Mode Split
John K. Abraham
CE 7630
CE 7630
4-steps of demand modeling
! Trip Generation (land-use & economic factors determine how many trips people will make) ! Trip Distribution (determining where trips start and end)
! Mode-choice (determination of how you get there, e.g., bus, walk, auto, etc.)
! Trip Assignment (determination of which path you take from your trip origin to your trip destination) TRANSIT MODES
CE 7630
Mode Split / Choice
CE 7630
! Several types of math models can be used ! Aggregate models (used for a group of trips)
trip end (used after trip generation and before trip distribution) trip interchange (used after trip distribution)
methods include: multiple regression, cross classification, and diversion curves
Mode Split / Choice
CE 7630
! Disaggregate models
used to model individual trip making behavior probabilistic (assumes a probability distribution of user behavior) based on the utility (or disutility) of a particular mode choice
e.g. probit or logit models are used
! All models require information on
- trip (type) - tripmaker preferences - transportation system (travel time and other characteristics)
CE 7630
Trip End Models:
These models are applied before trips are distributed, and predict transit share as a function of independent variables such as ...
Trip Generation Trip Distribution Modal Split Traffic Assignment
auto ownership in Trip Generation the zone residential density Modal Split (Trip End Model) distance from CBD (central Trip Distribution business district) income level Traffic Assignment others
CE 7630
CE 7630
Autos per household 0 1 2 or more
Trip End Model (PATS)
Cross classification
Zones <1mi from CBD 77 35.5 13.5
log YS = 3.30 0.91 log X
Y0 = 84.02 X 0.094 X 2 Y1 = 3.04 + 3.2 X 0.026 X 2 Y2+ = 16.4 + 3.6 X 0.0334 X 2
Zones > 1mi
NRD 12+ 88 53.5 31.5 NRD < 12 61 36 20.5
Trips to CBD School Other
Auto
Variables Auto Ownership Net residential Density Distance from CBD
Transit
Y=transit trips/1000 persons X=Net Residential density Yo=transit trips/1000 persons with no cars
CE 7630
Comments / Critique
! Trip end model tries to assign modes before knowing where they are going.. ! Based on aggregation of relationships exhibited in a region not transferable ! Changes to transit system in the future will have no effect on the split / is not incorporated
This may lead to continuing degradation of transit services / reduce funding ! Trip-end models are suitable for small urban areas where transit is basically a social service, or in developing countries where mode choice is almost completely determined by income & auto ownership
CE 7630
Trip interchange models:
! These models are applied after trip distribution, and predict transit share as a function of ...
relative travel time between competing modes relative travel cost economic status of tripmakers relative service
! All other variables (e.g. trip length, population density...) are assumed to be correlated to these characteristics
uses multiple regression must be recalibrated for different regions of Mode Split After Trip Distribution !! the country
CE 7630
CE 7630
Mode Split Function of:
Variables Relative Travel Time TTR Relative travel cost CR Economic status EC Relative travel service L
Work Trip Modal Split Curves
CE 7630
CE 7630
Comments / critique
! Some level of transferability offered ! only requires a small database to calibrate the model (for a metro area) ! Suggest fairly optimistic outlook for transit
Developed base don data from Bay Area, Philadelphia, Washington
CE 7630
Disaggregate models
! Individual choices ! Probability of choosing a mode among available alternatives ! Choice based on behavioral variables ! Predominant mathematical formulation is the LOGIT function
CE 7630
Concept of Utility
! Utility Function measures the degree of satisfaction that people derive from their choices. ! A disutility function represents the generalized cost associated with each choice
CE 7630
Example
12000
$ 2500 5000 7500 10000
U 6 10 12 13.5
10000
8000
6000
4000
2000
0 6 10 U 12 13.5
Example 2 -- 3 Utility Equations
CE 7630
Proposal: 30% chance of winning $500 70% chance of winning $2000
U=3X
3500 3000 2500 2000
U = 3X U=X
2 3
$
1500 1000 500 0
15 00 30 00 45 00 6 00 0 75 00 9 00 0
U = X 1 .2
EMV = 0.3*500+0.7*2000 = $1550 EUV=30% (U @ X=500)+70%(U @ X = 2000) = 4650 utils CE=X @ U=EUV = 4650/3 = 1550 RP = EMV-CE = 0
CE 7630
The Logit Model
Pi = eUi e
j
Pi= Probability of an individual choosing alternative i Ui=Utility function of mode i N=set of modal alternatives e = base of natural logarithm
model form :
U j
where
= f (X
j ,k
Typical utility structure:
U j = a0 + a1{travel time} + a2{wait time} + a3{cost}
CE 7630
Nested Logit Mode Choice
! A model that quantifies the decision on the type of trip (car, bus, walk) people will take in the region ! Some socioeconomic factors, used as input for the travel demand as well as logit models, will be predicted using a trend technique called the Holt model, within the OFM forecasts used for Growth Management ! This type of mode choice analysis is required by the Federal Transit Administration for studying light rail and other forms of public transportation ! In addition, we can also use these assumptions to model bike and pedestrian travel patterns
10
CE 7630
CE 7630
Nested Logit Mode Choice Structure Example: Airport Travel
Auto
Scheduled Bus Public Transit Rental Car Shared-Ride Van Off-Airport Terminal Drop-off Taxi Access Par Limousine Drop-off Park k Lot Taxi Short-Term Transit Long-Term Airport Bus Off-Airport Stop
Door-to-Door
11
CE 7630
Comments / Critique
! aggregation of choices to the zonal level taking an average of user's utilities and computing the probability of modal choice may produce different results than taking the average of computed probabilities ! model is sensitive to policy issues ! only requires a small database to calibrate the model (for a metro area) ! transferable to other, similar areas ! explains user behavior (aggregate models do not)
CE 7630
12