Links to:

2 Transport Modelling

2.1 The traditional four-step transport model

The traditional approach to the modelling of transport demand for person travel is the four step approach utilising the sub-models of trip generation, trip distribution, modal choice and traffic assignment. Each of the four steps addresses a reasonable question regarding the travel pattern:

How many trips will be made from to and from each zone (trip generation)?
What will be the number of trips between one zone and another zone, or within a zone (trip distribution)?
How many of the trips will be made by each mode of transport (modal choice)? and
What route will the trips follow (traffic assignment)?

It needs to be coordinated with other aspects of community planning, particularly land use planning.

2.2 Trip generation models

A trip is defined as a one-way person movement by one or more modes of travel.

Each trip has an origin and a destination.

Transport modelling studies usually consider trips in two broad categories:

Home-based (any trip having either the origin or the destination of the trip as the home), and
Non-home based (any trip having neither the trip origin nor the trip destination as the home).

Trip Generation is the sub-model of the transport planning model which predicts the total number of trips generated and attracted to each zone in the study area.

The two methods which are commonly used to build trip generation sub-models are:

Multiple linear regression, and
Category analysis.

2.2.1 Multiple linear regression

Past transport studies have made extensive use of the multiple linear regression method.

Transport studies have shown that residential land-use is an important trip generator, and non-residential land-use is a good attractor of trips.

A regression equation could therefore be developed which describes trips generated by a residential zone. For example:

Y		=	A + B1X1 + B2X2 + B3X3
where	Y	=	trips/household
	X1	=	car ownership
	X2	=	family income
	X3	=	family size
	A, B1, B2, B3	=	parameters derived by calibration

The variables to be used (X1, X2 and X3) in the equation above vary from one study area to another and are selected by using current trip information.

In developing regression equations it is assumed that:

• all the variables are independent of each other

• all the variables are continuous and normally distributed.

Usually these conditions cannot be met. For example in the equation X1, X2 and X3 are not continuous and they are not independent of each other. For these reasons regression analysis has attracted considerable criticism. It has however been widely used.

2.2.2 Category analysis

Difficulties with the use of regression equations for the study of trip generation led to the formulation of models based on the household or the person (i.e. disaggregate models. This approach is known as category analysis and has been used in a considerable number of transport studies. Category analysis uses the household as the fundamental unit generating trips and uses household characteristics to predict the number of trips generated. The household characteristics to be used must be easily measured.

The original work on category analysis in the 1960’s proposed that categories be established using 3 factors:

• car ownership – 3 classes – none, one, more than one

• income – originally 6 classes

• family structure – 6 classes.

	Adults employed	Adults not employed
1 2 3 4 5 6	None None 1 1 More than 1 More than 1	1 More than 1 1 More than 1 1 More than 1

These categories produced 108 household classes and associated with each class is a trip rate. One problem that occurs is that some classes may only have very small data sets from which to try and compute representative trip rates.

Some later studies have reduced the number of categories from 108 to 18 by the use of 3 car ownership groups and 6 household structure groups.

The advantages of category analysis over multiple linear regression analysis are:

•    computational processes are simpler

•    the use of disaggregate data may be expected to simulate human behaviour more realistically than zonal values

•    earlier home-interview data from other studies can be used, by conducting small studies to ensure trip rates by category are similar to those found in earlier studies.

One disadvantage of category analysis is that it assumes income and car ownership will increase in the future.

2.3 Trip Distribution

Trip distribution is used to determine the number of trips between two specific zones. The trip distribution sub-model is the second sub-model used in the transport planning model and builds off the data produced by the trip generation sub-model.

In mathematical terms, the trip generation sub-model will give a prediction of the total trips generated from zone i (Pi), and the total trips attracted to zone j (Aj). The trip distribution sub-model then allows the prediction of the interzonal trips between zones i and j (t_ij).

The methods used for trip distribution fall into two groups:

Growth Factor Methods, which include:

The Constant Factor Method,
The Average Factor Method,
The Fratar Method, and
The Furness Method.

Synthetic Methods, of which the most widely used is the Gravity Method.

2.3.1 Growth factor methods

Growth factor methods assume that the profile of future tripmaking will remain substantially the same as the present except that the volume of trips will increase according to growth in the generating or attracting zones. The methods are simpler than synthetic methods and where significant land-use change is not expected they give reasonable results.

The successful use of a growth factor method is dependent on the accurate assessment of the growth factor. The lack of any measure of travel impedance is a disadvantage and it is not possible to take into account new facilities or the restraining effects of congestion.

2.3.2 Synthetic methods

Synthetic methods allow the effect of differing planning strategies (in particular travel cost) to be incorporated in trip distribution. The methods seek to determine the causes of present day travel patterns, and assume that these causes will be unchanged in the future. The most widely used synthetic model is the gravity model.

Gravity model

The gravity model has been so named because of its similarity in form to Newton’s Law of Gravity.

The gravity model has as its basis that the trip interchange between zones is proportional to the attractiveness of the zones to trips, and inversely proportional to a function of the physical separation of the zones.

The form of the gravity model generally used in practice is:

where	tij	=	trips from zone i to zone j
	Pi	=	trips produced from zone i
	Aj	=	trips attracted to zone j
	Fij	=	travel time factor, or friction factor, where the zones are separated by a distance of zij. The factor Fij is often taken as
	Fij	=
			and a value of a around 2 is common. In practice zij may be a distance measure (km) or, more normally, a travel time (minutes).
	Kij	=	zone-to-zone adjustment factor known as a ‘socio-economic’ adjustment factor. An example of the need for such a factor would be in the location of a ‘slum’ area close to the CBD of a large city. Assuming the CBD was an entire traffic zone, it would provide many work attractions, the majority of which however would mostly be medium to high income white collar workers. Without zone to zone adjustment factors the gravity model would assign a large number of trips from the slum area to the CBD on the basis of its proximity.

2.4 Modal Choice

Trips may be made by differing methods or modes of travel e.g. car, walking, bus, train, etc. The determination of the choice of travel mode is known as mode choice or modal split. The basis of most modal choice models is the probability of a traveller selecting a particular mode of travel, and assesses the 'utility' of the transport service.

Utility is the ability of a service to satisfy human needs. It therefore depends on both the properties of the service and the desires of its users. It can be assumed that reasonable individuals (or households, or families, or firms, or other coherent groupings of people) will act to maximise their own utility, subject to any relevant external constraints (such as the law or social customs).

The opposite of utility is disutility, which relates to the cost the user experiences in obtaining the service. The approach used in most modal choice models is to consider disutility, where the disutility is proportional to generalised costs.

Various mathematical models are used for mode choice modelling, with the most commonly used being disutility curves, and probit and logit modelling techniques.

2.5 Traffic Assignment

The process of allocating trips to particular routes is known as traffic assignment. Traffic assignment is usually confined to road traffic as most other modes (except perhaps walking and cycling) are limited to a particular route. The basis of assignment is usually travel time, and as future trips are assigned to the network travel times can be expected to vary. In the traffic assignment modelling process it is not unusual to find that the proposed road network becomes overloaded and that some car trips may need to be restrained.

There are four common methods of traffic assignment:

All-or-nothing assignment - in this method the route of least cost between a pair of zone centroids is calculated and all future trips are then assigned to this link.
Use of diversion curves - a diversion curve is prepared to estimate the proportion of trips which will be diverted to a new or improved rout compared to other existing routes. The diversion curve may be prepared on the basis of travel time or the travel costs as perceived by the road users. The diversion curve is then used to provide adjustments to the all-or-nothing approach.
Capacity restrained assignment - the all-or-nothing approach may result in a link with favourable costs attracting a large number of trips and becoming overloaded. In practice this will not occur as drivers will transfer to less congested routes as congestion increases, and so the distribution of traffic between routes will adjust itself based on perceived travel times or costs. The capacity restrained approach attempts to model this situation by using speed-flow relationships for the various routes available.
Multipath proportional assignment - in urban areas many alternate travel paths may exist and travel is distributed over these many routes. In doing this drivers perceive the 'best' route for them in different ways e.g. minimum travel time, minimum congestion, minimum number of traffic signals, etc. Multipath proportional assignment attempts to simulate this situation by assigning trips between any two zones over all feasible routes.

Page last modified 28 June 2010.