You are on page 1of 5

MODAL SPLIT

The modal split stage is concerned primarily with the attempt to assign person-trips to the various
alternative modes available.

The basic rationale of modal split is the assumption that travellers make rational choices between
the available modes, based in part upon characteristics of these modes, and in part upon
characteristics of travellers.

There are basically two types of modal-split models. The first is the trip-end model, which precedes
trip distribution, and has as its function the prediction of the split of total trips generated by each
zone among the available modes. The trip-distribution phase, which follows the trip-end modal-split
model, involves the construction of separate distribution models for each mode. The second form of
modal-split model is the trip-interchange model, which follows the stage of trip distribution and has
as its function the splitting of specific zonal interchanges among available modes.

The underlying assumption of the trip-end models is that the transport patronage is relatively
insensitive to the service characteristics of the transport modes. In other words, trip-end models
assume that mode split is essentially determined by the socio-economic characteristics of the trip-
maker. Such an assumption is obviously most tenable in situations where virtually all transit ridership
is captive in nature.

Adam’s model is one of the first modal split models which was based on the following regression
equation comprised of several composite variables:

Y = -2.6466 + 3.71 lnP + 0.39 lnE + 2.38 lnT + 0.49 lnU – 0.97 lnM Travel time cost
accessibility infrastructure
Where Y = percentage of trips on transit

P = population over 5 years old

T = transit service factor

E = economic factor

U = land use distribution factor

M = urbanised land area in square miles

The Chicago Area Transportation Study isolated some (seven) variables which appeared to have a
significant effect on mass-transit ridership (public transport usage), namely land-use, the gender of
the trip-maker, the employment status of the trip-maker and the ability of the trip-maker to drive.

In this study, the following transit trip rates were determined for forecasting purposes as a function
of car ownership:

Number of cars per household Percent of transit trips


0 58
1 12
2-3 10

The Erie Transportation Study marks a significant change in this type of trip-end of models in that it
attempted to introduce into the modelling process a measure of the quality of transport services for
use in predicting the share of trips by mass-transit. This was done through a joint accessibility index
defined as follows:

Qik = Ej (Fik j)

Where Qik = accessibility to employment for zone i by mode k

Ej = employment in zone j

= travel time friction factor from zone i to j by mode k

These accessibilities were calculated for each of transit and highway, and it was found that the
transit system provided a lower accessibility to employment. An investigation of the ratio of transit
to highway accessibility was conducted. Percent transit usage was then determined by plotting
existing levels of transit usage against the accessibility-to-employment ratio comprising a diversion
curve.

In Puget Sound Transportation Study similar diversion curves were derived for different income
groups:

Critique of Trip-end Modal Split Models:

Unresponsiveness to transportation system changes in the form of modifications in the performance


of transport modes and their service characteristics.

Each model is constructed by examining the aggregate relationships exhibited by the specific study
area – therefore the lack of transferability.

Trip-end models are unable to deal with estimations related to possible major changes in the
transportation system, and therefore lack predictive capability.

They are not based on sound reasoning of the actual processes that determine choice of mode in
transportation.
Trip-interchange Models

Unlike trip-end models, trip interchange modal split models incorporate measures of the relative
service characteristics of competing modes, and seek to split each trip-interchange volume between
available modes. They use as input specific trip interchanges produced by a trip distribution model
and provide output in terms of volume by mode for an assignment process (the 4th stage of travel
forecasting process).

Primary reason for the development of these models is the relative unresponsiveness of trip-end
models to transit system changes. It was seen that the only available procedure for improving the
responsiveness of modal split models was to change the location of the model in the travel
forecasting process, so that it would utilise more specific information on trip making, hence on the
transportation system.

These models estimate mode split for individual zone pairs by relating transit ridership to specific
level-of-service comparisons.

Trip-interchange models are mostly used to predict mode split among choice riders, since these
riders presumably base their choice upon such typical service variable as in-vehicle times, walking
times, waiting times and transfer times, and travel costs.

Trip-interchange models predict mode split as a function of either ratios of differences between
modal service levels, as well as any relevant socio-economic characteristics of the traveller. Typically
this functional representation takes the form of a set of diversion curves.

For instance, in Bay Area Transportation Study, as a result of diversion curve-fitting exercise, time
ratio, a purpose split between work and other purposes, CBD and non-CBD orientation, and peak and
off-peak turned out to be the variables that were significant in explaining transit share of the trips.

The diversion curves derived were:


On the other hand, Washington, Toronto, and Philadelphia models used five basic parameters:

1) Relative overall travel time of transit & auto


2) Relative overall travel cost of transit & auto
3) Relative excess travel time (out-of-vehicle time) of transit & auto
4) Income of the worker
5) Trip purpose

As we see here, the first three variables are related with the relative performance of competing
modes, while the fourth one refers to the economic status of the trip maker. Each of these variables
were formulated as ratios. The two overall travel time and travel cost factors were each expressed as
ratios of the total door-to-door travel times and costs, respectively.

In Twin Cities modal split model, the following equation was developed for home-based work trips:

Tijwt / Tijw = 41.4 – 12.1 ln (travel time ratio) – 4.4. ln (income) + 8.0 ln (residential density) + 1.3
ln (employment density) + 363.5 (nine-hour parking cost)
(Explain the signs of the regression coefficients. They are in line with a priori reasoning)

Considerable advances have been made in the development of aggregate trip-interchange modal
split models by introducing the generalised cost concept. Generalised travel cost relationships are
usually developed from some weighted linear combination of the factors that create the
unpleasantness of trip making. Its main characteristic is that it reduces all items to a single index,
which is usually monetary.

The generalised cost index used in the modal split element of SELNEC study conducted in the late
1960s was of the following form:

Gijk = a1tijk + a2eijk + a3dijk + pjk + qk

Where Gijk is the generalised cost of travel by mode k between i and j

tij is the travel time from i to j by mode k (in minutes)

eij is the excess time for the trip from i to j (in minutes) (i.e. waiting for public transport)

dij is the distance from i to j which acts as a surrogate for the variable on the costs of trips

pjk is the terminal cost (parking charges) at j

q is a modal penalty reflecting the discomfort and lesser convenience associated with public
transport journeys

and a1, a2, a3 are parameters which value other cost items in monetary terms
A trip-interchange modal split model based on this generalised cost index would be of the form:

Tijk e-βCijk
=

Tijk e-βCijk

Where Cijk is the generalised cost of travel between zones i and j by mode k

This is actually the standard logit model which gives the proportion of trips between i and j by mode
k, which can be written in the two-mode (private-transit) binary choice as follows:

Tij1 1
=

Tijk 1+e-β(Cij2 - Cij1)

Where k=1 represents transit


k=2 represents private mode

Critique of Trip-interchange Models:

In contrast to trip-end models, trip-interchange modal split models have been developed on the basis
of careful hypothesis formulation and contain similarities which render them transferable.
Trip-interchange models suggest and optimistic outlook for mass transit since they imply that
increasing transit patronage is attainable by the provision of a sufficiently competitive level of
service. Trip-interchange models were mostly developed for studies in which the planning of a new
transit facility was the primary concern and for which, increasing patronage levels are predicted.

You might also like