Freitas and Rodrigues • TERRITORY SIZE AND HABITAT USE OF SERRA FINCH 
59 
observe significant seasonal changes in spatial 
distribution and general territory defense of mated 
pairs. We performed spatial analyses using Arc- 
View G1S Version 3.2 (ESR1 1998). The distance 
between nearest neighboring pairs was calculated 
from the arithmetic mean of the locations of each 
mapped pair. We calculated the density of in¬ 
dividuals in the study area by mapping territories 
■ Bibby et al. 2000). 
Territory size was calculated using the Home 
Range Extension Version 1.1 (Rodgers and Can" 
1998) We used the following estimators: 95% 
fixed kernel using the least-squares cross-valida¬ 
tion smoothing parameter (Worton 1989. Seaman 
and Powell 1996): 50% fixed kernel for the core 
areas (Powell 2000); and 100% minimum convex 
polygon (MCP; Mohr 1947). We chose the kernel 
method because it is the most recommended 
(Laver and Kelly 2008) and MCP for comparison 
because it is the most used and emphasized 
the boundaries of space used (Kernohan et al. 
2001), This is important for territory character 
ization based on the concept of use of an exclusive 
area (Pitelka 1959). We measured territory size 
only for pairs with >30 locations to avoid im¬ 
precision and bias in the size estimates (Seaman 
ctal. 1999, Kernohan et al. 2001). 
Habitat Selection.—Wo analyzed habitat selec¬ 
tion by comparing habitat use and that expected 
based on availability. We used the described ha¬ 
bitat classifications for availability and delineated 
habitats based on satellite images from Google 
forth (2010) in combination with direct observa¬ 
tions iri the field and measured with X-tools 
extension for ArcView Version 3.2 (ESRI 1998). 
We calculated habitat selection using three 
approaches. First, we compared the number ol all 
$erra Finch locations within each habitat with the 
Proportion of available habitats in the study area. 
Lbi-square tests were used to evaluate the null 
hypothesis that actual use of different habitat 
'ypes is directly proportional to their availability 
'Neu et al. 1974). Bonferroni confidence intervals 
11 = 0.05) were calculated from the observed 
Proportions of habitat use to identity which ha- 
bitat types were selected (Ncu et al. 1974, Byers 
and Steinhorst 1984). Second, we compared the 
Proportions of habitat within the 95% kernel 
territory boundaries to proportions of available 
habitats in the study area (second-order habitat 
election sens,, Johnson 1980). Third, we com¬ 
pared the proportion ol locations in each territory 
10 the proportion of available habitats within the 
95% kernel territory boundaries (third-order 
habitat selection sensu Johnson 1980). 
We used compositional analysis (Aebischer 
et al. 1993) for the second and third-order 
approach. This analysis used the Wilks' lambda 
statistic test for overall differences in habitat use. 
Comparisons of particular habitat types w'ere 
made with paired /-tests (Aebischer et al. 1993) 
if the Wilks' test suggested differential habitat 
use. Significance was set at P < 0.05. Habitat 
types were ranked from most to least selected 
using a matrix of mean and standard deviation of 
log ratio differences for all habitat types it 
selection was significantly non random. Missing 
values (zero) in the data matrix were replaced by 
O.OOI. The minimum number ot individuals 
needed for compositional analysis is six (Ae¬ 
bischer et al. 1993). Thus, all Sena Finches with 
>10 locations were used and those with <10 
locations were excluded from the analyses. Leban 
et al. (2001) noted that compositional analysis is 
primarily affected by the number of animals 
sampled and did not vary much with number ot 
locations (as few as 10 observations). We 
conducted the habitat selection analyses using 
Resource Selection for Windows, Version 1.0 
(Leban 1999). 
RESULTS 
Territory Distribution.— Twenty-one individual 
birds were marked with colored leg bands and 17 
were relocated distributed among 11 mated pairs. 
The four individuals not relocated included a 
juvenile and three possibly adult floaters. Six 
additional mated pairs of unmarked individuals 
were also detected (during at least 2 different 
days) occupying areas between some ot the 
marked pair’s territories. Those pairs were 
reliably identified because the adjacent pairs often 
sang duets at the same time. Thus, at those 
moments it was possible to hear up to three 
different pairs. Territories of 17 total pairs were 
mapped by 318 GPS locations in the entire 138-ha 
area during the study period (Fig. 1). This was a 
density of adults with established territories of 
0.25 birds/ha or 0.12 pairs/ha. Distance between 
territory centers was 162.38 ± 28.93 (range = 
119.95-233.85 m. n = 17). Territory size 
estimates (mean ± SD) of five pairs that were 
located most often ranged from 2.52 ± 0.77 ha 
(MCP) to 3.35 ± 0.90 ha (95% kernel) (Table I). 
The area of common use between territories 
(MCP) was 991.27 nr for pairs A (4.68% of 
