Third order selection
We first examined whether movement alone could predict third order
selection (null model), the selection of resources within the HR, or if
including habitat selection strengthened the model fit (model 1). We
found habitat selection significantly improved the model fits for most
individuals across seasons, except for 17 unique deer represented in 23
deer-seasons (23 models/160 total models showed the null model with
lowest ΔAIC). For these 17 deer, model 1 which incorporated habitat
selection, was <2 ΔAIC from the movement only models
indicating that both models adequately fit the data.
The top model describing movement and habitat selection within the HR
for male deer across pre-breeding, breeding, and post-breeding seasons
was model 4 (Tables 2, S3) providing support for speed and
directionality differences by time-of-day and habitat selected. The top
model for female deer was model 2 for breeding and post-breeding seasons
providing support for a time-of-day effect on habitat selection (Tables
2, S3). During the pre-breeding seasons, the numbers of female deer
showing model 2 and 3 as the best fit were equal. One male deer had over
95% of its observed steps in residential block-1, this animal was
removed from models because the extreme habitat specialization did not
support convergence. Deer selected natural over urban habitats most
frequently (124/160 models per deer-season), six models showed positive
selection of urban landcover in reference to natural habitat, and 30
models resulted in a neutral response to urban landcover indicated by a
non-significant coefficient for urban landcover selection in reference
to forest.
A subset of individuals (n = 27) was used to model resource
selection during breeding seasons (2016 – 2020) using fine thematic
resolution; most deer significantly avoided residential block-2 and
med-hi intensity development relative to forested habitat and
individuals showed high variation in selection strength (Figure 3b).
Deer showed stronger avoidance of increasingly developed landcover
(Figure 3b). Wetland-herb was significantly selected for in 10.34% of
models, selected against relative to forest in 41.38% of models, and
deer showed a neutral response in 48.42% of models. Low-dev was
selected for in 3.45% of models, selected against relative to forest in
55.17% of models, and 41.38% of models showed a neutral response.
Residential block-1 was selected for in 3.57% of models, selected
against in 64.29% of models, and 32.14% of models showed a neutral
response. Importantly, residential block-2 was never selected for
relative to forest, was selected against in 96.30% of models, and
3.70% of models showed a neutral response. Lastly, med-hi-dev was
selected against in 96.55% of models relative to forest and 3.45% of
deer showed a neutral response.
The simulations based on parameters from the animal with the largest
proportion of residential block-1 and low-dev within its HR (1151,
Figure 3a top) and the highest relative selection strength for block-1
(Figure 3b), showed high dispersal probability over block-1 and low-dev
and low dispersal probability over block-2 (Figure 3c left). Deer 1150’s
HR was mostly comprised of block-1 and forest (Figure 3a middle) and its
selection strength for non-forested landcover types were low or neutral
compared to forest (Figure 3b). The highest dispersal probability for
deer 1150 is centered at forested landcover and block-1 (Figure 3c
middle). Lastly, the animal with high wetland-herb landcover and the
highest proportion of block-2 in their HR (1093, Figure 3a bottom)
showed a neutral response to block-2 (Figure 3b) and high dispersal
tendency over block-2 (Figure 3c, right) with less contrast in dispersal
probability between block types when compared to deer 1151. Variation in
the simulation kernels’ size and shape results from individual variation
in step length and turning angle distributions conditioned by land use
type.
Males and females moved faster through natural than urban landcovers
across seasons, although the speed differences were greater for male
deer (Figures 4 and S8). Movement speeds for male individuals ranged
from 8m/2 hours over urban landcover to 541m/2 hours over natural
landcover while female speeds ranged from 46m/1 hour over urban
landcover to 142m/1 hour over natural landcover. The fastest average
speed was during the breeding season in natural habitats for males
(586m/2 hours) and females (156m/1 hour). Males moved an average of
175m/2 hours in natural areas with a maximum speed of 586m/2 hours and a
mean of 133m/2 hours in urban landcover with maximum movement of 252m/2
hours. Males moved slowest during the pre-breeding season through urban
landcovers. Females moved an average of 92m/1 hour and a maximum
movement of 156m/1 hour in natural landcovers and 82m/1 hour with a
maximum movement of 129m/1 hour in urban landcovers. Females moved
slowest during the breeding season through urban development.