Geoscience Reference
In-Depth Information
As we have seen,
GIS
distribution models rely on species-environment rela-
tionships to extrapolate distribution patterns based on the known distribution
of the environmental variables. We have also seen that the relationships reflect
the biological needs of the species. The extent to which we need to coarsen our
temporal and spatial scales depends on the stochastic events that must be min-
imized, which in turn depend essentially on the dynamics of the species under
investigation. To this extent, it is important to note that major population
dynamics events happen on different scales in both time and space. In figure
11.3 (modified from Wallin et al. 1992) the two axes indicate the increasing
temporal and spatial scale at which population dynamics events happen. In
accordance with the hypothesis formulated by other authors (O'Neill et al.
1986; Noss 1992), the figure shows a positive correlation between space and
time scales; that is, events that happen on a broader spatial scale are slower and
thus take more time.
As a tool for distribution modeling this graph can be of great help in defin-
ing scale thresholds toward both a minimum and a maximum scale for an
analysis. For instance, when considering cause-effect species-environment
relationships the processes involved (e.g., feeding behavior) must be analyzed
at an adequate scale (e.g., in our example, very detailed scale both in time and
space). On the other hand, if we need to overcome the stochasticity introduced
in our observation scheme by, for instance, individual foraging behavior we
must average our results on a coarser scale in both time and space.
Thus, in
GIS
distribution models, both temporal and spatial scales are gen-
erally broadened so that stochastic events can average to a null component and
thus be ignored. For instance, the stochasticity associated with the individual
selection of a particular site, which greatly influences the distribution at a local
scale, is overcome when dealing with distributions at regional scale averaging
the selection of different individuals. In a similar way, stochastic events such as
local fires, which influence regional distributions when measured over a short
time interval (e.g., 5-10 years), are considered outliers in an analysis that takes
into account the average vegetation cover over a longer time or a wider spatial
span. Similarly, we know that in short time intervals the population dynamics
status of a population is highly unpredictable, whereas it may be more easily
averaged on longer time scales (Levin 1992) to become scarcely predictable
again at even longer intervals.
A similar consideration is intrinsic in the minimum mappable unit (
MMU
),
a concept used largely to address spatial scale issues in
GIS
species distribution
models (Stoms 1992; Scott et al. 1993) that can be readily extended to the
time scale. MMU can be seen from two points of view. On one hand, it is a
