Spatial Capture-Recapture analysis
We first considered all candidate models ignoring habitat heterogeneity.
Among the 42 homogeneous models evaluated (Supplementary Table S2), the
top model supported session i.e. density varied by year and baseline
detection probability varied with sampling occasion, sex and year. The
movement parameter σ varied with the interaction of sex and year.
We included habitat covariates for density and detection probability in
the top model and evaluated 24 different inhomogeneous candidate models
(Supplementary Table S3). The homogeneous null model ranked as the ninth
model with an approximately 25 AIC difference relative to the top model
that included habitat characteristics (Table 2). According to the most
supported inhomogeneous model, white-tailed deer density was dependent
on landcover-type (agricultural areas, coniferous forests, mixed forests
and transitional woodland) and detection probability varied with
landcover-type (coniferous forest, mixed forests, transitional woodland)
as well as distance to agricultural areas (Table 2, Table 3). Distance
to water bodies or distance to artificial areas were not supported for
either density or detection probability (Supplementary Table S3).
White-tailed deer densities were highest in agricultural areas and mixed
forest and lowest in coniferous forests and transitional woodlands
during both years (Figure 2, Table 3). Density was session-dependent and
was higher during the second year compared to the first year (Figure 2,
Table 3).
Detection probability i.e. the probability of capturing an individual at
its home range center was highest in transitional woodlands, second
highest in mixed forests and lowest in coniferous forests (Table 4).
Detection probability decreased with distance to agricultural areas.
During the first year, detection probability was higher than during the
second year, which was expected, as the sampling interval was shorter in
the second year. Overall, females had a slightly higher probability of
being detected than males, with p0=0.10 for females and
p0=0.08 for males. The average detection probability was
0.09 and varied between 0.02 and 0.24 (Table 4).
The spatial scale parameter σ varied with the interaction of sex and
year. For females, σ was larger in the first sampling year compared to
the second. Male σ did not differ between years. Female σ was higher
than that of males in the first year but in the second year σ was
similar between the sexes. However, the confidence intervals generally
overlap (Figure 3). Because space use by males was smaller than that of
females, the estimated sex ratio (ψ) based on SCR under the top model
was about equal (0.52) despite the fact that more female than male
individuals were identified (Table 1).
Density estimates were higher for the heterogeneous landscape compared
with assuming a homogeneous landscape. The homogeneous top model
predicted that the overall density of white-tailed deer across the whole
state space was 111.7 (109.5-113.8) deer/1000 ha in 2016 and 254.9
(249.8-259.9) deer/1000 ha in 2017. The inhomogeneous top model with
habitat covariates predicted that the overall density was 131.0
(126.1-135.9) deer/ 1000 ha in 2016 and 317.0 (304.9-329.0) deer/ 1000
ha in 2017.