the san francisco model in fifteen minutes · the san francisco model... in fifteen minutes january...
TRANSCRIPT
The San Francisco Model... in Fifteen MinutesJanuary 2007
Billy Charlton, Principal [email protected]
(415) 522-4816
San Francisco County Transportation Authority100 Van Ness Avenue, 26th FloorSan Francisco, California 94102
http://www.sfcta.org/modeling
Plans for Further Development: CHAMP 4 and beyond
Tour Based Models Tour based models predict activities, locations, and times for individual travelers.
A tour is an entire chain of trips: from a primary origin to all destinations and then back to the primary origin. Using chains means trips now have relationships:
(1) Primary Destinations versus intermediate stops
(2) Consequences of mode availability: e.g. trip mode depends on the full tour mode: Leaving the car at home? Then no driving for any trip on that tour.
(3) Based on activity diaries for the household ... travel is implicit.
(4) More able to test key policy questions e.g. system efficiency, pricing strategies, demographic impacts.
(5) No more mysterious �non- home-based� trips
HOME
WORK
INTERMEDIATE STOP
EN ROUTE TO WORK
WORK-BASEDSUB-TOUR
DESTINATION
SECONDARYHOME-BASED TOUR
DESTINATION
1
7
6
5
4 3
2
= Tour
= Trip
Number indicates trip order
PRIMARY TOUR:Home-based Work
WORK-BASEDSUB-TOUR
SECONDARY HOME-BASED
TOUR HOME
WORK
INTERMEDIATE STOP
EN ROUTE TO WORK
WORK-BASEDSUB-TOUR
DESTINATION
SECONDARYHOME-BASED TOUR
DESTINATION
11
77
66
55
44 33
22
= Tour
= Trip
Number indicates trip order
= Tour
= Trip
Number indicates trip order
PRIMARY TOUR:Home-based Work
WORK-BASEDSUB-TOUR
SECONDARY HOME-BASED
TOUR
In this example, the household member made seven trips in their full day of activity, linked together into three main tours. A work tour started the day (with an intermediate stop on the way to work). A work-based tour (maybe for a lunch meeting) followed. Then, after returning home, a final home-based tour occurred. A busy day! Traditional trip-based models can't capture the richness of these interrelationships. SF-CHAMP can.
Sonoma Napa
Solano
Marin
Contra Costa
Alameda
Santa Clara
San Mateo
San Francisco
Full-Day TourPattern Models
Time of DayModels
WorkplaceLocation Model
Vehicle AvailabilityModel
Nonwork TourDestination Choice
Models
Tour Mode ChoiceModels
Intermediate StopChoice Models
Trip Mode ChoiceModels
HighwayAssignment byTime Period (5)
TransitAssignment byTime Period (5)
Visitor Trip ModeChoice Model
Visitor Trip andDestination Choice
Model
LogsumVariables
AccessibilityMeasures
Zonal DataPopulationSynthesizer
Regional TripTables for NonSF
Trips
LogsumVariables
All Models
NetworkLevel of Service
All RemainingModels
Trip Diary Records
Person Records
Trip Tables
The SF-CHAMP ModelOne of the first fully buzzword-enabled models in the U.S:
Activity-based model; Full-day pattern of tours with intermediate stops; Microsimulation of all San Francisco residents (about 800,000); Nonresident trips borrowed from MTC�s regional model (thanks, Chuck!);
Four primary tour purposes: work, work-based, school, other; Five time periods: AM, Midday, PM, Evening, Owl; Nine (!) transit modes: Muni local, express, LRT/Cablecar; BART; Caltrain/Ferry; Regional buses
Hardware
Three AMD64 workstations running in parallel; 4Gb RAM and lots of storage.
When first created, model ran on one PC in 36 hours. With introduction of CHAMP 2, three PC�s in parallel could run CHAMP in less than 10 hours. CHAMP 3 will take...
Software
Citilabs CUBE 4.1, for matrix manipulation and skims (using TRNBUILD)Core model components and postprocessing written in C++ and JavaESRI ArcMap 9.2 for GIS mapping
The dreaded model flowchart diagram.
Land Use Growth Model: UrbanSim at a parcel-level (150,000 parcels)Allocation of growth for households and employmentSynchronization with SF-CHAMP in 5 year increments(Summer 2007)
Congestion Pricing Feasibility Study:
Exploring the concept of congestion pricing a la London and Stockholm. Reducing congestion, improving transit alternatives, and preserving equity are paramount to the success of the study.
SF-CHAMP will require some �minor twea ks� though:
(1) Inclusion of destination choice logsums in tour/trip generation models(2) Enhanced time-of-day choice models, for peak-spreading effects(3) Implementation of toll/no-toll in tour and trip mode choice models(4) Better feedback between components(5) Generalized cost assignment (tolls)
and...(6) Expansion of modeled population to the entire nine-county Bay Area (!)
Expansion to nine counties will focus on trips to and from San Francisco. MTC also is actively developinga regional activity-based model, so the Bay Area willsoon have two activity-based models for comparisonand TRB papers!
This map of theBay Area shows San Francisco labeled with areally big font to make it look more important. But the reality is that San Francisco covers less than 1 percent of the land in the nine-county region � amazing when you realize San Francisco is home to 12 percent of the Bay Area population. Expanding the model to all nine counties is going to be quite a challenge!
Someday/Maybe:Bicycle and Pedestrian path building and assignmentModeling � Bus Rapid Transit� (BRT) as its own modeAdditional trip purposesVisitor model enhancements
SF-CHAMP Calibration and Validation
Trips Diverted from Van Ness Avenue, with BRT Project
Usage Example: Trip Diversions from Van Ness Avenue, with Bus Rapid Transit (BRT) Project
Van Ness Diversions, 2010 Base to 2010 Project
Change in Volume at California St. 1700 100%
Traffic diverted onto:Parallel streets in corridor 800 47%Transit or suppressed trips 73 4%Parallel streets outside corridor 827 49%
Of traffic on streets outside corridor:About 1/2 are way beyond corridor (19th Ave, Kezar, etc.) 24%About 1/2 are evenly distributed across grid network 25%
Trips Diverted from Van Ness Avenue, with BRT Project (PM)
Green shows streets with less traffic.
Purple shows streets with more traffic.
The removal of one lane of through-traffic from Van Ness is a 1/3 reduction in capacity on that roadway. The remaining two lanes are only slightly more congested � 71% of volume remains.
Due to the dense grid system and availability of wide parallel streets, the other streets within the corridor absorb almost 50% of the diverted traffic.
Other city streets absorb the remainder.
Regional Trips Local Trips Total
Divertible 1,823 (19%) 3,168 (33%) 4,991 (52%)Not Divertible 1,352 (14%) 3,239 (34%) 4,591 (48%)Total 3,175 (33%) 6.407 (67%) 9,582 (100%)
Linking SF-CHAMP to Microsimulation Tools for the Van Ness BRT Study
CHAMP 3.0: Incorporating new capabilities and the latest data availableMore zones: from 766 to 981 in San Francisco alone. Plus 350 new zones in non-SF Bay Area counties Done!
Regional Network Update:Imported the latest regional peak/offpeak transit data, and converted networks to AM/MD/PM/EV/OWL periods Done!
Land Use Inputs Update:Census data and Planning Department inputs just didn�t match Done!
Linking Roadway/Transit: GPS-based correlation bet-ween highway and transit speeds, instead of hard- coded transit times from unreliable schedules. Also includes explicit stop delay and separate relationships by vehicle type Done!
New Population Synthesis:SFCTA & MTC joint update to ARC (Atlanta) method, using 2000 Census data. Went from 110 classes to almost 600, and included group quarters for first time Done!
Recalibration: using latest MTC home-interview survey and 2004 onboard transit orig/dest survey In progress
Synthetic Population Classifications:
Type: Household / Group QuartersIncome: Four QuartilesAge: 0-64 / 65+HH Size: 1 / 2 / 3 / 4+Children: 0 / 1+Workers: 0 / 1 / 2 / 3+Units: Single / Multi-UnitRace: White / Minority
A unique connection between SF-CHAMP and simulation software.
The fine-grained spatial detail of SF-CHAMP includes every street and alley in the entire city. This allows new types of analyses such as the Van Ness traffic diversions study showcased above.
Microsimulation software provides more than just visual-izations for public outreach; a three-stage modeling framework helped Authority planners estimate travel time savings due to specific BRT elements.
Results were fed back into the SF-CHAMP model, and provided key performance indicators for roadway and transit systems.
Base year CHAMP outputs are compared to the forecast year
� No-Project� alternative to create growth factors for
Van Ness Avenue, the corridor as a whole, and cross streets.
These factors inform a Synchro traffic model which produces
traffic signal timings and approach volumes.
The timings and volumes are then fed into a VISSIM traffic
microsimulation model of the corridor, along with transit
service, transit volumes from CHAMP, and pedestrian inputs, to produce several measures of
effectiveness.
This model has been extensively poked, prodded, and pricked since its creation in 2001.
Original Calibration: Based on MTC's BATS 1990 and 1996 surveys. No onboard transit survey data was available.
2004 Recalibration: Mode-specific constants for tour and trip mode choice models recalibrated in support of FTA New Starts application. Embarked on transit onboard survey.
2006 Comparison with MTC's BAYCAST Model:Extensive side-by-side comparison of SF-CHAMP and MTC's BAYCAST model presented at TRB Activity Based ModelingConference in Austin, TX. Overall match was quite good but differences were revealed in nonmotorized trips and trip generation rates, especially in forecast years.
2006/7 Model Refresh and Recalibration:Major effort to update and refresh all model components.