Data analysis
Home ranges were estimated using HawthsTools, for ArcMap
9.2. For comparative purposes, home ranges were calculated using
two methods: the minimum convex polygon (MCP) using 100% of
locations, and fixed kernel using 95% of locations. Because of its
ease of calculation, many investigators use MCP to describe home
ranges, and we therefore present these analyses to facilitate comparison
with other studies. However, the kernel method provides a
more detailed home range map, identifying areas of intensive use
(Powell 2000) through concentric contour lines (95, 50, 20, 10 and
5%) indicating the probability of finding an animal in a given region
of its range (i.e., the utility density distribution, Powell (2000)). The
bandwidth (h) chosen for fixed kernel was 15 m, which we considered
as a good approximation of the perception radius (Powell
2000) of this species (see also Montgomery and Sunquist 1975).
Scaling factor used was 108 and raster cell size was 10 due to the
small areas under analysis.
Interpretation of aerial photographs provided by Instituto de
Estudos Socioambientais do Sul da Bahia (unpublished data) was
used as an initial source for the land cover of the region. However,
this method did not discriminate cabrucas from secondary logged
forest, due to similarities in canopy structure between these two
vegetation types. We therefore modified the original maps using
field observations in order to delimitate more accurately the extent
of cabrucas in the study site. The home ranges (MCP and kernel)
were clipped from the land cover map using the intersect function
from ArcMap 9.2. We then calculate the proportion of cabruca
falling within each home range using the resulting shape file.
We used contingency tables (Chi-square goodness-of-fit-tests)
complemented by residual analysis to assess if the radio-collared
sloths used the habitats according to their availabilities. First, we
contrasted the number of location fixes on each major habitat type
inside the MCPs (swamp/pasture, cabruca, early secondary forest
and late secondary forest) with the expected frequencies of
fixes based on the availability of these habitats in the MCPs. We
pooled the swamp with pasture as all fixes falling in pasture were
in fact in the very edge between pasture and swamp vegetation.
In a second Chi-square we contrasted the observed frequency of
fixes with the expected value based on habitat availability in a
buffer area surrounding the home ranges of the three sloths. To
calculate the width of this buffer area we measured the largest
distances between fixes inside each individual home range and
averaged them. The resulting width (487m) encompassed an area
of 214.7 ha (6.8 times larger than the combined area of the three
MCPs). Although this wide buffer may include land-use types differing
from the immediate study area, a narrower band would not
be representative of the larger landscape context. These tests were
performed using BioEstat version 5.0, a free statistical software
available at Instituto de Desenvolvimento Sustentável Mamirauá
(http://www.mamiraua.org.br).