Pointers for the Future - Environmental Modelling: Finding Simplicity in Complexity - page 435

27

Pointers for the Future

John Wainwright 1 and Mark Mulligan 2
1 Department of Geography, Durham University, UK
2 Department of Geography, King's College London, UK

can have a high degree of confidence. In this section, we
provide an overview of some of the ways in which this
process might happen and related limitations, drawn on
the preceding chapters of the topic.

27.1 What have we learned?

The task now returns to us to highlight the simplicity in the
complexity that has gone before. Are there ways in which
the complexity of environmental systems can be under-
stood, and if so what are the tools that are used to evaluate
them? As suggested by the title of one of the most com-
monly used texts on numerical computing (Press et al .,
1992; see also Cross and Moscardini, 1985), modelling
is as much of an art as a science. A number of discus-
sions in this topic (particularly the model-comparison
exercises discussed in Chapters 6 and 9) suggest that this
is especially the case for environmental models. Models
test our conceptualizations of our environment, so it is
not surprising, perhaps, that models do not always (ever?)
agree. What we are looking at is how best to represent
the environment, and 'best' will of course depend on why
it is we want to represent the environment at all. In the
same ways that artistic representations of the environ-
ment may modify the way it looks to tell us more about it
(and ourselves) than a simple photographic reproduction
could do (Figure 27.1), so too do our models attempt
to abstract meaning from the complexity we observe.
Many environmental scientists will be used to making
schematic sketches of their observations in the field. In
a lot of respects, the development of models attempts to
take this schematization further, within a more formal
framework that provides some means of testing our ideas.
Only by an iterative testing of our models - confronting
them with as wide a range of different datasets and sim-
ulation contexts as possible - can we hope to learn more
and provide syntheses of our understanding in which we

27.1.1 Explanation
As we note in Chapter 2, there are a number of situations
where different explanations of the same phenomena
are available. Favis-Mortlock (Chapter 4) related this
situation to the debate on equifinality as defined by Beven
(1996) (see also Cooke and Reeves, 1974, for an earlier
debate based on qualitative modelling). In a sense, these
debates relate to ideas of using multiple working hypothe-
ses (Chamberlain, 1890) to evaluate competing ideas.
Many see modelling as part of a methodology employing
Popperian falsification to test between competing ideas.
Yet we have seen that in a number of cases, our data are
insufficiently strong to allow us to use such an approach
(see below). Explanation comes as part of an iterative
process where we question both our models and our data
(see Wainwright et al ., 2000, for examples). In a number
of places - Fisher (Chapter 12), Engelen (Chapter 21),
Haraldsson and Sverdrup (Chapter 17), Mazzoleni
et al . (Chapter 14), Millington et al . (Chapter 18), and
others - it is suggested that there is a 'Nature' or a 'real
world' that we are trying to model (although these terms
may not be equivalent). In this sense, a lot of ongoing
modelling work employs a critical realist methodology
(cf. Richards, 1990; Bhaskar, 1997). There is an under-
lying assumption that there are real features that we
attempt to reproduce, structured by processes that we can
only observe via their effects. Modelling allows us to close