Geoscience Reference
In-Depth Information
biases in the sampling scheme or inaccuracy of the instrument used to locate
the species (e.g., radiotelemetry, global positioning system). A through discus-
sion of the sources of uncertainties in species distribution modeling can be
found in Stoms et al. (1992).
Once the sources and ranges of variability are identified, different input data
sets can be systematically produced, selecting the variates (sensu Sokal and
Rohlf 1995) from the variability range of the original input variables. These
alternative data sets are used to build alternative models that can be compared
with the original one, identifying the variability induced in the output by the
uncertainty of the input variables. The variability induced in the output is a
measure of the overall performance of the model and can be compared with a
predetermined acceptable significance level. As a general rule, when dealing
with great uncertainties in the measures of the input variables, a greater inertia
(less subject to changes in the results) of the model is generally preferable.
Sensitivity analysis does not replace validation but can be used at any stage
of the model-building process to identify the parameters that should be mon-
itored more carefully to maximize the reliability and the accuracy of the
results.
Discussion
j
The use of
GIS
in species distribution modeling should follow precise steps
in which each of the issues discussed in this chapter is accounted for. First
of all, we recommend more unambiguous use of some key terms such as
scale
and
habitat.
The latter seems to be a particular problem not limited to
GIS
applications but spanning the entire field of ecological studies (Hall et al.
1997). We believe that
GIS
can be a valid tool to overcome the current ambi-
guity between the species-related and land-related concepts of the term
habi-
tat.
As a matter of fact, the latter was introduced as a way of dealing with prob-
lems related to environmental mapping using traditional tools, and the
enhancements introduced by
GIS
do overcome those problems. In the mean-
time, however, we suggest replacing the word
habitat
with more unproblem-
atic terms such as
environment.
The ambiguity of the term
habitat
and most of the works on habitat use
and habitat selection have also given rise to the question of whether
GIS
-based
models can be used to explain the causal event of a species-environment rela-
tionship. In our opinion, use of
GIS
is not central to a better understanding of
causal effects in a species-environment relationship, especially if the quality of
