,999 #DIV/0! < 22% #N/A
Correlation Coefficient = 82% > 0.88 10,000 - 24,999 0.79 10,000 - 24,999 37% < 25% 1.52
%of Screenlines Within Caltrans Standard Dev. = 85% 100% 5,000 - 9,999 1.08 5,000 - 9,999 55% < 29% 0.99
Externals M/C Ratio = 2,500 - 4,999 1.61 2,500 - 4,999 116% < 36% 0.45
Externals % RMSE = 1,000 - 2,499 1.12 1,000 - 2,499 94% < 47%
Total Count 196 < 1,000 0.39 < 1,000 77% < 60%
Link Within Deviation 97
Link Outside Deviation 99
Remaining Total Needed
50 147
AM Peak Period ( 6 - 9 AM) MD Peak Period ( 10 AM - 2 PM) AM Peak Hour ( 7- 8 AM)
Model/Count Ratio = 1.43 Model/Count Ratio = 1.03 Model/Count Ratio = 0.77 Freeway Streets
Percent Within Caltrans Maximum Deviation = 43% > 75% Percent Within Caltrans Maximum Deviation = 26% > 75% Percent Within Caltrans Maximum Deviation = 45% > 75% 1.54 1.01
Percent Root Mean Square Error = 93% < 30% Percent Root Mean Square Error = 106% < 30% Percent Root Mean Square Error = 81% < 30% 1.86 1.41
Correlation Coefficient = 0.82 > 0.88 Correlation Coefficient = 0.78 > 0.88 Correlation Coefficient = 0.73 > 0.88 1.80 0.99
%of Screenlines Within Caltrans Standard Dev. = 77% 100% %of Screenlines Within Caltrans Standard Dev. = 77% 100% %of Screenlines Within Caltrans Standard Dev. = 75% 100% 1.92 1.16
Externals M/C Ratio = Externals M/C Ratio = Externals M/C Ratio = 0.60 0.54
Externals % RMSE = Externals % RMSE = Externals % RMSE = 1.78 0.71
Total Count 196 Total Count 196 Total Count 196 2.18 1.39
Link Within Deviation 84 Link Within Deviation 50 Link Within Deviation 88
Link Outside Deviation 112 Link Outside Deviation 146 Link Outside Deviation 108
Remaining Total Needed Remaining Total Needed Remaining Total Needed
63 147 97 147 59 147
PM Peak Period ( 3 - 7 PM) Off Peak Period ( 8 PM - 5 AM) PM Peak Hour ( 5 - 6 PM)
Model/Count Ratio = 1.20 Model/Count Ratio = 0.55 Model/Count Ratio = 1.43
Percent Within Caltrans Maximum Deviation = 49% > 75% Percent Within Caltrans Maximum Deviation = 21% > 75% Percent Within Caltrans Maximum Deviation = 41% > 75%
Percent Root Mean Square Error = 80% < 30% Percent Root Mean Square Error = 105% < 30% Percent Root Mean Square Error = 93% < 30%
Correlation Coefficient = 0.76 > 0.88 Correlation Coefficient = 0.75 > 0.88 Correlation Coefficient = 0.76 > 0.88
%of Screenlines Within Caltrans Standard Dev. = 92% 100% %of Screenlines Within Caltrans Standard Dev. = 85% 100% %of Screenlines Within Caltrans Standard Dev. = 83% 100%
Externals M/C Ratio = Externals M/C Ratio = Externals M/C Ratio =
Externals % RMSE = Externals % RMSE = Externals % RMSE =
Total Count 196 Total Count 196 Total Count 196
Link Within Deviation 97 Link Within Deviation 42 Link Within Deviation 80
Link Outside Deviation 99 Link Outside Deviation 154 Link Outside Deviation 116
Remaining Total Needed Remaining Total Needed Remaining Total Needed
50 147 105 147 67 147
MD Peak Period ( 10 AM - 2 PM)
PM Peak Period ( 3 - 7 PM)
Off Peak Period ( 8 PM - 5 AM)
AM Peak Hour ( 7- 8 AM)
PM Peak Hour ( 5 - 6 PM)
Highway
Expressway
Time Period Analyzed
DAILY Assignment
AM Peak Period ( 6 - 9 AM)
Appendix A - San Joaquin Valley Model Improvement Project (San
Joaquin Valley MIP) One-Way Volume Model Validation Results
Three County Model - San Joaquin
Arterial
Collector
ADT Model/Count by Functional Class
Model/CountFreeway Traffic vs. Local Traffic
Link Volume
Freeway
0
5,000
10,000
15,000
20,000
25,000
30,000
35,000
40,000
0 5,000 10,000 15,000 20,000 25,000 30,000 35,000 40,000
Mo
del
Count
ADT Model vs. Count
Upper Bound
No Deviation
Lower Bound
"Model versus Count"
0
1,000
2,000
3,000
4,000
5,000
6,000
7,000
8,000
9,000
10,000
0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000 10,000
Mo
del
Count
ADT Model vs. Count - Locations with Less Than 10,000 Vehicles/Day
Upper Bound
No Deviation
Lower Bound
"Model versus Count"
.
2/24/17 6:30 PM
DAILY Assignment Model/Count by ADT Volume Groups RMSE by ADT Volume Groups
Model/Count Ratio = 1.47 Link Volume M/C Link Volume %RMSE FHWA M/C
Percent Within Caltrans Maximum Deviation = 32% > 75% > 50,000 1.57 > 50,000 62% < 21% 1.62
Percent Root Mean Square Error = 89% < 30% 25,000 - 49,999 2.02 25,000 - 49,999 102% < 22% #N/A
Correlation Coefficient = 97% > 0.88 10,000 - 24,999 1.11 10,000 - 24,999 34% < 25% 1.44
%of Screenlines Within Caltrans Standard Dev. = 85% 100% 5,000 - 9,999 1.00 5,000 - 9,999 49% < 29% 0.79
Externals M/C Ratio = 2,500 - 4,999 #DIV/0! 2,500 - 4,999 #DIV/0! < 36% 0.75
Externals % RMSE = 1,000 - 2,499 #DIV/0! 1,000 - 2,499 #DIV/0! < 47%
Total Count 47 < 1,000 #DIV/0! < 1,000 #DIV/0! < 60%
Link Within Deviation 15
Link Outside Deviation 32
Remaining Total Needed
20 35
AM Peak Period ( 6 - 9 AM) MD Peak Period ( 10 AM - 2 PM) AM Peak Hour ( 7- 8 AM)
Model/Count Ratio = 1.82 Model/Count Ratio = 4.24 Model/Count Ratio = 1.47 Freeway Streets
Percent Within Caltrans Maximum Deviation = 8% > 75% Percent Within Caltrans Maximum Deviation = 0% > 75% Percent Within Caltrans Maximum Deviation = 27% > 75% 1.62 0.89
Percent Root Mean Square Error = 180% < 30% Percent Root Mean Square Error = 531% < 30% Percent Root Mean Square Error = 119% < 30% 1.84 0.59
Correlation Coefficient = 0.98 > 0.88 Correlation Coefficient = 0.68 > 0.88 Correlation Coefficient = 0.96 > 0.88 3.78 47.33
%of Screenlines Within Caltrans Standard Dev. = 77% 100% %of Screenlines Within Caltrans Standard Dev. = 77% 100% %of Screenlines Within Caltrans Standard Dev. = 75% 100% 2.64 37.28
Externals M/C Ratio = Externals M/C Ratio = Externals M/C Ratio = 1.50 18.04
Externals % RMSE = Externals % RMSE = Externals % RMSE = 1.49 0.70
Total Count 12 Total Count 12 Total Count 11 1.67 0.53
Link Within Deviation 1 Link Within Deviation 0 Link Within Deviation 3
Link Outside Deviation 11 Link Outside Deviation 12 Link Outside Deviation 8
Remaining Total Needed Remaining Total Needed Remaining Total Needed
8 9 9 9 5 8
PM Peak Period ( 3 - 7 PM) Off Peak Period ( 8 PM - 5 AM) PM Peak Hour ( 5 - 6 PM)
Model/Count Ratio = 3.04 Model/Count Ratio = 1.68 Model/Count Ratio = 1.66
Percent Within Caltrans Maximum Deviation = 8% > 75% Percent Within Caltrans Maximum Deviation = 42% > 75% Percent Within Caltrans Maximum Deviation = 18% > 75%
Percent Root Mean Square Error = 305% < 30% Percent Root Mean Square Error = 96% < 30% Percent Root Mean Square Error = 154% < 30%
Correlation Coefficient = 0.71 > 0.88 Correlation Coefficient = 0.68 > 0.88 Correlation Coefficient = 0.98 > 0.88
%of Screenlines Within Caltrans Standard Dev. = 92% 100% %of Screenlines Within Caltrans Standard Dev. = 85% 100% %of Screenlines Within Caltrans Standard Dev. = 83% 100%
Externals M/C Ratio = Externals M/C Ratio = Externals M/C Ratio =
Externals % RMSE = Externals % RMSE = Externals % RMSE =
Total Count 12 Total Count 12 Total Count 11
Link Within Deviation 1 Link Within Deviation 5 Link Within Deviation 2
Link Outside Deviation 11 Link Outside Deviation 7 Link Outside Deviation 9
Remaining Total Needed Remaining Total Needed Remaining Total Needed
8 9 4 9 6 8
Appendix B - San Joaquin Valley Model Improvement Project (San
Joaquin Valley MIP) One-Way Volume Model Validation Results
Three County Model - Stanislaus
Arterial
Collector
ADT Model/Count by Functional Class
Model/CountFreeway Traffic vs. Local Traffic
Link Volume
Freeway
Highway
Expressway
Time Period Analyzed
DAILY Assignment
AM Peak Period ( 6 - 9 AM)
MD Peak Period ( 10 AM - 2 PM)
PM Peak Period ( 3 - 7 PM)
Off Peak Period ( 8 PM - 5 AM)
AM Peak Hour ( 7- 8 AM)
PM Peak Hour ( 5 - 6 PM)
0
5,000
10,000
15,000
20,000
25,000
30,000
35,000
40,000
0 5,000 10,000 15,000 20,000 25,000 30,000 35,000 40,000
Mo
del
Count
ADT Model vs. Count
Upper Bound
No Deviation
Lower Bound
"Model versus Count"
0
1,000
2,000
3,000
4,000
5,000
6,000
7,000
8,000
9,000
10,000
0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000 10,000
Mo
del
Count
ADT Model vs. Count - Locations with Less Than 10,000 Vehicles/Day
Upper Bound
No Deviation
Lower Bound
"Model versus Count"
.
2/24/17 6:26 PM
DAILY Assignment Model/Count by ADT Volume Groups RMSE by ADT Volume Groups
Model/Count Ratio = 1.26 Link Volume M/C Link Volume %RMSE FHWA M/C
Percent Within Caltrans Maximum Deviation = 65% > 75% > 50,000 #DIV/0! > 50,000 #DIV/0! < 21% 1.47
Percent Root Mean Square Error = 75% < 30% 25,000 - 49,999 1.83 25,000 - 49,999 83% < 22% #N/A
Correlation Coefficient = 93% > 0.88 10,000 - 24,999 1.01 10,000 - 24,999 29% < 25% 0.93
%of Screenlines Within Caltrans Standard Dev. = 85% 100% 5,000 - 9,999 0.88 5,000 - 9,999 29% < 29% #N/A
Externals M/C Ratio = 2,500 - 4,999 1.18 2,500 - 4,999 42% < 36% #N/A
Externals % RMSE = 1,000 - 2,499 1.02 1,000 - 2,499 35% < 47%
Total Count 43 < 1,000 #DIV/0! < 1,000 #DIV/0! < 60%
Link Within Deviation 28
Link Outside Deviation 15
Remaining Total Needed
4 32
AM Peak Period ( 6 - 9 AM) MD Peak Period ( 10 AM - 2 PM) AM Peak Hour ( 7- 8 AM)
Model/Count Ratio = 1.90 Model/Count Ratio = 1.67 Model/Count Ratio = 2.28 Freeway Streets
Percent Within Caltrans Maximum Deviation = 26% > 75% Percent Within Caltrans Maximum Deviation = 21% > 75% Percent Within Caltrans Maximum Deviation = 35% > 75% 1.47 0.93
Percent Root Mean Square Error = 158% < 30% Percent Root Mean Square Error = 260% < 30% Percent Root Mean Square Error = 197% < 30% 2.24 1.43
Correlation Coefficient = 0.83 > 0.88 Correlation Coefficient = 0.92 > 0.88 Correlation Coefficient = 0.81 > 0.88 2.03 1.05
%of Screenlines Within Caltrans Standard Dev. = 77% 100% %of Screenlines Within Caltrans Standard Dev. = 77% 100% %of Screenlines Within Caltrans Standard Dev. = 75% 100% 1.46 0.94
Externals M/C Ratio = Externals M/C Ratio = Externals M/C Ratio = 0.42 0.35
Externals % RMSE = Externals % RMSE = Externals % RMSE = 2.81 1.51
Total Count 43 Total Count 43 Total Count 37 1.88 1.17
Link Within Deviation 11 Link Within Deviation 9 Link Within Deviation 13
Link Outside Deviation 32 Link Outside Deviation 34 Link Outside Deviation 24
Remaining Total Needed Remaining Total Needed Remaining Total Needed
21 32 23 32 15 28
PM Peak Period ( 3 - 7 PM) Off Peak Period ( 8 PM - 5 AM) PM Peak Hour ( 5 - 6 PM)
Model/Count Ratio = 1.24 Model/Count Ratio = 0.38 Model/Count Ratio = 1.55
Percent Within Caltrans Maximum Deviation = 67% > 75% Percent Within Caltrans Maximum Deviation = 14% > 75% Percent Within Caltrans Maximum Deviation = 38% > 75%
Percent Root Mean Square Error = 83% < 30% Percent Root Mean Square Error = 164% < 30% Percent Root Mean Square Error = 117% < 30%
Correlation Coefficient = 0.83 > 0.88 Correlation Coefficient = 0.79 > 0.88 Correlation Coefficient = 0.89 > 0.88
%of Screenlines Within Caltrans Standard Dev. = 92% 100% %of Screenlines Within Caltrans Standard Dev. = 85% 100% %of Screenlines Within Caltrans Standard Dev. = 83% 100%
Externals M/C Ratio = Externals M/C Ratio = Externals M/C Ratio =
Externals % RMSE = Externals % RMSE = Externals % RMSE =
Total Count 43 Total Count 43 Total Count 37
Link Within Deviation 29 Link Within Deviation 6 Link Within Deviation 14
Link Outside Deviation 14 Link Outside Deviation 37 Link Outside Deviation 23
Remaining Total Needed Remaining Total Needed Remaining Total Needed
3 32 26 32 14 28
MD Peak Period ( 10 AM - 2 PM)
PM Peak Period ( 3 - 7 PM)
Off Peak Period ( 8 PM - 5 AM)
AM Peak Hour ( 7- 8 AM)
PM Peak Hour ( 5 - 6 PM)
Highway
Expressway
Time Period Analyzed
DAILY Assignment
AM Peak Period ( 6 - 9 AM)
Appendix C - San Joaquin Valley Model Improvement Project (San
Joaquin Valley MIP) One-Way Volume Model Validation Results
Three County Model - Merced
Arterial
Collector
ADT Model/Count by Functional Class
Model/CountFreeway Traffic vs. Local Traffic
Link Volume
Freeway
0
5,000
10,000
15,000
20,000
25,000
30,000
35,000
40,000
0 5,000 10,000 15,000 20,000 25,000 30,000 35,000 40,000
Mo
del
Count
ADT Model vs. Count
Upper Bound
No Deviation
Lower Bound
"Model versus Count"
0
1,000
2,000
3,000
4,000
5,000
6,000
7,000
8,000
9,000
10,000
0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000 10,000
Mo
del
Count
ADT Model vs. Count - Locations with Less Than 10,000 Vehicles/Day
Upper Bound
No Deviation
Lower Bound
"Model versus Count"