Geoscience Reference
In-Depth Information
data does not support both high-resolution and large-extension analyses. Cur-
rently our analytical capabilities are limited by the lack of high-resolution,
global coverage data sets. Nevertheless, the use of GIS 's spatial analysis tools in
the framework of a controlled environment in which all the key variables are
monitored at an adequate resolution can increase our ability to assess causal
effects in species-environment relationships.
Apart from further generic considerations, we think that a few important
issues have been overlooked in these first years of application of GIS to the field
of ecological modeling and especially in the field of species distribution mod-
eling. There has been inadequate discussion and consideration of the assump-
tions underlying the model-building process and the related issues of spatial
and temporal scale, which are of paramount importance for sound scientific 11
use of GIS . Adequately discussed assumptions can justify the development of a
model. Whenever a hypothesis is stated and a model is built to test its congru-
ence, it should be regarded as a problem-solving tool. For instance, we will
never really know what will be the outcome of alternative management
options, but we can state different hypotheses, state the assumptions that must
be met to make each hypothesis hold true, and try to model the result of the
different options. In such cases we don't have direct control over the results of
the management action; we can only ensure that the assumptions are met. This
means that the output of the model will hold true if the assumptions are met
and if the model is built on the logical consequences of these assumptions. In
such cases validation, meant as an independent estimate of the truth, can to a
certain extent be neglected (Starfield 1997). Nevertheless, most of the time
assumptions are not adequately discussed and this is particularly evident when
dealing with the constraints of scale dependency of biological events. Probably
the issues of scale still suffer from inadequate support from the available tools.
For instance, we still lack convenient ways of handling spatiotemporal data in
GIS software packages, not to speak of analyzing the two components together.
If validation can be neglected somewhat when dealing with hypothesis-
testing models, it becomes a fundamental issue when building analytical mod-
els, which are built to assess species-environment relationships and ecological
processes. In such cases, validation steps must be included from the beginning
of the model building process, first assessing the quality and reliability of the
raw data used, then evaluating the limits of the relationships that drive the
process and finally analyzing the correspondence of the output with the truth.
Validation can be a costly exercise in model building, and efforts are being
made to find a cost-effective approach to this issue. Because the issue of vali-
dation is general to GIS modeling and especially GIS ecological applications, it
Search WWH ::




Custom Search