Transcript Slide 0

Application of Time-of Day Choice
Models Using EMME/2
Washington State DOT Congestion Relief Analysis
presented to
19th International EMME/2 Users’ Conference
presented by
Arun Kuppam, Cambridge Systematics
Maren Outwater, Cambridge Systematics, Inc.
Mark Bradley, MBRC
Larry Blain, PSRC
Robert Tung, RST International
Shuming Yan, WSDOT
October 19, 2005
Seattle, Washington
Transportation leadership you can trust.
Project Objectives
To capture variations in time of day by 30-minute time
periods
To develop an approach that is sensitive to pricing
scenarios
To capture travel behavior that reflects tendency to shift
to nearby time periods
1
Shortcomings of previous TOD Model
Five discrete time periods
Model calibration based on unweighted survey
Variation by income groups not captured
2
Characteristics of New TOD Model
A logit time of day choice model, applied after mode
choice to auto trips
32 time periods – half hours except first and last periods
Variables include demographics, trip characteristics
(carpool, bridge crossing), delay
Includes costs measured in units of time
Use of non-linear “shift” variable within 3 larger time
periods
3
Time Periods
AM Peak – 10 30-minute time periods from 5:00 a.m. to
10:00 a.m.
Midday – 10 30-minute time periods from 10:00 a.m. to
3:00 p.m.
PM Peak – 10 30-minute time periods from 3:00 p.m. to
8:00 p.m.
Evening – 1 3-hour time period from 8:00 p.m. to
11:00 p.m.
Night – 1 6-hour time period from 11:00 p.m. to 5:00 a.m.
4
Model Specification
Multinomial Logit Structure with 32 alternatives
U = ASC + C1*(Delay) +
C2*[(Delay.min.20 + sqrt(Delay–20).max.0)*Shift] +
C3*[(Delay.min.20 + sqrt(Delay–20).max.0)*(Shift^2)] +
C4*(Bridge Dummy) + C5*(Bridge Dummy*Shift) +
C6*(Carpool Dummy) + C7*(Carpool Dummy*Shift) +
C8*(Household Size) + C9*(Household Size*Shift) +
C10*(Income Group) + C11*(Income Group*Shift)
5
Where, Delay for AM = max[(AM GC – NI GC), 0]
Shift ‘Early’ for AM = (7.5 – T)
Shift ‘Later’ for AM = (T – 7.5)
T = Hour – 1, 2, 3, ……, 24
Bridge Dummy = 1 or 0
Carpool Dummy = 1 or 0
Household Size = min (hhsize, 4)
Income Group = <$45k, >$75k
TOD Modeling System
SOV, HOV2, and
HOV3+ Trip
Tables by Time
Period (32)
Auto Trip Tables
by Occupancy
and Purpose
Time-of-Day
Choice Model
Walk and Drive
Access Transit
Trip Tables by
Time Period (5)
Transit Trip Tables
for Walk and Drive
Access
Time-of-Day
Peaking Factor
Model
Commercial Vehicle
and External Trip
Tables
Light-, Medium-,
and HeavyTruck Trip
Tables by Time
Period (5)
Summary
Reports
Legend:
Input Files
6
Models/Processes
Report Output Files
Data Output Files
0500-0529
0530-0559
0600-0629
0630-0659
0700-0729
0730-0759
0800-0829
0830-0859
0900-0929
0930-0959
1000-1029
1030-1059
1100-1129
1130-1159
1200-1229
1230-1259
1300-1329
1330-1359
1400-1429
1430-1459
1500-1529
1530-1559
1600-1629
1630-1659
1700-1729
1730-1759
1800-1829
1830-1859
1900-1929
1930-1959
Evening
Night
Probabilities from TOD Model Application
Home to Work
Home to Work TOD Distribution as
a Function of AM Peak Delay
Probability
0.2
0.18
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
7
0 min
Time of Day
5 min
10 min
15 min
0500-0529
0530-0559
0600-0629
0630-0659
0700-0729
0730-0759
0800-0829
0830-0859
0900-0929
0930-0959
1000-1029
1030-1059
1100-1129
1130-1159
1200-1229
1230-1259
1300-1329
1330-1359
1400-1429
1430-1459
1500-1529
1530-1559
1600-1629
1630-1659
1700-1729
1730-1759
1800-1829
1830-1859
1900-1929
1930-1959
Evening
Night
Probabilities from TOD Model Application
Work to Home
Work to Home TOD Distribution as
a Function of PM Peak Delay
Probability
0.18
8
0 min
5 min
10 min
15 min
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
Time of Day
Probabilities from TOD Model Application
HBW Drive Alone Trips – Variation by Income Group
and Direction
Shares of Trips
0.200000
A-P Inc1
P-A Inc1
0.180000
A-P Inc2
P-A Inc2
A-P Inc3
P-A Inc3
A-P Inc4
P-A Inc4
0.160000
0.140000
0.120000
0.100000
0.080000
0.060000
0.040000
0.020000
Time of Day
9
19:30
19:00
18:30
18:00
17:30
17:00
16:30
16:00
15:30
15:00
14:30
14:00
13:30
13:00
12:30
12:00
11:30
11:00
10:30
10:00
9:30
9:00
8:30
8:00
7:30
7:00
6:30
6:00
5:30
5:00
-
Validation Results
Two-stage Validation
• Stage 1 – Validate TOD shares by trip purpose, mode of
travel, and direction, results within +/- 0.02
• Stage 2 –Validate VMT against traffic counts by TOD, results
within +/- 10%
Time Period
2000 Model
(VMT)
Percent
Difference
AM Peak
6 a.m. to 9 a.m.
4,428,739
4,480,908
1.2%
Midday
9 a.m. to 3 p.m.
8,555,459
8,337,805
-2.5%
PM Peak
3 p.m. to 6 p.m.
5,399,197
5,712,866
5.8%
Evening
6 p.m. to 10 p.m.
4,376,938
4,188,376
-4.3%
Night
10 p.m. to 6 a.m.
2,736,413
2,861,614
4.6%
25,496,746
25,581,569
0.3%
Total
10
2000 Counts
(VMT)
Conclusions
Time-of-day choice models can be estimated with 30+
time periods with existing data
Models are sensitive to time and cost tradeoffs, as well as
demographic factors and bridge constraints
Calibration by mode, trip purpose, and direction, as well
as for volumes provides more behavioral understanding
of results
Initial sensitivity tests indicate that models produce
reasonable results
11
Acknowledgements
Project was completed in support of model improvements
for
• Washington State Department of Transportation
• Puget Sound Regional Council
Expert Review Panel requested additional detail on time
periods
• University of Wisconsin, Milwaukee, WI
• North Central Texas Council of Governments, Dallas, TX
• Portland Metro, Portland, OR
• Sound Transit, Seattle, WA
• Atlanta Regional Commission, Atlanta, GA
12