Geoscience Reference
In-Depth Information
Generally, discrete models are built associating the presence of a species to
polygons of land unit types (e.g., vegetation categories), most often with a
deductive approach; in fact, transferring into the realm of
GIS
, the traditional
way of producing distribution maps is based on a similar but more arbitrary
partitioning of the study area (e.g., administrative boundaries, regular grids;
see also “Habitat Definitions and Use”). There are also some examples of
binary classifications of continuous environmental variables (e.g., slope,
aspect, elevation) using statistical techniques such as logistic regression (Pereira
and Itami 1991) or discriminant analysis (Corsi et al. 1999). Categorical-dis-
crete models do not account for species mobility and tend to give a static
description of species distribution. Nevertheless, this approach can be used to
address the problem of defining areas of occupancy (Gaston 1991) and thus
can be used successfully for problems of land management and administra-
tion. On the other hand, probabilistic models can describe part of the stochas-
ticity typical of locating an individual of a species and can be used to address
problems of corridor design and metapopulation modeling (Akçakaya 1993),
introducing the geographic dimension in the analysis of species viability.
LITERATURE REVIEW
Table 11.2 indicates the results of our bibliographic review. Papers are classi-
fied according to the categories described in the previous paragraph.
We have considered
GIS
and
RS
as two different views of the same tool, the
former being more devoted to spatial correlation analysis and the later more
concerned with basic data production. In fact, the two families of software
tools share many basic functions and are evolving toward integration into a
single system. It should be noted that the review includes not only papers that
use
GIS
or
RS
but also some that deal with
HSI
,
HEP
and general assessment of
species' ecological requirements. The papers in this last group do not generally
represent examples of spatial models (Scott et al. 1993), in the sense that their
products are not distribution maps, but they have been included because they
are considered to be just a few steps away from a real distribution model. In
fact, they describe the ecological requirements of the species in terms of map-
pable environmental conditions.
Most of the papers that use the deductive approach consider the a priori
knowledge sufficient to define the ecological requirements of the species under
investigation. This is especially true of papers that model distribution on the
basis of interpretation of remotely sensed data; in fact, 15 out of 16 papers per-
taining to the deductive group that used remotely sensed data to model species
