Geoscience Reference
In-Depth Information
pendent estimate of the distribution (either through comparison with an inde-
pendent set of observations or through comparison with the known distribu-
tion of the species); interestingly, 50 percent of these papers are based on the de-
ductive approach. In fact, it should be noted that because observation data sets
are the most expensive data to be collected within the general framework of set-
ting up a
GIS
species distribution model, the deductive approach is the most cost-
effective if seen from the validation point of view. In fact, to avoid bias, a model
developed with an inductive approach cannot be validated using the same data
set used to derive the species-environment relationship. Thus validation can be
performed either with a second, independent data set or by dividing the origi-
nal data set into two subsets, one of which is used to derive species-environment
relationships and the other to validate the resulting model.
Finally, it is interesting to note that the multidimensional power of
GIS
is still
not backed up by adequate quantity and quality of geographic data sets (Stoms
et al. 1992). This is reflected in the number of environmental variables used in
analysis. In the papers reviewed, the average is just below 4.8, and only 9 out of
82 analyze more than 9 environmental variables, whereas 23 papers base their
distribution models on only one environmental variable, generally vegetation.
Modeling Issues
j
Based on the results of the literature review, we have identified five major issues
that must be addressed to allow a sound
GIS
modeling of species distributions.
These range from uncertainties in the objectives of the research to the lack of
adequate support for the assumptions underlying the implementation of
GIS
models. A problem that is gaining awareness is that of scale, in both time and
space, but it still suffers from inadequate tools.
Slightly different is the issue of data availability, which is rarely addressable
by the biologist concerned with species distribution modeling but limits the
type of models that can be developed.
Finally, a review of sources of errors and ways of estimating the accuracy of
a
GIS
model addresses the problem of validation.
CLEAR OBJECTIVES
When setting up an ecological model, the very first step to be considered is
clear statement of the model's objective (Starfield 1997). There is great confu-
sion about the objectives of many published papers. This may caused by
overqualification of the tool, in the sense that use of the tool becomes the
