Geoscience Reference
In-Depth Information
Table 11.2
Classification of Reviewed Papers
Deductive
Inductive
Descriptive
Non-
spatial
Non-
spatial
GIS
GIS
and
RS
RS
GIS
GIS
and
RS
RS
9
8
7
8
32
3
1
0
0
4
36
Analytical
GIS
and
RS
Non-
spatial
Non-
spatial
GIS
RS
GIS
GIS
and
RS
RS
3
0
1
4
8
14
4
4
16
38
46
40
42
Papers are classified according to the approach used to define the species-environment relationship and
whether their approach was descriptive or analytical. Further subtopics indicate whether the author
considers the research to pertain to the domain of
RS
,
GIS
, or both.
Nonspatial
is used for papers that do
not contain an explicit distribution model but define species-environment relationship in terms of
mappable variables.
distributions fall within the descriptive group. In these papers, image classifi-
cation techniques tend to receive more emphasis, whereas the ecological appli-
cation is most often seen as an excuse to apply a specific classification algorithm.
The time trend of the papers published shows rather stable use of
RS
tech-
nology and increasing use of
GIS
. Up to 1986, no paper makes explicit reference
to the term
GIS
,
even though some of the papers dealing with the use of
RS
do
use raster
GIS
-style overlay procedures to define their distribution models (e.g.,
Lyon 1983) and others do use a spatial approach but do not mention
GIS
(e.g.,
Mead et al. 1981).
Little is generally said about model assumptions. Of the 82 papers
reviewed, only 21 discuss their assumptions. Those that do generally limit
their discussion to the statistical assumptions of the technique used to perform
the analysis. Very few deal with the biological and ecological assumptions and
tend to take them for granted. When dealing with ecological modeling, we
need to take into account both biological and methodological assumptions,
along with some general assumptions that may limit the applicability of the
results produced (Starfield 1997).
Validation, a step that is evidenced at different levels in the data flow dia-
gram (figure 11.2), is generally limited to the accuracy of the result of the
analysis (e.g., distribution map); nothing is said about the accuracy of the orig-
inal data sets (e.g.,
GIS
data layers, observation locations) and no consideration
is given to issues such as error propagation in
GIS
overlay (Burrough 1986).
Only 15 papers validate of the accuracy of their results based on an inde-
