Content uploaded by Robert C Lonsinger
Author content
All content in this area was uploaded by Robert C Lonsinger on Feb 12, 2018
Content may be subject to copyright.
Balancing sample accumulation and DNA degradation rates
to optimize noninvasive genetic sampling of sympatric
carnivores
ROBERT C. LONSINGER,* ERIC M. GESE,†‡ STEVEN J. DEMPSEY,‡BRYAN M. KLUEVER,‡
TIMOTHY R. JOHNSON§and LISETTE P. WAITS*
*Department of Fish and Wildlife Sciences, University of Idaho, 875 Perimeter Drive MS1136, Moscow, ID 83844-1136, USA,
†United States Department of Agriculture, Wildlife Services, National Wildlife Research Center, Logan, UT 84322-5230, USA,
‡Department of Wildland Resources, Utah State University, Logan, UT 84322-5230, USA, §Department of Statistical Science,
University of Idaho, 875 Perimeter Drive MS1104, Moscow, ID 83844-1104, USA
Abstract
Noninvasive genetic sampling, or noninvasive DNA sampling (NDS), can be an effective monitoring approach for
elusive, wide-ranging species at low densities. However, few studies have attempted to maximize sampling effi-
ciency. We present a model for combining sample accumulation and DNA degradation to identify the most efficient
(i.e. minimal cost per successful sample) NDS temporal design for capture–recapture analyses. We use scat accumula-
tion and faecal DNA degradation rates for two sympatric carnivores, kit fox (Vulpes macrotis) and coyote (Canis
latrans) across two seasons (summer and winter) in Utah, USA, to demonstrate implementation of this approach. We
estimated scat accumulation rates by clearing and surveying transects for scats. We evaluated mitochondrial
(mtDNA) and nuclear (nDNA) DNA amplification success for faecal DNA samples under natural field conditions for
20 fresh scats/species/season from <1–112 days. Mean accumulation rates were nearly three times greater for coyotes
(0.076 scats/km/day) than foxes (0.029 scats/km/day) across seasons. Across species and seasons, mtDNA amplifica-
tion success was ≥95% through day 21. Fox nDNA amplification success was ≥70% through day 21 across seasons.
Coyote nDNA success was ≥70% through day 21 in winter, but declined to <50% by day 7 in summer. We identified a
common temporal sampling frame of approximately 14 days that allowed species to be monitored simultaneously,
further reducing time, survey effort and costs. Our results suggest that when conducting repeated surveys for cap-
ture–recapture analyses, overall cost-efficiency for NDS may be improved with a temporal design that balances field
and laboratory costs along with deposition and degradation rates.
Keywords:Canis latrans, DNA degradation, genotyping error, noninvasive genetic sampling, scat deposition, Vulpes
macrotis
Received 15 September 2014; revision received 26 November 2014; accepted 28 November 2014
Introduction
Noninvasive genetic sampling, or noninvasive DNA
sampling (NDS), is increasingly being used to monitor
species that are rare, elusive or otherwise difficult to sur-
vey with traditional techniques (Waits & Paetkau 2005).
Genetic material obtained from noninvasive sources (e.g.
faeces, hair, feathers) can allow for species identification
and individual identification, population genetic
structure, genetic diversity, connectivity and sex ratios
(Beja-Pereira et al. 2009). Combining NDS with capture–
recapture and occupancy modelling approaches allows
researchers to estimate population demographic parame-
ters (Lukacs & Burnham 2005) and patterns of occur-
rence (Long et al. 2011). Many studies have opted for
NDS due to logistical and animal welfare considerations,
or improved cost-benefits (e.g. Prugh et al. 2005; Brøseth
et al. 2010; Stenglein et al. 2010b).
DNA degradation and genotyping errors can influ-
ence NDS results (Taberlet et al. 1999; Waits & Paetkau
2005; Beja-Pereira et al. 2009). Accordingly, researchers
have expended considerable effort to understand how
factors such as sample age (Piggott 2004; Murphy et al.
2007; Santini et al. 2007), environmental conditions (Pig-
gott 2004; Murphy et al. 2007; Santini et al. 2007; DeMay
et al. 2013), diet (Murphy et al. 2003; Panasci et al. 2011),
sample collection and storage techniques (Murphy et al.
Correspondence: Robert C. Lonsinger, Fax: 208-885-9080;
E-mail: Lons1663@vandals.uidaho.edu
©2014 John Wiley & Sons Ltd
Molecular Ecology Resources (2014) doi: 10.1111/1755-0998.12356
2002; Palomares et al. 2002; Piggott & Taylor 2003; Steng-
lein et al. 2010a; Panasci et al. 2011), locus length (Buchan
et al. 2005; DeMay et al. 2013) and species-specific differ-
ences (Piggott & Taylor 2003; Buchan et al. 2005) influ-
ence the degradation of DNA. Collectively these studies
indicate DNA degradation and genotyping errors vary
among species and environmental conditions. General
recommendations to reduce degradation and genotyping
errors included sampling the freshest scats and conduct-
ing surveys during the driest and/or coldest seasons
(Murphy et al. 2007; Santini et al. 2007).
While previous efforts to optimize NDS have focused
on ways to minimize DNA degradation and genotyping
errors, they have not explicitly incorporated sample
accumulation rates. Understanding sample accumulation
rates (i.e. the rate at which noninvasive genetic samples
accrue and can be obtained) is critical to designing effi-
cient sampling and may influence the optimal temporal
sampling frame. Faecal DNA is a common source of non-
invasive genetic samples, but sample accumulation rate
is probably affected by diet, behaviour, physiology and
environmental conditions. For example, seasonal varia-
tion in diet, behaviour and space use by carnivores can
influence scat deposition rates and patterns (Andelt &
Andelt 1984; Ralls et al. 2010). Additionally, heavy rain
or winds can remove scats, as can conspecifics (Living-
ston et al. 2005).
The temporal sampling design of NDS can be opti-
mized to maximize laboratory success while minimizing
overall cost per successful sample. Laboratory costs are
driven by the number of samples collected, polymerase
chain reaction (PCR) success rates and genotyping error
rates (Fig. 1). Scat accumulation rates, survey effort (spa-
tial coverage), desired sample size (number of samples
required to achieve objectives) and the number of sam-
pling events (temporal frequency) necessary to achieve
the desired sample size influence field costs (Fig. 1).
Thus, to optimize the temporal design for NDS, pilot
studies should consider both laboratory and field costs
by incorporating DNA degradation and sample accumu-
lation rates for each species, season and study site.
Here, we present a model for combining information
on sample accumulation and DNA degradation to opti-
mize (i.e. identify the most cost-effective) temporal sam-
pling design for capture–recapture studies employing
NDS. We use scat accumulation rates and faecal DNA
degradation rates for two sympatric carnivores, kit foxes
(Vulpes macrotis; hereafter foxes) and coyotes (Canis la-
trans), across two seasons in the Great Basin desert of
Utah, USA, to demonstrate how this approach can be
implemented. In regards to scat accumulation, we
hypothesized that (i) scat accumulation would be greater
for coyotes than foxes due to their more omnivorous diet
and higher abundance and (ii) seasonal variation in diets
would result in higher accumulation rates in summer
than winter for both species (Andelt & Andelt 1984; Arjo
et al. 2007; Kozlowski et al. 2008). Regarding DNA degra-
dation, we hypothesized that (i) due to its higher relative
abundance mitochondrial DNA (mtDNA) would have
higher PCR (or amplification) success rates than nuclear
DNA (nDNA), (ii) amplification success would decrease
over time for both nDNA and mtDNA, (iii) amplification
success would decrease more precipitously for nDNA
than mtDNA and (iv) amplification success for nDNA
would be higher for shorter microsatellite loci than
longer loci (Buchan et al. 2005; DeMay et al. 2013).
Materials and methods
Study area
Our investigation took place on the U.S. Army Dugway
Proving Ground (DPG), in western Utah. Located within
the Great Basin, DPG is characterized by basin and range
formations with elevations from 1228 to 2154 m (Arjo
et al. 2007). The site experiences cold winters and moder-
ate summers; coldest and warmest months are January
(mean high =3.3 °C, mean low =8.8 °C) and July
(mean high =34.7 °C, mean low =16.3 °C), respectively.
Mean annual precipitation is approximately 20 cm with
the greatest rainfall occurring in spring (Arjo et al. 2007).
Sampling seasons corresponded to periods preceding
breeding (January and February) and juvenile dispersal
(July and August) for target species and aligned with
periods of reduced precipitation in the region (Arjo et al.
2007).
Fig. 1 Conceptual diagram showing the major components
required to balance field and laboratory efficiency for optimiza-
tion of noninvasive genetic sampling for capture–recapture
analysis.
©2014 John Wiley & Sons Ltd
2R. C. LONSINGER ET AL.
Sample accumulation surveys
Scat accumulation surveys in which transects are cleared
and surveyed approximately 14 days later are commonly
used to estimate relative abundances of canids (Gese
2001; Schauster et al. 2002). Using this approach, we con-
ducted scat accumulation surveys between September
2010 and July 2012. Scat surveys were originally initiated
to evaluate relative abundance of foxes and coyotes and
therefore data were available not only for our winter and
summer sampling seasons, but also for spring. Fifteen
5 km transects along dirt or gravel roads were cleared
and surveyed for carnivore scats approximately 14 days
later (mean =13.9 0.51 SD, range =13–16). Each 5 km
transect was surveyed during two summers (2010, 2011),
two springs (2011, 2012) and one winter (2011). Addition-
ally, to expand the spatial coverage and ensure that stan-
dardized accumulation rates (scats/km/day) were
similar between sampling intervals of different durations,
we evaluated scat accumulation along eight shorter tran-
sects during one summer (2012), using a random starting
point, direction and length (mean =2.6 0.85 SD,
range =1–3.5 km) and surveying 7 days after clearing.
We determined species for each carnivore scat detected
during accumulation surveys based on overall appear-
ance, size and shape (Kozlowski et al. 2012).
Faecal DNA degradation
Faecal DNA degradation was assessed at DPG during
two seasons, winter (initiated 8 February 2012) and sum-
mer (initiated 11 July 2012), corresponding to proposed
field sampling seasons. In each season, 20 fresh scats
were collected per species. Fox scats were obtained from
live-captured, free-ranging individuals, and coyote scats
were obtained from the USDA/NWRC/Predator
Research Facility (Millville, UT, USA). Scats were frozen
within four hours of collection. On average, fox and coy-
ote scats were stored frozen for 18 months and
<1 month, respectively, before being transferred to the
study site, thawed and placed in the field and protected
from disturbance with a frame covered with wire mesh
(25 mm openings; 0.7 gauge wire). We collected faecal
DNA samples from each scat at days 1, 3, 7, 14, 21, 56
and 112, or until the scat was fully utilized. Day 1 sam-
ples were collected just prior to exposure to field condi-
tions. We added a day 5 time point during summer to
provide greater resolution, as a recent study detected a
significant decline in coyote faecal DNA quality as early
as 5 days postdeposition (Panasci et al. 2011). Addition-
ally, a severe wind event during winter buried experi-
mental plots after day 21, so day 56 and 112 time points
were only available for summer. Faecal DNA samples
were collected from the side of each scat following
procedures of Stenglein et al. (2010a), and scats were
considered fully utilized when no additional samples
could be collected in this manner. All samples were
stored in 1.4 mL of DET buffer (20% DMSO, 0.25 M
EDTA, 100 lMTris, pH 7.5 and NaCl to saturation; Seu-
tin et al. 1991). Due to natural variability in scat sizes,
some smaller scats were fully utilized before completion
of all time points, resulting in reduced sample sizes at
later time points. To maintain more equitable sample
sizes among time points during summer, we placed
three additional scats for each species out at the start of
the degradation study and sampled these scats in place
of fully utilized scats at later time points.
DNA extraction and PCR amplification
We conducted faecal DNA extraction and PCR amplifica-
tion in a facility dedicated to low-quality DNA. Faecal
DNA samples were extracted using the QIAamp DNA
Stool Mini Kits (Qiagen, Inc., Valencia, CA, USA) with
negative controls to monitor for contamination (Taberlet
& Luikart 1999; Beja-Pereira et al. 2009). We performed
mtDNA species identification tests by amplifying frag-
ments of the control region (Onorato et al. 2006; De Barba
et al. 2014). Species-specific PCR products lengths were
336–337 base pairs (bp) for foxes and 115–120 bp and
360–364 bp for coyotes (De Barba et al. 2014). Samples
that failed to amplify for mtDNA were repeated once to
minimize sporadic effects (Murphy et al. 2007). For indi-
vidual identification, we amplified fox and coyote sam-
ples with seven and nine nDNA microsatellite loci,
respectively (Appendix S1, Supporting information). We
conducted PCR on a Bio-Rad Tetrad thermocycler (Bio-
Rad, Hercules, CA, USA) including negative and posi-
tive controls. PCR conditions, including primer concen-
trations and thermal profiles, are presented in Appendix
S1 (Supporting information). We visualized results using
a 3130xl DNA Analyzer (Applied Biosystems, Foster
City, CA, USA) and scored allele sizes with Genemapper
3.7 (Applied Biosystems). Samples were considered suc-
cessful for species identification if amplification of
≥1 mtDNA fragment was achieved in either the first or
second amplification attempt. We calculated mtDNA
success rates as the proportion of successful samples
across each time point and season. We calculated nDNA
amplification success rates (number of successful ampli-
fications/total possible) and sample success rates (pro-
portion of samples that amplified at ≥50% of the loci) for
each time point and species.
Genotyping error rates
We combined replicates for each scat (i.e. all replicates
across time points with successful nDNA amplification)
©2014 John Wiley & Sons Ltd
OPTIMIZING NONINVASIVE GENETIC SAMPLING 3
to establish consensus genotypes (Taberlet et al. 1999;
Pompanon et al. 2005). To achieve a consensus genotype,
we required that heterozygote and homozygote alleles
be observed in two and three independent replicates,
respectively. Following the methods of Broquet & Petit
(2004), we classified the observation of an allele not pres-
ent in the consensus genotype as a false allele (FA) and
the amplification of only one allele in a heterozygous
consensus genotype as allelic dropout (ADO).
Data analysis
Scat accumulation results were standardized across tran-
sects and species as daily accumulation rates (scats accu-
mulated/days since clearing =scats/km/day). We
employed a generalized linear model to test effects of
season and species on scat accumulation (O’Hara & Ko-
tze 2010). We considered a Poisson regression model
with a log link function, but residuals indicated under-
dispersion so we based inferences on quasi-likelihood
with a free dispersion parameter. We used a likelihood
ratio test to compare models with and without interac-
tions. We compared the influence of main effects and fac-
tor levels with contrast analysis (Rpackage contrast;
Kuhn et al. 2011; R Core Team 2014).
We evaluated PCR success, FA and ADO as binary
response variables with mixed-effects logistic regression
models to assess DNA degradation rates, with sample
included as a random effect to resolve pseudoreplication
effects due to multiple observations per sample with SAS
9.3 (SAS Institute Inc. 2011). We included time since the
scat was placed in the field (log transformed), DNA type
(mtDNA vs. nDNA), species (fox vs. coyote), season
(winter vs. summer) and locus length as fixed effects in
the model for PCR success. We excluded DNA type from
models for FA and ADO as these pertain only to nDNA.
We categorized nDNA locus lengths based on the mid-
length of alleles per locus by species (range: 90–275 bp).
Optimization of NDS temporal design
Our goal was to optimize a NDS temporal design that
could be employed within a capture–recapture frame-
work for foxes and coyotes. To this end, we derived a
total cost per successful sample (i.e. sample that achieves
a consensus genotype for individual identification) at
sampling intervals from 1 to 56 days, where the interval
represented the number of days between clearing and
survey or between sequential surveys.
Both spatial survey effort and desired sample size
must be selected by the researcher, but may be informed
by previous research, power analyses and/or simula-
tions (Williams et al. 2002). We selected a survey effort of
150 km, a length of transect which we felt provided
reasonable coverage of our study site and encompasses
1350 km
2
within 2.5 km of transects, the radius of the
average fox home range at DPG (Dempsey 2013). We
identified desired sample sizes of 200 fox and 400 coyote
samples, values approximately three times the number
of individuals expected to be in our study area (Solberg
et al. 2006).
We determined the number of samples accumulated
and available for collection at each potential sampling
interval (1–56 days, hereafter interval), by calculating the
product of the daily accumulation rate (scats/km/day),
the number of kilometres surveyed (effort) and the num-
ber of days in the interval. We combined the number of
samples accumulated at each interval with our model-
predicted PCR success rates to calculate the number of
successful samples for each interval, considering that
each interval contained scats of varying ages and levels
of degradation. For example, for an interval of 3 days,
we assumed that 33.3% of the scats were 1, 2 and 3 days
old and that each age class was characterized by its
model-predicted PCR success.
Noninvasive samples commonly suffer from genotyp-
ing errors (Pompanon et al. 2005), which can influence
costs. For each interval, we summed the model-predicted
FA and ADO rates to determine the overall predicted
genotyping error rate. We then calculated the number of
genotyping errors expected for samples on each day as
the entrywise product of the number of successful sam-
ples and the predicted genotyping error rate for that day.
The total number of samples, with a genotyping error
within a given interval then, was the sum of the number
of samples with a genotyping error across all days con-
tributing to the interval. The cumulative genotyping
error rate for an interval was determined as the propor-
tion of successful samples with a genotyping error.
As genotyping errors increase, additional replicates
are required to reconcile differences among genotypes
(Pompanon et al. 2005). Within a capture–recapture
framework, errors in multilocus genotypes can result in
overestimates of abundance and bias survival estimates
(Lukacs & Burnham 2005). Consequently, we set a goal
of maintaining a probability of error ≤2% in our data set.
We assumed genotyping error rate was similar across
loci, and replicates were independent. We calculated the
probability of having an error in the consensus genotype
at a given interval as the cumulative genotyping error
rate raised to the number of replicates, then multiplied
by the number of loci. We estimated our laboratory costs
to be approximately $60/sample (including labour and
supplies for extraction, four independent amplifications
and finalization of the consensus genotype), based on
current laboratory expenses, with a 25% increase in cost
for each additional pair of replicates. Thus, when the
number of replicates required to maintain our goal of
©2014 John Wiley & Sons Ltd
4R. C. LONSINGER ET AL.
to establish consensus genotypes (Taberlet et al. 1999;
Pompanon et al. 2005). To achieve a consensus genotype,
we required that heterozygote and homozygote alleles
be observed in two and three independent replicates,
respectively. Following the methods of Broquet & Petit
(2004), we classified the observation of an allele not pres-
ent in the consensus genotype as a false allele (FA) and
the amplification of only one allele in a heterozygous
consensus genotype as allelic dropout (ADO).
Data analysis
Scat accumulation results were standardized across tran-
sects and species as daily accumulation rates (scats accu-
mulated/days since clearing =scats/km/day). We
employed a generalized linear model to test effects of
season and species on scat accumulation (O’Hara & Ko-
tze 2010). We considered a Poisson regression model
with a log link function, but residuals indicated under-
dispersion so we based inferences on quasi-likelihood
with a free dispersion parameter. We used a likelihood
ratio test to compare models with and without interac-
tions. We compared the influence of main effects and fac-
tor levels with contrast analysis (Rpackage contrast;
Kuhn et al. 2011; R Core Team 2014).
We evaluated PCR success, FA and ADO as binary
response variables with mixed-effects logistic regression
models to assess DNA degradation rates, with sample
included as a random effect to resolve pseudoreplication
effects due to multiple observations per sample with SAS
9.3 (SAS Institute Inc. 2011). We included time since the
scat was placed in the field (log transformed), DNA type
(mtDNA vs. nDNA), species (fox vs. coyote), season
(winter vs. summer) and locus length as fixed effects in
the model for PCR success. We excluded DNA type from
models for FA and ADO as these pertain only to nDNA.
We categorized nDNA locus lengths based on the mid-
length of alleles per locus by species (range: 90–275 bp).
Optimization of NDS temporal design
Our goal was to optimize a NDS temporal design that
could be employed within a capture–recapture frame-
work for foxes and coyotes. To this end, we derived a
total cost per successful sample (i.e. sample that achieves
a consensus genotype for individual identification) at
sampling intervals from 1 to 56 days, where the interval
represented the number of days between clearing and
survey or between sequential surveys.
Both spatial survey effort and desired sample size
must be selected by the researcher, but may be informed
by previous research, power analyses and/or simula-
tions (Williams et al. 2002). We selected a survey effort of
150 km, a length of transect which we felt provided
reasonable coverage of our study site and encompasses
1350 km
2
within 2.5 km of transects, the radius of the
average fox home range at DPG (Dempsey 2013). We
identified desired sample sizes of 200 fox and 400 coyote
samples, values approximately three times the number
of individuals expected to be in our study area (Solberg
et al. 2006).
We determined the number of samples accumulated
and available for collection at each potential sampling
interval (1–56 days, hereafter interval), by calculating the
product of the daily accumulation rate (scats/km/day),
the number of kilometres surveyed (effort) and the num-
ber of days in the interval. We combined the number of
samples accumulated at each interval with our model-
predicted PCR success rates to calculate the number of
successful samples for each interval, considering that
each interval contained scats of varying ages and levels
of degradation. For example, for an interval of 3 days,
we assumed that 33.3% of the scats were 1, 2 and 3 days
old and that each age class was characterized by its
model-predicted PCR success.
Noninvasive samples commonly suffer from genotyp-
ing errors (Pompanon et al. 2005), which can influence
costs. For each interval, we summed the model-predicted
FA and ADO rates to determine the overall predicted
genotyping error rate. We then calculated the number of
genotyping errors expected for samples on each day as
the entrywise product of the number of successful sam-
ples and the predicted genotyping error rate for that day.
The total number of samples, with a genotyping error
within a given interval then, was the sum of the number
of samples with a genotyping error across all days con-
tributing to the interval. The cumulative genotyping
error rate for an interval was determined as the propor-
tion of successful samples with a genotyping error.
As genotyping errors increase, additional replicates
are required to reconcile differences among genotypes
(Pompanon et al. 2005). Within a capture–recapture
framework, errors in multilocus genotypes can result in
overestimates of abundance and bias survival estimates
(Lukacs & Burnham 2005). Consequently, we set a goal
of maintaining a probability of error ≤2% in our data set.
We assumed genotyping error rate was similar across
loci, and replicates were independent. We calculated the
probability of having an error in the consensus genotype
at a given interval as the cumulative genotyping error
rate raised to the number of replicates, then multiplied
by the number of loci. We estimated our laboratory costs
to be approximately $60/sample (including labour and
supplies for extraction, four independent amplifications
and finalization of the consensus genotype), based on
current laboratory expenses, with a 25% increase in cost
for each additional pair of replicates. Thus, when the
number of replicates required to maintain our goal of
©2014 John Wiley & Sons Ltd
4R. C. LONSINGER ET AL.
degradation (Table 2). We detected significant interac-
tions between the fixed effects of time and both season
and locus length. PCR success for mtDNA and nDNA
declined more slowly in winter than summer, and
nDNA success declined more precipitously for longer
loci than shorter loci (Fig. 3). Significant interactions
Table 1 Generalized linear model and contrast analysis results with standard errors (SE) and lower (LL) and upper (UL) 95% confi-
dence bounds for scat accumulation samples collected from September 2012 to July 2012 at Dugway Proving Ground, Utah. Species lev-
els include coyote (Canis latrans) and kit fox (Vulpes macrotis). Season levels include spring, summer and winter
Estimate SE z-value P-value LL UL
Model parameters
(Intercept) 3.01 0.243 12.37 <0.001*3.52 2.56
Summer 0.66 0.277 2.38 0.019*0.13 1.22
Winter 0.47 0.349 1.36 0.177 0.23 1.16
Kit fox 0.97 0.253 3.83 <0.001*1.49 0.49
Contrasts
Coyote vs. Kit fox 1.08 0.119 9.09 <0.001*1.32 0.85
Summer vs. Winter 0.26 0.137 1.89 0.059 0.01 0.53
Summer vs. Spring 0.79 0.131 5.99 <0.001*0.53 1.04
Spring vs. Winter 0.53 0.167 3.16 0.002*0.85 0.19
Significant (*)P-values for zstatistic evaluated at a=0.05.
Table 2 Mixed-effects logistic regression model results for PCR success, allelic dropout and false alleles for kit fox (Vulpes macrotis) and
coyote (Canis latrans) faecal DNA samples collected in 2012 during winter and summer at Dugway Proving Ground, Utah. Reported
chi-square test statistics and P-values were generated with Type III tests of fixed effects
Fixed effect
PCR success Allelic dropout False alleles
Chi-square P-value Chi-square P-value Chi-square P-value
Time 4.93 0.0263*0.80 0.3706 0.09 0.7678
DNA type 224.06 <0.0001*————
Locus length 8.73 0.0031*0.03 0.8661 1.26 0.2614
Season 4.02 0.0449*4.11 0.0427*0.93 0.3337
Species 25.90 <0.0001*0.64 0.4237 7.95 0.0048*
Time 9Season 42.02 <0.0001*0.28 0.5966 5.91 0.0150*
Time 9Species 24.15 <0.0001*4.09 0.0432*4.94 0.0262*
Time 9Locus length 13.38 0.0003*1.03 0.3100 0.04 0.8386
Locus length 9Season 1.57 0.2100 1.22 0.2699 0.15 0.7020
Locus length 9Species 8.36 0.0038*1.57 0.2098 10.16 0.0014
Significance (*) was evaluated at a=0.05. Time was log-transformed days since the scat was placed in the field. DNA types included
mitochondrial and nuclear DNA. Locus length was based on the midpoint of each locus (range 90–275 base pairs).
Kit fox Co
y
ote
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
Summer Winter
0306090 0306090
Time (Days)
Probability of PCR success
nDNA (80 bp)
nDNA (130 bp)
nDNA (180 bp)
nDNA (230 bp)
nDNA (280 bp)
mtDNA
Fig. 3 Mixed-effects logistic regression
model results for PCR success for kit fox
(Vulpes macrotis) and coyote (Canis latrans)
faecal DNA samples collected in 2012
during winter and summer at Dugway
Proving Ground, Utah.
©2014 John Wiley & Sons Ltd
6R. C. LONSINGER ET AL.
were detected between species and both time and locus
length (Table 2).
Genotyping error rates
Overall genotyping error rates varied between species
(Fig. S2, Supporting information); across seasons and
sampling periods, overall ADO was lower for foxes
(18%) than coyotes (25%), while overall FA rate was
slightly higher for foxes (5%) than coyotes (2%). Win-
ter samples of both species had lower genotyping
error rates on average than summer samples. Fox win-
ter ADO rates ranged from 4% to 36%, whereas fox
summer ADO rates ranged from 15% to 42% (Fig. S2,
Supporting information). Coyote ADO rates ranged
from 10% to 29% in winter and 15% to 56% in sum-
mer (Fig. S2, Supporting information). In both seasons,
FA rates were low for both species (Fig. S2, Support-
ing information). Models for ADO and FA suggested
that season and species, respectively, were the only
main effects influencing each model (Table 2). Model
results for ADO were influenced by a significant inter-
action between time and species, while model results
for FA were influenced by significant interactions of
time with season and species, and locus length with
species (Table 2). Model-predicted cumulative genotyp-
ing error rates (combined ADO and FA rates across
loci and intervals) were lower for foxes (winter
mean =20.9 0.6% SE; summer mean =25.1 0.6%
SE) than coyotes (winter mean =31.5 0.6% SE; sum-
mer mean =37.4 0.5% SE) and higher in summer
than winter for both species.
Optimization of NDS temporal design
For fox, the predicted number of samples accumulated
ranged from 4.1 (interval =1 day) to 226.8 (inter-
val =56 days) in winter and 6.2 (interval =1 day) to
345.0 (interval =56 days) in summer. The predicted
number of coyote samples accumulated ranged from
12.5 (interval =1 day) to 697.2 (interval =56 days) in
winter and 13.5 (interval =1 day) to 756.0 (inter-
val =56 days) in summer. For both species, the number
of samples predicted to fail for nDNA microsatellite
amplification, however, increased as interval length
increased (Fig. S3, Supporting information). Across sea-
sons and time points, a greater proportion of accumu-
lated coyote samples were predicted to fail than fox
samples (Fig. S3, Supporting information).
Based on model-predicted genotyping error rates, our
goal of ≤2% probability of error in the data set could be
achieved for fox with five or fewer replicates at all inter-
vals, with four replicates being sufficient up to 34 days
in winter and 16 days in summer. To achieve this goal
for coyotes, up to seven replicates were required. In win-
ter, five replicates were required for intervals of 3–
16 days, six replicates for intervals of 17–49 days and
seven replicates for intervals ≥50 days. For summer coy-
ote samples, the minimum number of replicates required
was five (1–3 days). Six replicates were required for
intervals of 4–17 days and seven replicates for intervals
of ≥18 days.
The number of sampling events necessary to obtain
desired sample sizes was initially high due to the low
number of samples accumulating over shorter intervals,
but declined precipitously (Fig. 4). The number of sam-
pling events was higher initially in winter than summer
for both species due to seasonal differences in accumula-
tion. The number of sampling events required was typi-
cally greater for foxes than coyotes despite the smaller
desired sample size; this difference was greater in sum-
mer than winter (Fig. 4).
Overall cost per successful sample showed a similar
pattern across species and seasons, but with differ-
ences in the magnitude and timing of changes. Cost
per successful sample was highest for both species
and seasons at the shortest intervals and was higher
for foxes (Fig. 4a) than coyotes (Fig. 4b) at shorter
intervals. For both species, cost per successful sample
was higher in winter than summer at short intervals.
Summer cost per successful sample surpassed winter
costs at 7 days for coyotes and 16 days for foxes.
Costs per successful sample declined as the number of
required sampling events reduced field costs, until
genotyping errors were sufficiently high to require
additional replicates, increasing laboratory costs. The
overall lower cumulative genotyping error resulted in
smaller increases in overall cost for foxes (Fig. 4a) rela-
tive to coyotes (Fig. 4b). Sharp increases in cost associ-
ated with additional replicates occurred at a shorter
interval for foxes (35 days) than coyotes (50 days) in
winter. In summer, sharp increases in cost associated
with additional replicates occurred at the same inter-
val (17 days) for both species. When surveying species
simultaneously, overall cost per successful sample was
reduced (Fig. 4c) for each species, due to reduced field
costs for each species individually. Average annual
cost per successful sample suggested that a temporal
sampling frame of approximately 14 days would
reduce costs for each species and allow species to be
monitored simultaneously (Fig. 4c).
Discussion
Our study is among the first to incorporate DNA
degradation and sample accumulation rates to opti-
mize NDS design; a similar approach was recently
applied to ungulates (Woodruff et al. in press). Our
©2014 John Wiley & Sons Ltd
OPTIMIZING NONINVASIVE GENETIC SAMPLING 7
degradation (Table 2). We detected significant interac-
tions between the fixed effects of time and both season
and locus length. PCR success for mtDNA and nDNA
declined more slowly in winter than summer, and
nDNA success declined more precipitously for longer
loci than shorter loci (Fig. 3). Significant interactions
Table 1 Generalized linear model and contrast analysis results with standard errors (SE) and lower (LL) and upper (UL) 95% confi-
dence bounds for scat accumulation samples collected from September 2012 to July 2012 at Dugway Proving Ground, Utah. Species lev-
els include coyote (Canis latrans) and kit fox (Vulpes macrotis). Season levels include spring, summer and winter
Estimate SE z-value P-value LL UL
Model parameters
(Intercept) 3.01 0.243 12.37 <0.001*3.52 2.56
Summer 0.66 0.277 2.38 0.019*0.13 1.22
Winter 0.47 0.349 1.36 0.177 0.23 1.16
Kit fox 0.97 0.253 3.83 <0.001*1.49 0.49
Contrasts
Coyote vs. Kit fox 1.08 0.119 9.09 <0.001*1.32 0.85
Summer vs. Winter 0.26 0.137 1.89 0.059 0.01 0.53
Summer vs. Spring 0.79 0.131 5.99 <0.001*0.53 1.04
Spring vs. Winter 0.53 0.167 3.16 0.002*0.85 0.19
Significant (*)P-values for zstatistic evaluated at a=0.05.
Table 2 Mixed-effects logistic regression model results for PCR success, allelic dropout and false alleles for kit fox (Vulpes macrotis) and
coyote (Canis latrans) faecal DNA samples collected in 2012 during winter and summer at Dugway Proving Ground, Utah. Reported
chi-square test statistics and P-values were generated with Type III tests of fixed effects
Fixed effect
PCR success Allelic dropout False alleles
Chi-square P-value Chi-square P-value Chi-square P-value
Time 4.93 0.0263*0.80 0.3706 0.09 0.7678
DNA type 224.06 <0.0001*————
Locus length 8.73 0.0031*0.03 0.8661 1.26 0.2614
Season 4.02 0.0449*4.11 0.0427*0.93 0.3337
Species 25.90 <0.0001*0.64 0.4237 7.95 0.0048*
Time 9Season 42.02 <0.0001*0.28 0.5966 5.91 0.0150*
Time 9Species 24.15 <0.0001*4.09 0.0432*4.94 0.0262*
Time 9Locus length 13.38 0.0003*1.03 0.3100 0.04 0.8386
Locus length 9Season 1.57 0.2100 1.22 0.2699 0.15 0.7020
Locus length 9Species 8.36 0.0038*1.57 0.2098 10.16 0.0014
Significance (*) was evaluated at a=0.05. Time was log-transformed days since the scat was placed in the field. DNA types included
mitochondrial and nuclear DNA. Locus length was based on the midpoint of each locus (range 90–275 base pairs).
Kit fox Co
y
ote
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
Summer Winter
0306090 0306090
Time (Days)
Probability of PCR success
nDNA (80 bp)
nDNA (130 bp)
nDNA (180 bp)
nDNA (230 bp)
nDNA (280 bp)
mtDNA
Fig. 3 Mixed-effects logistic regression
model results for PCR success for kit fox
(Vulpes macrotis) and coyote (Canis latrans)
faecal DNA samples collected in 2012
during winter and summer at Dugway
Proving Ground, Utah.
©2014 John Wiley & Sons Ltd
6R. C. LONSINGER ET AL.
with winter samples showing less DNA degradation
than summer samples. Piggott (2004) documented higher
faecal DNA degradation rates in winter than summer
and attributed this to increased moisture during winter.
Previous studies indicate that environmental conditions
such as temperature, UV exposure and humidity influ-
ence DNA degradation rates (Nsubuga et al. 2004; Mur-
phy et al. 2007; Stenglein et al. 2010a). Winters and
summers at DPG receive less precipitation than other
seasons, but temperatures are significantly different (see
Study area) and UV exposure is highest in summer. Our
study design did not allow investigation of the influence
of weather on degradation. We placed all samples in the
field on the same day each season, and therefore,
weather and time were confounded. We suspect though,
that differences observed between seasons were related
to broad differences in environmental conditions.
Our observed ADO and FA rates were similar to those
reported in other canid studies (Piggott 2004; Santini et al.
2007; Stenglein et al. 2010b; Panasci et al. 2011). We were
unable to detect a significant effect of time on genotyping
errors, but this was likely due to small sample sizes associ-
ated with ADO and FA models. We observed a discern-
ible, but not statistically significant increase in model-
predicted ADO rates over time, but not in FA rates.
Optimization of NDS temporal design
By balancing sample accumulation and DNA degrada-
tion, an optimal NDS design can be selected that mini-
mizes cost per successful sample. The optimal interval
varies by species and season and is driven by sample
collection (field) and processing (laboratory) costs.
While the optimal interval is simply the interval that
minimizes the cost per successful sample, additional
factors should be considered such as the number of
target species and interspecific differences in sample
accumulation and DNA degradation. Initial costs per
successful sample were calculated for sampling species
independently (Fig. 4a,b). If a common interval is
selected for foxes and coyotes, both species can be sur-
veyed simultaneously on the same transects and over-
all field costs can be reduced (Fig. 4c). Additionally,
selection of the optimal interval should consider
downstream analyses. For example, demographic clo-
sure assumptions may be difficult to meet at extended
intervals and small reductions in the cost per success-
ful sample may be insufficient justification to select
extended intervals.
Our results indicate a range of intervals for foxes
and coyotes could be selected to improve efficiency,
and these intervals are shorter in summer than winter.
For example, summer cost per successful sample was
minimized for foxes at day 14 and coyotes at day 9,
but selection of an interval 2 days from these opti-
mal intervals changed the cost per successful sample
by <$1. The range of effective intervals was wider in
winter. Winter cost per successful sample was mini-
mized for foxes and coyotes at days 34 and 24, respec-
tively, yet the cost per successful sample changed <$1
for intervals up to 8 days shorter (25–33 days) for
foxes and for 24 intervals surrounding (16–40 days)
the optimal interval for coyotes. We were interested in
selecting a common interval that was effective for both
species and consistent across seasons. Summer cost
per successful sample limited the upper bound of the
common interval, as cost increased sharply for both
species after day 17. We thus identified an interval of
14 days as the common interval within our system
(Fig. 4c). At 14 days, winter cost per successful sample
was reduced and continuing to decline slowly for both
species and the number of sampling events was small
enough to conduct sampling over a single season.
Based on these results, we recommend NDS efforts
account for sample accumulation and DNA degradation
during the design phase (Fig. 1). Previous studies have
recommended sampling the freshest scats possible (Mur-
phy et al. 2007; Santini et al. 2007; DeMay et al. 2013).
Our results show that when sampling over time within a
capture–recapture framework, short intervals may be
cost-prohibitive if a substantial sample size is required.
Thus, we recommend sampling designs consider cost
per successful sample and minimize violations of
assumptions for downstream analyses.
Limitations and implications for research
Collection of fresh samples (e.g. samples known to be
≤1 day old) to evaluate DNA degradation is logistically
prohibitive, particularly when species are rare, elusive,
or difficult to capture. Consequently, many studies
comparing PCR success (e.g. between species, under
environmental variations, over time) have relied on
samples from captive populations (Murphy et al. 2002,
2003, 2007; Piggott 2004; Santini et al. 2007; DeMay et al.
2013). In our study, scats used to evaluate DNA degra-
dation varied between species in origin and length of
storage. We obtained fresh scats from free-ranging
foxes during live capture, but fresh scats from free-
ranging coyotes were unavailable. Consequently, fresh
coyote scats were obtained from a captive population.
Although scats were frozen upon collection, stored for
variable lengths of time and thawed prior to placement
in the field, we do not feel that storage time or the
freeze–thaw cycle significantly impacted PCR success.
While we did not explicitly test the influence of freez-
ing during this study, we previously evaluated PCR
success of canid scats stored in a standard freezer and
©2014 John Wiley & Sons Ltd
OPTIMIZING NONINVASIVE GENETIC SAMPLING 9
found no decline in PCR success for samples frozen for
up to 1 year, when the study ended (L. P. Waits &J. R.
Adams, unpublished data). Our results support this
conclusion. On average, fox and coyote scats were
stored frozen 18 months and <1 month, respectively.
Despite the longer storage time of fox scats, observed
PCR success rates were the same (mtDNA) or higher
(nDNA) for foxes in both seasons and scats of both spe-
cies produced high PCR success at the earliest time
points (Fig. S1, Supporting information). Additionally,
winter temperatures at our site fluctuate from below to
above freezing (night vs. day temperatures) and scats
naturally experience repeated freeze–thaw cycles, yet in
this experiment, we observed higher PCR success rates
for both species in winter relative to summer, suggest-
ing that freeze–thaw cycles were not the driving cause
of DNA degradation.
Variation in diets between captive and free-ranging
coyotes may also influence the generalization of results
to the wild population. Differences in diet could influ-
ence the rate of intestinal cell shedding or the amount
of inhibitors in faecal samples. However, we do not
believe that using captive coyote scats substantially
influenced our results. We have data on success rates
for free-ranging coyote samples collected in winter
and summer 2013, and results are comparable to
model-predicted results from our degradation experi-
ment. For example, for a 14-day interval our model
predicted mean nDNA success rates for coyote scats
of 64.6% (winter; range 46.5–80.7%; Fig. 3) and 47.7%
(summer; range 24.9–71.2%; Fig. 3), and success rates
for free-ranging coyotes sampled with a 14 day inter-
val were 78% (winter) and 55% (summer).
We analysed winter and summer degradation within
the same models for PCR success, ADO and FA to increase
sample size and statistical power, but winter samples were
only available through day 21. Model-predicted results for
winter intervals >21 days assume that trends in predicted
values continue in the same way beyond 21 days (i.e. that
the log odds of the outcome is linear in the log of time), and
consequently, these predictions should be interpreted with
caution. Missing winter data points do not affect the infer-
ences ≤21 days, and it is during this time that the most sub-
stantial changes occurred (Fig. 3).
Monitoring and management implications
This study presents a conceptual model for optimizing
NDS for capture–recapture analysis, which can be
extended to any species or system where estimates of
sample accumulation (e.g. hair snaring rate, scat accu-
mulation rate) and DNA degradation rates can be
quantified. We demonstrate that this novel optimiza-
tion approach can effectively reduce costs of NDS
monitoring programmes. By initiating a pilot study to
evaluate sample accumulation and DNA degradation
rates, NDS monitoring costs can be minimized, allow-
ing monitoring to occur over larger spatial extents and
at higher temporal resolutions than would be possible
otherwise. Differences observed in sample accumula-
tion and DNA degradation rates between species and
across seasons, at the same study site, reiterate the
importance of pilot studies for effectively implement-
ing NDS (Taberlet et al. 1999; Waits & Paetkau 2005).
We recommend that when possible pilot studies incor-
porating DNA degradation should use fresh scats col-
lected from target populations. Additionally,
practitioners optimizing NDS should compare field
collected data to model-predicted results to evaluate
model performance, particularly, when samples used
during pilot studies have an origin other than the
population being monitored.
Kit fox populations are believed to be declining,
and their contemporary distribution is unclear. High
mtDNA success suggests that NDS can be used to
explore presence or occupancy of elusive species, such
as kit fox, across large spatial areas. When employing
NDS for occupancy modelling (or similar approaches),
researchers should acknowledge that mtDNA amplifi-
cations may incorporate old samples violating closure
assumptions and should clear transects before survey-
ing or evaluate sample persistence (MacKenzie &
Reardon 2013). Nuclear DNA success rates were suffi-
cient to identify individuals and provide an effective
capture–recapture approach to estimate population
demographic parameters (Kohn et al. 1999; Marucco
et al. 2011). Both mtDNA and nDNA can be used for
monitoring communities or intraguild interactions and
provide a cost-effective means to monitor management
strategies.
Acknowledgements
Funding provided by the U.S. Department of Defense Environ-
mental Security Technology Certification Programme (12-EB-
RC5-006), Legacy Resource Management Programme (W9132T-
12-2-0050) and Army DPG Environmental Programme. Addi-
tional funding and logistical assistance provided by the U.S.
Department of Agriculture, Wildlife Services, National Wildlife
Research Center; and the Endangered Species Mitigation Fund
of the Utah Department of Natural Resources, Division of Wild-
life Resources. R Knight was essential to securing funding and
provided logistical support. J Adams assisted with laboratory
procedures.
References
Andelt WF, Andelt SH (1984) Diet bias in scat deposition-rate surveys of
coyote density. Wildlife Society Bulletin,12,74–77.
©2014 John Wiley & Sons Ltd
10 R. C. LONSINGER ET AL.
Arjo WM, Gese EM, Bennett TJ, Kozlowski AJ (2007) Changes in kit fox–
coyote–prey relationships in the Great Basin Desert, Utah. Western
North American Naturalist,67, 389–401.
Beja-Pereira A, Oliveira R, Alves PC, Schwartz MK, Luikart G (2009)
Advancing ecological understandings through technological transfor-
mations in noninvasive genetics. Molecular Ecology Resources,9, 1279–
1301.
Broquet T, Petit E (2004) Quantifying genotyping errors in noninvasive
population genetics. Molecular Ecology,13, 3601–3608.
Brøseth H, Flagstad Ø, W€
ardig C, Johansson M, Ellegren H (2010)
Large-scale noninvasive genetic monitoring of wolverines using
scats reveals density dependent adult survival. Biological Conserva-
tion,143, 113–120.
Buchan JC, Archie EA, Van Horn RC, Moss CJ, Alberts SC (2005) Locus
effects and sources of error in noninvasive genotyping. Molecular Ecol-
ogy Notes,5, 680–683.
De Barba M, Adams JR, Goldberg CS et al. (2014) Molecular species iden-
tification for multiple carnivores. Conservation Genetics Resources,6,
821–824.
DeMay SM, Becker PA, Eidson CA et al. (2013) Evaluating DNA degrada-
tion rates in faecal pellets of the endangered pygmy rabbit. Molecular
Ecology Resources,13, 654–662.
Dempsey SJ (2013) Evaluation of survey methods and development of
species distribution models for kit foxes in the Great Basin Desert.
M.S. Thesis, Utah State University.
Gese EM (2001) Monitoring of terrestrial carnivore populations. In: Carni-
vore Conservation(eds Gittleman JL, Funk SM, MacDonald D, Wayne
RK), pp. 372–396. Cambridge University Press, New York.
Kohn MH, York EC, Kamradt DA et al. (1999) Estimating population size
by genotyping faeces. Proceedings of the Royal Society of London, Series B:
Biological Sciences,266, 657–663.
Kozlowski AJ, Gese EM, Arjo WM (2008) Niche overlap and resource par-
titioning between sympatric kit foxes and coyotes in the Great Basin
Desert of western Utah. American Midland Naturalist,160, 191–208.
Kozlowski AJ, Gese EM, Arjo WM (2012) Effects of intraguild predation:
evaluating resource competition between two canid species with
apparent niche separation. International Journal of Ecology,2012,1–12.
Kuhn M, Weston S, Wing J, Forester J (2011) Contrast: A Collection of Con-
trast Methods. R package version 0.17. Available from http://CRAN.
R-project.org/package=contrast.
Livingston TR, Gipson PS, Ballard WB, Sanchez DM, Krausman PR
(2005) Scat removal: a source of bias in feces-related studies. Wildlife
Society Bulletin,33, 172–178.
Long RA, Donovan TM, MacKay P, Zielinski WJ, Buzas JS (2011) Predict-
ing carnivore occurrence with noninvasive surveys and occupancy
modeling. Landscape Ecology,26, 327–340.
Lukacs PM, Burnham KP (2005) Review of capture-recapture methods
applicable to noninvasive genetic sampling. Molecular Ecology,14,
3909–3919.
MacKenzie DI, Reardon JT (2013) Occupancy methods for conservation
management. In: Biodiversity Monitoring and Conservation: Bridging
the Gap Between Global Commitment and Local Action(eds Collen B,
Pettorelli N, Bailie JEM, Durant SM), pp. 248–264. John Wiley &
Sons, Hoboken.
Marucco F, Boitani L, Pletscher DH, Schwartz MK (2011) Bridging the
gaps between non-invasive genetic sampling and population parame-
ter estimation. European Journal of Wildlife Research,57,1–13.
Murphy MA, Waits LP, Kendall KC et al. (2002) An evaluation of long-
term preservation methods for brown bear (Ursus arctos) faecal DNA
samples. Conservation Genetics,3, 435–440.
Murphy MA, Waits LP, Kendall KC (2003) The influence of diet on faecal
DNA amplification and sex identification in brown bears (Ursus arctos).
Molecular Ecology,12, 2261–2265.
Murphy MA, Kendall KC, Robinson A, Waits LP (2007) The impact of
time and field conditions on brown bear (Ursus arctos) faecal DNA
amplification. Conservation Genetics,8, 1219–1224.
Nsubuga AM, Robbins MM, Roeder AD et al. (2004) Factors affecting the
amount of genomic DNA extracted from ape faeces and the identifica-
tion of an improved sample storage method. Molecular Ecology,13,
2089–2094.
O’Hara RB, Kotze DJ (2010) Do not log-transform count data. Methods in
Ecology and Evolution,1, 118–122.
Onorato D, White C, Zager P, Waits LP (2006) Detection of predator pres-
ence at elk mortality using mtDNA analysis of hair and scat samples.
Wildlife Society Bulletin,34, 815–820.
Palomares F, Godoy JA, Piriz A, O’Brien J, Johnson WE (2002) Faecal
genetic analysis to determine the presence and distribution of elu-
sive carnivores: design and feasibility. Molecular Ecology,11, 2171–
2182.
Panasci M, Ballard WB, Breck S et al. (2011) Evaluation of fecal DNA
preservation techniques and effects of sample age and diet on geno-
typing success. Journal of Wildlife Management,75, 1616–1624.
Piggott MP (2004) Effect of sample age and season of collection on the
reliability of microsatellite genotyping of faecal DNA. Wildlife Research,
31, 485.
Piggott MP, Taylor AC (2003) Extensive evaluation of faecal preservation
and DNA extraction methods in Australian native and introduced spe-
cies. Australian Journal of Zoology,51, 341.
Pompanon F, Bonin A, Bellemain E, Taberlet P (2005) Genotyping errors:
causes, consequences and solutions. Nature Reviews Genetics,6, 847–
859.
Prugh LR, Ritland CE, Arthur SM, Krebs CJ (2005) Monitoring coyote
population dynamics by genotyping faeces. Molecular Ecology,14,
1585–1596.
R Core Team (2014) R: A Language and Environment for Statistical Comput-
ing. R Foundation for Statistical Computing, Vienna. URL http://
www.R-project.org/.
Ralls K, Sharma S, Smith DA et al. (2010) Changes in kit fox defecation
patterns during the reproductive season: implications for noninvasive
surveys. Journal of Wildlife Management,74, 1457–1462.
Santini A, Lucchini V, Fabbri E, Randi E (2007) Ageing and environmen-
tal factors affect PCR success in wolf (Canis lupus) excremental DNA
samples. Molecular Ecology Notes,7, 955–961.
SAS Institute Inc. (2011) SAS/STAT
â
9.3 User’s Guide. SAS Institute Inc.,
Cary.
Scandura M, Capitani C, Iacolina L, Marco A (2006) An empirical
approach for reliable microsatellite genotyping of wolf DNA from
multiple noninvasive sources. Conservation Genetics,7, 813–823.
Schauster ER, Gese EM, Kitchen AM (2002) An evaluation of survey
methods for monitoring swift fox abundance. Wildlife Society Bulletin,
30, 464–477.
Seutin G, White BN, Boag PT (1991) Preservation of avian blood and
tissue samples for DNA analyses. Canadian Journal of Zoology,69,
82–90.
Solberg KH, Bellemain E, Drageset O-M, Taberlet P, Swenson JE (2006)
An evaluation of field and non-invasive genetic methods to estimate
brown bear (Ursus arctos) population size. Biological Conservation,128,
158–168.
Stenglein JL, DE Barba M, Ausband DE, Waits LP (2010a) Impacts of
sampling location within a faeces on DNA quality in two carnivore
species. Molecular Ecology Resources,10, 109–114.
Stenglein JL, Waits LP, Ausband DE, Zager P, Mack CM (2010b) Efficient,
noninvasive genetic sampling for monitoring reintroduced wolves.
Journal of Wildlife Management,74, 1050–1058.
Taberlet P, Luikart G (1999) Non-invasive genetic sampling and indi-
vidual identification. Biological Journal of the Linnean Society,68,41–
55.
Taberlet P, Waits LP, Luikart G (1999) Noninvasive genetic sam-
pling: look before you leap. Trends in Ecology and Evolution,14,
323–327.
Waits LP, Paetkau D (2005) Noninvasive genetic sampling tools for
wildlife biologists: a review of applications and recommendations
for accurate data collection. Journal of Wildlife Management,69,
1419–1433.
Williams BK, Nichols JD, Conroy MJ (2002) Analysis and Management of
Animal Populations. Academic Press, San Diego.
©2014 John Wiley & Sons Ltd
OPTIMIZING NONINVASIVE GENETIC SAMPLING 11
Woodruff SP, Johnson TR, Waits LP (in press) Evaluating the interaction
of faecal pellet deposition rates and DNA degradation rates to maxi-
mize sampling design for DNA-based mark-recapture analysis of Son-
oran pronghorn. Molecular Ecolog Resources.
R.C.L. performed data collection, laboratory procedures,
data analysis and interpretation and wrote the manu-
script. E.M.G., S.J.D. and B.M.K. provided scats for DNA
degradation experiments and assisted with data collec-
tion. T.R.J. assisted with statistical analyses and interpre-
tation. L.P.W. designed the study and assisted with data
interpretation. All authors assisted with the manuscript
preparation.
Data accessibility
Raw data (.csv) and analysis code for scat accumulation
(R script) and models of PCR success, ADO and FA (SAS
scripts) are available on Dryad, doi:10.5061/dryad.23k27.
Supporting Information
Additional Supporting Information may be found in the online
version of this article:
Fig. S1 Observed per cent PCR success for mitochondrial
(mtDNA) and nuclear (nDNA) DNA for kit fox (Vulpes macrotis)
and coyote (Canis latrans) faecal DNA samples.
Fig. S2 Observed nuclear DNA genotyping error rates (i.e. allelic
dropout and false alleles) for kit fox (Vulpes macrotis) and coyote
(Canis latrans) faecal DNA samples.
Fig. S3 Proportion of samples accumulated for kit fox (Vulpes
macrotis) and coyote (Canis latrans) in winter and summer that
were predicted to fail for individual identification across sam-
pling intervals.
Appendix S1PCR conditions, including primer concentrations and
thermal profiles, for mitochondrial and nuclear DNA amplification.
©2014 John Wiley & Sons Ltd
12 R. C. LONSINGER ET AL.
