Memorandum DRAFT SUBJECT TO CHANGE 1. BMC ABM Development Network Checks and Transit Path Tests

Transcription

1 \ Memorandum AECOM 2101 Wilson Boulevard Suite 800 Arlington, VA Tel Fax To Feng Liu, Tom Rossi, David Kurth, CS Page 1 CC Subject BMC ABM Development Network Checks and Path Tests From Date Srikanth Neelisetty, Pat Coleman November 30, 2015February 15, 2016 This memorandum documents the work done to perform FTA Tests for the Baltimore Metropolitan Council (BMC) Activity Based Travel Demand Model (TDM). This task involved checking transit network inputs, transit path parameters, and creating district-to-district transit flow summaries in order to compare with model transit results. Highway and transit skims were also developed based on the validated networks. Some of the transit tests were performed while developing highway and transit skims for model estimation in May Those tests were documented in an earlier memo Draft Skim Validation memo Final.docx. One of the major changes since May 2014 was that BMC s trip-based model has been updated with networks that include Traffic Analysis Zones (TAZs) structure from Metropolitan Washington Council of Governments (MWCOG) model for Prince George s County, Montgomery County, Frederick County and District of Columbia. Therefore, the hybrid skim approach used in May 2014 was no longer necessary to develop skims. Similar to earlier exercise, checks were done in the Baltimore area since travel to/from and within Baltimore area was the primary focus of the ABM. As an additional check, path parameters from Baltimore area were tested for the MWCOG counties by performing WMATA Metrobus and Metrorail survey trips assignments. The rest of this memorandum discusses the results from performing the tests mentioned above. Highway and Network Checks BMC updated its trip-based model networks to year 2012 conditions after May 2014 path parameter tests. Sabra Wang Associates (SWA) performed checks on highway network to make sure the network changes SWA performed on the 2010 validated highway network were incorporated in the 2012 version. Since transit travel times depend on highway travel times, modeled highway travel times were compared with observed INRIX traffic data to check for reasonableness. For this purpose, a representative sample of freeway and arterial highway segments were selected from the entire model region with input from BMC. Figure 1 shows the map of the selected highway segments. INRIX data for Traffic Message Channels (TMCs) in the Baltimore area that were averaged into 1 hour bins were provided by BMC for an average weekday in year For the segments in the DRAFT SUBJECT TO CHANGE 1

2 facilities identified in Figure 1, modeled times from the AM and Midday periods were compared with the average 7-8 AM and 12-2 PM average times estimated from the INRIX data. Figure 1: Map of Selected Highways for Times Comparison DRAFT SUBJECT TO CHANGE 2

3 Table 1 shows the comparison of the modeled times with the INRIX times for segments in the facilities identified in Figure 1. Figure 2 shows the same results graphically. Table 1: INRIX Times vs. Modeled Times for Select Highway Segments NB/EB/CW SB/WB/CCW Roadway From/To To/From Distance Inrix 7-8 Model AM Inrix 7-8 Model AM (mi) AM (min) Time(min) AM(min) Time(min) 1 PA Line Shawan Rd (in Huntvalley) I-83 2 Shawan Rd (in Huntvalley) I I-695 Downtown Baltimore I-95 ( in Rossville) I I-83 I I I-795 I I-70 I I-95 I I-97 I-95 ( in Rossville) I-795 I MD 140 (in Riestertown) I-270 (in Frederick) MD 940 (in Owings Mills) US MD 940 (in Owings Mills) US 40 I-695 I-695 (in Rossville) Susquehahha River I I-695 I I I-395 I-695 (in Halethrop) I-695 (in Halethrop) MD MD 32 I I US 50 MD MD 32 I I Arbutus BWI Airport I I95/I895 Split ( in Baltimore City) I- 95 (in Elkridge) MD 24 (in Edgewood) I-695 (in Rossville) US I-695 (in Rossville) MD 2 (in Downtown Baltimore) MD 2 (in Downtown Baltimore) US 29 (in Ellicott City) MD 24 ( IN Bel Air) Honeygo Blvd Honeygo Blvd I-695 (in Overlea) US 1 28 I-695 (in Overlea) I-83 (in Downtown Baltimore) I-83 (in Downtown Baltimore) I-695 (in Arbutus) I-695 (in Arbutus) MD MLK Blvd (in Downtown Baltimore) I MD I-695 MD MD 32 I-95/I-495 (In Greenbelt) MD 2 (in Annapolis) I US I -97 MD MD 3 I-95/I-495 (In Greenbelt) I -97 /MD3 I MD I-95 US US 29 MD US 29 I MD I -95 I I-97 MD 177 (in Annapolis) MD MD 100 I I -70 MD US MD 175 MD MD 216 I-95/I-495 (In Greenbelt) MD I-370 I MD-129 MD Garrison Forest Rd (in Owings Mills) Northern Pkwy Northern Pkwy MLK Blvd (in Downtown Baltimore MD-27 (in Mt Airy) I I-695 MD MD I-695 Cold Spring Ln (in Baltimore) MD MD-138 (in Monkton) I I-695 US-1 (in Downtown Baltimore) DRAFT SUBJECT TO CHANGE 3

4 Model AM Time (minutes) Figure 2: INRIX Times vs. Modeled Times for Select Highway Segments y = x R² = y = 0.983x R² = Inrix 7AM-8AM Time (minutes) NB/EB/CW SB/WB/CCW Linear (NB/EB/CW) Linear (SB/WB/CCW) Table 1 and Figure 2 show that aggregate model times and INRIX times are well correlated (more than 0.85 correlation coefficient) and that there is not much difference in correlation between freeways and arterials segments. It should be noted that the R 2 for both northbound and southbound highway segments improved compared to the May 2014 effort. AECOM incorporated the changes made by Gallop to the 2010 transit network in March 2014 into the 2012 transit network. A summary of transit network changes is shown in Table 2. Table 2: Network Changes Made to the 2010 Network in March 2014 Line Update MTA03X_SB Change Dwell_C from 1.4 to 0.7 MTA04_SB MTA04_NB MTA05X Change headway from 10 to minutes and Dwell_C from 1.4 to 0.7 MTA10X Change headway from 20 to 60 minutes and Dwell_C from 1.4 to 0.7 MTA15A_EB MTA15A_WB MTA15B_EB MTA15B_WB MTA15C_EB MTA15C_WB MTA15D_EB MTA15D_WB MTA15E_EB MTA15E_WB MTA15X and changed headway from 30 to 45 minutes and Dwell_C from 1.4 to 0.7 DRAFT SUBJECT TO CHANGE 4 Formatted Table Commented [MMR1]: Scheduled Headway is 20 min. Commented [SN2]: Fixed Commented [MMR3]: Headway is 30 minutes

5 MTA16_NB MTA16_SB MTA19X MTA44A_WB MTA44A_EB MTA44B_WB MTA44B_EB MTA46_EB MTA46_WB MTA50OP MTA58 Updated to split into MTA19XA and MTA19XB Changed headway from 30 to 15 minutes Changed headway from 30 to 15 minutes Changed headway from 30 to 45 minutes Changed headway from 30 to 45 minutes Updated to split into MTA58_EB and MTA58_WB MCE120_SB Change headway from 10 to 14 minutes and Dwell_C from 1.4 to 0.7 MCE120_NB Change headway from 10 to 14 minutes and Dwell_C from 1.4 to 0.7 MCE150_WB Change headway from 30 to 60 minutes and Dwell_C from 1.4 to 0.7 MCE150_EB Change headway from 30 to 60 minutes and Dwell_C from 1.4 to 0.7 MCE160A Change Dwell_C from 1.4 to 0.7 MCE160B Change Dwell_C from 1.4 to 0.7 MCE310 Change Dwell_C from 1.4 to 0.7 MCE311 MCE320_NB MCE320_SB and changed headway from 60 to 45 minutes and Dwell_C from 1.4 to 0.7 and changed headway from 60 to 45 minutes and Dwell_C from 1.4 to 0.7 MCE411 Change headway from 30 to 45 minutes and Dwell_C from 1.4 to 0.7 and changed headway from 20 to 45 MCE420 minutes and Dwell_C from 1.4 to 0.7 MCE915A Change Dwell_C from 1.4 to 0.7 MCE915B Change Dwell_C from 1.4 to 0.7 MCE921 MCE922A Change Dwell_C from 1.4 to 0.7 MCE922B Change Dwell_C from 1.4 to 0.7 MCE929A Change Dwell_C from 1.4 to 0.7 MCE929B_SB Change Dwell_C from 1.4 to 0.7 MCE929B_NB MCE950A Change Dwell_C from 1.4 to 0.7 MCE950B Change Dwell_C from 1.4 to 0.7 MCE995A Change Dwell_C from 1.4 to 0.7 MCE995B MCE995C Change Dwell_C from 1.4 to 0.7 and changed headway from 60 to 30 ANNBRO minutes ANNGOL Updated to split into ANNGOL_SB and ANNGOL_NB ANNGRE and changed headway from 60 to 30 DRAFT SUBJECT TO CHANGE 5 Commented [MMR4]: Did you change DWELL_C for consistency? Commented [SN5]: Yes

6 ANNORA ANNYEL CAWST1 CAWST2 HCTS1_WB HCTS1_EB HCTS2_WB HCTS2_EB HCTS6_WB HCTS6_EB CTCE_SB CTCE_NB UB601 UB602 UMBCIRV UMBCCAN UMBCHA UMB702 UMB703 JHUMTWA JHUEA_SB JHUEA_NB JHUMI_SB JHUMI_NB COLLB_SB COLLB_NB COLLR_SB COLLR_NB TOWLOCH TOWROD TOWGOU TOWKEN TOWTIM FTMEADE_WB FTMEADE_EB MTA38 MTA97 MTA98 MTA99SB MTA99NB minutes Change headway from 30 to 45 minutes, Circular Change headway from 120 to 60 minutes Change headway from 120 to 60 minutes Change headway from 120 to 60 minutes Change headway from 120 to 60 minutes Change headway from 120 to 90 minutes Change headway from 120 to 90 minutes DRAFT SUBJECT TO CHANGE 6 Commented [MMR6]: Are these private college routes? Are they for University trips only? Commented [SN7]: Removed all the private college routes Commented [MMR8]: What is this?

7 CCCBAN_NB CCCBAN_SB runtimes were checked to ensure modeled runtimes compared well with the observed scheduled times. The overall R 2 values for peak/off-peak were /0.87. Figure 3 show a plot of scheduled vs model times. Figure 3: 2012 Scheduled vs. Modeled Runtimes Comparison Peak Off-Peak Path Parameters The transit path parameters were checked using the same approach used in May 2014 effort. It involved preparing survey trip tables from Maryland Authority s (MTA) 2008 On-board survey and assigning them using the transit assignment procedures in the trip-based model. The model parameters were refined after each assignment and the process was repeated until the modeled boardings reasonably replicated the observed survey transit boardings. Peak and off-peak survey trip tables were assigned using procedures developed during previous efforts. Initial assignments showed slightly more than 1% of unassigned trips. Some of them were due to missing access/egress links and were later assigned after adding such links. path-building parameters were then modified iteratively to improve the modeled boardings. Since the parameters from 2010 validation effort were used as a starting point, only transfer penalties between local buses were modified to improve the model assignments. Table 3 shows the survey versus survey assignment boardings before and after path-building parameter adjustments. Mode Table 3: Survey vs. Modeled Boardings Detailed Submodes 2008 Survey Unlinked Trips 2008 Survey TT Assignment in 2010 Network (May 2014) 2008 Survey TT Assignment in Updated 2012 Network Peak Off-Peak Total Peak Off-Peak Total Peak Off-Peak Total MTA Local 86, , ,690 85, , ,931 84, , ,901 MTA Express(1) ,995-4,995 4,825-4,825 DRAFT SUBJECT TO CHANGE 7

8 MTA Commuter Bus 10, ,896 9,856-9,856 10,624-10,624 MTA Light Rail 12,983 8,502 21,485 8,415 8,911 17,326 9,252 8,914 18,166 MTA Metro Rail 19,259 23,776 43,035 22,520 24,307 46,827 23,053 24,531 47,584 MTA Commuter Rail 9,964 10,142 20,106 10,801 8,369 19,170 10,387 8,370 18,757 TOTAL 139, , , , , , , , ,857 (1) Survey doesn't have 150, 160. Model also includes express versions of local buses 3, 5, 10, 15,19. Modeled MTA local boardings are low for peak period while express boardings are high as MTA Express in the model has express versions of local routes 3, 5, 10, 15 and 19, while the survey total doesn t. Also, express routes 150 and 160 were not present in the survey. More detailed checks such as prediction-success tables and dot-density plots of transit trips will be performed during model validation. Table 4 summarizes the transit path parameters that were used to generate the results in Table 3. Table 4: Recommended Path Parameters Baltimore Area Walk - Peak Drive - Walk - Off-Peak Drive - Local IVT Weight Express Buses IVT Weight Commuter Bus IVT Weight LRT IVT Weight Metro Rail IVT Weight Commuter Rail IVT Weight Walk/Transfer Weight Drive Access weight Boarding Penalty (min) RECOSTMAX Transfer Penalty (min) To/From LRT Transfer Penalty (min) Rest XFERFACTOR Wait Factor Note: All values same as May 2014 version except the ones highlighted yellow. They were increased from 3 min in May 2014 version to 4 min. The MTA transit survey only covered routes in the Baltimore area. Therefore, to test the path parameters for the MWCOG area, transit trip tables were developed using the 2008 Washington DRAFT SUBJECT TO CHANGE 8

9 Metropolitan Area Authority (WMATA) Metrobus and Metrorail surveys and assigned using parameters in Table 4. The results were compared with the survey observations and the transit parameters were adjusted to closely reflect the observed boardings and transfers. Since the Metrobus and Metrorail surveys included portions of Northern Virginia, only trips that have both origin and destination within Prince George s County, Montgomery County, Frederick and DC were assigned in this exercise. Table 5 shows the initial survey assignment results Table 5: Initial Survey Assignments Summary WMATA Surveys 2008 Survey Linked Trips Survey Boardings Survey Transfer Ratio Path Peak Off-Peak Total Peak Off-Peak Total Peak Off-Peak Total Walk to 309, , , , , , Drive to 111,869 48, , ,114 52, , Total 420, , , , , , Survey Linked Trips Unassigned Trips Intrazonal Trips Survey Assignment Boardings Transfer Ratio Path Peak Off-Peak Total Peak Off-Peak Total Peak Off-Peak Total Peak Off-Peak Total Peak Off-Peak Total Walk to 309, , ,656 8,900 4,203 13,103 4,782 4,197 8, , ,159 1,384, Drive to 111,869 48, ,519 4,656 1,017 5, , , , Total 420, , ,175 13,556 5,220 18,777 5,145 4,412 9,557 1,094, ,111 1,820, Mode 2008 Survey Unlinked Trips 2008 Survey TT Assignment Peak Off-Peak Total Peak Off-Peak Total Local 288, , , , ,765 1,296,179 Express Commuter Bus Light Rail Metro Rail 250, , , , , ,331 Commuter Rail TOTAL 538, , ,657 1,094, ,111 1,820,510 Table 5 shows that a high percentage of trips were unassigned (2.6%), especially for walk mode. The transfer ratio is high compared to the survey observations, which is reflected in the total boardings from the model assignments. The survey data for some of the unassigned trip pairs were analyzed and the following observations were noted. - The survey suggested that that the maximum walk access/egress distances were higher than 1 mile and closer to 2.0 miles. - Drive access links to some park and rides were missing - Transfer links were missing, which was resulting in high bus boardings. - There were some survey records which indicated that transit was used where transit services did not exist, such bad survey records were deleted and survey was reweighted to produce survey trip tables. The transit skim parameters were modified to include maximum walk distance to 2.0 miles. All the missing PNRs in the Prince George s, Montgomery, and Frederick Counties were coded and drive access links were generated to those nodes. Transfer links were also generated. Survey records with bad geocoding were removed and the survey was reweighted. While the survey assignments improved with all those changes, the resulting boardings were still high probably because of availability of multiple transit options in the DC area. With headway combination, DRAFT SUBJECT TO CHANGE 9

10 transfers in the model were not as onerous as they were in reality. Therefore, transfer penalties were iteratively modified until the overall transfers were reasonable. Table 6 shows the assignments summary after all the path parameter adjustments. Table 6: Final Survey Assignments Summary WMATA Surveys 2008 Survey Linked Trips Survey Boardings Survey Transfer Ratio Path Peak Off-Peak Total Peak Off-Peak Total Peak Off-Peak Total Walk to 296, , , , , , Drive to 98,410 42, , ,389 46, , Total 395, , , , , , Survey Linked Trips Unassigned Trips Intrazonal Trips Survey Assignment Boardings Transfer Ratio Path Peak Off-Peak Total Peak Off-Peak Total Peak Off-Peak Total Peak Off-Peak Total Peak Off-Peak Total Walk to 296, , ,257 1, ,860 4,782 4,145 8, , , , Drive to 98,410 42, ,234 1, , ,220 45, , Total 395, , ,491 2, ,210 5,145 4,360 9, , , , Mode 2008 Survey Unlinked Trips 2008 Survey TT Assignment Peak Off-Peak Total Peak Off-Peak Total Local 249, , , , , ,855 Express Commuter Bus Light Rail Metro Rail 258, , , , , ,889 Commuter Rail TOTAL 508, , , , , ,744 As can be seen from Table 6, unassigned trips are now lower and the overall transfer ratio is closer to the observed ratio from the survey. Model assigned boardings are comparable to the unlinked trips from the survey. It should be noted that some of the difference could be attributed to the incorrect responses for transfers in the survey. There were several cases where the reported transfers were lower than the actual transfers required to make the transit trip. Table 7 shows the suggested transit parameters. Table 7: Recommended Parameters MWCOG area Walk - Peak Drive - Walk - Off-Peak Drive - Local IVT Weight Express Buses IVT Weight Commuter Bus IVT Weight LRT IVT Weight Metro Rail IVT Weight Commuter Rail IVT Weight Walk/Transfer Weight Drive Access weight Boarding Penalty (min) DRAFT SUBJECT TO CHANGE 10

11 RECOSTMAX Transfer Penalty (min) XFERFACTOR Wait Factor AECOM will check if separate path parameters are needed for the transit trips with both origin and destination within MWCOG counties during validation. AECOM will also modify path parameters as necessary during that time. Skims skims were generated for use in ABM using path parameters in Table 4 for Baltimore area and Table 6 for MWCOG area respectively. This was accomplished by generating skims for the entire model area using Table 4 parameters and then replacing the skims for ODs with both origin and destination within MWCOG area with those from the ones generated using Table 6 parameters. The set of transit skims (walk/drive, peak/off-peak) contained with the following tables- - INITWAIT - Initial wait time (min) - XFERWAIT - Transfer wait time (min) - WALKTIME - Walking time (min) - AUTOTIME - Auto access time (min) - AUTODIST - Auto access distance (mi) - TRANFARE - fare (cents) - NUMXFERS - Number of Transfers - BUSTIME - Local Bus In-Vehicle Time (min) - EXPTIME - Express Bus In-Vehicle Time (min) - CBUSTIME - Commuter Bus In-Vehicle Time (min) - LRAILTIME - Light Rail In-Vehicle Time (min) - METROTIME - Metro Rail In-Vehicle Time (min) - MARCTIME - Commuter Rail In-Vehicle Time (min) - BRTIME - BRT In-Vehicle Time (min) - WMATAACC - Washington Access Links Time (min) - WALKACC - Walk Access Time (min) - WALKEGR - Walk Egress Time (min) - XFERTIME - Transfer Time (min) - FIRST First transit node on the path - LAST Last transit node on the path District-to-District Flows District-to-District transit trip flows at the County level from the 2008 MTA transit on-board survey were prepared for use in model validation. They are summarized in a separate Excel file draft Survey County to County transit trip summaries. Next Steps Highway and transit runtimes as well as transit path parameter checks have not been done for the Frederick, Montgomery and Prince George s counties as the primary focus of the ABM model was to DRAFT SUBJECT TO CHANGE 11

12 model travel to/from the Baltimore area. Also, INRIX data for those counties were not readily available in the highway network. AECOM perform highway/transit runtimes checks upon receiving INRIX data for those three counties. ****** Please feel free to call or with questions or comments. DRAFT SUBJECT TO CHANGE 12