Geoscience Reference
In-Depth Information
variables that affect fitness. However, as pointed out by North and Reynolds
(1996), conclusions from studies of this sort are based on two fundamental,
but questionable, underlying assumptions: that used sites are in suitable habi-
tat and that unused sites are in unsuitable habitat. Consider the study by
Meyer et al. (1998), who observed significant habitat-related differences
among sites that were occupied versus unoccupied by spotted owls (
Strix occi-
dentalis
). They found, as expected, that these owls tended to select sites having
a large proportion and large patches of old-growth forest, but suggested that
the owls could tolerate certain amounts of clearcut and young forest (based on
the fact that used sites contained such areas). Their reproductive data were
inconclusive about what the owls could tolerate from a population standpoint,
but Meyer et al (1998:47) nonetheless asserted “that resource selection proba-
bility functions [which they generated using Manly et al.'s 1993 approach]
. . . are a reliable tool for assessing capability of landscapes to support northern
spotted owls.” Hence they concluded (abstract, p. 5) that their habitat selec-
tion model “can be used to predict the probability that a given landscape
mosaic will be a
suitable
spotted owl site” (emphasis mine). This is exactly what
I refer to as
fatal flaw #2
.
To avoid assumptions related to suitability in site attribute studies, North
and Reynolds (1996) proposed that instead of comparing habitat characteris-
tics of used sites with unused or random sites, used sites be categorized by their
intensity of use (e.g., 1-2 percent of locations = low use, 3-10 percent
medium use, more than 10 percent = high use), and habitat characteristics
compared among such categories (i.e., excluding unused sites) with a form of
logistic regression. Thus this is not really a site attribute design (according to
my definition), nor is it even applicable to the types of situations in which site
attribute studies are commonly used (i.e., in which an animal remains at a sin-
gle site for a fairly long period of time). Rather, this procedure represents just
another means of investigating habitat characteristics associated with differing
degrees of use. I previously discussed other studies of this sort (Litvaitis et al.
1986; Porter and Church 1987; Servheen and Lyon 1989; Clark et al. 1993)
as a special case of the use-availability design because, although lacking an
assessment of habitat availability, they are based on proportional use. Whereas
these methods may avoid many of the statistical and general assumptions of
site attribute studies, they digress further from the issue of habitat-related fit-
ness because the studied sites, instead of being of some special biological sig-
nificance, are simply points where the animal was located some number of
times. As North and Reynolds (1996) recognized, intensity of use may not be
a good indicator of habitat quality.
