Stratification entails modelling the network in question for a specific time of the
day, most often the morning peak hour but also, possibly, some critical off-peak
period, with trip purpose being stratified into work and non-work. For example,
the modeller may structure the choice sequence where, in the first instance, all
work-related trips are modelled during the morning peak hour. (Alternatively,
it may be more appropriate to model all non-work trips at some designated time
period during the middle of the day.) Four distinct traffic models are then used
sequentially, using the data obtained from the stratified grouping under scrutiny,
in order to predict the movement of specific segments of the area’s population
at a specific time of day. The models are described briefly as:
The trip generation model, estimating the number of trips made to and from
a given segment of the study area
The trip distribution model, estimating the origin and destination of each
trip
The modal choice model, estimating the form of travel chosen for each trip
The route assignment model, predicting the route selected for each trip.
Used in series, these four constitute what can be described as the basic travel
demand model. This sequential structure of traveller decisions constitutes a con-
siderable simplification of the actual decision process where all decisions related
to the trip in question are considered simultaneously, and it provides a sequence
of mathematical models of travel behaviour capable of meaningfully forecast-
ing traffic demand.
An overall model of this type may also require information relating to the
prediction of future land uses within the study area, along with projections of
the socio-economic profile of the inhabitants, to be input at the start of the mod-
elling process. This evaluation may take place within a land use study.
Figure 2.1 illustrates the sequence of a typical transport demand model.
At the outset, the study area is divided into a number of geographical seg-
ments or zones. The average set of travel characteristics for each zone is then
determined, base on factors such as the population of the zone in question. This
grouping removes the need to measure each inhabitant’s utility for travel, a task
which would in any case from the modeller’s perspective be virtually impossible
to achieve.
The ability of the model to predict future travel demand is based on the
assumption that future travel patterns will resemble those of the past. Thus the
model is initially constructed in order to predict, to some reasonable degree of
accuracy, present travel behaviour within the study area under scrutiny. Infor-
mation on present travel behaviour within the area is analysed to determine
meaningful regression coefficients for the independent variables that will predict
the dependent variable under examination. This process of calibration will gen-
erate an equation where, for example, the existing population of a zone, multi-
plied by the appropriate coefficient, added to the average number of workers at
present per household multiplied by its coefficient, will provide the number of
Forecasting Future Traffic Flows 17