DNA metabarcoding and morphological macroinvertebrate metrics reveal the same changes in boreal watersheds across an environmental gradient

Abstract

Cost-effective, ecologically relevant, sensitive, and standardized indicators are requisites of biomonitoring. DNA metabarcoding of macroinvertebrate communities is a potentially transformative biomonitoring technique that can reduce cost and time constraints while providing information-rich, high resolution taxonomic data for the assessment of watershed condition. Here, we assess the utility of DNA metabarcoding to provide aquatic indicator data for evaluation of forested watershed condition across Canadian eastern boreal watersheds, subject to natural variation and low-intensity harvest management. We do this by comparing the similarity of DNA metabarcoding and morphologically derived macroinvertebrate metrics (i.e. richness, % Ephemeroptera, Plecoptera and Trichoptera, % chironomid), and the ability of DNA metabarcoding and morphological metrics to detect key gradients in stream condition linked to forested watershed features. Our results show consistency between methods, where common DNA metabarcoding and morphological macroinvertebrate metrics are positively correlated and indicate the same key gradients in stream condition (i.e. dissolved oxygen, and dissolved organic carbon, total nitrogen and conductivity) linked to watershed size and shifts in forest composition across watersheds. Our study demonstrates the potential usefulness of macroinvertebrate DNA metabarcoding to future application in broad-scale biomonitoring of watershed condition across environmental gradients.

Full text

1
SCIentIfIC REPoRTS | 7: 12777 | DOI:10.1038/s41598-017-13157-x
www.nature.com/scientificreports
DNA metabarcoding and
morphological macroinvertebrate
metrics reveal the same changes
in boreal watersheds across an
environmental gradient
Caroline E. Emilson1, Dean G. Thompson1, Lisa A. Venier1, Teresita M. Porter1,2, Tom
Swystun1, Derek Chartrand1, Scott Capell1 & Mehrdad Hajibabaei2
Cost-eective, ecologically relevant, sensitive, and standardized indicators are requisites of
biomonitoring. DNA metabarcoding of macroinvertebrate communities is a potentially transformative
biomonitoring technique that can reduce cost and time constraints while providing information-rich,
high resolution taxonomic data for the assessment of watershed condition. Here, we assess the utility
of DNA metabarcoding to provide aquatic indicator data for evaluation of forested watershed condition
across Canadian eastern boreal watersheds, subject to natural variation and low-intensity harvest
management. We do this by comparing the similarity of DNA metabarcoding and morphologically
derived macroinvertebrate metrics (i.e. richness, % Ephemeroptera, Plecoptera and Trichoptera, %
chironomid), and the ability of DNA metabarcoding and morphological metrics to detect key gradients
in stream condition linked to forested watershed features. Our results show consistency between
methods, where common DNA metabarcoding and morphological macroinvertebrate metrics are
positively correlated and indicate the same key gradients in stream condition (i.e. dissolved oxygen, and
dissolved organic carbon, total nitrogen and conductivity) linked to watershed size and shifts in forest
composition across watersheds. Our study demonstrates the potential usefulness of macroinvertebrate
DNA metabarcoding to future application in broad-scale biomonitoring of watershed condition across
environmental gradients.
Ecological indicators are deeply embedded in sustainable use, forest management policy, and third party certica-
tion systems internationally1, highlighting the need for practical, inexpensive, ecologically relevant and sensitive
indicators in biomonitoring programs. Small freshwater streams are a ubiquitous feature of most forest regions
on the Boreal Shield and represent a direct link between the terrestrial and aquatic components of forested water-
sheds. Physical-chemical parameters of streams have been shown to reect disturbance history of the surround-
ing forested terrestrial ecosystem, for example through changes in sediment loads2, nutrients inputs3, leaf litter
inputs4, or quantity and quality of dissolved organic matter5. Aquatic biota living in these streams including mac-
roinvertebrates, sh, and algae have long been used as integrative indicators of watershed disturbance through
changes in stream physical-chemical condition6.
Macroinvertebrates are among the most common focal groups used in biomonitoring of lotic systems and
a wide array of assessment metrics that are useful in dierent biomonitoring scenarios have been developed
based on these communities. For example, metrics based on diversity, percent of the community comprised
of the taxa sensitive to disturbance (Ephemeroptera, Plecoptera, and Trichoptera; % EPT), and percent of the
community comprised of resilient taxa from the family Chironomidae7, are oen successful at detecting shis
1Natural Resources Canada, Canadian Forest Service, Great Lakes Forestry Centre, 1219 Queen St. East, Sault Ste.
Marie, P6A 2E5, Canada. 2Centre for Biodiversity Genomics @ Biodiversity Institute of Ontario & Department
of Integrative Biology, University of Guelph, 50 Stone Road East, Guelph, N1G 2W1, Canada. Correspondence
and requests for materials should be addressed to C.E.E. (email: caroline.emilson@gmail.com) or M.H. (email:
mhajibab@uoguelph.ca)
Received: 26 April 2017
Accepted: 19 September 2017
Published: xx xx xxxx
OPEN
Correction: Author Correction
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
2
SCIentIfIC REPoRTS | 7: 12777 | DOI:10.1038/s41598-017-13157-x
in stream condition associated with watershed disturbance. However, Baird and Hajibabaei (2012)8 identied
several constraints which have severely limited the utility of morphologically based macroinvertebrate metrics
in broad scale biomonitoring programs over the last y years. ese constraints include the extensive time
required for eld sample processing, the taxonomic expertise required to correctly identify each organism below
the family level of taxonomic resolution, the potential inconsistencies amongst various observers, and the general
lack of verication of morphology-based identications8. Metabarcoding presents a potential solution to these
morphological-based constraints, and these authors present DNA-based taxonomic identication of benthos
from mixed environmental samples based on the cytochrome c oxidase subunit 1 (CO1) barcode marker using
high-throughput DNA sequencing, herein referred to as DNA metabarcoding, as a potentially transformative
approach to biomonitoring, biodiversity discovery, and ecosystem health assessments9.
e ability of DNA metabarcoding to identify known aquatic macroinvertebrate communities identied by
morphological methods has been demonstrated in the literature9–13, and studies have highlighted the ability of
DNA metabarcoding to discriminate aquatic macroinvertebrate community alpha, beta, and gamma diversity
and ecological assessment metrics14,15. What is lacking, however, are direct validations comparing the ability of
DNA metabarcoding and morphological methods to indicate gradients in specic environmental characteristics,
especially gradual gradients in watershed and associated stream habitat characteristics resulting from natural
and forest management induced changes. Environmental biomonitoring requires not only the identication of
disturbed versus undisturbed conditions, but also the discovery of process and which environmental variables are
inuencing biota. In many cases disturbance is more gradual and associations between biota and the environment
can help identify critical thresholds in environmental change along with multiple-variable interactions in more
complex modelling scenarios. e main objective of our study is to explore the utility of DNA metabarcoding
to provide aquatic indicator data for evaluation of stream health and forest integrity across eastern boreal shield
watersheds subject to natural variation and the inuence of low-intensity harvest management. To address our
objective we: (1) Compare the similarity of DNA metabarcoding and morphological derived macroinvertebrate
metrics across streams (i.e. richness, % EPT, % chironomid), (2) Identify the key gradients in stream condition
that link our study streams to their forested watersheds, and (3) Compare the ability of DNA metabarcoding and
morphological macroinvertebrate metrics to indicate these key gradients in stream condition.
Results
Similarity of DNA metabarcoding and morphological metrics. We found that common macroinver-
tebrate metrics (i.e. for richness, % EPT, and % chironomid) were positively correlated across our study streams
for all methods (i.e. DNA metabarcoding and morphological), and taxonomic resolutions (i.e. genus and OTU)
(Fig.1). DNA metabarcoding provided higher richness values than morphological methods in most sites at the
genera level, and higher richness for all sites at the OTU level as would be expected. ese greater richness val-
ues, indicative of a greater number of unique taxa identied, inuenced community composition metrics in two
dierent ways, depending on the case example. In some cases, % EPT and % chironomid values decreased as
the result of an increased number of non-EPT and non-chironomid taxa being detected. In other cases, DNA
metabarcoding resulted in detection of comparatively more EPT and chironomid taxa, and in some cases where
Figure 1. Pearson correlations between common morphological and DNA metabarcoding macroinvertebrate
metrics for both genera and OTU taxonomic resolutions. e solid line represents the line of best t, and a 1:1
line of t is indicated (dotted line) to facilitate visual comparison between methods.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
3
SCIentIfIC REPoRTS | 7: 12777 | DOI:10.1038/s41598-017-13157-x
morphological methods detected none (see Supplementary FigsS1–S3 for site specic 1:1 taxonomic method
comparisons, and FigsS4–S6 for accumulation curve method comparisons). At the genus rank, the mean DNA
metabarcoding richness was 15 and 30% greater-, mean % EPT was 47.4 and 1% less-, and mean % chirono-
mid was 24.1 and 3.2% greater-than, mean morphologically derived genera indicator metrics of 11.5, 48.4%,
and 20.9%, respectively. Dierences between DNA metabarcoding and morphologically derived metrics were
even greater when DNA metabarcoding OTU richness was considered, with a mean richness of 52.1 being 353%
greater than mean morphological genera level richness, and with mean % EPT of 31.1 and mean % chironomid
of 16.7 being 17.3% and 4.2% less than means for these metrics calculated based on morphological genera level
taxonomic assignments.
Key gradients in stream condition linked to forested watershed features. e rst key gradi-
ent in stream condition linked to watershed features (RDA1; Fig.2) represented a shi from increased stream
conductivity, temperature and depth (loadings = −0.958, −0.635, and −0.505, respectively) to increased stream
dissolved organic carbon, total nitrogen (TN), and total phosphorus (TP) (loadings = 1.08, 1.07, and 0.737,
respectively) along a watershed gradient from higher-elevation forest with more jack pine (biplot scores = −0.773
jack pine, and −0.639 elevation) to lower elevation forest dominated by black spruce (biplot score = 0.502)
(Fig.2). e second key gradient in stream condition linked to watershed features (RDA2; Fig.2) was represented
by increased dissolved oxygen, ow, and temperature (loadings = 0.761, 0.792, 0.668 respectively) with increasing
total watershed area (biplot score = 0.819) (Fig.2). Collectively, these two watershed forest RDA axes explained
46.3% of the variation in stream physical-chemical characteristics across sites (Fig.2).
Utility of macroinvertebrate metrics to indicate key gradients in stream condition. Based
on hierarchical partitioning we found that dissolved oxygen and dissolved organic carbon were the stream
physical-chemical variables explaining the most variation inrichness, % EPT and % chironomid macroinver-
tebrate community composition metrics (Fig.3). Both physical-chemical variables were also main components
of the gradients in stream condition linked most strongly to forested watershed characteristics as noted above
(RDA 2 and RDA1 respectively: Fig.2). Dissolved oxygen had an independent contribution of 48.9%, 62.9%, and
43.7% for % EPT based on morphological taxonomic assignments at the level of genera, DNA metabarcoding
assignments at the level of genera, and DNA metabarcoding at the OTU resolution, respectively. Similarly, var-
iations in dissolved oxygen across streams independently contributed 72.9%, 67.9%, and 70.7% of the variance
explained for % chironomid as determined by morphological genera assignments, DNA metabarcoding genera
assignments, and DNA metabarcoding OTU assignments, respectively (Fig.3). All the above listed independent
contributions were found to be statistically signicant (randomization test p < 0.05). Dissolved organic carbon
was the only variable that had a signicant independent contribution (38.4%) for morphological genera rich-
ness. DNA metabarcoding measures of richness (i.e. genera & OTU resolutions) did not have statistically signif-
icant independent contributions from any one stream physical-chemical variable, but appeared to be inuenced
by multivariate combinations of 4 to 5 of the 8 variables with independent contributions ranging between 10
Figure 2. Redundancy analysis showing stream physical-chemical characteristics constrained by forested
watershed features across the 23 study sites. Both axes were found to be statistically signicant following a
permutation ANOVA test (p = 0.001, for both RDA1 and RDA2). *DOC stands for dissolved organic carbon.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
4
SCIentIfIC REPoRTS | 7: 12777 | DOI:10.1038/s41598-017-13157-x
and 30% (Fig.3). Additionally, for all measures of richness and for % EPT composition as determined by DNA
metabarcoding OTUs, joint contributions exceeded independent contributions (i.e. independent contribution/
joint contribution <1) for up to 3 of the 8 explanatory variables. In contrast, joint contributions did not exceed
independent contributions in any case (i.e. independent contribution/joint contribution >1 for all explanatory
variables) where global models of % chironomid and % EPT were evaluated at the genera level of taxonomic
resolution.
Using RDA to constrain macroinvertebrate metrics by stream physical-chemical characteristics, we found that
common macroinvertebrate metrics for all methods and taxonomic resolutions, were associated with the same
gradients in stream condition across sites (Fig.4). e rst two statistically signicant stream physical-chemical
RDA axes explained a total of 47.8% of the variation in all the macroinvertebrate metrics and the main compo-
nents of these gradients (i.e. dissolved organic matter, TN, conductivity and dissolved oxygen in Fig.4) were
the same as those identied as being most strongly linked to the key gradients found across the forest water-
sheds under study, namely jack pine and black spruce composition, watershed elevation, and watershed total area
(Fig.2).
e rst gradient in stream condition (RDA1; Fig.4), which was positively loaded with dissolved oxygen
(biplot score = 0.877), explained variation in all measures of % EPT based on morphological identications to
genera, DNA metabarcoding based identications to genera and DNA metabarcoding based identications to
OTU (loadings = 0.577, 0.858, and 0.765 respectively). Similarly, all measures of % chironomid composition in
the stream macroinvertebrate communities were strongly related to dissolved oxygen including % chironomid
at the level of genera based on both morphological and DNA metabarcoding data, and % chironomid based on
DNA metabarcoding OTU data (loadings = −0.8690, −1.07, and −1.04, respectively). Additionally, measures of
richness at the level of genera, whether determined morphologically or by DNA metabarcoding were also related
to dissolved oxygen (loadings = 0.582, 0.434, respectively). e second gradient in stream condition (RDA2;
Fig.4), was positively loaded with dissolved organic carbon (biplot score = 0.719) and TN (biplot score = 0.491),
and negatively loaded with conductivity (biplot score = −0.685), and explained variation in all measures of rich-
ness including morphological genera richness, DNA metabarcoding genera richness, and DNA metabarcoding
OTU richness (loadings = −0.511, −0.595, and −0.499 respectively) (Fig.4).
Discussion
We found that common macroinvertebrate metrics showed the same relative patterns across sites regardless of
method or resolution, and irrespective of the fact that DNA metabarcoding tended to identify a greater number
Figure 3. Percent independent contribution results for each macroinvertebrate metric based on hierarchical
partitioning. DO stands for dissolved oxygen, DOC for dissolved organic carbon, and TP for total phosphorus.
*Denotes a statistically signicant (p < 0.05) independent contribution based on a negative-log-likelihood
randomization test (n = 1000).
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
5
SCIentIfIC REPoRTS | 7: 12777 | DOI:10.1038/s41598-017-13157-x
of unique taxa than morphological methods. Furthermore, macroinvertebrate metrics from all methods and
taxonomic resolutions, were associated with the strongest gradients in stream condition linked to watershed fea-
tures across our study sites. ese results highlight the utility of macroinvertebrate DNA metabarcoding metrics
as sensitive bioindicators of environmental gradients across forested watersheds, and the potential for the CO1
BE marker to provide macroinvertebrate community data able to capture environmental change across dierent
regions. Since the primers used to target the CO1 BE region were specically developed to target macroinverte-
brate benthos, this method could be applied to other environments where this is the target assemblage12. In fact,
this primer set has successfully been applied in large-scale biomonitoring analysis of various sites (e.g. Gibson
et al.14 used it for taxa from wetlands in Alberta). e biggest challenge for application of metabarcoding, in
general, is the assumption that target taxa are already present in the CO1 reference sequence database to provide
a basis for taxonomic assignment16. As reference databases continue to grow, we can expect the proportion of
high condence taxonomic assignments to improve. Additionally, supplementing existing reference datasets with
CO1 barcode sequences obtained from locally-collected morphologically identied specimens can substantially
improve taxonomic assignment success17.
Across all methods and taxonomic resolutions measures of overall richness were strongly associated with dis-
solved organic carbon, TN and conductivity, which represented the strongest gradient in stream condition linked
to forested watershed features across our study sites. ese changes in dissolved organic carbon, TN and conduc-
tivity across streams were associated with shis in forest composition reective of watershed hydrology and soil
characteristics, where more lowland black spruce dominated stands were linked to higher dissolved organic car-
bon and nutrient concentrations, and more jack pine dominated stands linked to lower dissolved organic carbon
and nutrient concentrations and greater conductivity across streams. is is consistent with soil biogeochemical
dierences between these forest soil types18, as swampy areas typical of black spruce stands are high in organic
materials and thus capable of exporting greater amounts of dissolved organic carbon and nutrients to receiving
streams. Likewise, well-drained mineral soils typical of jack pine stands export more inorganic ions and less dis-
solved organic carbon and nutrients due to low organic matter content and the increased sorption of dissolved
organic carbon to mineral soil surfaces19. All methods and taxonomic resolutions suggest that swampy conditions
typical of black spruce stands result in higher stream water dissolved organic carbon and nutrients that reduce
macroinvertebrate richness, which agrees with previous research showing decreased macroinvertebrate diversity
in high dissolved organic carbon and nutrient streams ooding peatlands20.
Despite detecting the same macroinvertebrate responses to environmental gradients, DNA metabarcoding
generally found more taxa than morphological methods most likely related to the greater ability of DNA metabar-
coding to identify broken, early instar, pupal and chironomidae specimens11. Given the ecological properties of
our study system we were able to show the eectiveness of DNA metabarcoding in detecting the eects of mul-
tiple variables on macroinvertebrate richness. All measures of DNA metabarcoding and morphological richness
were associated with the same multivariate axis in RDA, and had explanatory variables with joint contributions
Figure 4. Redundancy analysis showing all macroinvertebrate metrics constrained by stream physical-
chemical characteristics across the 23 study sites. Both axes were found to be statistically signicant following
a permutation ANOVA test (p = 0.001 and p = 0.033, for RDA1 and RDA2 respectively). Meta stands for DNA
metabarcoding, Morph for morphological, DO for dissolved oxygen, and DOC for dissolved organic carbon.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
6
SCIentIfIC REPoRTS | 7: 12777 | DOI:10.1038/s41598-017-13157-x
that exceeded independent contributions in hierarchical partitioning, suggesting the inuence of multiple varia-
bles on overall richness measures. However, morphological genera richness identied the independent contribu-
tion of dissolved organic carbon as signicant in hierarchical partitioning, where DNA metabarcoding richness
measures did not nd the independent contribution of DOC, or of any other individual explanatory variable to be
signicant. is may be attributed to the higher resolution in taxonomic information provided by genetic-based
metabarcoding that is more sensitive to the detection of unique taxa and thus multiple variable gradients.
Comparatively, morphological methods identify a lower number of unique taxa, and therefore may be more
biased towards detecting only environmental gradients that inuence macroinvertebrates collectively at lower
taxonomic resolutions. Previous studies have documented the stronger discriminatory power of DNA metabar-
coding macroinvertebrate communities based on the high resolution taxonomic information provided14,21. In
our dataset, the highest resolution DNA metabarcoding OTU data detected the same environmental gradients as
DNA metabarcoding genera, and did not consistently show stronger associations with these gradients. However,
the increased statistical power of OTU resolution DNA metabarcoding data to detect gradients may be study
specic as previous research has shown considerable variation from one region to the next in the taxonomic
resolution of macroinvertebrate communities best suited to detect environmental gradients22. Additionally, the
resolution at genus level could be inuenced by the completeness of reference sequence libraries for a study site.
erefore, the high resolution OTU data that is derived from DNA metabarcoding is still an added benet that
may prove useful in teasing apart cumulative eects or multiple stressors in some cases.
All methods and taxonomic resolutions were also able to detect the same patterns in community composition
in response to natural gradients in stream condition, with indicators of sensitivity and tolerance (% EPT and %
chironomid) following dissolved oxygen gradients across streams. Dissolved oxygen was the second strongest
gradient in stream condition linked to watershed features, and was associated with variation in watershed size
and ow rate across sites. ese ndings are well supported in the literature, as dissolved oxygen concentrations
increase with ow rates because of increased absorption of oxygen from the atmosphere from increased water
movement23. Additionally, EPT taxa are known to be a sensitive indicator of environmental change in streams
and to require high dissolved oxygen concentrations (>5 mg/L)24, while taxa from the family chironomidae are
known to be more tolerant and thrive where EPT are in low abundance25. e ability of % EPT and % chironomid
indicator metrics to clearly detect a key gradient in stream condition is reective of the fact that these target taxa
are known to respond to the same environmental changes (albeit in opposite directions), thereby ltering out
noise that may result from including other taxa who might respond in a dierent or more gradual way. ese
ndings highlight the ability of DNA metabarcoding-based methods to detect the same patterns in metrics of
community tolerance and sensitivity as with morphologically derived metrics, and re-iterates the ability of DNA
metabarcoding to detect the same primary responses to even subtle environmental gradients, despite nding
greater richness.
Our data allowed us to demonstrate the potential use of DNA metabarcoding on a more sensitive scale
because the key gradients in stream and watershed condition across our study sites lacked the discrete inuence
of a high-impact disturbance. Across the boreal shield and forested watersheds in general, watersheds vary in
size, soil composition, topography, hydrology, and forest composition and structure. It is essential to character-
ize region-specic variation to dene a baseline from which to base comparisons across sites and regions when
assessing the eects of sustainable management practices. DNA metabarcoding of macroinvertebrate commu-
nities provides a powerful tool capable of detecting underlying gradients in forested watershed condition over
space and time. Given the demonstrated ability of DNA metabarcoding to detect variation across the watersheds
in this study that were subject to natural variation and low-intensity harvest management, the potential for DNA
metabarcoding to detect gradients across more intensively disturbed and managed watersheds is promising.
In this study, we focused on presence-absence data to make a direct comparison between methods. However,
future improvements in sequencing methods such as the optimization of mixed PCR template reactions to reduce
PCR bias26, or capture-based sequencing that remove the need for PCR entirely27 may add further value to DNA
metabarcoding data by allowing for abundance information to be used with greater condence. Previous studies
comparing the use of morphologically derived relative abundance with presence-absence macroinvertebrate data
have found that in some cases relative abundance has comparatively stronger associations with environmental
gradients22, while in other cases presence-absence data has been found to perform equally as well as relative abun-
dance, for example in discriminating pollution level thresholds28. Other important considerations for the use of
DNA metabarcoding data on large-scales, include the standardization of sample processing, DNA extraction and
bioinformatic methods29–31. e potential development and use of DNA metabarcoding based multimetrics that
integrate many dierent sensitivity and diversity metrics into one, as seen with stream and river macroinverte-
brates across a basin in China32, may also prove useful.
Conclusions
In conclusion, our results demonstrate that macroinvertebrate DNA metabarcoding data can capture the same
changes in community composition, and detect the same key gradients in stream and watershed condition across
sites as morphological data, ultimately leading to the same conclusions about the ecological integrity of forested
watersheds. Our study also highlights the advantages of DNA metabarcoding data to not only provide common
indicator response variables targeting sensitive and tolerant indicator group metrics, such as % EPT and % chi-
ronomid, but to also simultaneously provide high resolution, information rich data that may prove useful in
detecting cumulative eects of environmental variables, and potentially multiple stressor scenarios. Previous lit-
erature has also documented the cost and time eective, and veriable and reproducible nature of DNA metabar-
coding data8, further highlighting the potential future application of macroinvertebrate DNA metabarcoding data
in broad-scale biomonitoring of the ecological integrity of stream and associated forest ecosystems in Canada and
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
7
SCIentIfIC REPoRTS | 7: 12777 | DOI:10.1038/s41598-017-13157-x
internationally. Ongoing improvements and the standardization of sequencing and bioinformatic methods will
further add to the value of DNA metabarcoding in biomonitoring programs.
Methods
Study Sites. e 23 study sites are located within the 12 thousand km2 Hearst Sustainable Forest License,
located in northeastern Ontario (700060 E, 5490897 N 16U) (Fig.5). e study sites contain varying proportions
of black spruce (Picea mariana Mill. B.S.P.), cedar (uja occidentalis L.), tamarack (Larix laricina (Du Roi)
K. Koch), white spruce (Picea glauca (Moench) A. Voss), jack pine (Pinus banksiana Lamb.), balsam r (Abies
balsamea (L.) Mill), and trembling aspen (Populus tremuloides Michx.) as described by Penner et al. 201318, and
have varying histories of low-intensity, patchy harvest and natural regeneration typical of extensive management
strategies.
Watershed characteristics. Watershed characteristics included forest composition and structural data,
along with landscape features as outlined in Table1 and were derived using enhanced forest inventory infor-
mation18. Briey, compositional data were derived based on interpretation of high resolution (i.e. 40 cm) digital
stereo imagery, and structural characteristics and landscape features were derived from airborne light detecting
and ranging (LiDAR). LiDAR capture (0.82 point/cm2; <30 cm vertical accuracy) was completed in 2007 and
Figure 5. e Hearst Sustainable Forest License, showing the delineated boundaries of the 23 watersheds
involved in this study, stream layers and the location of stream sampling points. e map was created in
ArcMap version 10.4.1 using a base map from Land Information Ontario (https://www.ontario.ca/page/
land-information-ontario), a shapele provided by Hearst Forest Management Inc. to delineate the Hearst
Sustainable Forest License area, and the GPS coordinates of the study streams.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
8
SCIentIfIC REPoRTS | 7: 12777 | DOI:10.1038/s41598-017-13157-x
descriptor variables generated for 400 m2 cells allowing spatially explicit descriptions of forest structure at ne
scale. Streams and drainage areas of the watersheds were delineated based on the 5 m Digital Terrain-Model
derived from LiDAR data18.
Stream physical-chemical characteristics. We collected detailed stream physical-chemical charac-
teristics for the sampled reach of each of the study streams (Table1). Data collected included water chemis-
try as determined by single mid-water column grab sample (dissolved organic carbon, TN, and TP) analyzed
following standard protocol at the Canadian Forest Service water chemistry laboratory, Sault Sainte Marie33.
Water temperature, conductivity, and dissolved oxygen were collected using a YSI model 85 hand held device
(YSI Incorporated, Yellow Springs, Ohio, USA), pH collected using a HI 98127 handheld pH meter (Hanna
Instruments, Woonsocket, Rhode Island, U.S.A.), and ow measured using an Single Axis EM Flow Meter (Model
801, Valeport Ltd., Totnes, Devon, UK), at a midstream location. Additionally, for each stream, depth was meas-
ured four times along each of four transects where the sample collection took place. All stream physical-chemical
characteristics were collected once, at the time of macroinvertebrate sampling in 2013.
Macroinvertebrate communities. Sample collection. Macroinvertebrate samples were collected during
the last week of August 2013 following a modication of the standard Canadian Aquatic Biomonitoring Network
protocol for wadeable streams34. Briey, the samples were collected using a kick net (400 µm) over a 20 m length
of stream located at least 25 m upstream of a water crossing or culvert for a standardized time of 3 minutes. e
following modications were made to prevent DNA cross-contamination among the samples: All gear including
kick nets and collection containers were pre-washed with soapy water, rinsed with distilled water and ELIMINase
solution, followed by ethanol, and nally exposed to UV light for at least 24 hours. Prior to sample handling,
the eld crew wore clean nitrile gloves, and sample sorting gear was rinsed with ELIMINase, followed by clear
stream water and ethanol. Aer collection, the contents of the net were transferred to a bin where large inorganic
and organic materials (i.e. large pieces of gravel, cobble and sticks) were discarded aer rst being checked for
attached macroinvertebrates. e attached macroinvertebrates as well as all remaining contents of the bin were
then transferred to a 250 mL bottle, covered completely with 95% ethanol, and stored in a cool, dark place at
approximately 20 °C until the eld sample was processed.
Field sample processing and morphological identication. In the laboratory, eld samples were poured into
decontaminated enamel trays. Under a lighted magnifying glass and using tweezers macroinvertebrates were
Environmental characteristic Mean CV (%) Min Max VIF
Watershed area (km2) 47.1 162 0.842 278 —
Watershed elevation (m) 285 15.0 196 347 —
Watershed mean P90 8.01 29.6 3.74 13.4 —
Watershed mean VCI 0.598 14.2 0.450 0.768 —
Watershed mean CC2 66.0 20.4 45.5 91.9 —
Watershed mean CC10 21.9 54.7 1.11 49.2 —
Black spruce (%) 43.1 58.8 0.00 82.9 —
Jack pine (%) 3.25 148 0.00 14.6 —
Mixedwood conifer (%) 8.06 101 0.00 31.8 —
Mixedwood hardwood (%) 12.6 111 0.00 48.3 —
Upland spruce r (%) 7.55 105 0.00 32.7 —
Upland spruce pine (%) 8.50 99.4 0.00 36.9 —
Lowland conifer (%) 6.14 97.4 0.00 20.1 —
Intolerant hardwood (%) 3.75 133 0.00 14.7 —
Channel slope 0.0151 100 0.00322 0.0755 —
Stream depth (cm) 41.0 75.3 4.88 150 2.02
Stream ow (m s−1) 0.083 15.9 0.055 0.098 1.84
Stream temperature (°C) 19.8 13.5 15.5 23.6 1.69
Stream dissolved oxygen (mg L−1) 5.06 33.3 1.83 7.40 2.38
Stream pH 7.85 10.5 4.29 8.47 1.52
Stream conductivity (µmho cm−1) 211 31.1 111 343 2.62
Stream dissolved organic matter (mg L−1) 20.9 45.2 1.61 36.9 2.98
Stream total phosphorus (mg L−1) 0.009 47.1 0.002 0.019 1.68
Stream total nitrogen (mg L−1) 0.612 101 0.160 0.980 —
Table 1. Summary of environmental characteristics across the 23 study sites including mean, minimum,
maximum, and coecient of variation (CV). Variance ination factors (VIF) are included for stream physical-
chemical characteristics included in hierarchical partitioning. P90 is the 90th percentile of stand height, VCI
stands for the vertical complexity index and represents a more even stand height as you move towards 1, and
CC2 and CC10 represent percent crown closure at 2 and 10 m respectively.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
9
SCIentIfIC REPoRTS | 7: 12777 | DOI:10.1038/s41598-017-13157-x
picked from the stream matrix material and placed into clean vials. Subsequently, where possible, all macroin-
vertebrates from the class Insecta were identied down to genus under the microscope by Natural Resources
Canada taxonomists. In total 28.8% of the specimens identied as Insecta could not be assigned to a genera based
on morphology. Specimens that could not be identied down to genus included early instar or pupal specimens,
damaged or broken specimens, or specimens from the family chironomidae, as the identication of chironomidae
requires a comprehensive analysis by a chironomidae specialist, which is beyond the scope of routine biomon-
itoring analysis. Specimens from the family chironomidae were instead classied as being from the subfamily
Tanypodinae, the tribe Tanytarsini, or unclassied. Unclassied chironomidae comprised 42.9% of the specimens
unclassied at the genus level. To avoid DNA contamination nitrile gloves were changed, and enamel trays and
all sample handling instruments were cleaned with distilled water and ELIMINase solution, rinsed with ethanol,
and dried under UV light, prior to use on a dierent sample. e picked, morphologically identied samples were
subsequently submitted for DNA metabarcoding analysis.
DNA Metabarcoding analysis. Each macroinvertebrate sample was homogenized in a sterile blender one sample
at a time. e homogenized sample was transferred to a 50 mL falcon tube and incubated at 56 °C to evaporate
excess ethanol. e homogenate was subsampled into a lysing matrix tube, and further homogenized using a MP
FastPrep®-24 Instrument (MP Biomedicals Inc.; Santa Ana, California, USA) at speed 6.0 for 40 seconds. ree
times the reagents suggested in the Nucleospin Tissue extraction protocol were used during the chemical lysing
step, which were subdivided into three 1 mL microcentrifuge tubes. e total DNA from the homogenate was
extracted from these samples using a Nucleospin Tissue Kit (Macherey-Nagel Inc.; Duren, Germany), eluting
with 30 µL of molecular grade water then pooling the three reactions within each sample for a total of 90 µL
of DNA extract per macroinvertebrate sample. e CO1 BE marker from each sample was amplied through
a two-step polymerase chain reaction (PCR) regime using the B forward and E reverse35 with a standard mix
of 17.5 μL molecular grade water, 2.5 μL 10x reaction buer (200 mM Tris HCl, 500 mM KCl, pH 8.4), 1.0 μL
MgCl2 (50 mM), 0.5 μL dNTPs (10 mM), 0.5 μL forward primer (10 mM), 0.5 μL reverse primer (10 mM), 0.5 μL
Platinum Taq DNA polymerase (5 U/µL) (Life Technologies; Burlington, Ontario, Canada), and 2.0 μL DNA tem-
plate, for a total of 25 μL per reaction. One negative control (ie. reactions with 2 μL water instead of DNA) was
carried through from the rst PCR to ensure PCR products were free of contamination. Reactions underwent 35
cycles of 94 °C for 40 s, 46 °C for 60 s, 72 °C for 30 s using an Eppendorf Mastercycler ep gradient S thermal cycler.
PCR amplication success was visually conrmed through gel electrophoresis using a 1.5% agarose gel. Products
of the rst round of PCR were puried following the MinElute PCR Purication kit (Qiagen; Toronto, Ontario,
Canada) standard protocol, eluting with 30 μL molecular biology grade water. Puried PCR products underwent
a second round of PCR to attach the Illumina-tailed primer required for sequencing, this used the same reaction
volumes and PCR conditions as the rst round of PCR, with the exceptions of using Illumina-tailed primers, and
30 cycles. Illumina-tailed PCR products were then puried following the same protocol as the rst round PCR
products. Puried Illumina-tailed amplicons were dual indexed and sequenced on an Illumina MiSeq ow cell
using the v2 sequencing chemistry (250 × 2) (Illumina; San Diego, California, USA).
In total 9 124 153 sequences were acquired and processed using a semi-automated bioinformatics pipeline
described in detail in Supplementary Methods. Briey, raw paired-end sequences (n = 8 377 424) were assembled
using SeqPrep (https://github.com/jstjohn/SeqPrep), primers removed using cutadapt 1.1036, and sequences clus-
tered at 98% sequence similarity into Operational Taxonomic Units (OTUs) using USEARCH37. Quality ltering
of sequence data included minimum and maximum sequence lengths (i.e. cut os of >300 bp, <400 bp), and
removal of singletons, doubletons, sequences with 3 or more ambiguities, and chimeric sequences identied using
the UPARSE-OTU pipeline. Aer quality ltering, a total of 1406 OTUs (n = 5 296 344 sequences) remained with
an average length of 313 base pairs (See Supplementary TableS1 for summary of sequence statistics, and Fig.S7
for rarefaction curves). e Ribosomal Database Project (RDP) classier v2.1238, with a custom CO1 Arthropoda
training set (Porter and Hajibabaei, in prep), was chosen because of its speed and generation of condence scores
to help dene good taxonomic assignments and reduce false positive identications16. At the taxonomic reso-
lution of genera a bootstrap support cut o of <50% was implemented based on previously determined leave
one out testing thresholds16 (See Supplementary FileS1 for all taxonomic assignment results). Additionally, all
non-aquatic-insect OTUs were removed along with aquatic insect OTUs with bootstrap support cut os of <20%
on assignments to the class Insecta.
Statistical approach to assess the utility of DNA metabarcoding. To compare the similarity of
macroinvertebrate metrics derived from morphological and DNA metabarcoding datasets, the same metrics were
independently calculated for each dataset at the genus rank for each site (n = 23), and then subsequently com-
pared. e indicator metrics calculated included richness, and the two common community composition metrics
% EPT, and % chironomid calculated based on the proportion of unique taxa classied as EPT or chironomidae
divided by the total number of unique taxa for each sample. DNA Metabarcoding data at the nest level of res-
olution (OTU) were also used to calculate macroinvertebrate metrics to determine if higher resolution genetic
data provided evidence to support similar or dierent conclusions. To be conservative in our comparison of
methods and avoid any potential concerns with sampling inconsistencies, DNA extractions, and PCR biases39–41,
presence-absence data was used in both DNA metabarcoding and morphological calculations of macroinver-
tebrate metrics. e similarity of common DNA metabarcoding and morphological macroinvertebrate metrics
were then directly compared using Pearson correlation plotted with a line of best t, and a 1:1 line of t to visually
examine the magnitude and direction of potential dierences.
To identify the key gradients in stream condition that link the multivariate characteristics of our study streams
to the characteristics of the forested watersheds which they drain, redundancy analysis (RDA) in the ‘vegan’ R
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
10
SCIentIfIC REPoRTS | 7: 12777 | DOI:10.1038/s41598-017-13157-x
package42 was used. Permutation ANOVA was run to test the signicance of the RDA axes, and selected stream
and watershed variables are listed in Table1 along with summary statistics.
To compare the ability of DNA metabarcoding and morphological macroinvertebrate metrics to indicate
these key gradients in stream condition across sites, associations between macroinvertebrate metrics and stream
physical-chemical characteristics were examined using hierarchical partitioning in the ‘hier.part’ R package43,
complimented with RDA. Hierarchical partitioning allows for the exploration of the relative inuence of each
explanatory variable on a response variable by calculating the independent and joint contributions of each indi-
vidual explanatory variable in a global model, in all possible model combinations and then taking the mean44.
RDA compliments hierarchical partitioning because it allows for associations between all DNA metabarcod-
ing and morphological macroinvertebrate metrics and all stream physical-chemical metrics to be visualized and
explored simultaneously45. e function ‘rand.hp’ was used to test the signicance of independent contributions
using a randomization test with negative log-likelihood (n = 1000), and Permutation ANOVA was run to test the
signicance of the RDA axes.
Prior to hierarchical partitioning, the variance ination factors (VIF) of all explanatory variables were assessed
using the ‘vif’ function in the ‘car’ R package46, to avoid issues of multicollinearity47. All key explanatory stream
physical-chemical variables selected for use in hierarchical partitioning had VIF values of <3 (Table1). TN was
excluded from the hierarchical partitioning analyses because of its high VIF value (>5) indicative of its positive
association with dissolved organic carbon concentrations across streams. Prior to all analyses, variables were
transformed where necessary using log, square root, or logit transformations for percentages to meet normality
assumptions. All statistical analyses were run in R version 3.3.248.
Data availability. Raw sequence data was deposited at the National Centre for Biotechnology Information
(NCBI) Short Read Archive Project number SRP117571. Raw environmental data are also included
(Supplementary FileS2), along with fasta les of OTUs for each site (Supplementary FileS3).
References
1. Steenberg, J. W. N., Duiner, P. N., Van Damme, L. & Ziele, . Criteria and indicators of sustainable forest management in a
changing climate: an evaluation of Canada’s national framewor. J. Sustain. Dev. 6, 32–64 (2013).
2. reutzweiser, D. P., Good, . P., Capell, S. S. & Holmes, S. B. Leaf-litter decomposition and macroinvertebrate communities in boreal
forest streams lined to upland logging disturbance. J. North Am. Benthol. Soc. 27, 1–15 (2008).
3. reutzweiser, D. P., Hazlett, P. W. & Gunn, J. M. Logging impacts on the biogeochemistry of boreal forest soils and nutrient export
to aquatic systems: A review. Environ. ev. 16, 157–179 (2008).
4. Meyer, J. L., Wallace, J. B. & Eggert, S. L. Leaf litter as a source of dissolved organic carbon in streams. Ecosystems 1, 240–249 (1998).
5. Yamashita, Y., loeppel, B. D., noepp, J., Zausen, G. L. & Jaffé, . Effects of watershed history on dissolved organic matter
characteristics in headwater streams. Ecosystems 14, 1110–1122 (2011).
6. Hilsenho, W. L. An improved biotic index of organic stream pollution. Gt. Laes Entomol. 20, 31–40 (1987).
7. Li, Z., Zhang, Y., Yuan, J. P. & LU, L. eview of the role of benthic macro-invertebrates as indicators of river health assessment. South-
to-north Water Divers. Water Sci. Technol. 4, 22 (2011).
8. Baird, D. J. & Hajibabaei, M. Biomonitoring 2.0: a new paradigm in ecosystem assessment made possible by next-generation DNA
sequencing. Mol. Ecol. 21, 2039–44 (2012).
9. Hajibabaei, M., Shoralla, S., Zhou, X., Singer, G. aC. & Baird, D. J. Environmental barcoding: A next-generation sequencing
approach for biomonitoring applications using river benthos. PLoS One 6 (2011).
10. Hajibabaei, M., Spall, J. L., Shoralla, S. & van onynenburg, S. Assessing biodiversity of a freshwater benthic macroinvertebrate
community through non-destructive environmental barcoding of DNA from preservative ethanol. BMC Ecol. 12, 28 (2012).
11. Sweeney, B. W., Battle, J. M., Jacson, J. . & Dapey, T. Can DNA barcodes of stream macroinvertebrates improve descriptions of
community structure and water quality? J. North Am. Benthol. Soc. 30, 195–216 (2011).
12. Stein, E. D. et al. Does DNA barcoding improve performance of traditional stream bioassessment metrics? Freshw. Sci. 33, 302–311
(2014).
13. Carew, M. E., Pettigrove, V. J., Metzeling, L. & Homann, A. A. Environmental monitoring using next generation sequencing: rapid
identication of macroinvertebrate bioindicator species. Front. Zool. 10, 45 (2013).
14. Gibson, J. F. et al. Large-scale biomonitoring of remote and threatened ecosystems via high-throughput sequencing. PLoS One 10,
1–15 (2015).
15. Elbrecht, V., Vamos, E. E., Meissner, ., Aroviita, J. & Leese, F. Assessing strengths and weanesses of DNA metabarcoding-based
macroinvertebrate identication for routine stream monitoring. Methods Ecol. Evol. Early view (2017).
16. Porter, T. M. et al. apid and accurate taxonomic classication of insect (class Insecta) cytochrome c oxidase subunit 1 (COI) DNA
barcode sequences using a naive Bayesian classier. Mol. Ecol. esour. 14, 929–942 (2014).
17. Yang, J. et al. Zooplanton Community Proling in a Eutrophic Freshwater Ecosystem-Lae Tai Basin by DNA Metabarcoding. Sci.
ep. 7, 1773 (2017).
18. Penner, M., Pitt, D. G. & Woods, M. E. Parametric vs. nonparametric LiDA models for operational forest inventory in boreal
Ontario. Can. J. emote Sens. 39, 426–443 (2013).
19. Fellman, J. B., D’Amore, D. V., Hood, E. & Boone, . D. Fluorescence characteristics and biodegradability of dissolved organic matter
in forest and wetland soils from coastal temperate watersheds in southeast Alasa. Biogeochemistry 88, 169–184 (2008).
20. Astorga, A., Heino, J., Luoto, M. & Muota, T. Freshwater biodiversity at regional extent: Determinants of macroinvertebrate
taxonomic richness in headwater streams. Ecography (Cop.). 34, 705–713 (2011).
21. Jacson, J. . et al. Cryptic biodiversity in streams: a comparison of macroinvertebrate communities based on morphological and
DNA barcode identications. Freshw. Sci. 33, 312–324 (2014).
22. Heino, J. Taxonomic surrogacy, numerical resolution and responses of stream macroinvertebrate communities to ecological
gradients: Are the inferences transferable among regions? Ecol. Indic. 36, 186–194 (2014).
23. Cox, B. A. A review of dissolved oxygen modelling techniques for lowland rivers. Sci. Total Environ. 314–316, 303–334 (2003).
24. Wallace, J. B., Grubaugh, J. W. & Whiles, M. . Biotic indices and stream ecosystem processes: esults from an experimental study.
Ecol. Appl. 6, 140–151 (1996).
25. Arimoro, F. O. & Iomi, . B. Ecological integrity of upper Warri iver, Niger Delta using aquatic insects as bioindicators. Ecol. Indic.
9, 455–461 (2009).
26. Gohl, D. M. et al. Systematic improvement of amplicon marer gene methods for increased accuracy in microbiome studies. Nat.
Biotechnol. 34, 1–11 (2016).
Content courtesy of Springer Nature, terms of use apply. Rights reserved

www.nature.com/scientificreports/
11
SCIentIfIC REPoRTS | 7: 12777 | DOI:10.1038/s41598-017-13157-x
27. Shoralla, S. et al. Environmental DNA barcode sequence capture: Targeted, PC-free sequence capture for biodiversity analysis
from bul environmental samples. bioxiv 1–28 (2016).
28. Aylagas, E., Borja, Á. & odríguez-Ezpeleta, N. Environmental status assessment using DNA metabarcoding: Towards a genetics
based marine biotic index (gAMBI). PLoS One 9 (2014).
29. Bi, H. M. et al. Sequencing our way towards understanding global euaryotic biodiversity. Trends Ecol. Evol. 27, 233–243 (2012).
30. Nilsson, H. . et al. Towards standardization of the description and publication of next-generation sequencing datasets of fungal
communities. New Phytol. 191, 314–318 (2011).
31. Yilmaz, P. et al. Minimum information about a marer gene sequence (MIMAS) and minimum information about any (x)
sequence (MIxS) specications. Nat. Biotechnol. 29, 415–420 (2011).
32. Huang, Q. et al. Development and application of benthic macroinvertebrate-based multimetric indices for the assessment of streams
and rivers in the Taihu Basin, China. Ecol. Indic. 48, 649–659 (2015).
33. Beall, F. D., Semin, . G. & Jeries, D. S. Trends in the outputs of rst-order basins at Turey Laes Watershed, 1982–96. Ecosystems
4, 514–526 (2001).
34. Environment Canada. Canadian Aquatic Biomonitoring Networ eld manual- Wadeable streams (2012).
35. Hajibabaei, M., Spall, J. L., Shoralla, S. & van onynenburg, S. Assessing biodiversity of a freshwater benthic macroinvertebrate
community through non-destructive environmental barcoding of DNA from preservative ethanol. BMC Ecol. 12, 28 (2012).
36. Martin, M. Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet.journal 17, 10–12 (2011).
37. Edgar, . C. Search and clustering orders of magnitude faster than BLAST. Bioinformatics 26, 2460–1 (2010).
38. Wang, Q., Garrity, G. M., Tiedje, J. M. & Cole, J. . Naive Bayesian classier for rapid assignment of rNA sequences into the new
bacterial taxonomy. Appl. Environ. Microbiol. 73, 5261–5267 (2007).
39. Pompanon, F. et al. Who is eating what: Diet assessment using next generation sequencing. Mol. Ecol. 21, 1931–1950 (2012).
40. Lindahl, B. D. et al. Fungal community analysis by high-throughput sequencing of amplied marers – a user’ s guide. New Phytol.
199, 288–299 (2013).
41. Amend, A. S., Seifert, a & Bruns, T. D. Quantifying microbial communities with 454 pyrosequencing: does read abundance count?
Mol. Ecol. 19, 5555–65 (2010).
42. Osanen, J. et al. Vegan: community ecology pacage.  pacage version 2.4-0 (2016).
43. Walsh, C. & Mac Nally, . hier.part: hierarchical partitioning (2013).
44. Chevan, A. & Sutherland, M. Hierarchical partitioning. Am. Stat. 45, 90–96 (1991).
45. Bertini, S. C. B., Azevedo, L. C. B., de Carvalho Mendes, I. & Cardoso, E. J. B. N. Hierarchical partitioning for selection of microbial
and chemical indicators of soil quality. Pedobiologia (Jena). 57, 293–301 (2014).
46. Fox, J. & Weisberg, S. An  companion to applied regression, Second edition (2011).
47. O’Brien, . M. A caution regarding rules of thumb for variance ination factors. Qual. Quant. 41, 673–690 (2007).
48.  Core Team. : a language and environment for statistical computing. R-3.3.1 (2016).
Acknowledgements
We thank the Canadian Forest Service (CFS) of Natural Resources Canada, including leaders of the Forest
Ecological Integrity, Canadian Boreal Forest Agreement for joint funding. We also thank the Hearst Forest
Management Inc. for facilitating our work on the Hearst Sustainable Forest and contributing enhanced forest
inventory and mapping data, and colleagues Drs Doug Pitt (Canadian Forest Service) and Murray Woods
(Ontario Ministry of Natural Resources) for assisting in interpretation and use of the enhanced forest inventory
data. Finally, we thank Dr. Paul Hazlett and sta of the CFS Water Lab for analysing stream water chemistry
samples, and Shadi Shokralla and Shannon Eagle for help with DNA metabarcoding analysis.
Author Contributions
C.E.E. ran statistical analyses and wrote the manuscript. D.G.T. designed and implemented the experiment with
input from L.A.V., and both provided forestry expertise and editorial support. T.M.P. and M.H. provided DNA
metabarcoding data and expertise, and T.M.P. performed bioinformatic processing. T.S. and D.C. generated
watershed feature data, and S.C. performed morphological identications. All co-authors contributed to review
of the manuscript.
Additional Information
Supplementary information accompanies this paper at https://doi.org/10.1038/s41598-017-13157-x.
Competing Interests: e authors declare that they have no competing interests.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional aliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Cre-
ative Commons license, and indicate if changes were made. e images or other third party material in this
article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons license and your intended use is not per-
mitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
© e Author(s) 2017
Content courtesy of Springer Nature, terms of use apply. Rights reserved

1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com