the walk is never random: subtle landscape effects shape ...€¦ · the population. we found...
TRANSCRIPT
Molecular Ecology (2012) doi: 10.1111/j.1365-294X.2012.05681.x
The walk is never random: subtle landscape effectsshape gene flow in a continuous white-tailed deerpopulation in the Midwestern United States
STACIE J . ROBINSON,* MICHAEL D. SAMUEL,† DAVIN L. LOPEZ‡ and PAUL SHELTON§
*Department of Forest & Wildlife Ecology, University of Wisconsin-Madison, Rm 208 Russell Labs, 1630 Linden Dr., Madison,
WI 53706, USA, †U.S. Geological Survey, Wisconsin Cooperative Research Unit, University of Wisconsin-Madison, Rm 204
Russell Labs, 1630 Linden Dr., Madison, WI 53706, USA, ‡Wisconsin Department of Natural Resources, PO Box 7921,
Madison, WI 53707, USA, §Illinois Department of Natural Resources, 1 Natural Res. Way, Springfield, IL 62702, USA
Corresponde
E-mail: stacie
Published 20
Abstract
One of the pervasive challenges in landscape genetics is detecting gene flow patterns
within continuous populations of highly mobile wildlife. Understanding population
genetic structure within a continuous population can give insights into social structure,
movement across the landscape and contact between populations, which influence
ecological interactions, reproductive dynamics or pathogen transmission. We investi-
gated the genetic structure of a large population of deer spanning the area of Wisconsin
and Illinois, USA, affected by chronic wasting disease. We combined multiscale
investigation, landscape genetic techniques and spatial statistical modelling to address
the complex questions of landscape factors influencing population structure. We
sampled over 2000 deer and used spatial autocorrelation and a spatial principal
components analysis to describe the population genetic structure. We evaluated
landscape effects on this pattern using a spatial autoregressive model within a model
selection framework to test alternative hypotheses about gene flow. We found high
levels of genetic connectivity, with gradients of variation across the large continuous
population of white-tailed deer. At the fine scale, spatial clustering of related animals
was correlated with the amount and arrangement of forested habitat. At the broader
scale, impediments to dispersal were important to shaping genetic connectivity within
the population. We found significant barrier effects of individual state and interstate
highways and rivers. Our results offer an important understanding of deer biology and
movement that will help inform the management of this species in an area where
overabundance and disease spread are primary concerns.
Keywords: generalized linear model, landscape genetics, model selection, spatial ordination,
white-tailed deer
Received 27 January 2012; revision received 13 April 2012; accepted 27 April 2012
Introduction
Examining gene flow patterns for large continuous
wildlife populations presents a substantial challenge
within the landscape genetics field. Yet there is increas-
ing need to understand genetic processes within these
nce: Stacie J. Robinson, Fax: 608 262 9922;
12. This article is a US Government work and is in the p
populations, particularly in cases with overabundant
wildlife (Garroway et al. 2008), disease risks (Cross
et al. 2008), human–wildlife conflicts (Henderson et al.
2000) and negative ecological impacts of overpopulation
(Cote et al. 2004). Understanding population genetic
structure within a continuous population can provide
insights into social structure, movement, reproductive
dynamics, ecological interactions or pathogen transmis-
sion. While many landscape genetics studies focus on
ublic domain in the USA.
2 S . J . R O B I N SO N E T A L.
identifying clusters of individuals and measuring
genetic differentiation among distinct subpopulations
(Storfer et al. 2010), there is a growing interest in large-
scale continuous animal populations (Balkenhol et al.
2009; Schwartz & McKelvey 2009; Segelbacher et al.
2010; Storfer et al. 2010). Several novel methods have
recently been developed or applied in landscape genet-
ics to address these subtle questions about population
structure (Borcard et al. 2004; Dray et al. 2006; Jombart
et al. 2008; Epperson et al. 2010).
The white-tailed deer (Odocoileus virginianus) popula-
tion in the Midwestern United States provides an excel-
lent case study for application of these emerging
landscape genetic methods. The population is burgeon-
ing, widespread and continuously distributed over the
landscape. Deer are highly mobile, habitat generalists
and adapted to human populations throughout the
region (Hirth 1977; Lesage et al. 2000; Kilpatrick et al.
2001). Deer densities frequently exceed 40 per km2,
impact natural ecosystems (Bartelt et al. 2003) and are
important to local economies (over 600 000 hunting
licences annually in Wisconsin alone; Wisconsin Depart-
ment of Natural Resources, 2003). But overpopulation
also has negative impacts including forest disturbance
(Rooney 2001; Cote et al. 2004), agricultural damage
(Fagerstone & Clay 1997; Retamosa et al. 2008), elevated
risk of deer–vehicle collisions (�1.5 million annually in
the US; Mastro et al. 2008), and transmission and
spread of bovine tuberculosis and chronic wasting dis-
ease (O’Brien et al. 2002; Williams et al. 2002).
Given the high mobility and complex social structure
of white-tailed deer, different processes act to shape
population structure at different scales. At the fine
scale, deer populations are organized by social groups
of related female deer, with successive generations of
females establishing home ranges overlapping their
mother’s (as described by the ‘rose petal hypothesis’
refering to the expanding pattern of home ranges of
related females; Porter et al. 1991). This social organiza-
tion can produce localized clusters of genetically related
animals within populations (Mathews & Porter 1993;
Purdue et al. 2000; Wang & Schreiber 2001; Grear et al.
2006). Local habitat quality and arrangement can influ-
ence the home range requirements of individuals and
the degree of overlap between social groups (Kohn &
Mooty 1971; Larson et al. 1978; Schauber et al. 2007).
At a broader scale, different biological processes and
landscape features affect genetic connectivity and
admixture beyond the localized groups. In white-tailed
deer, males disperse at higher rates than females and
may move considerable distances, homogenizing
genetic composition (Hawkins & Klimstra 1970; Nixon
et al. 1991; Nelson 1993; Rosenberry et al. 1999).
Although average male dispersal distances in the Mid-
Published 2012. This article is a
west range from 5 to 10 km, and commonly up to
20 km (Larson et al. 1978; Wozencraft 1978; Ishmael
1984; Skuldt et al. 2008), both direction and distance of
dispersal may depend on habitat quality and arrange-
ment (Marchinton & Hirth 1984; Wiegand et al. 1999;
Wiens 2001; Long et al. 2005). Landscape features such
as roads, fences and rivers might additionally restrict
deer movement (Coulon et al. 2006; Perez-Espona et al.
2008).
We employed a novel combination of multiscale
investigation, multivariate techniques and spatial mod-
elling to address the complex questions of how land-
scape factors affect the degree of genetic clustering
among social groups at the local scale and shape
genetic connectivity at the landscape scale. We
addressed three research questions: (i) whether the pop-
ulation structure of white-tailed deer in the Wisconsin–
Illinois area is best described by distinct population
subdivisions or a continuous gradient of genetic varia-
tion; (ii) what habitat characteristics affect local genetic
structure at the scale of social groups; and (iii) what
habitat factors or potential dispersal barriers shape gene
flow within the population at a landscape scale. We
used a combination of spatial statistics and ordination
to describe landscape genetic patterns, but these analy-
ses fall short of explaining the mechanisms behind the
observed genetic structure. Therefore, we constructed
spatial autoregressive models within a model selection
framework to test alternative hypotheses about land-
scape influences on gene flow.
Materials and methods
Study area and landscape variables
Our research took place in chronic wasting disease
management zones of southern Wisconsin (approxi-
mately 23 300 km2 in 15 counties) and northern Illinois
(approximately 8800 km2 in six counties) (Fig. 1) where
deer tissues were collected for disease surveillance
(Shelton & McDonald 2010; Wisconsin Dept. Natural
Resources, 2010). Our study area encompassed a hetero-
geneous landscape covering several ecoregions with
varying levels of urbanization and potential natural and
anthropogenic barriers to animal movement. The region
was comprised of the heavily forested Western Coulee
and Ridge (WCR) and grassland Southwest Savanna
ecoregions (SWS) in the west, and the agriculture-domi-
nated Southeast Glacial Plains (EGP) in the east (Omer-
nik 1987). The EGP was the most urbanized landscape
in the region, containing Madison, WI, and Rockford,
IL, as well as other metropolitan areas along interstate
corridors. A continental climate dominates the study
area with warm summers and cold, snowy winters (WI
US Government work and is in the public domain in the USA.
Fig. 1 Study area for landscape genet-
ics analysis of white-tailed deer in the
Midwestern (Wisconsin and Illinois),
USA. Sample townships (black) are
shown relative to the systematic grid
used to ensure even sample coverage
over the broad study area. Data from a
pilot area (dashed outline) were used to
determine the sampling design, which
included approximately 30 individuals
from each of 64 townships.
LANDSCAPE G ENETICS IN A C ONTINUOUS DEER POPULATION 3
State Climatology Office, 2008). But winters were mild
enough that deer were not seasonally migratory, as seen
in more severe climates (Nelson 1995).
We investigated habitat characteristics and dispersal
barriers previously shown to impact deer movement
(Long 2006; Blanchong et al. 2008). Landscape variables
were summarized at the scale of Public Land Survey
System townships (9.6 km · 9.6 km) (classifications for
each township shown in Fig. 2). We used three vari-
ables to describe the broad-scale ecological characteris-
tics and site-specific habitat in each township.
Ecoregion classification (ECO) described the ecological
community in and surrounding the township (based on
the ecoregions defined above). Per cent forest canopy
cover (CAN) provided a measure of potential deer habi-
tat in a township (Rolley 2007; Storm 2011) and can
affect both population density and dispersal patterns
(Long et al. 2005; Joly et al. 2006). CAN was calculated
from the national landcover database as the per cent of
each township covered by deciduous forest canopy (Fry
et al. 2009). Because habitat fragmentation has been
related to deer space use and dispersal (Nixon et al.
1991, 2008; Etter et al. 2002; Long et al. 2005), we used
the landscape metric per cent like adjacency (ADJ) to
indicate the level of continuity or fragmentation of for-
est patches within each township (calculated as the pro-
portion of deciduous forest patches adjacent to another
forest patch vs. nonforest, using FRAGSTATS; McGarigal &
Marks 1995). We also evaluated three variables fre-
quently suggested as impediments to deer gene flow:
interstates, state-level highways and major rivers (Long
et al. 2005; Coulon et al. 2006; Joly et al. 2006; Blan-
chong et al. 2008). Interstates comprised large divided
Published 2012. This article is a US Government work and is in the p
(federally managed) highways with 2–3 traffic lanes in
each direction, whereas state-managed highways were
typically not divided and consisted of 1 (sometimes 2)
traffic lane in each direction. Major rivers were desig-
nated by the Wisconsin Department of Natural
Resources as those that maintained permanent high-vol-
ume water flow. We classified townships based on their
location on opposite sides of interstate highways I39 ⁄ 90,
I43 and I94 (INT); state highways 12, 14, 18, 20, 51, 151
and (HWY); and the Wisconsin, Rock and Yahara Riv-
ers (RIV), thus creating categorical variables represent-
ing geographic zones as shown in Fig. 2.
Sampling
We conducted a pilot study using microsatellite data
from over 1300 samples collected from 2001 to 2006
(pilot area shown in Fig. 1, data from microsatellites
obtained as described for the full data set below; Grear
et al. 2010) to determine sampling needed to represent
the genetic variation in the local population and detect
subtle genetic patterns in a large, dense, continuous
deer population with relatively high gene flow (Grear
et al. 2006; Blanchong et al. 2008). We found no tempo-
ral trend in fine-scale genetic patterns (no significant
change in allele frequencies from 2001 to 2006, calcu-
lated in GENEPOP; Raymond & Rousset 1995) and there-
fore pooled deer sampled across several years. We also
found no subdivision of populations within the four
township area considered (performed in STRUCTURE, Prit-
chard et al. 2000, K = 1 within the 310-km2 pilot area),
and thus we concluded that a single township could
adequately represent a neighbourhood of at least four
ublic domain in the USA.
(A)
(B)
(C)
(D)
(E)
(F)
Fig. 2 Study area maps show variation
in the predictor variables used in spatial
genetic models for white-tailed deer in
the Midwestern USA. Variables mea-
sured at each sample township described
potential dispersal barriers [INT
(A) = zones defined by intervening inter-
state highways, 4–6 lanes divided roads
with high traffic, HWY (B) = zones
defined by intervening state highways,
typically 2 lanes with moderate traffic,
RIV (C) = zones defined by intervening
rivers of class four or five water volume]
and habitat features [ECO (D) = ecore-
gion classification, CAN (E) = % forest
canopy, ADJ (F) = % like adjacency mea-
suring continuity of forest habitat).
4 S . J . R O B I N SO N E T A L.
townships. Our pilot analysis (rarefaction curve of allele
numbers for samples of 5–500 per township) further
indicated that the allelic composition of a township was
adequately captured by 30 deer.
We employed a hierarchical design to achieve a sam-
ple intensity representing fine-scale genetic patterns
and also sufficiently widespread to capture landscape-
scale patterns in gene flow. We divided the study
region into a grid with cells of 19.3 km · 19.3 km
(encompassing four townships; Fig. 1). We then ran-
domly sampled deer from one township from most grid
cell (in the few cases, samples were inadequate from
the randomly selected township, the township contain-
ing the most deer samples was selected, and grid cells
in which no township contained adequate samples were
skipped). This created a systematic coverage of the area
with 64 sampled townships separated by a variety of
distances. Within each township, we attempted to sam-
ple 40 adult female deer to characterize the local gene
Published 2012. This article is a
pool. We used only females to maximize the detection
of population structure and minimize noise introduced
by male migrants (Rassmann et al. 1997).
Deer were sampled between 2001 and 2010 as part of
disease monitoring programmes conducted by the WI
and IL Departments of Natural Resources (DNRs).
Lymph nodes were collected by DNR staff when animals
were harvested or registered. Samples were georefer-
enced to the Public Land Survey System section
(2.56 km2) where the animal was harvested. All WI sam-
ples were collected from hunter-harvested deer. Harvest
intensity was lower in IL, so samples were collected from
hunter-harvest and sharpshooting by IL DNR (part of
CWD monitoring and control efforts). Because sharp-
shooting was highly localized, with many deer collected
from a single shooting station, we sampled individuals
from as many shooting stations as possible and only used
animals culled on different dates to avoid sampling-
related groups that might arrive at sharpshooting
US Government work and is in the public domain in the USA.
LANDSCAPE G ENETICS IN A C ONTINUOUS DEER POPULATION 5
stations together and to distribute samples over the land-
scape similar to hunter-harvested samples.
Genetic data
Whole-genomic DNA was extracted from approxi-
mately 10 mg of lymph node using the Qiagen DNeasy
tissue extraction kit (Qiagen Ltd., Crawley, West Sussex,
UK). Extracted DNA was quantified using a Sapphire II
UV spectrometer, then diluted to 25 ng DNA ⁄ lL H2O.
Individuals were genotyped using a suite of 15 highly
variable microsatellite markers (Table S1, Supporting
information). All PCR were performed using a Qiagen
multiplex kit containing a pre-optimized mix of Taq
enzyme, dNTPs, MgCl2 and buffering chemicals. Reac-
tions were run using the thermoprofile prescribed by
the Qiagen kit, with marker-specific annealing tempera-
tures and primer concentrations detailed in Table S1
(Supporting information). The labelled fragments were
visualized using an ABI 3730 capillary sequencer
(Applied Biosystems, Foster City, CA, USA) and analy-
sed using GENEMARKER software from Softgenetics (Soft-
genetics, State College, PA, USA). We replicated PCR
and genotype analysis for approximately 10% of all
samples and determined error rates between replicates.
As preliminary analysis, we tested for deviations from
Hardy–Weinberg equilibrium (HWE) and linkage equi-
librium (LE) using GENEPOP (Raymond & Rousset 1995).
Detection of population subdivision
We used a spatially explicit Bayesian population assign-
ment algorithm to determine whether the study area con-
stituted a single population or should be subdivided into
genetically distinct groups. The test was performed in
BAPS v5.1 (Corander et al. 2008), which finds the most
likely grouping of individuals to maximize HWE and LE
through stochastic optimization. Unlike nonspatial popu-
lation assignments, which use uninformative priors, the
spatial test assumes that populations should be spatially
cohesive with higher probability of nearby samples being
assigned to the same population (Corander et al. 2008).
Data for the assignment test consisted of each individ-
ual’s multilocus genotype and spatial coordinates of the
harvest location (based on the centroid of the PLSS sec-
tion (2.6 km · 2.6 km) identified by the hunter). We con-
ducted five iterations of tests for each potential number
of subpopulations (K) varying from 1 to 10 to extend far
beyond numbers of populations that might be expected.
Fine-scale social structure
At the fine scale, we used correlograms to explore the
patterns of spatial clustering among genetically related
Published 2012. This article is a US Government work and is in the p
animals. Correlograms have been used to explore con-
tinuous spatial genetic patterns across the full range of
a species (Quemere et al. 2010) or spanning heteroge-
neous habitats within a species range (Gauffre et al.
2008) (see overviews in Wagner et al. 2005; Guillot et al.
2009). We constructed correlograms showing the degen-
eration of spatial association with increasing distance
classes (using the ‘r’ statistic in the ‘Multiple D Class’
routine in GENALEX 6; Peakall & Smouse 2006). We used
distance classes of 2 km because this approximates the
PLSS section (1 mi2 or 2.6 km2) at which samples were
collected, and ensured an adequate number of pairs
within each distance class. We plotted separate correlo-
grams for each ecoregion to assess the influence of
broad-scale landscapes on local patterns of relatedness
and to identify genetic neighbourhoods.
We hypothesized that local genetic structure would
be correlated with habitat characteristics within the
township that affect social group organization and
genetic neighbourhood size (Hirth 1977; Wang & Schre-
iber 2001). We constructed generalized linear models
(glm) in R (R Development Core Team, 2008) to relate
habitat variables to the average level of genetic related-
ness (rxy) between all deer pairs within each township
(calculated in SpagEdi; Hardy & Vekemans 2007; Quel-
ler & Goodnight 1989). These models used ECO, CAN
and ADJ as the habitat predictor variables. The best
model was chosen based on AIC score, and for closely
competing models (DAIC £ 2), we averaged model coef-
ficients (Burnham & Anderson 2002). Fit of the final
model was assessed using residual deviance to measure
improvement in likelihood between the selected and
the null models.
Broad-scale population connectivity
To identify broad-scale spatial genetic patterns, we
employed spatial principal component analysis (sPCA)
(see overviews in Jombart et al. 2008, 2009), which com-
bines multivariate allele frequencies and spatial auto-
correlation, to summarize genetic variation and its
spatial pattern. This multivariate ordination technique
can summarize a large multivariate data set using a few
highly informative and uncorrelated factors, the ‘princi-
pal components’ (McGarigal & Cushman 2000). Our
data matrix consisted of allele frequencies in each sam-
ple township, and the resulting sPCA component axes
were scaled to preserve this genetic relationship among
townships, as well as spatial dependence between them.
This is accomplished by optimizing the maximum
eigenvector scores based on the product of genetic vari-
ance and Moran’s I values. We plotted the principal
component eigenvalues to evaluate their relative infor-
mation and determine the number of informative sPCA
ublic domain in the USA.
6 S . J . R O B I N SO N E T A L.
axes (Legendre & Legendre 1998; Jombart et al. 2008,
2009). Additionally, we used factor loading plots to
judge the contribution of each allele and locus to the
sPCA and to assess the degree of clinal or clustered
genetic structure in the population. We performed the
sPCA using the adegenet package in R; we chose the
inverse distance weighting option to include all possible
township pairs in the spatial neighbourhood matrix
(Jombart et al. 2008).
Each sampled township was assigned a score on each
sPCA axis representing its position in the multidimen-
sional space of genetic variability relative to all other
townships. These sPCA scores provided site-specific
measures of genetic structure that simultaneously
accounted for genetic variability and broad-scale spatial
structure. We constructed regression models to test the
association of landscape variables with this genetic
structure using sPCA scores on each axis as response
variables for each township. These site-based measures,
as opposed to pairwise genetic distances, allowed us to
determine which landscape variables predicted genetic
structure (Jombart et al. 2009), were essential to meeting
the independence assumptions of regression analysis
and facilitated the use of model selection methods
(Burnham & Anderson 2002).
We hypothesized that landscape-scale genetic struc-
ture might be influenced by habitat characteristics at
each township, barriers to movement between town-
ships or a combination of these factors. As an explor-
atory step, we computed partial Mantel tests to assess
whether each variable (ECO, CAN, ADJ, INT, HWY,
RIV) explained significant genetic variation after
accounting for isolation by distance. We considered
genetic distance measured as the Euclidean distance
between sampled township coordinates on the two
sPCA axes. We measured landscape distance as the dif-
ference in numerical measures (CAN, ADJ) or as cate-
gorical zones (Fig. 2B) for barrier variables (i.e.
distance = 0 between townships within the same barrier
zone, distance = 1 between townships separated by a
putative barrier, tests calculated using the ZT software;
Bonnet & Van de Peer 2007).
While the sPCA scores represent the broad-scale spa-
tial patterns of genetic variation, they do not specifically
account for the isolation-by-distance process (Jombart
et al. 2008). Therefore, we constructed spatial autore-
gressive models to simultaneously account for isolation
by distance and evaluate the effect of landscape features
on the observed genetic structure. We used a spatially
lagged regression to assess the effect of multiple param-
eters while accounting for spatial autocorrelation in the
response variable (Legendre & Legendre 1998). Using
the lagsarlm routine in the spdep package (Bivand et al.
2011) within R, we fit the model:
Published 2012. This article is a
Gi¼q�Wij � Gj þ b � X þ e ðeqn 1Þ
where Gi represents the genetic measure (sPCA score)
at township i and Gj represents scores measured at
other townships. Wij is the weight matrix that deter-
mines the relative influence of township j on township
i. The weight matrix was constructed for all possible
pairs of townships based on the inverse of the distance
between them. We chose the inverse distance weighting
because this best approximates the expectation of spa-
tial autocorrelation from isolation by distance. In the
spatially lagged model, q accounts for the lack of inde-
pendence among townships. The contribution of predic-
tor variables is determined by fitting the linear function
b *X and estimating the effects (b) of the included envi-
ronmental variables (X). Finally, e is the residual error.
We used a hierarchical model selection procedure
based on an information theoretic approach. We sepa-
rately evaluated the contribution of either habitat vari-
ables (ECO, CAN and ADJ) or barrier variables (INT,
HWY and RIV) in explaining the sPCA genetic pattern.
Alternative models from either set were selected based
on the lowest AIC score (or within DAIC < 2 from the
lowest). We then considered the best models from each
set and combinations thereof. The final set of candidate
models containing habitat and barrier variables were
also evaluated based on AIC score (and ⁄ or averaged
based on AIC weights in the event of closely competing
models with DAIC < 2) (Burnham & Anderson 2002).
We evaluated model goodness of fit by residual devi-
ance (as above) and R2. While inclusion in the model
indicated that a predictor variable impacted genetic
structure, we also tested whether specific landscape fea-
tures (individual interstates, highways, rivers) consti-
tuted significant barriers to gene flow. To accomplish
this, we used z tests to compare model regression coeffi-
cients between zones separated by a single potential bar-
rier (i.e. INT zones A vs. B separated by interstate
I90 ⁄ I94). If regression coefficients differed significantly
between zones on either side of a putative barrier, we
concluded that sample townships were more similar to
one another within zones and differed significantly
between zones, thus indicating a barrier effect differenti-
ating townships on either side of that landscape feature.
Results
Genetic data
In total, we genotyped 2069 deer from 64 townships,
averaging 32 deer per sampled township (range 10–72).
We sampled adult female deer with an average age of
2.76 years. The locus OBCAM performed poorly with
US Government work and is in the public domain in the USA.
LANDSCAPE G ENETICS IN A C ONTINUOUS DEER POPULATION 7
15% missing data and >5% error, so it was dropped from
the data set, leaving 14 loci for analysis. Error rates for all
other loci were low (1–3%) and missing data were mini-
mal (Table 1). Five locus pairs (RT23-RT7, BM1225–
BM4208, RT27-BM4208, RT27-OarFCB193, and RT9-Oar-
FCB193) showed linkage disequilibrium after Bonferroni
correction for multiple tests (0.05 ⁄ 91 pairwise tests signif-
icance = a < 0.0006). However, Wilson et al. (1997) dem-
onstrated no linkage among these loci; thus, our result
likely relates to structure within the population rather
than physical linkage between loci. Six loci were out of
HWE after correcting for multiple tests (0.05 ⁄ 14 locus-
based tests significance = a < 0.0036). Deviations from
HWE were also likely attributable to population sub-
structure as genotyping error rates were not elevated for
these loci and they showed strong spatial structuring
based on loading scores for the sPCA axes (below).
Detection of population subdivision
The Bayesian assignment test indicated that deer in our
study area were not divided into genetically distinct
subpopulations (posterior probability of K = 1 was 1),
indicating a single genetic population, continuous over
a broad geographic scale. Thus, our remaining land-
scape genetic analyses examine subtle genetic patterns
within a single genetic population rather than between
separate populations.
Fine-scale social structure
Correlograms showed that clustering of genetic similar-
ity was greatest at short distances (2–6 km), although
Table 1 Microsatellite characteristics and data quality for genotypes
of data quality, we present the number of alleles for each polymorph
cients (Fis) and observed vs. expected ratios of homozygotes (HO), a
exact test of Hardy–Weinberg equilibrium (Pval HWE)
Locus No. of alleles % Missing data Fis
BM 6506 14 4.592 0.099
BM 1225 12 3.142 0.040
N 24 5.607 0.038
RT 23 20 3.287 0.020
BM 4107 17 1.982 0.017
Oar 193 18 6.235 0.015
RT 27 18 4.592 0.006
BM 6438 13 7.347 0.010
BM 4208 19 6.815 0.085
Cervid 1 19 2.127 0.011
Cervid 2 15 4.447 0.166
IGF 1 15 2.320 0.003
RT 7 17 3.335 0.029
RT 9 11 2.465 0.121
*Corrected a = 0.0036.
Published 2012. This article is a US Government work and is in the p
differences in fine-scale population structure were
apparent among ecoregions (Fig. 3). The degree of
genetic autocorrelation declined with increasing dis-
tance beyond which individuals became mixed more
randomly. The EGP and SWS had similar correlograms,
while genetic autocorrelation was substantially higher
in the WCR region. While the correlograms indicated
differences across ecoregions, linear models identified
the amount and arrangement of forest habitat as the
best predictors of relatedness within townships
(CAN + ADJ was the model with lowest AIC, DAIC = 3
for the second best model including ECO). Genetic
relatedness was higher in heavily forested townships
(CAN coefficient = 0.0007, P = 0.001) and also increased
with higher forest fragmentation (ADJ coeffi-
cient = )0.001, P = 0.0002).
Broad-scale population connectivity
The sPCA revealed clinal patterns in allele frequencies
across our study landscape. The first two sPCA axes
accounted for most of the spatial genetic pattern in the
data. The first two eigenvalues were 0.020 and 0.011,
while the remaining eigenvalues were <0.004, indicating
that the majority of the data structure was accounted
for on the first two axes. Factor scores for the first two
sPCA axes displayed a different, orthogonal gradient in
the genetic structure (Fig. 4B,D) and were retained for
further modelling. sPCA1 showed a genetic cline from
the northwest to southeast, and alleles from Cervid1, N
and RT 7 were the most informative (Fig. 4A). sPCA2
showed northeast to southwest cline and was related to
variation in BM4107, BM6438 and RT7 allele frequencies
generated for Midwestern white-tailed deer. As basic measures
ic locus, the amount of data missing for each, inbreeding coeffi-
nd heterozygotes (HE), and finally, P values associated with an
HOexp : HOobs HEexp : HEobs Pval HWE
0.518 0.878 <0.000*
0.890 1.041 <0.000*
0.716 1.061 <0.000*
0.822 1.021 0.014
0.912 1.022 0.040
0.986 1.002 0.013
0.901 1.019 0.135
0.873 1.027 0.035
0.484 1.122 <0.000*
0.889 1.029 <0.000*
0.522 1.212 <0.000*
0.998 1.000 0.902
0.839 1.025 0.069
0.936 1.017 0.140
ublic domain in the USA.
Fig. 3 Variation among ecoregions in
fine-scale population genetic structure
of white-tailed deer in the Midwestern
USA. Correlograms for each ecoregion
show the distance-based decline in spa-
tial clustering of related genotypes.
8 S . J . R O B I N SO N E T A L.
(Fig. 4C). sPCA1 had a more consistent and gradual
gradient, while sPCA2 showed abrupt spatial changes
in factor scores.
As expected for a spatial ordination, the genetic pat-
tern depicted by each sPCA axis was strongly related to
distance, so accounting for local spatial autocorrelation
in our model of sPCA scores was crucial in distinguish-
ing the role of landscape features from spatial distance.
Inclusion of the spatial autocorrelation parameter (q)
substantially improved model performance (sPCA 1
q = 0.827, SE = 0.115, DAIC = 8.350 from nonspatial lin-
ear model: sPCA 2 q = 0.796, SE = 0.137, DAIC = 4.560
from nonspatial linear model). Noting the strong spatial
cline in sPCA scores, we can expect that the q parame-
ter alone should have substantial power to explain the
data; we judged this by looking at the R2 for the q-only
model (sPCA1, R2 = 0.794; sPCA2, R2 = 0.787). To deter-
mine the contribution of landscape factors, we then
evaluated the improvement of model fit achieved with
the addition of explanatory landscape variables. Land-
scape variables substantially improved model fit (see
AIC scores in Table S2, Supporting information) and
predictive performance (increases to R2 for sPCA1 =
0.112 and for sPCA2 = 0.103).
Our spatial autoregressive models indicated that,
beyond isolation by distance, potential impediments to
dispersal played the biggest role in shaping broad-scale
genetic connectivity among Midwestern deer. For both
sPCA1 and sPCA2, all three barrier variables were
included in all of the final models (for sPCA1:
q + INT + HWY + RIV, q + INT + HWY + RIV + ECO,
q + INT + HWY + RIV + ECO + CAN all with DAIC < 2
Published 2012. This article is a
and for sPCA2: q + INT + HWY + RIV, q + INT +
HWY + RIV + ECO + ADJ, q + INT + HWY + RIV + E-
CO + ADJ + CAN all with DAIC < 2). Parameter esti-
mates were averaged among the top models (Table S2,
Supporting information). After accounting for model
uncertainty through model averaging, some of the
included parameters were no longer significant. The
regression effect estimates were highest and were signifi-
cant for the barrier variables, whereas ECO effects were
not significant in either the sPCA1 or sPCA2 models, and
CAN and ADJ had very minor effects (near 0 after model
averaging). This indicated that, while habitat features may
have a minor impact on genetic structure, the barriers we
measured made the most important contribution to pre-
dicting genetic structure across the study landscape. Par-
tial Mantel tests confirmed that habitat and barrier
variables were significantly correlated with genetic differ-
entiation after accounting for isolation by distance (ADJ:
r = )0.070, P = 0.130; ECO: r = )0.228, P = 0.0001; CAN:
r = )0.154, P = 0.004; INT: r = 0.252, P = 0.0001; HWY:
r = 0.147, P = 0.0001; RIV: r = 0.011, P = 0.006).
Subsequent analysis of regression coefficients showed
that there were specific gene flow impediments of
importance within the general INT, HWY and RIV cate-
gories (Table 2, Fig. 5). A primary north–south barrier
was formed by the coincidence of the I39 ⁄ 90 corridor,
the Rock River and WI highway 51S. While the INT was
only marginally significant in the sPCA1 model, it was
highly significant in the sPCA2 model. Conversely, the
Rock River was highly significant on sPCA1. In an east–
west direction, WI highway 18E-W formed a significant
barrier to gene flow in both sPCA models. The coinci-
US Government work and is in the public domain in the USA.
Fig. 4 sPCA of genetic structure in a white-tailed deer population in the Midwestern USA. Bar graphs show the allelic contributions
(weights) to loading on spatial principal component axes sPCA1 (A) and sPCA2 (C). Height of bars indicates the weight of each
allele (based on 14 microsatellite loci) on each sPCA axis. The map plots show factor loading scores for each sample unit on sPCA1
(B) and sPCA2 (D). Each square represents a sample township. Dark squares indicate positive values, light squares indicate negative
values, and size of the square indicates magnitude of the score on each axis, representing each township’s position relative to the
overall genetic structure of the population. Contour lines illustrate the orthogonal genetic gradients based on interpolation of the
sPCA scores across the study area.
LANDSCAPE G ENETICS IN A C ONTINUOUS DEER POPULATION 9
dence of HWY 18E with interstate I94 may have com-
pounded the effects of these two barriers. The concentra-
tion of major roadways in the eastern and southeastern
portion of the study area led to several significant road
barrier effects (interstates I39 ⁄ 90, I43 and I94 and state
highways 12E, 18E, 20E and 151N, Fig. 5) and coincided
with the steepest portions of the genetic gradient dis-
played in the sPCA axes (Fig. 4B,D).
Discussion
Landscape genetics of white-tailed deer
Our analysis revealed that landscape processes shape
white-tailed deer population structure at varying
scales. At the ecoregion scale, we found that genetic
relatedness was most clustered in the highly forested
WCR and least in the more agricultural ⁄ grassland EGP
Published 2012. This article is a US Government work and is in the p
and SWS regions (Fig. 3). We further found that
genetic relatedness within sample townships increased
with both the amount and fragmentation of forested
habitat. We expect this reflects higher density and
philopatry of related social groups in areas with dense
forest habitat. Previous research has shown that home
range requirements are smaller in contiguous high-
quality habitat (Lesage et al. 2000), leading to higher
population density and greater spatial clustering of
genetically related individuals (Wang & Schreiber
2001). Female deer are also more likely to disperse in
open landscapes where habitat resources are sparse
(Nixon et al. 1991; Nelson & Mech 1992). Dispersal
distances also tend to be greater in areas with less for-
est cover or more fragmented forest habitat (Long
et al. 2005; Nixon et al. 2008). Animals might also shift
between summer- and winter-use areas in agricultural
habitat as in the EGP, where fields become unsuitable
ublic domain in the USA.
Tab
le2
Eff
ects
esti
mat
esre
late
dto
gen
efl
ow
bar
rier
sd
etec
ted
thro
ug
hsp
atia
lau
tore
gre
ssiv
em
od
els
of
gen
etic
pat
tern
sin
wh
ite-
tail
edd
eer
inth
eM
idw
este
rnU
SA
Est
imat
eS
E
(a)
sPC
A1
q0.
697
0.25
8
Inte
rcep
t)
0.07
90.
088
CA
N0.
000
0.00
0
AD
JN
ot
inm
od
el
Wo
fI3
9⁄9
0E
of
I39
⁄90
NE
of
I39
⁄90
C
Eo
fI3
9⁄9
0N
)0.
065
0.04
70.
163
Eo
fI3
9⁄9
0C
0.02
60.
046
0.57
20.
048
Eo
fI3
9⁄9
0S
)0.
052
0.04
70.
268
0.09
6
20W
—18
W18
W—
14W
14W
—12
W12
W—
151E
151E
—18
E18
E—
12E
12E
—14
E14
E—
20E
18W
—14
W0.
067
0.03
90.
085
14W
—12
W0.
057
0.05
00.
827
12W
—15
1E0.
027
0.04
50.
509
151E
—18
E)
0.05
90.
060
0.15
0
18E
—12
E0.
122
0.08
40.
031
12E
—14
E0.
127
0.07
80.
947
14E
—20
E0.
141
0.08
10.
861
So
f20
)0.
062
0.05
60.
267
0.00
0
Eo
fR
ock
No
fW
isc
Yah
-Ro
ck
No
fW
isc
0.10
50.
094
Yah
-Ro
ck0.
120
0.10
00.
016
Wis
c-Y
ah0.
123
0.09
10.
007
0.84
00.
975
EG
PS
WS
SW
S0.
021
0.01
40.
134
WC
R)
0.00
30.
023
0.88
90.
294
(b)
sPC
A2
q0.
752
0.31
2
Inte
rcep
t0.
056
0.16
1
CA
N0.
000
0.00
0
AD
J)
0.00
30.
001
Wo
fI3
9⁄9
0E
of
I39
⁄90
NE
of
I39
⁄90
C
Eo
fI3
9⁄9
0N
0.20
20.
086
0.01
9
Eo
fI3
9⁄9
0C
0.17
80.
083
0.03
10.
776
Eo
fI3
9⁄9
0S
0.05
20.
053
0.32
70.
017
20W
—18
W18
W—
14W
14W
—12
W12
W—
151E
151E
—18
E18
E—
12E
12E
—14
E14
E—
20E
10 S . J . ROB I N SO N ET AL.
Published 2012. This article is a US Government work and is in the public domain in the USA.
Tab
le2
Con
tin
ued
Est
imat
eS
E
18W
—14
W0.
028
0.04
30.
514
14W
—12
W0.
069
0.06
30.
521
12W
—15
1E0.
101
0.06
90.
645
151E
—18
E)
0.03
30.
067
0.04
5
18E
—12
E0.
288
0.13
10.
014
12E
—14
E0.
106
0.08
50.
031
14E
—20
E0.
062
0.07
80.
574
So
f20
)0.
076
0.06
70.
257
0.04
0
Eo
fR
ock
No
fW
isc
Yah
-Ro
ck
No
fW
isc
0.16
80.
115
Yah
-Ro
ck0.
161
0.12
10.
153
Wis
c-Y
ah0.
068
0.09
40.
397
0.28
90.
325
EG
PS
WS
SW
S)
0.01
80.
021
0.38
7
WC
R0.
027
0.03
60.
448
0.20
5
WC
R
AD
J,ad
jace
ncy
;W
CR
,W
este
rnC
ou
lee
and
Rid
ge.
Reg
ress
ion
mo
del
par
amet
eres
tim
ates
and
stan
dar
der
ror
ind
icat
eth
eef
fect
of
each
bar
rier
zon
eo
rh
abit
atv
aria
ble
on
gen
etic
dif
fere
nti
atio
n.
Th
em
atri
xp
ort
ion
sho
ws
Pv
alu
es
ind
icat
ing
the
stat
isti
cal
sig
nifi
can
ceo
fea
chp
ote
nti
alg
ene
flo
wb
arri
ers
bas
edo
nst
atis
tica
lco
mp
aris
on
of
par
amet
eres
tim
ates
bet
wee
nre
gio
ns
on
eith
ersi
de
of
the
bar
rier
.
Bo
ldan
dit
alic
ized
tex
tin
dic
ates
sig
nifi
can
tef
fect
s(a
<0.
05);
ital
ics
on
lyin
dic
ates
mar
gin
alsi
gn
ifica
nce
(a<
0.1)
.
LANDSCAPE GENETICS IN A CO NTI NUOUS DEER POPULATION 11
Published 2012. This article is a US Government work and is in the public domain in the USA.
Fig. 5 Landscape patterns illustrating
population genetic structure of white-
tailed deer in the Midwestern USA. Bar-
riers to genetic connectivity were
detected through multiple regression in
a spatial autoregressive format. This
map combines factors that contributed
significantly to explaining genetic gradi-
ents portrayed by both sPCA1 and
sPCA2. Sample townships are shown in
colour gradients representing relative
positions on the sPCA axes; gradation
of red indicates sPCA1; gradation of
green indicates sPCA2.
12 S . J . ROB I N SO N ET AL.
once crops are cut. Higher dispersal or seasonal move-
ment could lead to looser organization of kin groups,
which might produce lower autocorrelation in the less
forested ecosystems and the positive association of for-
est cover with fine-scale genetic relatedness.
At the broader scale, landscape connectivity among
deer was determined primarily by natural and anthro-
pogenic impediments to dispersal (Fig. 5). These land-
scape features may exert different influences on deer
movement (whether altering behaviour, habitat choice
or survival as in road mortality; Etter et al. 2002) over
different temporal scales (rivers pose a historic barrier
to gene flow compared with anthropogenic features).
We note that because our data were female biased, the
genetic structure we detected may emphasize the
impact of these dispersal barriers. Although barrier
effects were statistically significant, the assignment test
indicated that sufficient gene flow occurred across these
barriers to maintain connectivity within the population.
We found particularly strong effects where several
impediments occurred in near proximity. For example,
the potential combined effects of the interstate I39 ⁄ 90
corridor along with WI highway 51S and the Rock
River may present substantial limits to gene flow. Such
confounding of features may be common on the land-
scape, for example river valleys frequently facilitate
highway construction, and interstates may often aug-
ment existing highway routes. The coincidence of land-
scape features makes it difficult to quantify the separate
contribution of each feature.
The combination of landscape genetic techniques and
spatial modelling provided a powerful tool for under-
standing connectivity and movement within a dense and
continuous population of white-tailed deer. The sPCA
Published 2012. This article is a
method provided genetic metrics that could be assigned
to each sample township and used in a general linear
model framework allowing us to simultaneously evalu-
ate several landscape variables without subjective, a
priori assignment of weights to any particular feature
(e.g. cost path analysis). Multivariate and regression-
based methods have been recommended as powerful
alternatives to linear distance-based relationships (as in
the Mantel test framework) (Legendre & Fortin 2010).
Such tests are widely used in ecology (Borcard & Legen-
dre 2002; Borcard et al. 2004; Dray et al. 2006) and are
gaining popularity in landscape genetics (Geffen et al.
2004; Guinand 2008; Jombart et al. 2009). The model
selection framework using AIC helped us identify impor-
tant landscape features affecting gene flow and the
spatial association parameter (q) allowed us to account
for isolation by distance while assessing the additional
impact of landscape variables and specific features
(highways, rivers or interstates) that were influential in
differentiating adjacent groups of deer.
Our results complement previous white-tailed deer
research in the Midwest and other areas in North Amer-
ica and offer a valuable expansion or our knowledge of
genetic structure at both fine and broad scale. We found
fine-scale spatial genetic clustering similar to Grear et al.
(2010), who found that first- or second-order relatives
were more likely to be in close proximity than unrelated
deer. Blanchong et al. (2008), using a subset of our study
area, found that highways WI 18W and WI 14W and the
Wisconsin River were barriers to local gene flow. Our
analysis of the wider area confirmed that highways and
rivers act as gene flow barriers. We also found that large
INT in the eastern and western portions of our study
region acted as a major impediment to gene flow between
US Government work and is in the public domain in the USA.
LANDSCAPE GENETICS IN A CO NTI NUOUS DEER POPULATION 13
deer. Rogers et al. (2011) found that certain mitochon-
drial DNA sequences from a portion of our study area
(363 deer from 12 of the WI sample pops) varied across
the landscape similarly to either the sPCA1 or sPCA2
gradients in our findings; though, mtDNA haplotypes
were more clustered on the landscape than our microsat-
ellite data. Our results in the Midwest were similar to
those for white-tailed deer spanning the Canadian prov-
inces of Alberta and Saskatchewan, where Cullingham
et al. (2011) found little evidence for population subdivi-
sion, but detected significant isolation by distance. How-
ever, we found that roads or rivers are impediments to
dispersal and gene flow, as previously suggested for sev-
eral ungulate species (Nussey et al. 2005; Coulon et al.
2006; Perez-Espona et al. 2008; Long et al. 2010), and
these landscape features have become an important focus
of animal movement studies (Forman & Alexander 1998;
Jackson 2000), as well as in landscape genetics (Balkenhol
& Waits 2009).
Our results, however, contrast with those of Kelly
et al. (2010), who found significant population substruc-
ture for a study area of similar size, including a portion
overlapping our study area in north-central Illinois.
Kelly et al. (2010) identified four genetic populations,
including two admixed populations in the central part
of our study area and differentiated populations in spa-
tially disjoint areas to the west and south. Difference in
sample designs appears to be the primary cause of the
different conclusions about population subdivision.
Whereas we sampled deer in random locations within a
systematic grid, Kelly et al. (2010) sampled areas sepa-
rated by 100s of km. Because isolation by distance is a
prevailing genetic pattern for white-tailed deer popula-
tions, sampling clusters of individuals at far-spaced
intervals along the genetic cline may over-represent the
degree of population subdivision (Frantz et al. 2009;
Schwartz & McKelvey 2009). Additionally, we used
more spatially random samples from hunter-harvest,
while Kelly et al. (2010) relied on groups of deer har-
vested at DNR sharpshooting stations. This method is
likely to collect related animals that arrive at baited sta-
tions together and fawns constituted over 38% of their
data set, indicating a significant portion of the data may
have represented undispersed relatives. Studies of alter-
native sampling design have shown that if local areas
of relatedness are confounded with sampling patterns,
assignment programmes may find an artificial associa-
tion between sampling blocks and population clusters
(Schwartz & McKelvey 2009; Storfer et al. 2010).
Management implications
Understanding deer population structure on the Mid-
west landscape has important implications for the
Published 2012. This article is a US Government work and is in the p
future management of this abundant species. In general,
the large spatial scale of gene flow and high connectiv-
ity within the deer population suggests that large man-
agement units should be used to determine harvest and
population goals. It is also important to note that
although we found impediments to gene flow, there
were no absolute barriers subdividing the population.
Thus, the study area contains many potential deer man-
agement units that are clearly interconnected through
high levels of gene flow. Thus, emigration and immi-
gration between management units could impact demo-
graphics used by wildlife managers to establish
population goals. It will be important to treat manage-
ment units as open populations to accurately design
population control measures or harvest regulations.
Infectious disease is increasingly a priority in cervid
management because chronic wasting disease, brucello-
sis, bovine tuberculosis and bovine viral diarrhoea
affect wild cervid populations and raise concern for
livestock health. The processes investigated here rela-
tive to population connectivity and dispersal relate
directly to epidemiological processes of disease trans-
mission between individuals at the fine scale (Altizer
et al. 2003) and spread of disease among populations
with broad-scale connectivity (Ostfeld et al. 2005; Joly
et al. 2006; Real & Biek 2007). Our study area in WI and
IL contained two distinct CWD outbreak areas in the
northwest (WCR between highways 14W and 18W) and
southeast (EGP east of interstate I39 ⁄ 90 and the Rock
River). Our results indicate that fine-scale population
structure differs among these regions with more cluster-
ing in the WCR around the western CWD outbreak.
This level of overlap among social groups indicated
higher potential contact rates among individuals that
could increase local transmission of disease around the
western outbreak area. Conversely, disease may spread
faster across the EGP landscape where more frag-
mented forest habitat lead to longer dispersals. At the
broader scale, dispersing animals are likely to spread
disease beyond social groups to other areas across the
landscape. Knowing that highways, interstates and riv-
ers provide some level of impediment to deer dispersal
might give management agencies the opportunity to
maximize such barriers in efforts to contain or other-
wise manage disease spread.
Acknowledgements
We thank the Wisconsin Department of Natural Resources,
particularly those at the Black Earth sample archive facility,
and the Illinois Department of Natural Resources, particularly
those working at deer check stations. We appreciate pilot data
and information from Dr. Julie Blanchong and Dr. Dan Grear.
Thanks also to all laboratory technicians and staff at the UW
Biotechnology Core Facility. Our manuscript benefited from
ublic domain in the USA.
14 S . J . ROB I N SO N ET AL.
comments on previous drafts from Dr. Lisette Waits and Dr.
Kim Scribner. Funding was provided by the U.S. Geological
Survey, a U.S. Department of Agriculture Hatch grant and the
Wisconsin Department of Natural Resources. Thanks to the
University of Wisconsin Department of Forest & Wildlife Ecol-
ogy for assistance with publication costs. Note that any use of
trade, product or firm names is for descriptive purposes and
does not imply endorsement by the US Government.
References
Altizer S, Nunn CL, Thrall PH et al. (2003) Social organization
and parasite risk in mammals: integrating theory and
empirical studies. Annual Review of Ecology Evolution and
Systematics, 34, 517–547.
Balkenhol N, Waits LP (2009) Molecular road ecology:
exploring the potential of genetics for investigating
transportation impacts on wildlife. Molecular Ecology, 18,
4151–4164.
Balkenhol N, Gugerli F, Cushman SA et al. (2009) Identifying
future research needs in landscape genetics: where to from
here? Landscape Ecology, 24, 455–463.
Bartelt G, Pardee J, Thiede J (2003) Environmental Impact
Statement on Rules to Eradicate Chronic Wasting Disease in
Wisconsin’s Free-Ranging White-tailed Deer Herd. Wisconsin
Department of Natural Resources, Madison, Wisconsin.
Bivand R, Altman M, Anselin L et al. (2011) R Package
‘spdep’: spatial dependence: weighting schemes, statistics
and models. Available from http://cran.r-project.org/web/
packages/spdep/index.html.
Blanchong JA, Samuel MD, Scribner KT et al. (2008) Landscape
genetics and the spatial distribution of chronic wasting
disease. Biology Letters, 4, 130–133.
Bonnet E, Van de Peer Y (2007) zt: a software tool for simple
and partial Mantel tests.
Borcard D, Legendre P (2002) All-scale spatial analysis of
ecological data by means of principal coordinates of
neighbour matrices. Ecological Modelling, 153, 51–68.
Borcard D, Legendre P, Avois-Jacquet C, Tuomisto H (2004)
Dissecting the spatial structure of ecological data at multiple
scales. Ecology, 85, 1826–1832.
Burnham KP, Anderson DR (2002) Model Selection and
Multimodel Inference: A Practical Information-Theoretic
Approach. Springer Verlag, New York, NY, USA.
Corander J, Marttinen P, Siren J, Tang J (2008) Enhanced
Bayesian modelling in BAPS software for learning genetic
structures of populations. BMC Bioinformatics, 9, 539.
Cote SD, Rooney TP, Tremblay JP, Dussault C, Waller DM
(2004) Ecological impacts of deer overabundance. Annual
Review of Ecology, Evolution, and Systematics, 35, 113–147.
Coulon A, Guillot G, Cosson J-F et al. (2006) Genetic structure
is influenced by landscape features: empirical evidence from
a roe deer population. Molecular Ecology, 15, 1669–1679.
Cross P, Drewe J, Patrek V et al. (2008) Wildlife population
structure and parasite transmission: implications for disease
management. In: Management of Disease in Wild Mammals
(eds Delahay R, Smith G, Hutchings M), pp. 9–29. Springer,
New York City, New York.
Cullingham C, Merrill E, Pybus M et al. (2011) Broad and fine
scale genetic analysis of white-tailed deer populations:
Published 2012. This article is a
estimating the relative risk of chronic wasting disease
spread. Evolutionary Applications, 4, 116–131.
Dray S, Legendre P, Peres-Neto PR (2006) Spatial modelling: a
comprehensive framework for principal coordinate analysis
of neighbour matrices (PCNM). Ecological Modelling, 196,
483–493.
Epperson BK, McRae BH, Scribner K et al. (2010) Utility of
computer simulations in landscape genetics. Molecular
Ecology, 19, 3549–3564.
Etter D, Hollis K, Van Deelen T et al. (2002) Survival and
movements of white-tailed deer in suburban Chicago,
Illinois. Journal of Wildlife Management, 66, 500–510.
Fagerstone KA, Clay WH (1997) Overview of USDA animal
damage control efforts to manage overabundant deer.
Wildlife Society Bulletin, 25, 413–417.
Forman RTT, Alexander LE (1998) Roads and their major
ecological effects. Annual Review of Ecology and Systematics,
29, 207–231.
Frantz AC, Cellina S, Krier A, Schley L, Burke T (2009) Using
spatial Bayesian methods to determine the genetic structure
of a continuously distributed population: clusters or isolation
by distance? Journal of Applied Ecology, 46, 493–505.
Fry JA, Coan MJ, Homer CG, Meyer DK, Wickham JD (2009)
Completion of the National Land Cover Database (NLCD)
1992–2001 Land Cover Change Retrofit product. U.S.
Geological Survey Open-File Report 2008–1379, 18 p.
Garroway CJ, Bowman J, Carr D, Wilson PJ (2008)
Applications of graph theory to landscape genetics.
Evolutionary Applications, 1, 620–630.
Gauffre B, Estoup A, Bretagnolle V, Cosson J (2008) Spatial
genetic structure of a small rodent in a heterogeneous
landscape. Molecular Ecology, 17, 4619–4629.
Geffen E, Anderson MJ, Wayne RK (2004) Climate and habitat
barriers to dispersal in the highly mobile grey wolf.
Molecular Ecology, 13, 2481–2490.
Grear DA, Samuel MD, Langenberg JA, Keane D (2006)
Demographic patterns and harvest vulnerability of chronic
wasting disease infected white-tailed deer in Wisconsin.
Journal of Wildlife Management, 70, 546–553.
Grear DA, Samuel MD, Scribner KT, Weckworth BV, Langenberg
JA (2010) Influence of genetic relatedness and spatial proximity
on chronic wasting disease infection among female white-
tailed deer. Journal of Applied Ecology, 47, 532–540.
Guillot G, Leblois R, Coulon A, Frantz AC (2009) Statistical
methods in spatial genetics. Molecular Ecology, 18, 4734–4756.
Guinand B (2008) Use of a multivariate model using allele
frequency distributions to analyse patterns of genetic
differentiation among populations. Biological Journal of the
Linnean Society, 58, 173–195.
Hardy O, Vekemans X (2007) SPAGeDi 1.2: a program for Spatial
Pattern Analysis of Genetic Diversity, Available from http://
www.ulb.ac.be/sciences/ecoevol/docs/manual_SPAGeDi_1-
2.pdf.
Hawkins RE, Klimstra WD (1970) A preliminary study of the
social organization of white-tailed deer. Journal of Wildlife
Management, 34, 407–419.
Henderson DW, Warren RJ, Cromwell JA, Hamilton RJ (2000)
Response of urban deer to a 50% reduction in local herd
density. Wildlife Society Bulletin, 28, 902–910.
Hirth DH (1977) Social behavior of white-tailed deer in relation
to habitat. Wildlife Monographs, 53, 3–55.
US Government work and is in the public domain in the USA.
LANDSCAPE GENETICS IN A CO NTI NUOUS DEER POPULATION 15
Ishmael WE (1984) White-Tailed Deer Ecology and Management in
Southern Wisconsin. University of Wisconsin, Madison, WI,
USA.
Jackson SD (2000) Overview of transportation impacts on
wildlife movement and populations. In: Wildlife and
Highways: Seeking Solutions to an Ecological and Socio-Economic
Dilemma. The Wildlife Society, Attachment to EPIC Comment
Letter on 2006 RMP dated 10-23-06., pp. 7–20.
Joly DO, Samuel MD, Langenberg JA et al. (2006) Spatial
epidemiology of chronic wasting disease in Wisconsin white-
tailed deer. Journal of Wildlife Diseases, 43, 578–588.
Jombart T, Devillard S, Dufour A-B, Pontier D (2008) Revealing
cryptic spatial patterns in genetic variability by a new
multivariate method. Heredity, 101, 92–103.
Jombart T, Pontier D, Dufour A (2009) Genetic markers in the
playground of multivariate analysis. Heredity, 102, 330–341.
Kelly AC, Mateus-Pinilla NE, Douglas M et al. (2010) Utilizing
disease surveillance to examine gene flow and dispersal in
white-tailed deer. Journal of Applied Ecology, 47, 1189–1198.
Kilpatrick HJ, Spohr SM, Lima KK (2001) Effects of population
reduction on home ranges of female white-tailed deer at
high densities. Canadian Journal of Zoology, 79, 949–954.
Kohn BE, Mooty JJ (1971) Summer habitat of white-tailed deer
in north-central Minnesota. Journal of Wildlife Management,
35, 476–487.
Larson TJ, Rongstad OJ, Terbilcox FW (1978) Movement and
habitat use of white-tailed deer in southcentral Wisconsin.
Journal of Wildlife Management, 42, 113–117.
Legendre P, Fortin M (2010) Methodological advances –
inference of population structure: comparison of the Mantel
test and alternative approaches for detecting complex
multivariate relationships in the spatial analysis of genetic
data. Molecular Ecology Resources, 10, 831–844.
Legendre P, Legendre L (1998) Numerical Ecology. Elsevier
Science, Amsterdam, Netherlands.
Lesage L, Crete M, Huot J, Dumont A, Ouellet J-P (2000) Seasonal
home range size and philopatry in two northern white-tailed
deer populations. Canadian Journal of Zoology, 78, 1930–1940.
Long ES (2006) Landscape and Demographic Influences on
Dispersal of White-Tailed Deer. The Pennsylvania State
University, State College, PA, USA.
Long ES, Diefenbach DR, Rosenberry CS, Wallingford BD,
Grund MD (2005) Forest cover influences dispersal distance
of white-tailed deer. Journal of Mammalogy, 86, 623–629.
Long ES, Diefenbach DR, Wallingford BD, Rosenberry CS
(2010) Influence of roads, rivers, and mountains on natal
dispersal of white-tailed deer. The Journal of Wildlife
Management, 74, 1242–1249.
Marchinton RL, Hirth DH (1984) Behavior. In: White-Tailed
Deer Ecology and Management (ed. Halls LK), pp. 129–168.
Stackpole Books, Harrisburg, Pennsylvania.
Mastro LL, Conover MR, Frey SN (2008) Deer–vehicle collision
prevention techniques. Human-Wildlife Conflicts, 2, 80–92.
Mathews NE, Porter WF (1993) Effect of social structure on
genetic structure of free-ranging white-tailed deer in the
Adirondack Mountains. Journal of Mammalogy, 74, 33–43.
McGarigal K, Cushman S (2000) Multivariate Statistics for
Wildlife and Ecology Research. Springer-Verlag, New York
City, New York.
McGarigal K, Marks BJ (1995) FRAGSTATS: Spatial Pattern
Analysis Program for Quantifying Landscape Structure. Gen.
Published 2012. This article is a US Government work and is in the p
Tech. Report PNW-GTR-351. USDA Forest Service, Pacific
Northwest Research Station, Portland, Oregon.
Nelson ME (1993) Natal dispersal and gene flow in white-
tailed deer in northeastern Minnesota. Journal of Mammalogy,
74, 316–322.
Nelson ME (1995) Winter range arrival and departure of
white-tailed deer in northeastern Minnesota. Canadian Journal
of Zoology, 73, 1069–1076.
Nelson ME, Mech LD (1992) Dispersal in female white-tailed
deer. Journal of Mammalogy, 73, 891–894.
Nixon CM, Hansen LP, Brewer PA, Chelsvig JE (1991) Ecology
of white-tailed deer in an intensively farmed region of
Illinois. Wildlife Monographs, 118, 3–77.
Nixon CM, Mankin PC, Etter DR et al. (2008) Migration behavior
among female white-tailed deer in central and northern
Illinois. The American Midland Naturalist, 160, 178–190.
Nussey DH, Coltman DW, Coulson T et al. (2005) Rapidly
declining fine-scale spatial genetic structure in female red
deer. Molecular Ecology, 14, 3395–3405.
O’Brien DJ, Schmitt SM, Fierke JS et al. (2002) Epidemiology of
Mycobacterium bovis in free-ranging white-tailed deer,
Michigan, USA, 1995–2000. Preventive Veterinary Medicine, 54,
47–63.
Omernik JM (1987) Ecoregions of the conterminous United
States. Annals of the Association of American Geographers, 77,
118–125.
Ostfeld RS, Glass GE, Keesing F (2005) Spatial epidemiology:
an emerging (or re-emerging) discipline. Trends in Ecology &
Evolution, 20, 328–336.
Peakall R, Smouse PE (2006) GenAlEx 6: genetic analysis in
Excel. Population genetic software for teaching and research.
Molecular Ecology Notes, 6, 288–295.
Perez-Espona S, Perez-Barberia FJ, McLeod JE et al. (2008)
Landscape features affect gene flow of Scottish Highland red
deer (Cervus elaphus). Molecular Ecology, 17, 981–996.
Porter WF, Mathews NE, Underwood HB, Sage RWJ, Donald AB
(1991) Social organization in deer: implications for localized
management. Environmental Management, 15, 809–814.
Pritchard JK, Stephens M, Peter D (2000) Inference of
population structure using multilocus genotype data.
Genetics, 155, 945–959.
Purdue JR, Smith MH, Patton JC (2000) Female philopatry and
extreme spatial genetic heterogeneity in white-tailed deer.
Journal of Mammalogy, 81, 179–185.
Queller DC, Goodnight KF (1989) Estimating relatedness using
genetics markers. Evolution, 43, 258–275.
Quemere E, Crouau-Roy B, Rabarivola C, Louis E, Chikhi L
(2010) Landscape genetics of an endangered lemur
(Propithecus tattersalli) within its entire fragmented range.
Molecular Ecology, 19, 1606–1621.
R Development Core Team (2008) R: A Language and
Environment for Statistical Computing. R Foundation for
Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0,
URL http://www.R-project.org.
Rassmann K, Tautz D, Trillmich F, Gliddon C (1997) The
microevolution of the Galapagos marine iguana
Amblyrhynchos cristatus assessed by nuclear and
mitochondrial genetic analyses. Molecular Ecology, 6, 437–452.
Raymond M, Rousset F (1995) GENEPOP (Version 1.2):
population genetics software for exact tests and
ecumenicism. Journal of Heredity, 86, 248–249.
ublic domain in the USA.
16 S . J . ROB I N SO N ET AL.
Real LA, Biek R (2007) Spatial dynamics and genetics of
infectious diseases on heterogeneous landscapes. Journal of
the Royal Society Interface, 4, 935–948.
Retamosa MI, Humberg LA, Beasley JC, Rhodes OE Jr (2008)
Modeling wildlife damage to crops in northern Indiana.
Human-Wildlife Conflicts, 2, 225–239.
Rogers KG, Robinson SJ, Samuel MD, Grear DA (2011)
Diversity and distribution of white-tailed deer mtDNA
lineages in CWD outbreak areas in southern Wisconsin,
USA. Journal of Toxicology and Environmental Health, 74, 1521–
1535.
Rolley RE (2007) White-tailed Deer Population Status.
Department of Natural Resources, Wisconsin, Madison, WI,
USA.
Rooney TP (2001) Deer impacts on forest ecosystems: a North
American perspective. Forestry, 74, 201.
Rosenberry CS, Lancia RA, Conner MC (1999) Population
effects of white-tailed deer dispersal. Wildlife Society Bulletin,
27, 858–864.
Schauber EM, Storm DJ, Nielsen CK (2007) Effects of joint
space use and group membership on contact rates among
white-tailed deer. Journal of Wildlife Management, 71, 155–163.
Schwartz MK, McKelvey KS (2009) Why sampling scheme
matters: the effect of sampling scheme on landscape genetic
results. Conservation Genetics, 10, 441–452.
Segelbacher G, Cushman SA, Epperson BK et al. (2010)
Applications of landscape genetics in conservation biology:
concepts and challenges. Conservation Genetics, 11, 375–385.
Shelton P, McDonald P (2010) Illinois Chronic Wasting Disease:
2009–2010 Surveillance ⁄ Management Summary Forest Wildlife
Program. Illinois Department of Natural Resources, Rockford,
Illinois.
Skuldt LH, Mathews NE, Oyer AM (2008) White-tailed deer
movements in a chronic wasting disease area in south-
central Wisconsin. Journal of Wildlife Management, 72, 1156–
1160.
Storfer A, Murphy MA, Spear SF, Holderegger R, Waits LP
(2010) Landscape genetics: where are we now? Molecular
Ecology, 19, 3496–3514.
Storm D (2011) Chronic Wasting Disease in White-Tailed Deer:
Evaluation of Aerial Surveys; Age-Estimation; and the Role of
Deer Density and Landscape in Disease Transmission. University
of Wisconsin, Madison, Wisconsin.
Wagner HH, Holderegger R, Werth S et al. (2005) Variogram
analysis of the spatial genetic structure of continuous
populations using multilocus microsatellite data. Genetics,
169, 1739–1752.
Wang M, Schreiber A (2001) The impact of habitat
fragmentation and social structure on the population
genetics of roe deer (Capreolus capreolus L.) in Central
Europe. Heredity, 86, 703–715.
WI State Climatology Office (2008) Available from http://
www.aos.wisc.edu/~sco/clim-history/state/index.html.
Wiegand T, Moloney KA, Naves J, Knauer F (1999) Finding the
missing link between landscape structure and population
dynamics: a spatially explicit perspective. American
Naturalist, 154, 605–627.
Published 2012. This article is a
Wiens JA (2001) The landscape context of dispersal. In:
Dispersal (eds Clobert J, Danchin E, Dhondt AA, Nichols JD),
pp. 96–109. Oxford University Press, Oxford.
Williams ES, Miller MW, Kreeger TJ, Kahn RH, Thorne ET
(2002) Chronic wasting disease of deer and elk: a review
with recommendations for management. The Journal of
Wildlife Management, 66, 551–563.
Wilson GA, Strobeck C, Wu L, Coffin JW (1997)
Characterization of microsatellite loci in caribou, Rangifer
tarandus, and their use in other artiodactyls. Molecular
Ecology, 6, 697–699.
Wisconsin Department of Natural Resources (2003) Wisconsin
deer harvest and license sales, 1966–2002. Retrieved March
19, 2004, from State of Wisconsin, Department of Natural
Resources Available from: http://dnr.wi.gov/org/land/
wildlife/hunt/deer/harvest02.htm.
Wisconsin Dept. Natural Resources (2010) Wisconsin’s Chronic
Wasting Disease Response Plan: 2010–2025. Wisconsin
Department of Natural Resources, Madison, Wisconsin.
Wozencraft WC (1978) Investigations Concerning a High-Density
White-Tailed Deer Population in South Central Wisconsin.
University of Wisconsin, Madison, WI, USA.
S.J.R. is interested in the intersection of landscape genetics and
landscape epidemiology; she is currently a postdoctoral fellow
in the Department of Forest & Wildlife Ecology at the Univer-
sity of Wisconsin-Madison. M.D.S. is an Associate Professor of
Wildlife Ecology and assistant unit leader of the Wisconsin
Cooperative Wildlife Research Unit; his research focuses on
wildlife epidemiology and modeling. D.L. and P.S. work for
the Wisconsin and Illinois Departments of Natural Resources
respectively.
Data accessibility
R code and microsatellite data: DRYAD entry doi:10.5061/
dryad.p7639.
Supporting information
Additional supporting information may be found in the online
version of this article.
Table S1 Microsatellite primers and PCR conditions for geno-
types generated for Midwestern USA white-tailed deer.
Table S2 Model selection determining landscape variables that
best explain genetic patterns in white-tailed deer in the Mid-
western USA.
Please note: Wiley-Blackwell is not responsible for the content
or functionality of any supporting information supplied by the
authors. Any queries (other than missing material) should be
directed to the corresponding author for the article.
US Government work and is in the public domain in the USA.