computerized databases. One important feature of GIS is rapid and simple construction
of tailor-made “maps” that are readily accessible from a computer screen. This allows
users to rapidly sift through complex spatial information in a visual context. Just as
important, GIS allows the user to identify and measure spatial interrelationships among
variables that would be exceedingly difficult to perform in the field. For example,
one can rapidly calculate the size of forest stands of similar species composition,
measure the distance of each of these stands from the nearest road, and calculate
what fraction of the stands fall within the home range of a wildlife species of interest.
From the point of view of assessing habitat selection, GIS also offers a convenient
means of random sampling of geographic features across complex landscapes. GIS is
clearly a technological breakthrough in the analysis of wildlife habitat needs that is
transforming the way we think about conservation and management issues.
The logical basis of virtually all measures of selective use is comparison between
the frequency of use of a particular resource (habitat) and its availability in the envi-
ronment. We surmise that a resource (habitat) is preferred when its use by animals
exceeds its availability and conversely that a resource (habitat) is avoided when its
use is less than that expected from its availability in the environment. Note our
purposeful intermingling of “resource” and “habitat.” That is because we can use a
similar analysis for determining whether animals preferentially choose diets as for
determining whether animals show preferential habitat selection.
Perhaps the easiest way to understand the resource selection procedure is to walk
through an example. The rufous bristlebird is a threatened passerine species living
in coastal areas of Australia. Gibson et al. (2004) used GIS to evaluate critical
habitat needs for bristlebirds in a site with competing land use interests (biodiver-
sity values versus mining). Along a series of trails bisecting the study area, Gibson
et al. recorded the presence (scored with a 1) or absence (0) of bristlebirds, noting
the exact geographic coordinates of each positive identification made. They later trans-
ferred these sightings to a GIS, overlaying digitized topographic data on aspect, slope,
and elevation as well as spatially explicit data on hydrology and vegetation complexity
derived from multispectral remote sensing imagery. The probability that a habitat is
used, w(x), is given by the following logistic regression model:w(x) =where the logistic regression coefficients β 1 to βkmeasure the strength of selection
for the k habitat variables replicated over the full set of sample units. The function
w(x) is bounded between 0 and 1 and represents a probability of usage, given the
set of habitat characteristics within a spatial unit. Given the descriptive nature of both
the data on bristlebird presence or absence as well as habitat variables derived from
the GIS, Gibson et al. elected to use model evaluation (Chapter 15) (Burnham and
Anderson 1998) rather than classical hypothesis testing (Chapter 16). They found
that there was a positive association between bristlebird presence and vegetation
vertical complexity, but negative associations between bristlebird presence and
“elevation,” “distance to creek,” “distance to the coast,” and “sun incidence.” This
suggests that bristlebirds require densely vegetated stands in close proximity to coastal
fringes and drainage lines. Such habitats composed approximately 16% of the study
area, demonstrating how resource selection can help in the assessment of land useexp(β 0 +β 1 X 1 +... βkXk)
1 +exp(β 0 +β 1 X 1 +... βkXk)THE ECOLOGY OF BEHAVIOR 71
WECC05 18/08/2005 14:42 Page 71