Geoscience Reference
In-Depth Information
Geoghegan (2001) and the empirical-statistical models of Millington et al. (2007)
rely, largely, on confrontational approaches in that they compare 'real' world obser-
vations with model predictions. The tools used in this confrontation vary but include
visual comparison of predicted and observed spatial patterns, comparative statistical
measures (
r
2
and likelihood methods) and pixel-by-pixel comparisons (e.g., the
kappa statistic,
). Jenerette and Wu's 2001 model of urbanisation in Phoenix was
evaluated by comparison of the model's predictions with various measures of spatial
pattern in the landscape. As their model was stochastic they used Monte Carlo
methods (i.e., where did 'real' world observations fall in relation to model esti-
mates?) and avoided pixel-by-pixel confrontation. The agent-based 'Artifi cal
Anasazi' models are evaluated through both confrontation and experiment. The
population dynamics produced by the models are visually compared to population
changes inferred from archæological reconstructions, and are experimentally evalu-
ated by the researchers 'tinkering' (
sensu
Dowling, 1999) with the model until some
adequate resemblance is reached (similar to pattern-oriented modelling).
The case studies also highlight the diffi culties in establishing an adequate
typology of models and modelling, whether based on methodology or purpose.
Methodologically,
all
of the models considered above blur the boundaries between
analytical, empirical-statistical and simulation modelling. For example, based on
the outcomes of the (empirical-statistical) models developed by Bockstael (1996)
and Millington et al. (2007), maps of possible future change may be produced using
stochastic simulation. A typology based on purpose is no clearer: all of the exam-
ples presented above contain elements of consolidative, integrative and exploratory
modelling, and all in some way attempt to improve understanding and to make
predictions.
κ
Evaluating the Role of Models in Environmental Geography
A discussion of models and modelling in geography would be incomplete without
some mention of the debates about their place in the discipline
4
. During geography's
(so-called) 'quantitative revolution', quantitative modelling was embraced as a
methodology, peaking in the aftermath of Chorley and Haggett's seminal
Models
in Geography
(1967). While models and modelling remain key components in much
geographic research (especially in physical geography), geographers continue to
debate the appropriate place and use of modelling. Critics of modelling range in
position from those who view it as being a worthwhile, but typically poorly done,
enterprise, through to those who see it as having little or no place in geography
(Flowerdew, 1989). In the following discussion, I will focus on the criticisms put
forward by human geographers. This is not because physical geographers all agree
about the use and role of modelling, but rather because their debate(s) tend to be
rather narrower and methodological (e.g., concerning the appropriateness, or
otherwise, of specifi c techniques and representational assumptions).
In essence, the debate over modelling in geography is an extension of the long-
running debate over the usefulness or otherwise of positivism and the scientifi c
method in the discipline (Rhoads, 1999; Demeritt and Wainwright, 2005). Haines-
Young (1989) identifi es three common critiques of science and positivism in, but
not limited to, geography. First, some human geographers complain that modelling,
based on abstract quantitative theorising, cannot address the fundamental questions
of human geography relating to uniqueness of place, individuality, imagination,