Geoscience Reference
In-Depth Information
Table 20.1
A typology of modelling approaches
Model type
Description
Conceptual
Description of some system or process using narrative or
graphical tools.
Analytical (mathematical)
Formal description of some system or process using the
language of mathematics; can take many different forms
including both deterministic and stochastic approaches.
Note, however, that the term 'mathematical' is somewhat
misleading as models almost invariably contain, to a
greater or lesser degree, mathematical elements in some
guise (Guisan and Zimmermann, 2000).
Empirical (statistical)
Models based on observed data (usually, but not necessarily,
quantitative); includes statistical models.
Simulation
In a loose sense simulation simply involves 'building a
likeness' (Kleindorfer et al., 1998). In general, however, it
is usually taken to mean computer-based or
in silico
(see
page 341) activity. Simulation modelling encompasses a
multitude of activities ranging from the numerical solution
of analytically intractable systems of equations to attempts
to produce faithful
in silico
mimics or surrogates of specifi c
'real' world systems and processes (Winsberg, 2003;
Küppers and Lenhard, 2005).
Note: Falling outside this typology are 'hardware' models, that is, scaled physical reconstructions
such as fl umes and wind tunnels.
are based on observations and focus on prediction of a system's dynamics; they do
not
consider why a change will occur, only what the nature of the change will be.
Conversely, simulation models tend to consider the dynamics of the system and the
processes that explain those dynamics; they consider what the response of the system
might be to change and what processes explain that response. Thus, simulation
models are often also referred to as 'mechanistic' or 'process-based' (Guisan and
Zimmermann, 2000). In many cases the boundaries between the methodologies are
blurred; for example, nearly all simulation models contain mathematical elements
and some empirical component.
Another view is to consider models as being either 'top-down' or 'bottom-up'
(Grimm, 1999). Bottom-up modelling is an atomistic approach, motivated by the
belief that the dynamics and organisation of complex systems arise from, and can
be explained by, interactions between the units that comprise that system. In
environmental geography, agent-based models (ABMs) epitomise bottom-up model-
ling (Parker et al., 2003; Brown et al., 2004; Brown, 2006). In ABMs, the agents
are autonomous, goal-seeking entities. Although agents often represent individuals,
they may also represent aggregate structures such as family units, tribes, settlements
or business organisations. Schelling's (1978) segregation model provides a famous
example of a bottom-up, agent-based approach. In Schelling's model, householders
are divided into two groups and have preferences regarding how many of each type
of neighbour they prefer to live next to. 'Unhappy' households move to new sites
in an effort to improve their situation. Over time the model produces broad-scale
