increasing numbers of random locations (Robertson et al. 1998). Whichever way
one assesses the stability of outlines estimated round home ranges, the numbers of
locations needed increases with the precision of fit to the locations. If an ellipseis
plotted round, say, 95% of a bivariate distribution centered on the arithmetic
mean of all the locations (E 95 ), stability can occur after 12–15 locations. In this
case the area estimate can expand or contract, depending on whether the locations
add peripherally or close to the center (Figure 6.3(b)). E 95 gives a stable area
estimate with a minimal number of locations, but has low precision in analyses of
interactions with resources or other animals. At the other extreme, high preci-
sion is obtained by adding records in a grid of cells at the size of the tracking
resolution, but hundreds of locations are required for stability.
Two methods that give reasonable precision with moderate numbers of
locations are contours based on kernel functions (Worton 1989), for instance
“fixed”kernel Kf 95 (Figure 6.3(c)), and cluster analysisthat defines convexpolygons
round groups of locations, for example, Cx 95 (Figure 6.3(d)). In tests of ability to
answer biological questions, the best results were usually with one of these meth-
ods (Kenward et al.2001). It may therefore be wise to collect enough locations for
both, because that is also enough for ellipses and X 100. Indeed, stability of X 100 is
quite a good criterion, because the other methods (except grid cells) have then
gone through their initial increase phase (Figure 6.3). Provided the same standard
number of locations is used throughout, there may be some added variation in the
least stable methods, but there should not be bias.
Having decided how many locations are needed to estimate a seasonal home
range, the frequency of interval-sampling will depend on how many different ani-
mals are to be tracked at once. If it takes all day to visit 10–20 animals, then inter-
vals may be daily. It is then very important to vary the sampling order so that
individuals that may be timetabling are not always recorded at the same time each
day. Alternatively, locations of 10–20 animals that live densely may be recorded in
1–2 h, enabling 3–5 sample sessions each day. Autocorrelation analysis of spatio-
temporal correlation between consecutive locations (Swihart and Slade 1985)
may also have value for identifying optimal sampling intervals, although the orig-
inal “time to independence” tends to be overestimated.
6.4.3Using location data
Location data can be used for purposes other than home range definition. For
example, resources can be estimated in circles or ellipses of a size that either reflect
uncertainty or define availability (Arthur et al. 1996). Interactions can be deter-
mined solely from the distances between animals. However, home range outlines
also provide overlap indices, identify neighbors for sociality analyses and estimate
154 |Radio-tagging