Transit Estimation and Mode Split CE 451/551 Source: NHI course on Travel Demand Forecasting (152054A) Session 7.

Transit Estimation and Mode Split

CE 451/551

Source: NHI course on Travel Demand Forecasting (152054A) Session 7

Terminology

HOV Light Rail Portland; Florence Heavy rail Commuter rail Local bus service Express bus service Paratransit service Busways Headways/frequency Transit captive

Factors Affecting Mode Split

Person/household characteristics– Auto availability, income, HH size, life cycle

Trip characteristics– Purpose, chaining, time of departure, OD, length

Land use characteristics– Sidewalk/ped facilities, mix of uses at both ends, distance to

transit, parking and costs at both ends, density at both ends Service characteristics

– Facility design (HOV, bikes), frequency, congestion, cost (parking, tolls, fares, out-of-pocket costs), stop spacing

Mode Split Model Applications

Route or service changes– effect of changes in cost, frequency, transfer

system, more or less service and routes– Not usually modeled with TDF (use analogy or

elasticity) Major investment studies, e.g. HOV, New rail

or other capital investment project design Policy changes

– Parking, urban growth boundaries, congestion pricing

Mode Split Strategies

Analogy Elasticity Analysis Direct Estimation of Transit Share Disaggregate Mode Split

Choosing a Mode Split Technique

Application Time and budget constraints Project costs Existing data availability Existing service?

– if none, have to “borrow” a model

Selecting Analogy Routes

Selection based on similarities in:– Household characteristics– Transit service

Adjustments– Service area household characteristics– Service differences– Fare differences

Elasticities: ratio of change in demand over change in system

Example of Elasticity

If transit fares are raised from $1.00 to $1.25 and there is a resulting drop in daily transit ridership from 8,000 to 7,200, the elasticity, as calculated below, would be -0.40

Elasticity analysis example

What does the –0.4 factor mean? typical values for cities range from

-0.15 to -0.4 Is this elastic, or inelastic? Do you think larger cities would have larger

or smaller elasticity? Why?

Direct Estimation of Transit Share

In small-to-medium regions with limited transit use Particularly when transit use is limited to specific

populations (zero-car household, students, and elderly)

Generally estimate district-to-district transit share– Find relationship between SE&D and %transit– Calibrate for base year– Assume relationship will hold in future

Subtract resulting transit trips from person trip table.

Disaggregate Mode Split Models

Travel is a result of choices Elasticity, analogy, and direct estimation of

transit share are limited, particularly in policy analysis

Output– Share of person trips using each mode (by trip

purpose) for each production-attraction cell.

Disaggregate Mode Split Models

Utility functions– Building blocks for DMS models– Rank desirability of the alternate transportation modes– Deterministic equations

Probability models (overcomes limitations of deterministic utility functions)

– Logit the most common– Incorporate utility equations into probabilistic equations

Binomial logit models– Predict choice between two alternatives

Multinomial logit models– Predict choice between more than two alternatives

Disaggregate mode split using Utility Functions and Probabilistic Models

Input: Individual responses on mode desirability and usage to develop “Utility functions”

Preference and usage data may be from census or special home surveys.

System data such as travel time and cost generally from network data

usually don’t have the kind of data needed to know all users preferences

Observation v. prediction

If we wish to estimate transit by income level (or other detailed variable) in the future we need to be able to forecast the population characteristic in each group.

The more disaggregate the data set for modeling, the more difficult the prediction of future.

Just like trip generation and distribution … can you give examples?

Probability Equations

Auto Utility Equation: UA= -0.025(IVT) -0.050(OVT) - 0.0024(COST)

Transit Utility Equation: UB= -0.025(IVT) -0.050(OVT) – 0.10(WAIT) – 0.20(XFER) - 0.0024(COST)

Where:IVT= in-vehicle time in minutesOVT = out of vehicle time in minutesCOST = out of pocket cost in centsWAIT = wait time (time spent at bus stop waiting for bus)XFER = number of transfers

Question: what is the implied cost of IVT? OVT? WAIT? XFER?

Binomial Logit Model Example

References

Transit Fact Book, 50th ed, American Public Transit Association, Washington, D.C. January 1999.

Federal Highway Administration. Traveler Response to Transportation System Changes. 2nd ed, U.S. Department of Transportation, Washington, D.C., July 1981.

Federal Transit Administration, A Self-Instructinf Course in Disaggregate Mode Choice Modeling. Report No. DOT-T-93-19. U.S. Department of Transportation, Washington, D.C., December 1986

Meyer, M.D., and E.J. Miller. Urban Transportation Planning, A Decision-Oriented Approach. 2nd ed. McGraw-Hill, 2001.

Homework

Network DataIn-vehicle Time

Out of Vehicle Time Cost

Calculate Mode Shares

Mode OVT IVT Cost (cents)

1 person 5 17 200.0

2-person carpool 5 21 100.0

3-person carpool 5 23 66.6

4-person carpool 5 25 50.0

Transit 7 33 160.0

Part 1: CALCULATE MODE PROBABILITIES BY MARKET SEGMENT

Overview: Calculate the mode probabilities for the trip interchanges. Use the tables on the next pages.

Part A: Calculate the utilities for transit as follows:– Insert in the table the appropriate values for OVT, IVT, and COST.– Calculate the utility relative to each variable by multiplying the variable

by the coefficient which is shown in parenthesis at the top of the column; and

– Sum the utilities (including the mode-specific constant) and put the total in the last column.

Part B: Calculate the mode probabilities as follows:– Insert the utility for transit in the first column;– Calculate eU for transit– Sum of eU for transit and put in the “Total” column; and– Calculate the probability for transit using the formula:– Sum the probabilities (they should equal 1.0)

Say, from trip distribution, the number of trips was 14,891. Calculate the number of trips by mode using the probabilities calculated.

Solo Driver

2-Person Carpool

3-Person Carpool

4-Person Carpool

Transit

Total 14, 891

Mode Trips (Zone 5 to Zone 1)

If we had time …

Source: publicpurpose.com

Cheaper to lease cars than provide new transit?

http://www.publicpurpose.com/ut-2000rail.htm

Transit share dropping?

http://www.publicpurpose.com/ut-intlmkt95.htm

Where rail transit works

http://www.publicpurpose.com/utx-rails.htm

You can see an alternative view here:

http://www.sprawlwatch.org/