Geoscience Reference
In-Depth Information
patterns of segregation, arising purely from decisions made by individual house-
holders; macro-level patterns (segregation) 'emerge' from micro-level (individual)
decisions.
By contrast, top-down modelling focuses on aggregate entities (e.g., entire popu-
lations) and on representing system-level relationships between aggregate variables
with the goal of fi nding relationships between those variables. As such, it involves
the application of general frameworks to particular problems (Grimm, 1999). The
classical models of population dynamics, such as the exponential (d
N
/d
t
=
rN
) and
logistic (d
N
/d
t
N
/
K
]) models, represent top-down approaches. These
models assume that while all populations behave in the same general way, as is
encoded in the functional form of the equation, the specifi c nature of their behaviour
will vary from case to case, and this is specifi ed by the exact parameter values
used.
A fi nal way to classify models is according to their use. Models serve three broad
purposes in environmental geography: (i) predicting the future state of some system
or phenomenon, (ii) making inferences about how a system or phenomenon is
structured and changes, and (iii) integrating and synthesising knowledge and data
from disparate sources. Bankes (1993) identifi es two basic purposes of modelling:
=
rN
[1
−
1.
consolidation
: modelling based on compiling all available information about a
system with the goal of creating a realistic and faithful surrogate of it. In this
context prediction will be important, whether to test the realism of the model
or to inform management and policy decisions about the actual system being
modelled; and
2.
exploration
: modelling in the face of epistemic uncertainty, where the model is
used experimentally to reduce this uncertainty by investigating the consequences
of various assumptions about the modelled object. The goal of such modelling
is heuristic.
This classifi cation does
not
represent a rigid either-or division. Exploration and
consolidation are synergistic. Improving our
understanding
of a process or system
should enable us to predict its behaviour better (or determine whether it has the
quality of predictability). Likewise,
reliable prediction
may lead to better under-
standing (Brown et al., 2006).
Consolidation: models for prediction
The desire to predict a system's or phenomenon's behaviour is a common motiva-
tion for modelling. Making predictions and testing them is central to the 'conven-
tional' deductive-nomological model of scientifi c inquiry. Predictive models take
many forms, from simple deterministic analytical models to complicated stochastic
simulation models. In geography, predictive modelling is often equated with empiri-
cal-statistical models (e.g., regression models); indeed statistical modelling is prob-
ably the most commonly applied and most criticised form of modelling used by
geographers (Macmillan, 1989a). As outlined above, empirical-statistical models are
formalised descriptions based on observed characteristics of the entity of concern.
While they may describe the links between components in a system, they do not
consider the underlying mechanisms. This approach has often been denounced for
