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