Transcript Slide 1
14th TRB National Planning Applications Conference
May 5-9, 2013, Columbus, Ohio
A TEST OF TRANSFERABILITY:
THE SE FLORIDA ACTIVITY-BASED MODEL
Rosella Picado
Parsons Brinckerhoff
Background
Southeast Florida is home to 5.5 million people, spanning
Miami-Dade, Broward and Palm Beach counties
Relatively disperse travel patterns with significant
congestion on Turnpike and north-south freeways
Growing interest in improving transit, expand toll and
managed lane infrastructure, mitigate adverse EJ impacts
Trip-based model reaching its limits, especially regarding
variably-priced tolling, fare policies, spatial detail, EJ
analysis
SE Florida ABM
Coordinated Travel – Regional Activity-based Modeling
Platform Family of ABMs
Main features:
Explicit intra-household interactions
Continuous temporal dimension (half-hour time periods)
Fine spatial dimension (12,000 MAZs)
Faithful transit access coding
Distributed values of time
Integration of location, time-of-day, and mode choice models
Model Development Strategy
Transfer the San Diego ABM
Adopt CT-RAMP structure and sub-models
Adopt model parameters estimated with San Diego household
survey data
Update certain models to reflect SE Florida conditions:
input data availability (employment, population controls)
modal supply
trip assignment methods
ancillary models
Calibrate models to SE Florida travel patterns
Re-specify models that fail to perform well
Why Model Transfer?
Schedule:
To use the ABM in the development of the 2013 Long Range
Transportation Plan
Approximately 18 months available for model development
was insufficient time to estimate & validate all models
Data:
Quantity and quality of NHTS SE Florida sample may preclude
statistically significant estimation of some models and/or
population effects
Largely sufficient for calibration, with caveats
Data Limitations
Small sample size – 2,000 households
Some subareas within model region under
represented
Retired households over-sampled
College students and children under-represented
Missing data, ‘ungeocodable’ activity locations, etc.
Incomplete transit on-board survey
Assessing the Model Transfer Outcome
Evaluate initial estimated travel patterns against model
calibration targets
Regional targets for important person markets
Sub-regional where data allow
Assess the magnitude of constant or parameter
adjustments to match targets
Importance of model calibration targets
Based on NHTS and supplemented with other sources
Evaluated for reasonableness
Compared to targets from other regions
Work Location Model - initial results
Tour Frequency (%)
10%
9%
8%
Person
Type
Full-time
Part-time
All
7%
6%
5%
4%
3%
2%
1%
0%
0
5
10
15
20
25
30
Distance (miles)
Observed
Estimated
35
40
45
50
Avg. Length (mi.)
Obs.
10.6
7.5
9.9
Est.
9.4
5.3
8.7
Work Location Model - calibrated
Tour Frequency (%)
10%
9%
8%
5%
Person
Type
Full-time
Part-time
4%
All
7%
6%
3%
2%
1%
0%
0
5
10
15
20
25
30
Distance (miles)
Observed
Estimated
35
40
45
50
Avg. Length (mi.)
Obs.
Est.
10.6
10.2
7.5
7.0
9.9
9.7
School Location Model - initial
Tour Frequency (%)
40%
35%
30%
25%
20%
15%
10%
5%
0%
0
1
2
3
4
5
6
7
8 9 10 11 12 13 14 15 16 17 18 19 20
Distance (miles)
Observed
Estimated
School Location Model - calibrated
Tour Frequency (%)
40%
35%
30%
25%
20%
15%
10%
5%
0%
0
1
2
3
4
5
6
7
8 9 10 11 12 13 14 15 16 17 18 19 20
Distance (miles)
Observed
Estimated
Eating Out Location Model - initial
Tour Frequency (%)
30%
25%
20%
15%
10%
5%
0%
0
2
4
6
8
10
12
Survey
14 16 18
Distance
20
Model
22
24
26
28
30
Daily Activity Pattern Model
Target DAP
Model Initial DAP
Mandatory
Non
Mandatory
Home
Mandatory
Non
Mandatory
Home
Full-time worker
80%
14%
7%
81%
12%
7%
Part-time worker
55%
37%
8%
63%
26%
11%
University student
78%
18%
4%
63%
25%
12%
Non-working adult
0%
76%
24%
0%
74%
26%
Non-working senior
0%
72%
28%
0%
75%
25%
Driving age student
89%
4%
6%
92%
3%
5%
Pre-driving student
94%
3%
2%
96%
2%
2%
Pre-school
35%
43%
22%
43%
41%
15%
Person type
Daily Activity Pattern Model
Target DAP
Model Initial DAP
Mandatory
Non
Mandatory
Home
Mandatory
Non
Mandatory
Home
Full-time worker
80%
14%
7%
81%
12%
7%
Part-time worker
55%
37%
8%
63%
26%
11%
University student
78%
18%
4%
63%
25%
12%
Non-working adult
0%
76%
24%
0%
74%
26%
Non-working senior
0%
72%
28%
0%
75%
25%
Driving age student
89%
4%
6%
92%
3%
5%
Pre-driving student
94%
3%
2%
96%
2%
2%
Pre-school
35%
43%
22%
43%
41%
15%
Person type
Non-Mandatory Tour Frequency
Estimated Tour
Frequency (%)
30%
Calibrated
Initial
30%
25%
25%
20%
20%
15%
15%
10%
10%
5%
5%
0%
0%
0%
10%
20%
Observed Tour Frequency
Full-time Workers
University students
30%
Part-time workers
0%
10%
20%
Observed Tour Frequency
Full-time Workers
University students
Part-time workers
30%
Non-Mandatory Tour Frequency
Estimated Tour
Frequency (%)
30%
Calibrated
Initial
30%
25%
25%
20%
20%
15%
15%
10%
10%
5%
5%
0%
0%
0%
5%
10%
15%
20%
25%
Observed Tour Frequency
Driving students
Pre-school children
Pre-driving students
30%
0%
5%
10%
15%
20%
25%
Observed Tour Frequency
Driving students
Pre-school children
Pre-driving students
30%
Work Departure and Arrival Times
18%
Initial
16%
14%
12%
10%
8%
6%
4%
2%
0%
Work Departure Observed
Work Departure Estimated
Work Arrival Observed
Work Arrival Estimated
Shop Tour Departure Time
Initial
12%
10%
8%
6%
4%
2%
0%
Shopping Departure Observed
Shopping Departure Estimated
Shop Tour Departure Time
Calibrated
14%
12%
10%
8%
6%
4%
2%
0%
Shopping Departure Observed
Shopping Departure Estimated
Work Tour Mode Choice
Target
Tour Mode
Drive-Alone
Shared 2
Shared 3+
Walk
Bike
Walk-Transit
PNR-Transit
KNR-Transit
Toll
Local Bus
Express Bus
BRT
Urban Rail
Com Rail
auto sufficiency
no veh.
insuf.
suf.
0%
49%
78%
13%
30%
13%
8%
11%
6%
11%
3%
0%
5%
1%
0%
62%
5%
1%
0%
1%
1%
2%
1%
0%
Initial Estimate
total
67%
18%
8%
1%
0%
4%
1%
0%
13%
68%
5%
1%
21%
5%
no veh.
0%
24%
12%
31%
18%
15%
0%
0%
auto sufficiency
insuf.
suf.
51%
67%
25%
17%
13%
13%
6%
1%
2%
0%
3%
1%
0%
0%
0%
0%
total
60%
20%
13%
4%
1%
2%
0%
0%
14%
51%
9%
13%
23%
3%
Work Ahead
Finalize model calibration
Validation to traffic counts and transit boardings
Future year forecast and sensitivity tests
Conclusions / Lessons Learned
SANDAG CT-RAMP ABM is able to reproduce most
regional travel patterns in SE Florida
Largest differences between observed and initial
model forecasts:
non-mandatory tour location
CDAP and tour frequency for college students, part-time
workers, pre-school children
Modest constant adjustments sufficient to calibrate
the model
Conclusions / Lessons Learned
Supplemental data sources important to validate
calibration targets and selected model outputs
Unable to observe transferability at high levels of
disaggregation
Acknowledgments
Shi-Chiang Li, Florida DOT
Paul Larsen, Palm Beach MPO
Paul Flavien, Broward MPO
Larry Foutz, HNTB (formerly Miami-Dade MPO)
Ken Kaltenbach, The Corradino Group
Sung-Ryong Han, BCC Engineering
Bill Davidson, Ben Stabler, Jinghua Xu
Questions?
Rosella Picado
Parsons Brinckerhoff
Seattle, WA
[email protected] | (206) 382-5227