Content uploaded by Naiara Sales
Author content
All content in this area was uploaded by Naiara Sales on Mar 11, 2020
Content may be subject to copyright.
J Appl Ecol. 2020;00:1–10. wileyonlinelibrary.com/journal/jpe
|
1© 2020 British Ecological Society
Received: 3 Octob er 2019
|
Accepted: 21 January 2020
DOI : 10.1111/136 5-2664.13592
RESEARCH ARTICLE
Fishing for mammals: Landscape-level monitoring of terrestrial
and semi-aquatic communities using eDNA from riverine
systems
Naiara Guimarães Sales1 | Maisie B. McKenzie1 | Joseph Drake2 |
Lynsey R. Harper3 | Samuel S. Browett1 | Ilaria Coscia1 | Owen S. Wangensteen4 |
Charles Baillie1 | Emma Bryce5 | Deborah A. Dawson6 | Erinma Ochu1 |
Bernd Hänfling3 | Lori Lawson Handley3 | Stefano Mariani1,7 | Xavier Lambin5 |
Christopher Sutherland2,8 | Allan D. McDevitt1
1Environm ent and Ecosyste m Research Centre, Scho ol of Science, Engi neeri ng and Environment, Universit y of Salfo rd, Salford, UK ; 2Department of
Environmental C onser vatio n, Unive rsity of Massachuset ts-Am herst, Amherst, USA; 3Department of Bio logical and Marine Sciences, U niversity of Hu ll,
Kings ton upon H ull, UK ; 4Norwegian College of Fis hery Science , Univer sity of Troms ø, Tromsø, Nor way; 5School of Biologic al Sciences, University of Aberdeen,
Aberdeen, UK; 6Department of Animal an d Plant Sciences , Univer sity of S heff ield, Sheffield, UK; 7School of Natur al Sciences and Psychology, Liverpool John
Moores University, Liver pool, UK and 8Centre for Research into Ecological an d Environ mental Modelling, University of St Andrews, St Andrews, UK
Naiara G uimarães Sa les, Ma isie B. McKenzie an d Joseph Drake co ntributed eq ually to t his work.
Correspondence
Allan D. McDevitt
Email: a.mcdevitt@salford.ac.uk
Funding information
British Ecological Society, Grant/Award
Number : SR17/1214; University of Salfor d;
University of Massachusetts
Handling Editor: Brittany Mosher
Abstract
1. Environmental DNA (eDNA) metabarcoding has revolutionized biomonitoring in
both marine and freshwater ecosystems. However, for semi-aquatic and terres-
trial animals, the application of this technique remains relatively untested.
2. We first assess the efficiency of eDNA metabarcoding in detecting semi-aquatic
and terrestrial mammals in natural lotic ecosystems in the UK by comparing
sequence data recovered from water and sediment samples to the mammalian
communities expected from historical data. Secondly, using occupancy mod-
elling we compared the detection efficiency of eDNA metabarcoding to mul-
tiple conventional non-invasive survey methods (latrine surveys and camera
trapping).
3. eDNA metabarcoding detected a large proportion of the expected mammalian
community within each area. Common species in the areas were detected at the
majority of sites. Several key species of conservation concern in the UK were de-
tected by eDNA sampling in areas where authenticated records do not currently
exist, but potential false positives were also identified.
4. Water-based eDNA metabarcoding provided comparable results to conventional
survey methods in per unit of survey effort for three species (water vole, field vole
and red deer) using occupancy models. The comparison between survey ‘effort’ to
reach a detection probability of ≥.95 revealed that 3–6 water replicates would be
equivalent to 3–5 latrine surveys and 5–30 weeks of single camera deployment,
depending on the species.
2
|
Journal of Applied Ecology
SALES E t AL.
1 | INTRODUCTION
Environmental DNA (eDNA) metabarcoding (the simultaneous iden-
tification of multiple taxa using DNA extracted from an environmen-
tal sample, e.g. water, soil, based on short amplicon sequences) has
revolutionized the way we approach biodiversity monitoring in both
marine and freshwater ecosystems (Deiner et al., 2017; Valentini
et al., 2016). Successful applications include tracking biological in-
vasions, detecting rare and endangered species and describing en-
tire communities (Holman et al., 2019). Most eDNA metabarcoding
applications on vertebrates to date have focused on monitoring
fishes and amphibians (Hänfling et al., 2016; Valentini et al., 2016).
What has become apparent from studies in lentic systems (ponds
and lakes) is that semi-aquatic and terrestrial mammals can also be
detected (Hänfling et al., 2016; Harper et al., 2019). As a result, there
has been an increasing focus on the use of both vertebrate (Harper
et al., 2019) and mammal-specific primer sets (Leempoel, Herbert, &
Hadly, 2020; Sal es, Kaizer, et al., 2020; Ushio et al., 2017) for detect-
ing mammalian communities using eDNA metabarcoding.
Mammals include some of the most imperiled taxa, with over
one-fifth of species considered to be threatened or declining
(Visconti et al., 2011). Monitoring of mammalian biodiversit y is
therefore essential. Given that any optimal survey approach is likely
to be species-specific, ver y few species can be detected at all times
when they are present. This imperfect detection (even greater for
elusive and rare species) can lead to biased estimates of occurrence
and hinder species conservation (Mackenzie et al., 2002). For mam-
mals, repeated sur veys using several monitoring methods are usually
applied. These include indirect observations such as latrines, faeces,
hair or tracks, or direct observations such as live-trapping or cam-
era trapping sur veys over short time intervals such that closure/in-
variance can be assumed and detectability estimated (Nichols et al.,
2008). Each of these methods has associated efficiency, cost and
required expertise trade-offs, which become more challenging as
the spatial and temporal scales increase.
Environmental DNA sampling yields species-specific presence/
absence data that are likely to be most valuable for inferring spe-
cies distributions using well-established analytical tools such as
occupancy models (MacKenzie et al., 2002). These models resolve
concerns around imperfect detection of difficult to obser ve species.
When coupled with location-specific detec tion histories, these can
be used to infer true occurrence states, factors that influence oc-
cupancy rates, colonization-extinction probabilities and estimates
of detection probability (MacKenzie et al., 2017). The use of eDNA
sampling to generate species-specific detection data has unsurpris-
ingly increased in recent years, and in many cases has outperformed
or at least matched conventional survey methods (Lugg, Griffiths,
van Rooyen, Weeks, & Tingley, 2018; Tingley, Greenlees, Oertel,
van Rooyen, & Weeks, 2019). Although comparisons between eDNA
analysis and conventional surveys for multi-species detection are
numerous (see table S1 in Lugg et al., 2018), studies focusing on de-
tection probability estimates for multiple species identified by me-
tabarcoding are rare (Abrams et al., 2019; Valentini et al., 2016).
The aim of this study was to assess the efficiency of eDNA me-
tabarcoding for detecting semi-aquatic and terrestrial mammals in
natural lotic systems in the UK. We conducted eDNA sampling in
rivers and streams in two areas (Assynt, Scotland and Peak District
National Park, England). Together these locations have the major-
ity of UK semi-aquatic and terrestrial mammalian species present
(Table S1). Our objectives were twofold: first, we sought to estab-
lish whether eDNA metabarcoding is a viable technique for moni-
toring semi-aquatic and terrestrial mammals by comparing it to the
mammalian communities expected from historical data, a group for
which eDNA sampling has rarely been evaluated in a natural set-
ting. Secondly, we evaluate the detection ef ficiency of water- and
sediment-based eDNA sampling in one of these areas (Assynt) for
multiple species compared to multiple conventional non-invasive
survey methods (latrine surveys and camera trapping).
2 | MATERIALS AND METHODS
2.1 | Latrine surveys
Assynt, a heather-dominated upland landscape in the far north-
west of the Scottish Highlands, UK (Figure 1a), is the location of
5. Synthesis and applications. eDNA metabarcoding can be used to generate an initial
‘distribution map’ of mammalian diversity at the landscape level. If conducted dur-
ing times of peak abundance, carefully chosen sampling points along multiple river
courses provide a reliable snapshot of the species that are present in a catchment
area. In order to fully capture solitary, rare and invasive species, we would cur-
rently recommend the use of eDNA metabarcoding alongside other non-invasive
surveying methods (i.e. camera traps) to maximize monitoring efforts.
KEY WORDS
biomonitoring, camera trapping, eDNA metabarcoding, latrine surveys, mammals, occupancy
modelling, rivers
|
3
Journal of Applied Ecolog
y
SALES E t AL.
an ongoing 20 -year met apopulation study of water voles Arvicola
amphibius led by the University of Aberdeen (Figure S1). Here, we
mainly focus only on data collected in 2017. The metapopulation is
characterized by 116 discrete linear riparian habitat patches (ranging
from 90 m to nearly 2.5 km) distributed sparsely (4% of waterway
network) throughout the 140 km2 study area (Sutherland, Elston, &
Lambin, 2014). Water voles use promin ently placed latrines for terri-
tory marking (Figure S2a). Using latrine surveys, a reliable method of
detection (Sutherland et al., 2014), water vole occupancy status was
determined by the detection of latrines that are used for territory
marking (Sutherland, Elston, & Lambin, 2013). During the breed-
ing season (July and August), latrine surveys were conducted twice
at each site. In addition to water vole latrines, field vole Microtus
agrestis pellets are also easily identifiable, and so field vole detec-
tions were also recorded along waterways as a formal part of the la-
trine survey protocol. Live-trapping was then carried out at patches
deemed to be occupied by water voles according to latrine surveys
to determine their abundances (this was used to determine which
sites were sampled for eDNA; Figure 1a).
2.2 | Camera trap data
Camera traps were deployed at the beginning of July and thus over-
lapped temporally with the latrine survey in Assynt. Data were col-
lected from cameras deployed at seven of these patches. Within
each of these patches, cameras were deployed at the midpoint of
the areas where active signs (latrines, grass clipping, burrows) were
detected, and if no signs were detected, at the midpoint of historical
water vole activity (J. Drake, C. Sutherland and X. Lambin, pers.
comm.). These will also capture images of any species present in the
area that come within close proximity of the camera (Figure S3a–f).
Camer as were deployed app roximately 1 m above-gro und on iron
‘u-posts’ to avoid flooding, prevent knock-down by wind/wildlife and
optimize both depth of field and image clarity. Cameras (Bushnell
HD Trophy Cam) were set at normal detection sensitivity (to reduce
false-triggers from grass/shadows), low night time LED intensity (to
prevent image white out in near depth of field), three shot burst
(to increase chance of capturing small, fast moving bodies) and 15-
min intervals between bursts (to increase temporal independence
of captures and decrease memory burden). The area each camera
photographed was approximately 1–2 m2. Animals were identified
on images and information was stored as metadata tags using the
r (R Core Team, 2018) package cam trapr following the procedures
described in Niedballa, Courtiol, and Sollmann (2018). Independence
between detections was based on 60 -min intervals between species-
specific detections.
2.3 | eDNA sampling
A total of 18 potential water vole patches were selected for eDNA
sampling in Assynt from 25 to 27 October 2017. The time lag be-
tween the latrine/live-trapping and eDNA surveys was because of
two main reasons: (a) legitimate concerns around cross-site DNA
contamination during latrine/live-trapping where researchers moved
on a daily basis between sites as well as regularly handled and pro-
cessed live animals (for decontamination procedures see Supporting
FIGURE 1 (a) Environmental DNA (eDNA) sampling sites in A ssynt, Scotland; the size of sites corresponds to abundance categories based
on summer live-trapping. (b) A bubble graph representing presence/absence and categorical values of the number of reads retained (after
bioinformatic filtering) for eDNA (water in blue and sediment in orange) from each wild mammal identified in each site in Assynt (A1–A18)
(a) (b)
4
|
Journal of Applied Ecology
SALES E t AL.
Information) and (b) the selec tion of eDNA sampling sites was based
on the latrine surveys and abundance data provided by live-trapping
so could only occur af ter this was completed by August 6th. Water
and sediment samples were collected from patches where water
voles were determined to be absent (five sites; A1–A5); with 1–2 in-
dividuals present (three sites; A9, A16 and 18); 3–5 individuals (five
sites; A6, A8, A11, A14 and A17); and 7–11 individuals (five sites;
A7, A10, A12, A13 and A15; Figure 1a). Each of these streams/rivers
differed in their characteristics (in terms of width, depth and flow)
and a representation of the sites is depicted in Figure S4a–d. Three
water (two litres each) and three sediment (~25 ml) replicates were
taken at each patch (further details of sample collection are provided
in Appendix S1).
In addition to Assynt, eDNA sampling was also conducted on a
smaller scale in the Peak District National Park, England (Figure S5)
to incorporate additional mammals that are not known to be present
in Assynt ( Table S1). Here, the occurrence of water vole was identi-
fied by the presence of latrines in two sites (P1 and P2) at the time of
eDNA sampling (Figure S2a), whilst no latrines were identified at one
site (P3). At site P1, an otter Lutra lutra spraint was identified at the
time of eDNA sampling (Figure S2b). These three sites were sampled
in March 2018 using the same methodology as in Assynt but were
taken in close proximity (<50 cm) to water vole latrines where pres-
ent (Figure S2a).
2.4 | eDNA laboratory methods
DNA was extracted from the sediment samples using the DNeasy
PowerMax Soil kit and from the water samples using the DNeasy
PowerWater Kit (both QIAGEN Ltd.) following the manufacturer's
instructions in a dedicated eDNA laboratory in the University of
Salford. In order to avoid the risk of contamination during this step,
DNA extraction was conducted in increasing order of expected
abundance of water voles in the eDNA samples (all field blanks
were extracted first, followed by the sites with supposedly zero
water vole abundance, up to the highest densities last). Along with
field blanks (Assynt = 8, Peak District = 2), six lab extraction blanks
were included (one at the end of each daily block of extractions).
A decontamination stage using a Phileas 25 Airborne Disinfection
Unit (Devea SAS) was undertaken before processing samples from
different locations. Additional information regarding decontamina-
tion measures and negative controls can be found in the Supporting
Information.
A complete description of PCR conditions, library preparation
and bioinformatic analyses is provided in Appendix S1. Briefly, eDNA
was amplified using the MiMammal 12S primer set (MiMammal-U-F,
5′-GGGTTGGTAAATTTCGTGCCAGC-3′; MiMammal-U-R, 5′-CATA
GTGGGGTATCTAATCCCAGTTTG-3′; Ushio et al., 2017) targeting
a ~170 bp amplicon from a variable region of the 12S rRNA mito-
chondrial gene. A total of 147 samples, including field collection
blanks (10) and laboratory negative controls (12, including six DNA
extraction blanks and six PCR negative controls), were sequenced
in two multiplexed Illumina MiSeq runs. To minimize bias in individ-
ual reactions, PCRs were replicated three times for each sample and
subsequently pooled. Illumina libraries were built using a NextFlex
PCR-free library preparation kit according to the manufacturer's
protocols (Bioo Scientific) and pooled in equimolar concentrations
along with 1% PhiX (v3, Illumina). The libraries were run at a final
molarit y of 9 pM on an Illumina MiSeq platform using the 2 × 150 bp
v2 chemistry.
Bioinformatic analysis was conducted using OBItOOls metabar-
coding package (Boyer et al., 2016) and the taxonomic assignment
was conduc ted using ecotag against a custom reference database
(see Appendix S1). To exclude MOTUs/reads putatively belonging to
sequencing errors or contamination, the final dataset included only
MOTUs that could be identified to species level (>98%), and MOTUs
cont aining <10 reads and with a similarity to a sequence in the refer-
ence database lower than 98% were discarded (Cilleros et al., 2019).
The maximum number of reads detected in the controls for each
MOTU in each sequencing run was removed from all samples (Table
S7). For water voles, field voles and red deer (the most abundant wild
mammals in terms of sequence reads in our dataset), this equated
to a sequence frequency threshold of ≤0.17%, within the bounds
of previous studies on removing sequences to account for contam-
ination and tag jumping (Cilleros et al., 2019; Hänfling et al., 2016;
Schnell, Bohmann, & Gilbert, 2015).
2.5 | Occupancy/detection analysis in Assynt
The data collection from the different sur vey types described above
(water-based eDNA, sediment-based eDNA, latrine and camera
traps) produced the following site-specific detection/non-detection
data:
1. Latrine: two latrine surveys at 116 patches.
2. w-eDNA: three water-based eDNA samples at 18 of the 116
patches surveyed.
3. s-eDNA: three sediment-based eDNA samples at 18 of the 116
patches surveyed.
4. Camera: six 1-week occasions of camera trapping data at seven of
the 18 patches sur veyed by both Latrine and eDNA (w-eDNA +
s-eDNA) surveys.
We chose to focus on three species that were detected by at
least three of the four methods: water voles, field voles and red
deer Cervus elaphus. Water voles and field voles were recorded
using all four survey methods and had detection histories for 14
surveying events ((Latrine × 2) + (w-eDNA × 3) + (s-eDNA × 3) +
( C am e ra × 6) ). Re d d e er w e r e no t r e co rd ed d u r in g l a tr i n e su r ve ys a n d
had detection histories for 12 surveying events ((w-eDNA × 3) +
(s-eDNA × 3) + (Camera × 6)). To demonstrate the relative effi-
cacy of the four sur veying methods, we restricted the analyses
to the 18 sites where both latrine surveys were conducted and
eDNA samples were taken, seven of which had associated camera
|
5
Journal of Applied Ecolog
y
SALES E t AL.
trapping data. Although each surveying method differs in terms
of effort and effective area sur veyed, each is a viable surveying
method that is readily applied in practice. A unit of survey effort
here is defined as one latrine survey, one w-eDNA replicate, one
s-eDNA replicate or 1 week of c amera tra pping. So, while the spe-
cific units of effort are not directly comparable, the relative de-
tection efficacy per surveying method-specific unit of effort is of
interest and will provide important context for designing future
monitoring studies and understanding the relative merits of each
surveying method. Analysing the data using occupancy models al-
lowing for method-specific detectability enables such a compari-
son in per unit effort efficacy between eDNA metabarcoding and
multiple conventional survey methods.
A single season occupancy model (MacKenzie et al., 2002) was
applied to the ensemble data where detection histories were con-
structed using each of the surveying events as sampling occasions
(MacKenzie et al., 2017). The core assumption here is that the un-
derlying occupancy state (i.e. occupied or empty) is constant over
the sampling period, and therefore, every sampling occasion is a
potentially imper fect observation of the true occupancy status.
Because occasions represent method-specific surveying events,
we used ‘surveying method’ as an occasion-specific covariate on
detection (Latrine, w-eDNA, s-eDNA and Camera). Our primary
objective was to quantify and compare method-specific detect-
ability, so we did not consider any other competing models. For
comparing the methods, we compute accumulation curves as
(MacKenzie & Royle, 20 05):
where
p∗
smk
is the cumulative probability of detecting species s, when
species s is present, using method m after k sur veying events based
on the estimated surveying method-specific detection probability for
each species (
̂
psm
). We vary k from 1 to a large number and find the
value of k that results
p∗
smk
≥ .95. We con ducted the same analysis sepa-
rately for water voles, field vole s and red deer. An alysis was conducted
in r (R Core Team, 2018) usi ng the package unmarked (Fis ke & Chandler,
2011).
3 | RESULTS
3.1 | Mammal detection via eDNA metabarcoding
The two sequencing runs generated 23,276,596 raw sequence
reads and a total of 15,463,404 sequences remained following
trimming, merging and length filtering. After bioinformatic analy-
sis, the final ‘filtered’ dataset contained 23 mammals (Tables S2
and S3). For mammals, ~12 million reads were retained after ap-
plying all quality filtering steps (see Appendix S1). Reads from hu-
mans, cat tle Bos taurus, pig Sus scrofa, horse Equus ferus, sheep
Ovis aries and dog Canis lupus familiaris, were not considered fur-
ther as the focus of this study was on wild mammals (Table S4).
Felis was included because of the potential of it being wildcat Felis
silvestris or domestic cat F. catus/wildcat hybrids. A final dataset
comprising ~5.9 million reads was used for the downstream analy-
ses (Table S4).
In Assynt, the wild species identified were the red deer (18/18
sites); water vole (15/18); field vole (13/18); wood mouse Apodemus
sylvaticus—9/18; pygmy shrew Sorex minutus—4/18; wild/domes-
tic cat Felis spp.—4/18; mountain hare Lepus timidus—4/18; rabbit
Oryctolagus cuniculus—3/18; water shrew Neomys fodiens—3/18;
common shrew Sorex araneus—2/18; edible dormouse Glis glis—
2/18; grey squirrel Sciurus carolinensis—1/18; pine marten Martes
martes—1/18; brown rat Rattus norvegicus—1/18; red fox Vulpes
vulpes—1/18 and badger Meles meles—1/18 (Figure 1b). All of these
species are distributed around/within Assynt (Table S1), with the
exception of the edible dormouse and the grey squirrel. These are
unequivocally absent from the region. The edible dormouse is only
present in southern England and the grey squirrel is not distributed
that far north in Scotland (Mathews et al., 2018).
Of the wild mammals in the Peak District, the water vole, field
vole, wood mouse and otter were found in two sites (P1 and P2).
The red deer, pygmy shrew, common shrew, water shrew, red squir-
rel Sciurus vulgaris, grey squirrel, pine marten and badger were each
found at a single site (Figure S5). Only rabbit was found in site P3. All
species identified are currently distributed within the Park (Table S1),
except the red squirrel and pine marten. The pine marten, which is
critically endangered in England, has only two reliable records that
have been confirmed in the Park since 2000 and the red squirrel
has not been present for over 18 years (Alston, Mallon, & Whiteley,
2012).
Overall, water samples yielded better results than sediment
samples regarding species detection and read count for both areas
sampled (Figure 1b; Figure S5). In Assynt, only the wild/domestic cat
was exclusively detected in sediment samples (four sites), whereas
water samples recovered eDNA for ten additional species not found
in the sediment samples. The red deer, water vole, field vole, moun-
tain hare and pygmy shrew were also found in sediment samples
in Assynt (Figure 1b), and water vole and wood mouse in the Peak
District sediment samples (Figure S5).
3.2 | Occupancy analysis
Of the 18 sites where both latrine and eDNA surveys were con-
ducted, water voles were detected at 13 and field voles were de-
tected at 11. A total of seven wild mammals were recorded at the
seven sites with a camera trap from 10 July to 25 October 2017
(Figure S3; Table S5). There were several incidences where a shrew
could not be identified to species level using camera traps. For cam-
era traps, water voles were recorded at all sites, red deer at five out
of seven, field voles and weasels at three sites, water shrews and
otters at two and a red fox at a single site.
For the 18 sites in Assynt, estimated site occupancy (with 95%
confidence intervals) from the combined surveying methods was
p
∗
smk
=1−(1−
̂
p
sm
)
k,
6
|
Journal of Applied Ecology
SALES E t AL.
0.91 (0.63–0.98) for water voles and 0.88 (0.57–0.98) for field
voles. Red deer were observed at every patch by at least one of
the methods, and therefore occupancy was 1 (Table 1). For all three
species, per sample detection probability was higher for eDNA
taken from water than for eDNA taken from sediment (Table 1;
Figure 2). The surveying method-specific efficacy pattern was sim-
ilar for water voles and field voles (Table 1; Figure 2): latrine sur-
veys had the highest probability of detecting the species (.77 and
.52 respectively), followed by eDNA from water (.57 and .40 re-
spectively), then camera trapping (.50 and .20 respectively) and
finally eDNA from sediment (.27 and .02 respectively). Detection
probability was higher for water voles than field voles using all four
methods (Table 1; Figure 2). No effort was made to record red deer
pr ese nce duri ng la tri n e sur veys . Lik e the wate r vol es an d fiel d vol es,
red deer detection was higher using eDNA from water (0.67, CI:
0.53–0.78) compared to eDNA from sediment (0.10, CI: 0.04–0.21).
Un li ke th e vol es, wh ich were more dete c tab l e by ca mer a s tha n sed i-
ment eDNA, red deer detection on cameras was similar to sediment
eDNA (0.10, CI: 0.04–0.24).
The patterns described above detail surveying event-specific de-
tectability. We also computed the cumulative detection probabilit y
for each method and each species (
̂
psm
). The cumulative detection
curves over 15 surveying events are shown in Figure 2. The num-
ber of surveying event s, k, required to achieve
p∗
psm
≥ .95 for water
voles was three surveys, four samples, 10 samples and 5 weeks, for
latrines, water eDNA, sediment eDNA and cameras respectively.
The number of surveying events, k, required to achieve
p∗
psm
≥ .95 for
field voles was five sur veys, six samples, 141 samples and 14 weeks,
for latrines, water eDNA, sediment eDNA and cameras respectively.
The number of surveying events, k, required to achieve
p∗
psm
≥ .95 for
red deer was three samples, 30 samples and 29 weeks, for water
eDNA, sediment eDNA and cameras respectively (see also Figure 2).
TABLE 1 Estimated site occupancies and detection probabilities, with associated 95% confidence inter vals in brackets, obtained for
water-based eDNA (w-eDNA), sediment-based eDNA (s-eDNA) and conventional survey methods (Latrine and Camera) in Assynt, Scotland
Species Occupancy
Detection probability
Latrine w-eDNA s-eDNA Camera
Water vole 0.91 (0.63–0.98) 0.77 (0.59–0.89) 0.57 (0.43–0.71) 0.27 (0.16–0.41) 0.50 (0.35–0.65)
Field vole 0.89 (0.57–0.98) 0.52 (0.34–0.69) 0.40 (0.26–0.55) 0.02 (0.00–0.14) 0.20 (0.10–0.37)
Red deer 1.00 (1.00–1.00) —0.67 (0.53–0.78) 0.10 (0.04–0.21) 0.10 (0.09–0.24)
FIGURE 2 Figures on the left show
estimated detection probabilities of
each sur vey method for each of three
focal species; the vertical lines are 95%
confidence intervals. Figures on the right
show the method- and species-specific
cumulative detection probability with
increasing number of sampling events; the
horizontal dashed line shows a probability
of .95 for reference
Red deer
Field vole
Water vole
SedimentWater Latrine Camera
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
Sampling method
Detection probability
Red deer
Field vole
Water vole
4812
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
0.00
0.25
0.50
0.75
1.00
Events
Method
Sediment
Water
Latrine
Camera
|
7
Journal of Applied Ecolog
y
SALES E t AL.
4 | DISCUSSION
Despite the increasing potential of eDNA metabarcoding as a bio-
monitoring tool (Deiner et al., 2017), its application has largely been
focused on strictly aquatic or semi-aquatic animals, thus restrict-
ing management and conservation efforts of the wider ecosystem
(Williams, Huyvaert, Vercauteren, Davis, & Piaggio, 2018). Here, we
demonstrate the ability of eDNA metabarcoding to provide a valu-
able ‘terre strial divide nd’ for mammals from freshwater lotic ecosys-
tems, with a large proportion of the expected species from the wider
landscape being detected in each of the two study locations. In par-
ticular, we have demonstrated that water-based eDNA sampling
offers a promising and complementary tool to conventional survey
methods for the detection of whole mammalian communities.
4.1 | Detection of mammalian communities using
eDNA metabarcoding
Of the species known to be common in both Assynt and the Peak
District, eDNA metabarcoding readily detected the water vole,
field vole and red deer at the majority of sites surveyed (Figure 1b;
Figure S5). Pygmy, common and water shrews, wood mice and moun-
tain hares were also detected by eDNA metabarcoding at multiple
sites in Assynt (Figure 1b). A higher eDNA detection rate is expected
for aquatic and semi-aquatic mammals compared to terrestrial mam-
mals in aquatic environments due to the spatial and temporal stochas-
ticity of opportunities for terrestrial mammals to be in contact with
the water (Ushio et al., 2017). The semi-aquatic water vole was gen-
erally detected by eDNA metabarcoding where we expected to find
it and at relatively high read numbers (Figure 1b; Figures S1 and S5).
This is in line with previous studies in lentic systems (Harper et al.,
2019). However, the red deer was the only terrestrial species de-
tected by eDNA sampling at all sites in Assynt, and the terrestrial
field vole at over 70% of surveyed sites.
In addition to lifestyle (semi-aquatic or terrestrial), the number
of individuals of each species (i.e. group-living) may be important
for eDNA detection (Williams et al., 2018). As a counter example
to this, otters and weasels were notably absent in the eDNA sam-
ples in Assynt despite being captured by camera traps (Figure S3;
Table S5). Otters were present in the water eDNA samples at two
sites in the Peak District, albeit at a lower number of reads in com-
parison to most of the other species detected (Figure S5; Table S2).
This mirrors previous studies where eDNA analysis has performed
relatively poorly for otter detection in captivity and the wild (Harper
et al., 2019; Thomsen et al., 2012). Carnivores were generally de-
tected on fewer occasions (e.g. red foxes, badgers and pine mar tens;
Figure 1b; Figure S5) or not at all (e.g. stoats and American mink in
addition to those discussed above) in comparison to smaller mam-
mals and red deer, and a similar pattern has been shown with North
American carnivores in a recent study using eDNA from soil samples
(Leempoel et al., 2020). For some of these species, species ecology/
behaviour such as a relatively large home range and more solitary
nature (e.g. red foxes) may go some way towards explaining a lack of,
or few, eDNA records. Furthermore, as demonstrated by Ushio et al.
(2017) poor efficiency for amplifying some mammal species might be
associated to suboptimal experimental conditions (e.g. inadequate
primer design, primer bias, DNA concentration, species masking
and/or annealing temperatures).
Regarding the sampling medium for eDNA, we demonstrated
that water is a more effective method for detection of mammal
eDNA than sediment (Table 1; Figure 1b; Figure S5). For one of our
foc al species, the water vole, 75% of site s which were deemed un oc-
cupied by latrine surveys and those with ≤2 individuals (eight sites) in
Assynt, returned a non-detection for sediment eDNA as opposed to
37.5% of sites for wate r (F igu re 1a,b ; Fi gur e S1) . Disti nct temp ora l in -
ferences are provided by eDNA recovered from water and sediment
samples. DNA bound to sediments can remain detectable for a lon-
ger period (i.e. up to hundreds of years) and provide historical data,
whereas, eDNA retrieved from water samples provide more contem-
porary data due to a faster degradation in the water column (Turner,
Uy, & Everhart, 2015). It is worth investigating further if sediment
eDNA could indicate the presence of a more ‘established’ popula-
tion, where a cert ain threshold of individuals and long-term occupa-
tion (i.e. historical) is required for detection in sediment (Figure S1;
Leempoel et al., 2020; Turner et al., 2015).
Importantly, sparse or single eDNA records should be carefully
verified. The edible dormouse and grey squirrel sequences identi-
fied within the Assynt samples (Figure 1b) and red squirrel within the
Peak District (Figure S5) highlight the caveats associated with this
technique. If management decisions had relied on eDNA evidence
alone, false positives for these species could lead to unnecessary
resources being allocated for management/eradication programmes
as the edible dormouse and grey squirrel are classified as invasive
species within Great Britain. These potentially arose due to sample
carryover from a previous sequencing run on the same instrument
(a known issue with Illumina sequencing platforms; Nelson, Morrison,
Benjamino, Grim, & Graf, 2014) which included those species for the
reference database construction. Controlling for false positives is
certainly a huge challenge in eDNA metabarcoding and the need
to standardize and optimize thresholds for doing so is an ongoing
debate (Ficetola et al., 2015; Harper et al., 2019).
Even with these concerns around false positives highlighted, two
records are potentially notewor thy in a conservation contex t for UK
mammals because of the relatively high read number associated with
these records (Tables S2 and S3). The first of these is the Felis records
in sediment samples in multiple sites in Assynt (Figure 1b). Even with
a ‘pure’ F. silvestris as a reference sequence, it was not possible to
distinguish between the wild and domesticated species for this 12S
fragment (data not shown). Despite ongoing conservation efforts,
there may now be no ‘pure’ Scottish wildcats left in the wild in the UK
but isol at ed popu la tions (pe rhaps of hybr id origin) may exis t in this re-
gion (Sainsbury et al., 2019). Given that these eDNA detections were
all from sediment samples, it is possible that they may be historical
rather than contemporary (see above). The other significant eDNA
record was the pine mar ten in the Peak District. The pine mar ten
8
|
Journal of Applied Ecology
SALES E t AL.
Mar tes martes is know n to occur in the Scot tis h Hi ghlan ds bu t ha d dis-
appeared from most of the UK and recently has been recovering from
historical persecution, including a potential expansion of its range.
Still, authentic records from northern England are scarce or lacking
altogether (Alston et al., 2012; Sainsbury et al., 2019). However, a
record of a recent roadkill exists from just outside the Park's bound-
ary (BBC News, 2018). The high number of reads recovered for the
Peak District sample (4,293 reads vs. 25 in the Assynt sample) adds
credence to this positive eDNA detection but further investigations
are warranted into the potential presence of this species in the area.
4.2 | Comparisons between surveying methods
Comparisons of species detection by traditional survey approaches and
eDNA analysis are now numerous in the literature, and mainly focus on
what is and what is not detected within and across different methods
(Hänfling et al., 2016; Leempoel et al., 2020). Yet, there has been grow-
ing incorporation of occupancy modelling to estimate the probability of
detecting the focal species, in comparison to one other survey method,
either for a single species (Lugg et al., 2018; Tingley et al., 2019) or mul-
tiple species (Abrams et al., 2019; Valentini et al., 2016). Simultaneous
multi-method comparisons for multiple species have been lacking and
this study directly addresses this for the first time.
The probability of detecting the water vole and field vole was
higher for the latrine surveys than eDNA sampling (both water and
sediment) and camera traps (Table 1; Figure 2). However, when consid-
ering confidence intervals, there was considerable overlap between
latrine, water-based eDNA metabarcoding and camera traps for both
species, with only sediment-based eDNA metabarcoding yielding
a low probability of detection (Table 1). Detection probabilities for
water-based eDNA metabarcoding and camera traps were similar for
water voles, with camera tr aps less likely to detec t the field vole than
water-based eDNA . For the red deer (for which no latrine survey was
undertaken), water-based eDNA metabarcoding had a much higher
probability of detection than either sediment-based eDNA metabar-
coding or camera traps (which performed similarly; Table 1). Despite
the increasing adoption of camera traps in providing non-invasive de-
tections for mammals (Hofmeester et al., 2019), camera traps were
outperformed by water-based eDNA metabarcoding for the three
focal species in this component of the study. Here, camera traps were
deployed so as to sample the habitat of the water vole (see Figure
S3), which may explain lower detection for other terrestrial species
in comparison to eDNA metabarcoding (see above). Studies focusing
on a single species often repor t that eDNA analysis outperforms the
conventional survey method in terms of detec tion probabilities (e.g.
Lugg et al., 2018). For metabarcoding, there is clearly a need to care-
fully consider the potential for cross contamination between samples
an d ho w fal se posi tives (an d neg at ive s) cou ld im pac t det ect ion pro ba-
bilities using occupancy modelling with eDNA data (Brost, Mosher, &
Davenport, 2018; Lahoz-Monfort, Guillera-Arroita, & Tingley, 2016).
Amon g th e re co mm endations mad e by Lahoz-Monfort et al. (2 016) to
account for these uncertainties, one was the simultaneous collection
of data from more conventional sur veying methods. Here, we have
demonstrated general congruence between surveying methods for
the water vole (Table S5; Figure S1) and using certain sp ecies to app ly
a multiple detection methods model would be appropriate in further
studies (Lahoz-Monfort et al., 2016). Alternatively, using repeated
samplin g and known neg at ive controls in oc cupanc y models that ful ly
incorporate false-positive errors could be applied in the absence of
other surveying data (Brost et al., 2018). Overall, multi-species me-
tabarcoding studies may trade-off a slightly lower (but comparable)
detection probability than other survey methods for individual spe-
cies (Figure 2) in favour of a better overall ‘snapshot’ of occupancy of
the whole mammalian community (Ushio et al., 2017).
The comparison between survey ‘effort’ for the four methods
to reach a probability of detection of ≥.95 is highly informative
and provides a blueprint for future studies on mammal monitoring.
Focusing on the water vole for example, three latrine surveys would
be required. A total of four water-based and 10 sediment-based
eDNA replicates or 5 weeks of camera trapping would be required
to achieve the same result (Figure 2). This increases for the field vole
in the same habitat, with five latrine surveys and six water-based
eDNA replicates. Sediment-based eDNA metabarcoding would
be impractical for this species and camera trapping would take
14 weeks. What is important here is the spatial component and the
amount of effort involved in the field. Taking 4–6 water-based eDNA
replicates from around one location within a patch could provide
the same probability of detecting these small mammals with three
latrine surveys. In many river catchments, there may be 100 s to
1,000s of kilometres to survey that would represent suitable habi-
tat, and only a fraction of that may be occupied by any given species.
This is par ticularly relevant in the context of recovery of water vole
populations post-translocation or in situations where remnant pop-
ulations are bouncing back after invasive American mink Neovison
vison control has been instigated. On a local scale, finding signs of
water voles through latrine surveys is not necessarily dif ficult, but
monitoring the amount of potential habitat (especially lowland) for a
species which has undergone such a massive decline nationally is a
huge undertaking (Morgan, Cornulier, & Lambin, 2019).
The use of eDNA metabarcoding from freshwater systems to gen-
erate an initial, coarse and rapid ‘distribution map’ for vertebrate bio-
diversity (and at a relatively low cost) could transform biomonitoring at
the landscape level. For group-living (i.e. deer) and small mammal spe-
cies, carefully chosen sampling points (with at least five water-based
replicates) along multiple river courses could provide a reliable indica-
tion of what species are present in the catchment area if conducted
during times of peak abundance (i.e. Summer and Autumn). Then, on
the basis of this, practitioners could choose to further investigate spe-
cific areas for confirmation of solitary, rare or invasive species (e.g. car-
nivores) with increased effort in terms of both the number of sampling
sites and replicates taken. At present, we would recommend the use of
eDNA metabarcoding alongside other non-invasive surveying meth-
ods (e.g. camera traps) when monitoring invasive species or species of
conservation concern to maximize monitoring efforts (Abrams et al.,
2019; Sales, Kaizer, et al., 2020).
|
9
Journal of Applied Ecolog
y
SALES E t AL.
It is clear that eDNA metabarcoding is a promising tool for moni-
toring semi-aquatic and terrestrial mammals in both lotic (this study)
and lentic systems (Harper et al., 2019; Ushio et al., 2017). We de-
tected a large proportion of the expected mammalian community
(Table S1). Water-based eDNA metabarcoding is comparable or
outperforms other non-invasive survey methods for several species
(Figure 2). However, there remain challenges for the application of this
technique over larger spatial and temporal scales. Technical issues of
metabarcoding in laboratory and bioinformatic contexts have been
dealt with elsewhere (Harper et al., 2019) but understanding the dis-
tribution of eDNA transport in the landscape and its entry into natural
lotic systems is at an early stage (and incorporating such variables in
occupancy modelling approaches). This clearly requires more detailed
and systematic eDNA sampling than undertaken here, particularly in
an interconnected river/stream network with organisms moving be-
tween aquatic and terrestrial environments. Leempoel et al. (2020)
recently demonstrated the feasibility for detecting terrestrial mammal
eDNA in soil samples but this study has shown that sampling a few
key areas in freshwater ecosystems (e.g. larger rivers and lakes) within
a catchment area could potentially provide data on a large propor-
tion (if not all) of the mammalian species within it, even when some
species are present at low densities (Deiner et al., 2017). In this re-
gard, future studies might also investigate the value of citizen science,
where trained volunteers can contribute to data collection at key sites,
thus scaling up the reach of research whilst raising public awareness
and the significance of mammalian conservation concerns (Parsons,
Goforth, Costello, & Kays, 2018).
ACKNOWLEDGEMENTS
The eDNA component of this project was funded by the British
Ecological Society (grant no. SR17/1214) and a University of Salford
Internal Research Award awarded to A.D.M. J.D. was supported by
the University of Massachusetts Organismal and Evolutionary Biology
Research Grant and Spring 2018 Graduate School Fieldwork Grant.
We thank Kristy Deiner for enlightening conversations about these
results. We are grateful to Jerry Herman and Andrew Kitchener for the
tissue samples from National Museums Scotland. Christine Gregory,
Douglas Ross and Sarah Proctor provided water vole and otter infor-
mation for sampling in the Peak District and Sara Peixoto provided
sequence assemblies. We thank the various landowners for permis-
sion to sample on their property. We thank Brittany Mosher and the
anonymous reviewers for significantly improving the manuscript. The
authors declare that no conflict of interest exists.
AUTHORS' CONTRIBUTIONS
A.D.M., X.L., C .S., O.S.W., I.C., S.M., N.G.S., S.S.B., E.O., B.H. and
L.L.H. conceived the study; Monitoring and live-trapping of water
voles was pa rt of X.L ., C. S., E. B. and J.D.'s ongoing work in As synt;
J.D. analysed the camera trap data; D.A.D. advised on primer set/
data validation and provided information and data on mammals in
the Peak District; A.D.M., N.G.S., S.S.B. and M.B.M. carried out
the eDNA sampling; M.B.M., N.G.S., S.S.B., C.B. and A.D.M. per-
formed the laboratory work; N.G.S., O.S.W., L.R.H., M.B.M., C .B.
and A. D.M. c arried out the bioinfo rmatic analyses; N.G .S., A. D.M.,
I.C. and M.B.M. analysed the eDNA data; C.S. and J.D. conducted
the occupancy modelling; A.D.M., N.G.S., C.S., J.D., M.B.M. and
L. R .H. wr ot e the pape r, wit h all auth ors con tri butin g to ed iting and
discussions.
DATA AVA ILAB ILITY STATE MEN T
Data are available via the Dryad Digit al Repository ht tps://doi.
org/10.5061/dryad.d51c5 9zzf (Sales, McKenzie, et al., 2020).
ORCID
Naiara Guimarães Sales https://orcid.org/0000-0002-2922-3561
Joseph Drake https://orcid.org/0000-0003-0458-3533
Lynsey R. Harper https://orcid.org/0000-0003-0923-1801
Stefano Mariani https://orcid.org/0000-0002-5329-0553
Christopher Sutherland https://orcid.org/0000-0003-2073-1751
Allan D. McDevitt https://orcid.org/0000-0002-2677-7833
REFERENCES
Abrams, J. F., Hoerig, L., Brozovic, R., Axtner, J., Crampton-Platt, A.,
Mohamed, A., … Wilting, A. (2019). Shifting up a gear with iDNA: From
mammal detection events to standardized surveys. Journal of Applied
Ecology, 56, 1637–1648. https://doi.org/10.1111/1365-2664.13411
Alston, D., Mallon, D. P., & Whiteley, D. (2012). The mammals of der-
byshire. England: Derbyshire Mammal Group and Sorby Natural
His tory So ci et y.
BBC News. (2018). Mystery of pine marten found dead ‘100 miles from
home’. Retrieved from https://www.bbc.co.uk/news/uk-engla nd-derby
shire-45331872
Boyer, F., Mercier, C., Bonin, A., Le Bras, Y., Taberlet , P., & Coissac, E.
(2016). obitools: A unix inspired software package for DNA me-
tabarcoding. Molecular Ecology Resources, 16, 176–182. https://doi.
org /10.1111/1755 -0 998.12428
Brost, B. M., Mosher, B. A., & Davenport, K. A . (2018). A model-based
solution of observational errors in the laboratory. Molecular Ecology
Resources, 18, 5 80 –589. htt ps ://doi.or g/10.1111/1755- 0998.12765
Cilleros, K., Valentini, A., Allard, L., Dejean, T., Etienne, R., Grenouillet, G.,
… Brosse, S. (2019). Unlocking biodiversity and conservation studies
in high diversity environments using environmental DNA (eDNA): A
test with Guiane se freshwater fishes. Molecular Ecology Resources,
19, 27–46 . ht tp s://doi.o rg /10.1111/1755-0 998 .12 90 0
Deiner, K., Bik, H. M., Mächler, E., Seymour, M., Lacoursière-Roussel, A.,
Altermatt, F., … Bernatchez, L. (2017). Environmental DNA metabar-
coding: Transforming how we survey animal and plant communities.
Molecular Ecology, 26, 5872–5895. https://doi.org/10.1111/mec.14350
Ficetola, G. F., Pansu, J., Bonin, A., Coissac, E., Giguet-Covex, C., De
Barba, M., … Taberlet , P. (2015). Replication levels, false presences
and the estimation of the presence/absence from eDNA metabar-
coding data. Molecular Ecology Resources, 15, 543–556. https://doi.
org /10.1111/1755 -0 998.12 33 8
Fiske, I., & Chandler, R. (2011). unmarked: An R package for fit ting hi-
erarchical models of wildlife occurrence and abundance. Journal of
Statistical Software, 43(10), 1–23.
Hänfling, B., Handley, L. L., Read, D. S., Hahn, C., Li, J., Nichols, P., …
Winfield, I. J. (2016). Environmental DNA metabarcoding of lake fish
communities reflects long-term data from established survey meth-
ods. Molecular Ecology, 25(13), 3101–3119. ht tps://doi .org/10.1111/
mec.13660
10
|
Journal of Applied Ecology
SALES E t AL.
Harper, L. R., Handley, L. L., Carpenter, A. I., Ghazali, M., Di Muri, C.,
Macgregor, C. J., … Hänfling, B. (2019). Environmental DNA (eDNA)
metabarcoding of pond water as a tool to survey conservation and
management priority mammals. Biological Conservation, 238, 108225.
https://doi.org/10.1016/j.biocon.2019.108225
Hofmeester, T. R., Cromsigt, J. P. G. M., Odden, J., Andrén, H., Kindberg,
J., & Linnell, J. D. C. (2019). Framing pictures: A conceptual framework
to identify and correct for biases in detec tion probability of camera
traps enabling multi-species comparison. Ecology and Evolution, 9,
2320–2336. https://doi.org/10.1002/ece3.4 878
Holman, L. E., de Bruyn, M., Creer, S., Car valho, G., Robidart, J., & Rius,
M. (2019). The detection of introduced and resident marine species
using environmental DNA metabarcoding of seawater and sediment.
Scientific Reports, 9, 115 59.
Lahoz-Monfort, J. L., Guillera-Arroita, G., & Tingley, R. (2016). Statistical
approaches to account for false-positive errors in environmental
DNA samples. Molecular Ecology Resources, 16 , 673–685. https://doi.
org /10.1111/1755 -0 998.12486
Leempoel, K ., Herber t, T., & Hadly, E. A . (2020). A comparison of eDNA
to camer a trapping for assessment of terrestrial mammalian diver-
sit y. Proceedings of the Royal Society of London. Series B: Biological
Sciences, 287, 20192353. https://doi.org/10.1098/rspb.2019.2353
Lugg, W. H., Griffiths, J., van Rooyen, A. R., Weeks, A. R., & Tingley,
R. (2018). Optimal survey designs for environmental DNA sam-
pling. Methods in Ecology and Evolution, 9, 1049–1059. htt ps://doi.
org /10.1111/2041-210X .12951
MacKenzie, D. I., Nichols, J. D., Lachman, G. B., Droege, S., Royle, J. A., &
Langtimm, C. A. (2002). Estimating site occupancy rates when detec-
tion probabilities are less than one. Ecology, 83, 2248–2255. https://
doi.org/10.1890/0012-9658(2002)083[2248:esorw d]2.0.co;2
MacKenzie, D. I., Nichols, J. D., Royle, J. A., Pollock, K. H., Bailey, L., &
Hines, J. E. (2017). Occupancy estimation and modeling: Inferring pat-
terns and dynamics of species occurrence. London, UK: Elsevier.
MacKenzie, D. I., & Royle, J. A . (2005). Designing occupanc y studies: General
advice an d allocating survey effort . Journal of Applied Ecology, 42(6),
1105–1114. htt ps://doi. or g/10 .1111/ j.13 65-2664.20 05.010 98 .x
Mathews, F., Kubasiewicz, L. M., Gurnell, J., Harrower, C. A., McDonald,
R. A., & Shore, R. F. (2018). A Review of the Pop ulation and Cons ervation
Status of British Mammals: Technical Summary. Peterborough, UK:
Natural England.
Morgan, W., Cornulier, T., & Lambin, W. (2019). Colonisation dynamics
during range expansion is poorly predicted by dispersal in the core
range. Ecography, 42, 1142–1151. https://doi.org/10.1111/ecog.04180
Nelson, M. C., Morrison, H. G., Benjamino, J., Grim, S. L., & Graf, J. (2014).
Analysis, optimization and verification of Illumina-generated 16s
rRNA gene amplicon surveys. PLoS ONE, 9, e0094249. htt ps://doi.
org/10.1371/journ al.pone.0094249
Nichols, J. D., Bailey, L. L., O'Connell, A. F. Jr, Talancy, N. W., Grant, E. H. C.,
Gilbert, A. T., … Hines, J. E. (2008). Multi-scale occupancy estimation and
modelling using multiple detection methods. Journal of Applied Ecology,
45(5), 1321–1329. https://doi.org/10.1111/j.1365-2664.2008.01509.x
Niedballa, J., Courtiol, A., & Sollmann, R. (2018). camtrapR: Camera trap
data management and preparation of occupancy and spatial capture-
recapture analyses. R Package version 1.0.
Parsons, A. W., Gofor th, C., Costello, R., & Kays, R. (2018). The value
of citizen science for e cological monitoring of mammals. PeerJ, 6,
e4536. https://doi.org/10.7717/peerj.4536
R Core Team. (2018). R: A language and environment for s tatistical com-
puting. V ienna, Austria: R Foundation for Statistical Computing.
Retrieved from https://www.R-proje ct.org/
Sainsbury, K. A., Shore, R. F., Schofield, H., Croose, E., Campbell, R. D.,
& McDonald, R. A. (2019). Recent histor y, current status, conser-
vation and management of native mammalian carnivore species in
Great Britain. Mammal Review, 49, 171–188 . http s://doi.o rg /10.1111/
mam.12150
Sales, N. G., Kaizer, M. D. C., Coscia, I., Perkins, J. C., Highlands, A.,
Boubli, J. P., … McDevitt, A. D. (2020). Assessing the potential of
environmental DNA metabarcoding for monitoring Neotropical
mammals: A case study in the Amazon and Atlantic Forest, Brazil.
Mammal Review. https: //doi.org/10 .1111/mam.1218 3
Sales, N. G., McKenzie, M. B., Drake, J., Harper, L. R., Browett, S. S.,
Coscia, I., … McDevit t, A. D. (2020). Data from: Fishing for mammals:
Landscape-level monitoring of terrestrial and semi-aquatic commu-
nities using eDNA from lotic ecosystems. Dryad Digital Repository,
ht tps ://doi .o rg/10.5 0 61/d r yad .d51c 5 9 zzf
Schnell, I. B., Bohmann, K., & Gi lbert, M. T. P. (2015). Tag jumps illumi nat-
ed-reducing sequence-to-sample misidentifications in metabarcod-
ing studies. Molecular Ecology Resources, 15(6), 1289–1303. ht tps://
doi .org/10.1111/1755- 09 98.124 02
Sutherland, C., Elston, D. A., & Lambin, X. (2013). Accounting for false
positive detection error induced by transient individuals. Wildlife
Research, 40(6), 490–498. https://doi.org/10.1071/WR12166
Sutherland, C. S., Elston, D. A., & Lambin, X. (2014). A demographic, spa-
tially explicit patch occupancy model of metapopulation dynamics and
persistence. Ecology, 95, 3149–3160. https://doi.org/10.1890/14-
0384.1
Thomsen, P. F., Kielgast, J., Iversen, L. L., Wiuf, C., Rasmussen, M., Gilbert,
M. T. P., … Willerslev, E. (2012). Monitoring endangered freshwa-
ter biodiversity using environmental DNA . Molecular Ecology, 21,
2565–2573. https://doi.org/10.1111/j.1365-294X.2011.05418.x
Tingley, R., Greenlees, M., Oer tel, S., van Rooyen, A. R., & Weeks, A. R.
(2019). Environmental DNA sampling as a sur veillance tool for cane
toad Rhinella marina intro ductions on offshore islands . Biological
Invasions, 21, 1–6. ht tps ://doi.org/10 .1007/s10530- 018-1810-4
Turner, C. R., Uy, K. L., & Everhart, R. C. (2015). Fish environmental
DNA is more concentrated in aquatic sediments than surface water.
Biological Conservation, 183, 93–102. https://doi.org/10.1016/j.biocon.
2014.11.017
Ushio, M., Fukuda, H., Inoue, T., Makoto, K., Kishida, O., Sato, K., … Miya,
M. (2 017). Environ me ntal DNA en ab les detec ti on of ter re st rial mam-
mals from f orest pond water. Molecula r Ecology Resources, 17(6), e6 3–
e75. htt ps://doi. org/10 .1111/1755 -099 8.1 2690
Valentini, A ., Taberlet , P., Miaud, C., Civade, R., Herder, J., Thomsen, P.
F., … Dejean, T. (2016). Next-generation monitoring of aquatic biodi-
versity using environmental DNA metabarcoding. Molecular Ecology,
25(4), 929–942. https ://doi.or g/10.1111/m ec .13 428
Visconti, P., Pressey, R. L ., Giorgini, D., Maiorano, L., Bakkenes, M.,
Boitani, L., … Rondinini, C . (2011). Future hot spots of terrestrial
mammal loss. Philosoph ical Transactions of the Royal Societ y of London.
Series B: Biological Sciences, 366 (1578), 2693–2702. https://doi.
org /10.1098/rstb. 2011. 0105
Williams, K. E., Huy vaer t, K. P., Vercauteren, K. C., Davis, A. J., & Piaggio,
A. J. (2018). Detec tion and persistence of environmental DNA from
an invasive, terrestrial mammal. Ecology and Evolution, 8(1), 688–695.
htt ps://doi.org/10.10 02 /ece3 .3698
SUPPORTING INFORMATION
Additional supporting information may be found online in the
Supporting Information section.
How to cite this article: Sales NG , McKenzie MB, Drake J,
et al. Fishing for mammals: Landscape-level monitoring of
terrestrial and semi-aquatic communities using eDNA
from riverine systems. J Appl Ecol. 2020;00:1–10.
https://doi .org /10.1111/1365-26 64.13592
