Does hunting or hiking affect wildlife communities in
protected areas?
Roland Kays1,2,3,*, Arielle W. Parsons1, Megan C. Baker4, Elizabeth L. Kalies5,
Tavis Forrester4, Robert Costello3, Christopher T. Rota6,7, Joshua J. Millspaugh6 and
William J. McShea4
1North Carolina Museum of Natural Sciences, 11 West Jones St, Raleigh, NC 27601, USA; 2Department of Forestry
& Environmental Resources, North Carolina State University, Raleigh, NC 27695, USA; 3Smithsonian’s National
Museum of Natural History, 10th & Constitution Av NW, Washington, DC 20560, USA; 4Smithsonian Conservation
Biology Institute, 1500 Remount Rd., Front Royal, VA 22630, USA; 5The Nature Conservancy, Durham, NC 27701,
USA; 6Department of Fisheries and Wildlife Sciences, University of Missouri, Columbia, MO 65203, USA; and
7Division of Forestry and Natural Resources, Wildlife and Fisheries Resources Program, West Virginia University,
Morgantown, WV 26506, USA
Summary
1. Managed public wild areas have dual mandates to protect biodiversity and provide
recreational opportunities for people. These goals could be at odds if recreation, ranging
from hiking to legal hunting, disrupts wildlife enough to alter their space use or community
structure.
2. We evaluated the effect of managed hunting and recreation on 12 terrestrial wildlife spe-
cies by employing a large citizen science camera trapping survey at 1947 sites stratified across
different levels of human activities in 32 protected forests in the eastern USA.
3. Habitat covariates, especially the amount of large continuous forest and local housing
density, were more important than recreation for affecting the distribution of most species.
The four most hunted species (white-tailed deer, raccoons, eastern grey and fox squirrels)
were commonly detected throughout the region, but relatively less so at hunted sites.
Recreation was most important for affecting the distribution of coyotes, which used
hunted areas more compared with unhunted control areas, and did not avoid areas used by
hikers.
4. Most species did not avoid human-made trails, and many predators positively selected
them. Bears and bobcats were more likely to avoid people in hunted areas than unhunted pre-
serves, suggesting that they perceive the risk of humans differently depending on local hunt-
ing regulations. However, this effect was not found for the most heavily hunted species,
suggesting that human hunters are not broadly creating ‘fear’ effects to the wildlife commu-
nity as would be expected for apex predators.
5. Synthesis and applications. Although we found that hiking and managed hunting have
measureable effects on the distribution of some species, these were relatively minor in com-
parison with the importance of habitat covariates associated with land use and habitat frag-
mentation. These patterns of wildlife distribution suggest that the present practices for
regulating recreation in the region are sustainable and in balance with the goal of protecting
wildlife populations and may be facilitated by decades of animal habituation to humans. The
citizen science monitoring approach we developed could offer a long-term monitoring proto-
col for protected areas, which would help managers to detect where and when the balance
between recreation and wildlife has tipped.
Key-words: camera trap, citizen science, hiking, hunting, mammals, park, protected area,
protected forest, recreation, wildlife communities
*Correspondence author. E-mail: Roland.Kays@ncsu.edu
© 2016 The Authors. Journal of Applied Ecology © 2016 British Ecological Society
Journal of Applied Ecology 2016 doi: 10.1111/1365-2664.12700
Introduction
Most protected areas have a double mandate to protect
natural resources while also allowing recreation (Hammitt,
Cole & Monz 2015). This includes nonconsumptive recre-
ation such as hiking, biking and horseback riding, and, in
many parts of the world, also consumptive recreation such
as managed hunting and trapping. Recreation is important
for maintaining public support for protected areas, con-
necting people with nature (Louv 2005), and is the third
largest component of the United States economy, with
$646 billion spent annually (Outdoor Industry Association
2012). Recreation could also be a major disturbance to
wildlife within protected areas, potentially reducing biodi-
versity, and thus be counter to natural resource manage-
ment goals (Hammitt, Cole & Monz 2015).
Unregulated hunting can quickly lead to population
declines and extinction (Schipper et al. 2008), but the
impacts of regulated hunting on wildlife communities are
less severe. In North America, restrictions on harvest
methods, bag limits and hunting seasons are managed
locally with the goal of sustainable harvests (Mahoney &
Jackson 2013). Nonetheless, hunting is widespread on the
continent with over 13 million participants (Outdoor
Foundation 2014), such that human hunters outnumber
wolves (Canis lupus) and cougars (Puma concolor) 165 to
1 (Mech & Peterson 2003; Caso et al. 2008). Thus, human
hunters could be ecologically acting as predators, having
direct (population) and indirect (behavioural) effects on
wildlife. Managed hunting has been shown to affect popu-
lations of the targeted game species (Behrend et al. 1970;
Vucetich, Smith & Stahler 2005), but population effects
on sympatric species are rarely evaluated. Another unan-
swered question is the importance of indirect effects of
human predators through fear mediated behavioural mod-
ifications (Creel & Christianson 2008). Although studies
have shown deer change behaviour during short hunting
seasons (e.g. Little et al. 2015), the longer term beha-
vioural effects of hunting have not been the subject of
much study (Cromsigt et al. 2013).
Nonconsumptive recreation should have fewer impacts
on wildlife in protected areas than hunting or trapping
activities, but is still a concern because there are so many
more participants (13 million hunters vs. 376 million non-
aquatic, nonconsumptive recreationalists in the United
States in 2013 (Outdoor Foundation, 2014)). Although
some studies have documented the avoidance of hikers by
wildlife (Erb, McShea & Guralnick 2012; Hammitt, Cole &
Monz 2015), others found prey species attracted to busier
trails, presumably using humans as shields against preda-
tors (Muhly et al. 2011). There are numerous potentially
mediating factors, including predator–prey dynamics and
habituation to humans, that make it difficult to scale up
individual responses to community-wide effects (Tablado &
Jenni 2015). One study evaluated the effects of quiet recre-
ation on predators and found an alarming fivefold decline
in abundance of four carnivore species in areas that allowed
hikers (Reed & Merenlender 2008). The effects of consump-
tive recreation (hunting and trapping) on wildlife are pre-
sumably greater than the quiet non-consumptive recreation
studied by Reed & Merenlender (2008), raising the spectre
of a potentially large conflict between the preservation and
recreation mandates of protected areas. More broad-scale
surveys are needed to test the generality of a recreation
effects in other regions and species.
We evaluate the relative importance of consumptive
and nonconsumptive recreation on the wildlife communi-
ties within public forests of eastern North America using
camera traps to compare animal use of 1947 sites strati-
fied by their recreation use levels. We surveyed a balanced
pairing of public forests with similar habitat but differing
in whether they allowed hunting. Within an area, we
stratified sites by placing cameras on, near and far from
hiking trails. If human hunting activity has strong effects
on wildlife communities, we expect to see large differences
in their use of paired sites, and if animals avoid humans,
we expect them to avoid trails and to be less common in
heavily used areas. Most of our data come outside of the
main fall hunting season, allowing us to test for the ulti-
mate long-term effects of hunting without the proximate
avoidance of active hunters. Finally, we test the hypothe-
sis that there will be an interaction between hunting and
hiking leading to animals being more sensitive to all recre-
ating humans in areas where they are hunted.
Materials and methods
SITE SELECTION
We surveyed wildlife at 1947 sites within 32 protected areas
across six states (Fig. 1, Table S1 in Supporting Information).
We targeted larger protected areas (29 sites were >10 km2) to
ensure that most animals within the area were subjected to the
given treatment (hunted or not hunted), and not moving between
areas. Most of our sites were selected as 13 pairs having similar
habitat and landscapes but differing in whether hunting was
allowed or not (Fig. 1). Rock Creek Park conducted their first
ever deer cull during our study, but we considered it unhunted
because of the restricted hunt area and use of sharpshooters over
bait. Two other unhunted areas had no comparable hunted area.
All sites were predominantly forested, but varied in elevation (4–
1152 m) and in degree of development of the surrounding land-
scape (0–187 houses km
2).
CIT IZEN SCIENCE CAMERA TRAPPING
We recruited and trained 352 volunteer citizen scientists and uni-
versity students to deploy camera traps within our study areas from
2012 to 2014. Camera surveys were generally run from April to
June or August to November, with only 23 camera runs extending
into the main deer gun-hunting season in late November. Cameras
were deployed in groups of three, with one camera placed on a hik-
ing trail, one 50 m from the trail and one about 200 m from the
trail (Fig. 2). Volunteers used Reconyx (RC55, PC800 and PC900)
© 2016 The Authors. Journal of Applied Ecology © 2016 British Ecological Society, Journal of Applied Ecology
2 R. Kays et al.
and Bushnell (Trophy Cam HD) camera traps that were equipped
with an infrared flash. These cameras all function similarly in hav-
ing highly sensitive triggers and quick trigger times, allowing them
to record animals passing in front of the camera without the addi-
tion of bait. Volunteers attached cameras to trees at 40 cm above
the ground and returned after 3 weeks to retrieve images and move
the cameras. Cameras were set on maximum trigger sensitivity and
recorded multiple photographs per trigger, re-triggering immedi-
ately if the animal was still in view.
Volunteers used the eMammal software to identify all wildlife
species in camera trap images and uploaded pictures to the
eMammal Expert Review Tool, where we confirmed or corrected
all volunteer species identifications (McShea et al. 2016). We
grouped consecutive photographs into sequences if they were
<60 seconds apart, and used these sequences as independent
records for subsequent analysis of detection rate (sequences per
day) and occupancy patterns.
COVARIATES
We obtained covariates for each site to test the relative impor-
tance of habitat, recreation and land management on wildlife. We
also included nuisance covariates in all models to account for
variation not directly related to our main hypotheses. We initially
considered 46 covariates (Table S1) but reduced those by remov-
ing any that were correlated >060 and those that performed
poorly in univariate exploratory analyses (Table S2). All covari-
ates were mean-centred. The resulting 17 covariates were used in
our analyses (Table 1) along with 4 interaction terms.
We used ARCMAP (Version 10.1) to obtain habitat covariates
for each of our camera sampling points (Table 1). We calculated
average housing density in a 5-km radius using the Silvis housing
density data set (Hammer et al. 2004). We used the LANDSCAPE
FRAGMENTATION TOOL v2.0 (Vogt et al. 2007) and the NLCD
(Fry et al. 2011) to create landcover layers representing the per
cent of large core forest and edge habitat in a 5-km radius. We
also used the NLCD (Fry et al. 2011) to calculate the per cent of
agricultural and recently disturbed habitat in a 5-km radius
around each camera point. We calculated the distance to the
nearest camera site (Nearest_neighbor, NN) to take into account
potential spatial autocorrelation effects (Dormann et al. 2007).
We used the Env-Data tool (Dodge et al. 2013) to obtain camera
site-level Normalized Difference Vegetation Index (MODIS Land
Terra Vegetation Indices 1 km monthly NDVI daily) and average
daily camera site-level temperature (ECMWF Interim Full Daily
SFC Temperature (2 m above-ground)), precipitation (NCEP
NARR Precipitation Rate at Surface) and cloud cover (NCEP-DOE
Surface Total Cloud Cover Entire Atmospheric Column).
As a covariate for nonconsumptive recreational use, we used
the detection rate for groups of people recorded by our camera
traps on trails as a quantitative measure of human use for a pro-
tected area. We coded hunting as a categorical yes/no and coded
Fig. 1. Map of the 32 protected areas surveyed with inset showing the details of camera trap placements at one pair of sites.
© 2016 The Authors. Journal of Applied Ecology © 2016 British Ecological Society, Journal of Applied Ecology
Hunting and hiking effects on wildlife 3
camera distance as a continuous variable describing cameras on
(0 m), near (50 m) or far (200 m) from a hiking trail.
To gauge the overall level of hunting for different species in the
region, we obtained one year of harvest data from two states
(North Carolina Wildlife Resources Commission 2012; Virginia
Department of Game and Inland Fisheries 2012). These values
were used to rank and understand the overall hunting pressure
(high, medium, low) but were not used as covariates in any model.
ANALYTICAL APPROACH
To test hypotheses about the effect of recreation on wildlife, we
created models that predict the space use of animals based on
data describing the habitat, recreation and management of the
protected areas surveyed. As indicators of space use, we analysed
two complementary measures from the camera trap data: occu-
pancy and visit frequency (VF) (Fig. 2). Occupancy results
describe the probability a site is occupied by a given species and
is analysed using a hierarchical model structure that also
accounts imperfect detection (MacKenzie et al. 2006). VF is sim-
ply the detection rate (sequences per day) for a species at a site
and is comparable across our sites because we did not use bait,
and selected locations in a stratified random design. These mea-
sures are complementary in that occupancy describes the geo-
graphical spread of a species across the landscape while VF
describes the relative degree of use of areas as a reflection of local
habitat preference.
For both model types, we evaluated the importance of recre-
ation to the distribution of wildlife by comparing the relative
importance of predictor covariates in two ways, using the all-sub-
sets approach. First, we considered the results of restricted mod-
els using only covariates from one class of predictor variables
(habitat, recreation or management) and compared model perfor-
mance to test which class of covariates alone was most impor-
tant. Secondly, we considered combinations of all classes of
covariates to create multivariate models that show the combined
effects and allowed for interactions (Fig. 2).
OCCUPANCY MODELS
We used the single season occupancy modelling framework of
MacKenzie et al. (2006) and estimated detection probability (P),
defined as the probability of detecting an occurring species at a
camera site, and occupancy (w), defined as the expected proba-
bility that a given camera site is occupied, for each species. We
hypothesized that eight covariates could cause heterogeneity in
detection probability (hereafter, ‘detection covariates’; Table 1).
By modelling day and week of sampling, we capture variability
in weather and other attributes that are difficult to measure (e.g.
changing resource availability), all of which can affect ani-
mal behaviour and detectability. We fixed nuisance covariates
and tested two detection models, one accounting for people use
of a site and one without, coupled with the global occupancy
model, to identify the top model of detection per species.
We included the nuisance variables LatbyLong and Near-
est_neighbor (NN) in all occupancy models, then used the
remaining 8 covariates and 4 interactions that we hypothesized
could cause heterogeneity in occupancy (hereafter, ‘occupancy
covariates’), to apply an all-subsets approach using the highest
ranked detection model and all combinations of occupancy
covariates.
We constructed these models (n = 717 per species) using the
RMark package (Laake 2011) and MuMIn package (Barto
n
2014) in R (R Development Core Team 2015). We used quasi-
Akaike’s Information Criterion (QAIC) for each species due to
overdispersion of the global model. For each model, we com-
puted QAIC, difference in QAIC (DQAIC) and Akaike weights
(wij, weight of covariate i for species j; Burnham & Anderson
2002) and used these values to assess model fit. We ranked rela-
tive covariate importance by summing wij, across all models in
Hunted
1947 camera points
32 protected areasData
Metrics
Study
design
12 species
Occupancy (VF) Visit
frequency
On trail
Near trail
Off trail
Restricted
models
Habitat
vs.
Recreaon
vs.
Management
Full models
Habitat
+
Recreaon
+
Management
Analyses
Not hunted
Fig. 2. Schematic of our approach to evaluate the importance of
consumptive and nonconsumptive recreation on wildlife in
protected areas.
© 2016 The Authors. Journal of Applied Ecology © 2016 British Ecological Society, Journal of Applied Ecology
4 R. Kays et al.
which a given covariate occurred and used cumulative weights to
rank relative covariate importance for each species. Larger values
of wij are indicative of greater importance for covariate i relative
to other covariates in the model (Burnham & Anderson 2002).
We considered wij ≥ 050 indicative of a strong occupancy
response to the covariate and wij < 050 a weak response. We
also calculated model-averaged parameter ð~bjÞ estimates and
unconditional standard errors (SE) for each covariate across
models with DQAIC <4 to assess the direction of response by
each species. For each species, we estimated w using the top
ranked model.
VIS IT FREQUENCY MODELS
We assumed the number of detections of each species obtained at
camera trap site i was a Poisson random variable:
yiPoissonðkiÞ eqn 1
We modelled the expected number of photographs at site i as a
loglinear model:
logðkiÞ ¼ xibþ offseti þ ei;
where xi is a vector of covariates and b is a conformable vector
of slope parameters; the offset term is equal to the log of the
number of days camera trap i operated; and e ~ iid N(0, r2) and
is meant to capture additional variation in the number of detec-
tions of each species. We assumed independent normal prior
distributions for the slope parameters (b ~ multivariate normal
(0, 10I), and we assumed a uniform prior distribution for the ran-
dom error standard deviation parameter (r ~ Uniform (0, 10)).
We assessed model fit with posterior predictive checks (PPC)
(K
ery & Schaub, 2012, Gelman et al., 2014). Briefly, we calcu-
lated the sum of squared Pearson residuals from observed data
(T(y)) and from data simulated assuming model (1) was the data-
generating model (T(ysim)). We calculated a Bayesian P-value as
pB = Pr(T(ysim) > T(y)) from posterior simulations and assumed
adequate fit if 01 < pB < 09.
We ran a set of 5 visit frequency (VF) models for each species
(Table S8). We fit VF models in OPENBUGS v3.2.3 (Lunn et al.,
2000) via R2OPENBUGS v3.2 (Sturtz, Ligges & Gelman, 2005) in
R v3.1.0 (R Core Team, 2015). We based inference on posterior sam-
ples generated from three markov chains. We used trace plots to
determine an adequate burn-in phase. After discarding burn-in
samples, we saved every 10th sample and ran simulations until all
chains adequately converged (R^ ≤ 11 (Gelman et al., 2014, p. 287)).
Results
CAMERA SURVEY DATA
Over 42 872 camera nights, we obtained 30 975 detections
of people and 53 372 detections of wildlife. There were 12
species with at least 200 detections, which we considered
the minimum necessary to ensure model convergence:
eastern chipmunk (Tamias striatus, hereafter ‘chipmunk’),
Table 1. Covariates used in the visit frequency (VF) and occupancy analyses. Each variable is categorized according to our main
hypotheses, including nuisance covariates included to account for variation not directly related to our main hypotheses
Category Covariate Description Type Model type
Habitat Ag_5 km % Agriculture (crop fields) in 5 km neighbourhood† GIS 5 km Ψ, VF
Habitat LC_5 km % Large core (cont. forest frag >5 acre) in 5 km
neighbourhood‡
GIS 5 km Ψ, VF
Habitat Edge_5 km % Edge (large core) and Perforated (small core) pixels in
5 km neighbourhood‡
GIS 5 km Ψ, VF
Habitat HDens_5 km Average Housing Density (houses km
2) in 5 km radius§ GIS 5 km Ψ, VF
Recreation Cam_distance Distance camera was placed from a trail (0, 50 or 250 m) Camera site Ψ, VF
Recreation Hunting Categorical covariate for hunting or no hunting permitted in
the park (0,1)
Camera site Ψ, VF
Recreation People_site Total # of people recorded by the camera over the sampling
period
Camera site Ψ, p, VF
Management Dist_5 km % Disturbed (burned, logged, grassland conversion) in 5 km
neighbourhood†
GIS 5 km Ψ, VF
Nuisance LatbyLong Latitude 9 longitude Camera site Ψ, VF
Nuisance NN Distance to nearest camera neighbour (m) Camera site Ψ, VF
Nuisance NDVI_site Moderate Resolution Imaging Spectroradiometer Land Terra
Vegetation Indices 1 km monthly NDVI daily; averaged at
the site level for the days sampled*
Camera site p, VF
Nuisance Cloud National Center for Environmental Prediction-DOE Surface
Total Cloud Cover Entire Atmospheric Column*
Camera site/day p, VF
Nuisance Temp ECMWF Interim Full Daily SFC Temp (2 m above-ground)* Camera site/day p, VF
Nuisance Precip NCEP NARR Precipitation Rate at Surface* Camera site/day p, VF
Nuisance Week Week of the year Camera site/day p, VF
Nuisance Year Year sampled (year 1 or 2) Camera site p, VF
Nuisance Det_dist Maximum distance at which camera detects animals (m) Camera site p, VF
*EnvData (Dodge et al. 2013).
†GAP landcover data set 2006 (http://gapanalysis.usgs.gov/gaplandcover/).
‡Landscape Fragmentation Tool 2006 (http://clear.uconn.edu/%5C/tools/lft/lft2/index.htm).
§Hammer et al. Landscape and Urban Planning 69 (2004) 183–199 (http://silvis.forest.wisc.edu/old/Library/HousingData.php).
© 2016 The Authors. Journal of Applied Ecology © 2016 British Ecological Society, Journal of Applied Ecology
Hunting and hiking effects on wildlife 5
coyote (Canis latrans), American black bear (Ursus ameri-
canus, hereafter ‘bear’), bobcat (Lynx rufus), gray fox
(Urocyon cinereoargenteus), red fox (Vulpes vulpes), white-
tailed deer (Odocoileus virginianus, hereafter ‘deer’), wild
turkey (Meleagris gallopavo, hereafter ‘turkey’), northern
raccoon (Procyon lotor, hereafter ‘raccoon’), eastern gray
squirrel (Sciurus carolinensis, hereafter ‘gray squirrel’),
eastern fox squirrel (Sciurus niger, hereafter ‘fox squirrel’)
and Virginia opossum (Didelphis virginiana, hereafter
‘opossum’).
HARVEST DATA
State harvest records for the region show high
(>100 000 killed year
1) hunting levels for deer and gray
squirrels, medium intensity (10 000–100 000) harvest of
raccoon, fox squirrel, coyote and turkey, and a low
(<10 000) harvest of foxes, bobcat and bear (Fig. S1).
Opossums can be harvested in both states but no records
are kept, and hunting pressure is presumed to be low.
Chipmunks are not game species in the region and can
only be legally taken through special nuisance permits.
RESTRICTED MODEL RESULTS
The restricted model sets present the results of using vari-
ables from one class of covariates for both occupancy and
VF and showed that habitat alone explained the distribu-
tion of most species better than recreation or management
alone (Table 2).
The top occupancy models for most species (bear, bob-
cat, chipmunk, coyote, fox squirrel, gray fox, opossum,
red fox, Turkey) performed well as they ranked higher
than the associated null models (which had DQAIC values
ranging from 11 to 202). Occupancy models did not per-
form as well (DQAIC < 3 compared with null model) for
the three more ubiquitous species (gray squirrel, deer and
raccoon) because there was little variation to be explained
(i.e. the species occupied most sites).
Our models of VF showed strong relationships and
good fit (01 < pB < 09) for all species, most of which
had had strong habitat associations. Coyotes, deer and
fox squirrel had recreation as the top VF model, while the
null model, including only nuisance variables, was best
for gray fox and chipmunk.
FULL MODEL RESULTS
The full multivariate occupancy (Table 3, Tables S3–S6)
and VF (Table 4, Tables S7–S11) models for most species
included habitat and recreation covariates, showing that
both types of variables have some effects on animal distri-
bution. Hunting had a negative effect for the four most
hunted species (deer, raccoon, gray and fox squirrels) but
a positive effect for coyotes and turkeys. Hiking trails
were not a significant factor for most species, but were
positively associated with coyote and bobcat VF.
The most important habitat factors across the animal
community were an interaction effect between housing
density and large core forests, which was detected with
both modelling approaches (Tables 3–4). In all cases, this
interaction was due to a change in the relationship at
higher housing density where there were no large core for-
ests, and some more sensitive species were absent. A num-
ber of species moderately well adapted to humans (deer,
raccoons, fox squirrels and opossums) were negatively
associated with large core forests, positively associated
with low-density housing and negatively associated with
high-density housing (Table 5). Red foxes were the most
specialized urban species in our study, with positive asso-
ciations towards houses at both scales, and an avoidance
of large natural areas. Bobcats were the most specialized
wilderness species in our study, with a positive association
with large core forests and a negative association with
houses. Similar relationships were seen in the occupancy
results, except that deer and red fox were positively asso-
ciated with high housing density, turkeys were positively
associated with large core forests, and fox squirrels were
Table 2. Results from restricted model sets predicting the occupancy and visit frequency of sites within protected areas by twelve wildlife
species. These models use only covariates related to habitat, recreation or management of the area. Values reported here are the differ-
ences in QAIC (occupancy models) or DIC (visit frequency models) scores compared with the top restricted model; thus, zero represents
the best restricted model and those that performed worse have higher values. Covariates used in each model set are described in Table 1
Gray
squirrel Deer Raccoon
Fox
squirrel Coyote Turkey Gray fox Red fox Bobcat Bear Opossum Chipmunk
Occupancy models
Habitat 0 0 0 0 14 2 0 0 0 0 0 0
Recreation 4 2 3 111 0 0 9 37 84 92 11 35
Management 5 2 4 44 17 10 15 56 94 35 10 35
Null 3 0 3 203 17 36 81 116 135 133 12 60
Visit frequency models
Habitat 0 23 0 32 93 0 29 0 0 0 0 10
Recreation 93 0 62 0 0 26 52 107 74 9 176 3
Management 24 20 7 33 31 30 35 95 156 36 84 13
Null 28 25 54 22 168 24 0 72 51 117 47 0
© 2016 The Authors. Journal of Applied Ecology © 2016 British Ecological Society, Journal of Applied Ecology
6 R. Kays et al.
significantly associated with houses and negatively associ-
ated with large core forests (Table S11).
For three carnivores, we also found an interaction
between the species response to people on trails and the
hunting status of a site (Fig. 3, Tables 3–4). Red foxes
had higher occupancy and VF at sites with high use by
people, and this relationship was stronger in hunted areas.
Bear and bobcat both avoided trails more strongly at
hunted sites; these sites had substantial difference in the
rate that people used the trails, which we suspect may
have contributed to the significance of this interaction
term. Four species also showed statistical interactions
between trails and people, reflecting different slopes in
their response to the drastically different detection rates
of people on and off trails. Turkeys, chipmunks and bob-
cats had lower occupancy rates at heavily hiked trails
(Table 3), while gray squirrels had higher visit frequency
on trails heavily used by people (Table 4).
Table 3. Full multivariate occupancy model results for 12 species summarized to show which environmental covariates had significant
positive (+) or negative (
) relationships with the model-averaged coefficients. Significant interaction terms are indicated with a (*); see
the text and graphs for a discussion of the direction of their effects. Cell shading separates the predictor covariates into habitat (top),
recreation (mid) and land management categories (bottom)
Gray
squirrel Deer Raccoon
Fox
squirrel Coyote Turkey
Gray
fox Red fox Bobcat Bear Opossum Chip.
HDens_5km 

Ag_5km + 
 + + + +
LC_5km
Edge_5km 
 + 

LC_5km X
HDens_5km
* * * * * * * *
Trail + +
Hunting 
 + +
People_site 

HuntingXPeople_site * * *
TrailXPeople_site * * *
Managed Habitat
(Dist_5km)

 + 
 
 
 +
Table 4. Full multivariate visit frequency model results for 12 species summarized to show which environmental covariates had signifi-
cant positive (+) or negative (
) relationships in the model average coefficients. Significant interaction terms are indicated with a (*); see
the text and graphs for a discussion of the direction of their effects. Cell shading separates the predictor covariates into habitat (top),
recreation (mid) and management categories (bottom)
Gray
squirrel Deer Raccoon
Fox
squirrel Coyote Turkey
Gray
fox
Red
fox Bobcat Bear Opossum Chip.
HDens_5km 

Ag_5km + 
 + 
 + + + +
LC_5km + +
Edge_5km + + 
 + 
 +
LC_5km X HDens_5km * * * * * * * *
Trail + +
Hunting 
 
 
 
 + + +
People_site
HuntingXPeople_site * * *
TrailXPeople_site *
Managed Habitat (Dist_5km) 
 
 + 
 
 +
Table 5. Significant positive (+) or negative (
) relationships
between wildlife, housing density and large core forests from visit
frequency models accounting for an interaction between the two
covariates. Similar results were found in occupancy models
(Table S11)
Low regional housing
density High regional
housing density
Housing
density
Large core
forests Housing density
Deer + 
 

Bear 
 + 

Fox squirrel + 
 

Opossum + 
 

Red fox + 
 +
Bobcat 
 + +
Gray squirrel + + +
Turkey + 
 

© 2016 The Authors. Journal of Applied Ecology © 2016 British Ecological Society, Journal of Applied Ecology
Hunting and hiking effects on wildlife 7
The results of our full occupancy and VF models were
generally similar. Of 50 significant variable associations,
only one had opposite predictions from the different mod-
elling approaches, edge habitat for gray squirrel. Across
all species our occupancy models detected seven signifi-
cant relationships not present in the VF models, the VF
approach produced 11 significant relationships not found
in the occupancy models, and the two approaches have
similar significant predictions for 29 covariates.
Discussion
Although other studies have found isolated negative
impacts of recreation on animal behaviour (Hammitt,
Cole & Monz 2015), there have been few studies on com-
munity-wide impacts (Reed & Merenlender 2008), and
none that integrated the evaluation of both consumptive
and nonconsumptive recreation. Our broad-scale survey,
in collaboration with citizen scientists, shows that the
impact of recreational use on wildlife communities in pub-
lic areas is relatively minor. For most species, habitat fac-
tors were more important than recreation in models
predicting their distribution and habitat preferences. Com-
paring types of recreation, hunting appeared to have more
influence than hiking on wildlife species, as hunted sites
were correlated with a decrease in activity of the four
most hunted species, while fewer species avoided hiking
trails. Nonetheless, these hunted species remain common
throughout the region. Although we did not consider finer
points of population vital rates, animal stress or changes
to species interactions, our broad comparisons of animal
distribution suggest that recreational use of our public
areas, as presently managed in the region, is not having a
widespread harmful effect on wildlife communities.
HABITAT ASSOCIATIONS
Our modelling results support previous work of species
sensitivity to habitat type and development (Lesmeister
et al. 2015). Across species, the most important habitat
features were large unfragmented forests and the housing
density surrounding the protected area. Bears and bobcats
were primarily associated with the more wild areas with
few houses and large unfragmented forests. Contrasting
this, most other wildlife species (deer, raccoon, fox squir-
rel, gray squirrel, opossum, red fox) had higher occupancy
and a preference for using protected areas within more
fragmented habitat and surrounded by moderate densities
of houses. The amount of agriculture surrounding pro-
tected areas was positively associated with a number of
wildlife species, while heavily managed lands more often
had a negative association for forest species.
HUNTING EFFECTS
Results from our restricted model sets and variable
weights suggest that, across this animal community, hunt-
ing is secondary in importance to habitat covariates in
determining animal occupancy and habitat use. However,
there was some important effects as the four most heavily
harvested species in our study (gray squirrel, deer, rac-
coon and fox squirrel) had negative relationships between
hunting and their visit frequency, although this was not a
predictor of their occupancy (except fox squirrel). None
of these widespread species are of conservation concern;
indeed, deer overabundance is more likely to cause dam-
age to vegetation communities, or cause conflict with peo-
ple (McShea 2012). Many protected areas encourage
hunting with the goal of reducing the impact of deer on
Fig. 3. Visit frequency (VF) model results
showed that three species of wildlife
responded to the human use of trails dif-
ferently in hunted and unhunted areas.
Red fox increased their use of trails with
high human traffic (groups of people
detected during 3-week survey) in all sites,
but this was stronger in hunted areas.
Bobcats and bears both had a stronger
avoidance of people in hunted areas. Dot-
ted lines are 95% credible intervals for VF
models.
© 2016 The Authors. Journal of Applied Ecology © 2016 British Ecological Society, Journal of Applied Ecology
8 R. Kays et al.
natural resources of the public area or the surrounding
landscape, and our results suggest this is occurring. How-
ever, the effect was variable across the region, and of
moderate biological significance, with an average increase
of 31 deer (SE = 23) detected per 3 week deployment
and no reduction in their occupancy.
The species showing the strongest relationships with
recreation covariates for both occupancy and VF was the
coyote. Surprisingly, these were positive associations, with
more coyote activity in hunted areas and a preference for
using hiking trails. Some coyotes are harvested for fur,
but the primary motivation by most coyotes hunters is to
reduce the local coyote population (Stevens, More &
Glass 1994). Coyote removal experiments suggest this
strategy is not effective, probably because it disrupts their
social system, which encourages dispersing animals to set-
tle in the area (Kilgo et al. 2014). Although we did not
estimate coyote density, our higher values for coyote
occupancy and VF in hunted areas support this hypothe-
sis and suggest that managers seeking fewer coyotes in an
area should encourage stable packs which, in turn, might
discourage dispersing animals from settling in an area
(Maletzke et al. 2014).
IMPACTS OF RECREATIONAL TRAIL USE
Our comparison of animal activity on, near and far from
trails did not find strong or consistent avoidance of hiking
trails by most wildlife. Indeed, most predatory species
were actually detected most often on trails, although at
night, when few human are using trails. This trail prefer-
ence of predators has been noted for numerous tropical
species (Harmsen et al. 2010; Kays et al. 2010). Although
they did not avoid trails per se, four species (raccoon,
bear, turkey and bobcat) avoided the most heavily used
trails while (red foxes and gray squirrels) actually had
higher VF at busy trails, perhaps using humans as shields
against predators (Muhly et al. 2011). A previous study
of wildlife in the same region found similar results for
black bears and red foxes (Erb, McShea & Guralnick
2012). However, even our most heavily used trails
(>100 groups of people day
1) were used by wildlife
including deer (11/15 sites), coyotes (8/15), gray squirrels
(7/15) and red fox (5/15).
Our lack of a consistent strong relationship between
trail use and wildlife communities is counter to the star-
tling decrease in abundance of four predator species in
recreational areas in California found by Reed & Meren-
lender (2008). While all of our sites had some public
recreation, their study contrasted public recreation sites
with private land where recreation was forbidden, which
could explain the difference in our results. Additionally,
the eastern forests we surveyed are denser than the oak
woodland habitat of the California study, which might
offer animals more cover and seclusion from hikers.
Finally, there could be differences in level that wildlife
has habituated to people across the country.
INTERACTIONS BETWEEN HUNTING AND HIKING
Although a variety of studies have shown behavioural
responses of wildlife to hikers (Hammitt, Cole & Monz
2015) and hunters (e.g. Little et al. 2015), none have con-
sidered both at once, and none have evaluated the impli-
cations of these responses on the community composition
and habitat preference outside of the hunting season. The
absence of strong spatial avoidance of recreationalists sug-
gests that many species are habituated to humans in pro-
tected areas, even where hunted. This habituation was
also shown by a recent analysis of animal behaviour from
the same camera trap data set we analysed here, which
showed decreased vigilance for deer in areas with high
levels of recreation (S.G. Schuttler, A.W. Parsons, T.
Forrester, M.C. Baker, W.J. McShea, R. Costello & R.
Kays, personal communication).
We predicted that if hunters maintain a landscape of
fear for wildlife towards people (sensu Laundre, Hernan-
dez & Ripple 2010), habituation of wildlife to people
would be greater in unhunted areas. We found some sup-
port for this in our two most wilderness-dependent spe-
cies, bobcats and bears, which had a moderately sharper
avoidance of people in hunted areas. The slight nature of
this relationship, and its absence in other more heavily
harvested species, suggests that human hunters are not
having a strong indirect effect on most prey behaviours as
would be expected from native apex predators, probably
because of the limited hunting season (Cromsigt et al.
2013).
OCCUPANCY AND VIS IT FREQUENCY
To evaluate the effect of recreation on wildlife, we used
two separate metrics of space use derived from the same
camera trap data set: occupancy and visit frequency (VF).
While occupancy modelling is well-established (MacKen-
zie et al. 2006), using VF (i.e. the raw detection rate) is
more controversial (Jennelle, Runge & MacKenzie 2002).
However, there are two factors that differ between our
approach and some previous efforts. First, we used a con-
sistent field methodology, with no bait, in a stratified ran-
dom design to avoid many of the potential biases
discussed by Sollmann et al. (2013). Secondly, we inter-
pret the VF metric not as a measure of abundance, but as
an indication of local habitat preference that offers a sim-
ple measure of relative habitat use. This innovative
approach uses these two metrics as complementary mea-
sures of animal distribution (occupancy) and habitat pref-
erence (VF).
Our models using the same sets of covariates to predict
occupancy and VF gave similar results for most species.
However, the VF models performed better (vs. null
model) for the most ubiquitous species, which had very
high occupancy rates at all study sites and therefore less
variation in the presence/absence context of occupancy
modelling. Disparities between the two modelling
© 2016 The Authors. Journal of Applied Ecology © 2016 British Ecological Society, Journal of Applied Ecology
Hunting and hiking effects on wildlife 9
approaches may in part reflect a reduced sensitivity of the
occupancy approach caused by the reduction of a contin-
uous variable used in VF (detection rate) into a categori-
cal variable (occupied or not), or the hierarchical nature
of occupancy, which helps account for imperfect detec-
tion. Additionally, environmental factors could have dif-
ferent influences on occupancy vs. local habitat
preferences. Regardless, both approaches were consistent
in showing that recreation had relatively minor effect on
the distribution and habitat preferences for most wildlife
species. We recommend this complementary analytical
approach for quantifying the relative importance of envi-
ronmental factors affecting wildlife species that cannot be
identified to the individual for use with capture–recapture
density estimators.
MANAGEMENT IMPLICATIONS
The impact of recreation on wildlife in protected areas is an
important question that relates not only to management of
protected areas, but also to how modern society connects
to nature through recreation. Given the huge economic
contributions of the outdoor industry ($646 billion annu-
ally, Outdoor Industry Association 2012), this issue also
has significant economic implications. Our large-scale
study, enabled by citizen science surveys, is broadly relevant
because the 32 protected areas we included had a wide vari-
ety of management practices and degree of use by humans.
These areas are typical for the country in either permitting
seasonal hunting or not, and in allowing trail-based recre-
ation by hikers, bicyclists and/or horseback riders. We
found human trails were not widely avoided by most spe-
cies of wildlife, suggesting that the relatively sparse recre-
ation trail networks typical of this region are not having
negative effects on the broader wildlife community. Our
results show the managed wildlife harvest of some pro-
tected areas can marginally reduce the activity levels (VF)
of the most hunted species (deer, raccoons and gray squir-
rels), but do not affect their occupancy levels, and have lit-
tle impact on the distribution of most other wildlife species.
Hunting is often encouraged by managers to reduce the
negative ecological impacts of deer populations; our results
suggest the present level of harvest in most areas is resulting
in variable but significant impacts on local deer habitat
preferences (VF) but not on their occupancy. Furthermore,
our results showing an increase in coyote activity in hunted
areas suggest that direct persecution aimed to reduce their
populations may have the reverse effect, possibly due to
more immigrants filling the territories vacated by hunted
animals (Kilgo et al. 2014).
We acknowledge that recreation could still be affecting
more subtle aspects of the wildlife populations, such vital
rates or animal stress; however, we suggest that if these
resulted in important population level differences in distri-
bution or abundance, we would have detected them with
our analyses. Additionally, we think the citizen science
monitoring approach we developed could offer a
sustainable long-term monitoring protocol for protected
areas, which would help them detect where and when the
balance between recreation and wildlife has tipped.
Our results suggest that present levels of managed wild-
life harvest and nonconsumptive recreation in the region
are sustainable, without significant negative impacts to the
distribution patterns of the broader wildlife community.
Connecting the public with their natural resources
through encouraging its recreational use is not contradic-
tory, within the bounds of existing regulations. This find-
ing offers hope for maintaining wildlife diversity in the
United States and for the future of wildlife in any part of
the world where animal harvest is managed and people
look to protected areas as a place they can escape to
enjoy nature without harming it in the process.
Acknowledgements
We thank all of our 352 volunteers for their hard work collecting data for
this study, including Master Naturalists and North Carolina State Univer-
sity undergraduate classes. For their field assistance and volunteer coordi-
nation, we thank the staff at the protected areas where we worked. We
thank R. Montgomery for input on study design. This work was con-
ducted with funding from the National Science Foundation grant
#1232442 and #1319293, the VWR Foundation, the US Forest Service,
the North Carolina Museum of Natural Sciences and the Smithsonian
Institution. The authors have no conflict of interest to declare.
Data accessibility
All camera trap data used in this study are available through eMam-
mal.org and the Dryad Digital Repository, http://dx.doi.org/10.5061/
dryad.gv1dq (Kays et al. 2016).
References
Barto
n, K. (2014) MuMIn: Multi-Model Inference, version 1.10.0. https://
cran.r-project.org/web/packages/MuMIn/index.html.
Behrend, D.F., Mattfeld, G.F., Tierson, W.C. & Wiley, J.E. III (1970)
Deer density control for comprehensive forest management. Journal of
Forestry, 68, 695–700.
Burnham, K.P. & Anderson, R.D. (2002) Model Selection and Multi-
Model Inference. Springer, New York.
Caso, A., Lopez-Gonzalez, C., Payan, E., Eizirik, E., de Oliveira, T.,
Leite-Pitman, R., Kelly, M., Valderrama, C. & Lucherini, M. (2008)
Puma Concolor. The IUCN Red List of Threatened Species, Version
20.
Creel, S. & Christianson, D. (2008) Relationships between direct predation
and risk effects. Trends in Ecology & Evolution, 23, 194–201.
Cromsigt, J.P.G.M., Kuijper, D.P.J., Adam, M., Beschta, R.L., Churski,
M., Eycott, A. et al. (2013) Hunting for fear: innovating management
of human-wildlife conflicts. Journal of Applied Ecology, 50, 544–549.
Dodge, S., Bohrer, G., Weinzierl, R., Davidson, S.C., Kays, R., Douglas,
D. et al. (2013) The environmental-data automated track annotation
(Env-DATA) system: linking animal tracks with environmental data.
Movement Ecology, 1, 1–14.
Dormann, C.F., McPherson, J.M., Araujo, M.B., Bivand, R., Bolliger, J.,
Carl, G. et al. (2007) Methods to account for spatial autocorrelation in
the analysis of species distributional data: a review. Ecography, 30, 609–
628.
Erb, P.L., McShea, W.J. & Guralnick, R.P. (2012) Anthropogenic influ-
ences on macro-level mammal occupancy in the Appalachian trail corri-
dor. PLoS One, 7, e42574.
Fry, J., Xian, G., Jin, S., Dewitz, J., Homer, C., Yang, L., Barnes, C.,
Herold, N. & Wickham, J. (2011) Completion of the 2006 National
Land Cover Database for the Conterminous United States. Photogram-
metric Engineering and Remote Sensing, 77, 858–864.
© 2016 The Authors. Journal of Applied Ecology © 2016 British Ecological Society, Journal of Applied Ecology
10 R. Kays et al.
Gelman, A., Carlin, J.B., Stern, H.S., Dunson, D.B., Vehtari, A. & Rubin,
D.B. (2014) Bayesian Data Analysis, 3rd edn. CRC Press.
Hammer, R.B., Stewart, S.I., Winkler, R., Radeloff, V.C. & Voss., P.R.
(2004) Characterizing spatial and temporal residential density patterns
across the U.S. Midwest, 1940–1990. Landscape and Urban Planning, 69,
183–199.
Hammitt, W.E., Cole, D.N. & Monz, C.A. (2015) Wildland Recreation:
Ecology and Management. John Wiley & Sons, Oxford, UK.
Harmsen, B.J., Foster, R.J., Silver, S., Ostro, L. & Doncaster, C.P. (2010)
Differential use of trails by forest mammals and the implications for
camera-trap studies: a case study from Belize. Biotroica, 42, 126–133.
Jennelle, C.S., Runge, M.C. & MacKenzie, D.I. (2002) The use of
photographic rates to estimate densities of tigers and other cryptic
mammals: a comment on misleading conclusions. Animal Conservation,
5, 119–120.
Kays, R., Tilak, S., Kranstauber, B., Jansen, P.A., Carbone, C., Row-
cliffe, M., Fountain, T., Eggert, J. & He, Z. (2010) Monitoring wild
animal communities with arrays of motion sensitive camera traps. Inter-
national Journal of Research and Reviews in Wireless Sensor Networks,
1, 19–29.
Kays, R., Parsons, A.W., Baker, M.C., Kalies, E.L., Forrester, T., Cost-
ello, R., Rota, C.T., Millspaugh, J.J. & McShea, W.J. (2016) Data
from: Does hunting or hiking affect wildlife communities in protected
areas? Dryad Digital Repository, http://dx.doi.org/10.5061/dryad.gv1dq.
Kery, M. & Schaub, M. (2012) Bayesian Population Analysis using Win-
BUGS: A Hierarchical Perspective. Academic Press, Waltham, MA,
USA.
Kilgo, J.C., Vukovich, M., Scott Ray, H., Shaw, C.E. & Ruth, C. (2014)
Coyote removal, understory cover, and survival of white-tailed deer
neonates. The Journal of Wildlife Management, 78, 1261–1271.
Laake, J. (2011) RMark: R Code for MARK Analysis, version 2.0.7.
https://cran.r-project.org/web/packages/RMark/index.html.
Laundre, J.W., Hernandez, L. & Ripple, W.J. (2010) The landscape of
fear: ecological implications of being afraid. The Open Ecology Journal,
3, 1–7.
Lesmeister, D.B., Nielsen, C.K., Schauber, E.M. & Hellgren, E.C. (2015)
Spatial and temporal structure of a mesocarnivore guild in midwestern
north America. Wildlife Monographs, 191, 1–61.
Little, A.R., Demarais, S., Gee, K.L., Webb, S.L., Riffell, S.K., Gaskamp,
J.A. & Belant, J.L. (2015) Does human predation risk affect harvest sus-
ceptibility of white-tailed deer during hunting season? Wildlife Society
Bulletin, 38, 797–805.
Louv, R. (2005) Last Child in the Woods, Saving Our Children from Nat-
ure-Deficit Disorder. Algonquin Books of Chapel Hill, NC, USA.
Lunn, D.J., Thomas, A., Best, N. & Spiegelhalter, D. (2000) WinBUGS –
a Bayesian modelling framework: concepts, structure, and extensibility.
Statistics and Computing, 10, 325–337.
MacKenzie, D., Nichols, J.D., Royle, J.A., Pollock, K.H., Bailey, L. &
Hines, J.E. (2006) Occupancy Estimation and Modeling: Inferring Pat-
terns and Dynamics of Species Occurrence. Academic Press, New York,
USA.
Mahoney, S.P. & Jackson, J.J. (2013) Enshrining hunting as a foundation
for conservation – The North American Model. International Journal of
Environmental Studies, 70, 448–459.
Maletzke, B.T., Wielgus, R., Koehler, G.M., Swanson, M., Cooley, H. &
Alldredge, J.R. (2014) Effects of hunting on cougar spatial organization.
Ecology and Evolution, 4, 2178–2185.
McShea, W.J. (2012) Ecology and management of white-tailed deer in a
changing world. Annals of the New York Academy of Sciences, 1249,
45–56.
McShea, W.J., Forrester, T., Costello, R., He, Z. & Kays, R. (2016) Vol-
unteer-run cameras as distributed sensors for macrosystem mammal
research. Landscape Ecology, 31, 55–66.
Mech, L.D. & Peterson, R.O. (2003) Wolf-prey relations. Wolves: Behav-
ior, Ecology, and Conservation (eds L.D. Mech & L. Boitani), pp. 131–
160. University of Chicago Press, Chicago, Illinois.
Muhly, T.B., Semeniuk, C., Massolo, A., Hickman, L. & Musiani, M.
(2011) Human activity helps prey win the predator-prey space race.
PLoS One, 6, e17050.
North Carolina Wildlife Resources Commission (2012) 2011-12 Harvest
Survey of North Carolina Hunters.
Outdoor Foundation (2014) Outdoor Participant Report.
Outdoor Industry Association (2012) The Outdoor Recreation Economy.
Reed, S.E. & Merenlender, A.M. (2008) Quiet, nonconsumptive recreation
reduces protected area effectiveness. Conservation Letters, 1, 146–154.
R Development Core Team. (2015) R: A Language and Environment for
Statistical Computing, version 3.1.2. R Foundation for Statistical Com-
puting, Vienna, Austria.
Schipper, J, Chanson, J.S., Chiozza, F, Cox, N.A., Hoffmann, M, Katariya,
V. (2008) The status of the world’s land and marine mammals: diversity,
threat, and knowledge. Science (New York, N.Y.), 322, 225–230.
Sollmann, R., Mohamed, A., Samejima, H. & Wilting, A. (2013) Risky
business or simple solution – relative abundance indices from camera-
trapping. Biological Conservation, 159, 405–412.
Stevens, T.H., More, T.A. & Glass, R.J. (1994) Public attitudes about coy-
otes in New England. Society & Natural Resources, 7, 57–66.
Sturtz, S., Ligges, U. & Gelman, A. (2005) R2WinBUGS: a package for
running WinBUGS from R. Journal of Statistical Software, 12, 1–16.
Tablado, Z. & Jenni, L. (2015) Determinants of uncertainty in wildlife
responses to human disturbance. Biological Reviews of the Cambridge
Philosophical Society, doi: 10.1111/brv.12224.
Virginia Department of Game and Inland Fisheries (2012) 2011-2012 Hun-
ter survey report.
Vogt, P., Ritters, K.H., Estreguil, C., Kozak, J., Wade, T.G. & Wickham,
J.D. (2007) Mapping spatial patterns with morphological image process-
ing. Landscape Ecology, 22, 171–177.
Vucetich, J.A., Smith, D.W. & Stahler, D.R. (2005) Influence of harvest,
climate and wolf predation on Yellowstone elk, 1961-2004. Oikos, 111,
259–270.
Received 21 January 2016; accepted 16 May 2016
Handling Editor: Johan du Toit
Supporting Information
Additional Supporting Information may be found in the online version
of this article.
Figure S1. Annual harvest estimates showing the relative intensity
of hunting across species.
Table S1. List of protected areas surveyed and their characteristics.
Table S2. Complete list of covariates initially considered in VF and
occupancy analyses.
Table S3. Full results for occupancy model selection.
Table S4. Model averaged occupancy coefficient estimates for
predator species.
Table S5. Model averaged occupancy coefficient estimates for non-
predator species.
Table S6. Model averaged occupancy variable importance weights
for all species.
Table S7. Full VF model results for each species.
Table S8. Model weights for variable used in full multivariate
occupancy models for 12 species.
Table S9. Global model VF coefficient estimates for predator
species.
Table S10. Global model VF coefficient estimates for non-predator
species.
Table S11. Significantpositive (+) or negative (-) relationshipsbetween
wildlife, housing density, and large core forests from occupancy
models accounting for an interaction between the two covariates.
© 2016 The Authors. Journal of Applied Ecology © 2016 British Ecological Society, Journal of Applied Ecology
Hunting and hiking effects on wildlife 11