Integrating models and geographical information systems

by
Roger Bivand and Anne Lucas

Introduction
Since geographic information systems (GIS) currently dominate our perception
of how computing and geography should interface, and since GeoComputation
(GC) is providing analysts of spatial phenomena with ever more powerful
computing tools, it may be helpful to examine the experience that has accrued
concerning links between them. Our examination is both empirical and
normative, and the reader may find it useful to repeat at least some of our
literature surveys, since new papers and articles are accumulating rapidly.
Searching on the key words ‘GIS’ and ‘model*’ or ‘integral*’, where ‘*’ is the
wild card, led to a wide range of hits both in ISI Science and Social Science
Citations Indices, and in OCLC-FirstSearch. These sources primarily contain
journal articles, while conference proceedings may be searched at the Ohio State
University GIS Master Bibliography Project, and more recently through the
web-sites of conference organizers, such as NCGIA and GISDATA in Europe.
Adding these resources to what we already knew about the issues involved, we
were able to scan the field for interesting regularities, trends, and citation
clustering.
In our initial hypotheses, we believed that there existed a common
understanding of what integrating models and GIS entailed. In fact, our search
illustrated that views on GIS-model integration were disparate, and no
consensus has yet emerged through either practice or theory in the GIS or
modelling communities. This early disappointment proved to be a useful pointer
towards a more fruitful way forward, by problematizing both the contents and
contexts of the concepts involved, together with the practices in which they are
entrained. Views that we had expected to resonate through the literature can be
quoted from presentations at the Boulder, Colorado, conference on
environmental modelling and GIS held in 1991. Something big, interesting, and
synergistic was to be had by combining GIS and environmental models. Both
use maps (or map-like display) as an expression of reality, both have lots of data
to deal with, and both require analysis of these data in some map-like form. GIS
is presented as a general purpose technology for satisfying the following specific
needs (Goodchild, 1993):
• preprocessing data from large stores into a form suitable for analysis including reformatting, change of projection, resampling, and generalization;
• supporting analysis and modelling, such that analysis, calibration of models, forecasting and prediction are all handled through instructions to the GIS;
• postprocessing of results including reformatting, tabulation, report generation, and mapping (these operations are expected to be available under a graphical user interface or GUI).
Dangermond (1993) saw the future as bringing more user-friendly GIS, with less
need for GIS specialist knowledge, because both software and hardware would
be more powerful, more graphical, less expensive, and easier to use. In addition,
artificial intelligence would support users behind the scenes, there would be
more input/output formats, the natural environment and its variability would be
easier to represent as hardware prices fell, animation would become feasible for
displaying results, and in ESRI, the trend towards integration into a single
system: raster with vector, CAD with GIS, image processing with GIS, would
continue. All of these should enhance the integration of models and GIS and be
offered as the rational solution.
Maybe this is a solution looking for a problem, but it is not necessarily a
viable solution to problems faced by many modellers. The fact that there is still
a need to discuss integration of models and GIS in such ‘general’ terms suggests
that either this expectation has not been fulfilled to the anticipated extent, and/or
that there is something inhibiting ‘easy’ integration. Enough experience with
models should have been accumulated after 30 years; GIS has been around just
as long in one form or another. A review of the literature reveals a lack of clarity
in the definition of ‘integration’, which can take place at many levels and in
many forms. This issue needs to be re-examined. Integration has been promoted
as a ‘good thing’, and that with better integration will come better, faster, easier
systems and enhanced analysis. One would then expect to be able to follow the
trend in the literature, culminating in some ‘well integrated’ system. We need to
examine the literature, to determine how integration has proceeded over the
years.
We have found that the majority of the models discussed in the literature are
environmental, rather than social or economic. It seems that social and economic
topics are mostly covered in the integration of spatial and exploratory data
analysis and statistical tools in GIS. Consequently, we will by and large restrict
this presentation to environmental models, based on theory drawn from physics,
chemistry, or biology. We begin with a survey of the background concerning
GIS, what integration in GIS has meant, and modelling. Continuing to develop
the theme of integrating models and GIS, we examine the reasons for such steps,
the problems encountered, how integration may be defined, and what alternative
system architectures may be employed. We find that the issues involved in data
modelling are of considerable importance for integration projects, and discuss a
range of requirements for the viability of such projects. To illustrate these
points, we describe an integration project with data from the Baltic Sea. In
conclusion, we discuss some consequences not only for integrating
environmental models and GIS, but possibly also for GC in more general terms,
addressing data modelling in particular.

This entry was posted in Articles, Gis-RS Books and tagged , . Bookmark the permalink.

Comments are closed.