Deer resource selection analysis
Urban deer strongly selected large residential lots and areas of high
vegetation productivity (or greenness; NDVI). Model validation did not
indicate any model misspecification (Dsq = 10.21; overdispersion = 0.98;
K-fold Δ = 0.17). Deer were more than twice as likely to use an area
with each unit increase in residential lot size (β = 0.89, s.d. = 0.04,
OR = 2.44) and NDVI (β = 0.80, s.d. = 0.06, OR = 2.21) (Fig. 3). Deer
also selected areas closer to public parks (β = 0.68, s.d. = 0.04, OR =
1.97) and golf courses (β = 0.68, s.d. = 0.05, OR = 1.97) (Fig. 3). Deer
showed a weaker selection for small (β = 0.35, s.d. = 0.05, OR = 1.41)
and medium-sized residential lots (β = 0.21, s.d. = 0.04, OR = 1.23).
Deer strongly avoid areas with high road densities (β = -0.03, s.d. =
0.04, OR = 0.74). Deer did not select or avoid area of treed cover (β =
0.01, s.d. = 0.05, OR = 1.01). Effect sizes varied among landscape
features, and error was small (Fig. 4). Extrapolation of observed deer
responses (i.e., β coefficients) to natural and urban landcover
covariates across our study area highlights affluent neighbourhoods in
Oak Bay are most strongly selected by urban black-tailed deer, along
with golf courses, followed by parks (Fig. 5).