Pointers for the Future - Environmental Modelling: Finding Simplicity in Complexity - page 437

there is no such thing as 'not my field', as the criticism
of increased specialization as a function of reductionist
perspectives in a number of the chapters has illustrated.
Communication between different modellers is impor-
tant, as the discussion by Millington et al . has illustrated.
There is no reason to prefer one form of working over
another. There are benefits of both hypothetic-deductive
and inductive approaches, while Thornes (Chapter 24)
highlights the use of a heuristic framework. All can pro-
vide powerful means of reaching conclusions in different
contexts. If one mode of working leaves us at an impasse,
we should consider whether an alternative might provide
a way out.
Most explanations in environmental science are based
on a tension between parsimony and generality as noted
for fluvial systems and for catchments by Mulligan and
Wainwright in Chapter 11. As we pointed out in the
introduction, complex systems theory is essentially a
rewording of Occam's razor (there is nothing new under
the sun, as Newton might have said!). Although we
might often talk about 'laws of nature', environmental
science deals with a higher level of aggregation where
fundamental laws are not appropriate. Thus, it is difficult
to produce models with a sufficient level of generality to
be suitable for all applications (even assuming sufficient
computer power were available). In this vein, Mulligan
and Wainwright, in Chapter 2, question how easy it is
to interpret holistic results. Such a question relates to
perceptual problems related to scale as pointed out in
many of the chapters including those by Perry and Bond,
Mazzoleni et al ., Nearing (Chapter 22), Millington et al .
and Zhang et al . (Chapter 5). It is often assumed that
we simply need to find the right model components and
link them together to tackle this problem but there are
different ways of linking them together too. To reach our
ultimate explanatory goals, we thus need to provide the
means of finding optimal model structures.

an iterative approach, as discussed above. The level to
which we can represent the environment depends on our
understanding and computational power. But as noted
by Engelen, the very process of producing models in this
way forces us to refine our ways of thinking about prob-
lems and produce tools that assist our thought process.
There is a tendency not to question the use of calculators
in everyday life (for example in shop tills or indeed in
sophisticated laboratory equipment) - why should there
be so much resistance to using models as appropriate
tools to solving questions of environmental understand-
ing? Without such tools, our explanations are reduced to
the level of analogies, as pointed out by Favis-Mortlock.
The limitations of such approaches are well understood
by every archaeologist, and considered in relation to
environmental problems in Meyer et al . (1998).
The model-building process often provides a means
of collecting information from 'non-scientific' sources
about the ways in which specific systems operate. Enge-
len points to an intermediate stage in integrated analysis
where qualitative models can be built up from knowl-
edge acquired from a wide range of sources. Twery and
Weiskittel (Chapter 23) also demonstrate how rule-based
approaches to modelling can allow the codification of
institutional knowledge. Such knowledge is often lost as
individuals move on, retire or die (or goes out of disci-
plinary fashion). The loss of such information often leads
to the reinvention of modelling wheels.
In a related sense we should beware of assuming that
models provide a totally objective means of tackling prob-
lems. Often, there are hidden assumptions in the ways
different people approach the modelling process. Wright
and Hargreaves (Chapter 6) discuss this problem in rela-
tion to a comparison of different applications of the
same model to the same problem (see also the compari-
son of erosion models in Favis-Mortlock, 1998). Models
are sensitive to boundary conditions, discretizations and
parameterizations as discussed in Chapter 1, so we should
not be surprised at this result. Such comparisons allow
us to investigate more robust approaches and the extent
to which knowledge and interpretations are embedded
within our models.

27.1.2 Qualitative issues
As we noted above, modelling can often be considered
to be as much of an art as a science (consider Penrose's,
1989, discussion of 'inspirational flashes'). Integrated
models are considered to be a fundamental way forward
to improving our understanding, but as noted by Enge-
len, their development remains a relatively subjective
process. The production of such models often throws
up a number of ambiguities and inconsistencies. Thus,
their development provides another means of furthering
our understanding of environmental systems, following

27.1.3 Reductionism,holismandself-organized
systems

Favis-Mortlock notes that 'We are not doomed to
ever more complex models!' The coupling together of
simple models can provide significant insights, even
in modelling the global climate, as noted by Harvey.