https://orcid.org/0000-0003-3514-2634
Figures
Abstract
Large-scale monitoring of wildlife populations is a cornerstone of effective conservation and resource management, particularly for highly mobile taxa whose distributions shift rapidly across space and time. In this paper we explore the possibility of using environmental DNA-based surveys as a cost-efficient, complementary tool to estimate populations of North American waterfowl species. To achieve this, we first evaluated the performance of all currently available avian metabarcoding primers and compared them to newly designed primers targeting the mitochondrial ND2 gene within the Anatidae tribes of North America. All the existing avian assays showed strong cross-priming amplification with other vertebrates. In contrast, in-silico analyses of our waterfowl targeted assays showed a high degree (>90%) of avian specificity, encompassing all the 132 Anatidae species sequenced thus far. We used this targeted metabarcoding approach to track the temporal variation in the relative abundance of waterfowl species during the fall migration at Montezuma National Wildlife Refuge, New York, a major resting area for waterfowl on their journey to and from North American nesting areas. We compared eDNA results with visual surveys conducted by us and from those reported on eBird, a community science database. Our results showed that eDNA detected all waterfowl species (n = 25) observed during the visual surveys. Positive correlations existed between standardized amplicon sequence variant (ASV) counts and the relative abundance of waterfowl species as reported in eBird on the day of sampling and up to five days prior. Importantly, eDNA metabarcoding captured temporal shifts in community composition and species turnover during fall migration, highlighting its utility for tracking relative changes in waterfowl assemblages through time. As is often reported in metabarcoding studies, eDNA did not provide a good metric for absolute abundance of species; accordingly, only 8 out of 25 waterfowl species showed significant correlations between the number of eDNA reads and the total abundance of birds. Overall, while eDNA-targeted metabarcoding has not yet been applied to study bird communities extensively, our results demonstrate that this technique can be used as an effective complementary tool for assessing species composition of waterfowl communities and estimating relative abundance of species within those communities.
Citation: Guimaraes de Andrade L, Bogdanowicz SM, Klinck H, Lodge D, Andrés JA (2026) Environmental DNA monitoring of waterfowl reveals community changes during migration. PLoS One 21(4): e0337508. https://doi.org/10.1371/journal.pone.0337508
Editor: Mizanur Rahman, UTRGV: The University of Texas Rio Grande Valley, UNITED STATES OF AMERICA
Received: November 9, 2025; Accepted: March 23, 2026; Published: April 28, 2026
Copyright: © 2026 Guimaraes de Andrade et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Sequencing Data is available in Genbank: Accession to cite for these SRA data: PRJNA1420233 Temporary Submission ID: SUB15806852 https://www.ncbi.nlm.nih.gov/sra/PRJNA1420233 Metadata: Dryad Reviewer sharing link: http://datadryad.org/share/LINK_NOT_FOR_PUBLICATION/ub0QHIf0Cq-gtZRTNddJwp4v7E2Rnh-dBtPby9WRZK4.
Funding: Funding for this study was provided by the Department of Defense (DoD) Strategic Environmental Research and Development Program (SERDP) Grant/Award Number: RC19-1004 and the Huyck Research Grants for the research at the Huyck Preserve. L. Andrade was a postdoc associate in the Department of Ecology and Evolutionary Biology at Cornell University. The funder had no role in this study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Large-scale wildlife monitoring programs play a crucial role in meeting the management and conservation objectives for waterfowl (Anatidae), a taxonomic group that is ecologically, economically, and culturally significant [1]. To estimate population sizes and trends, traditional waterfowl monitoring efforts cover extensive geographic areas using a combination of visual surveillance techniques [2–9]. Ground visual surveys are labor-intensive and constrained by limited access to private lands [10]. Thus, aerial surveys have become the preferred tool in these large-scale surveys (e.g., [11]).
Despite their advantages, aerial surveys also have important limitations beyond the disturbance associated with low-altitude flights [12]. Depending on the sampling design, extrapolation to unsampled areas may be difficult and spatial or temporal shifts in waterfowl distributions may lead to underrepresentation of areas with high use [6]. Additionally, the infrequent timing of many aerial surveys constrains their ability to capture short-term variation in abundance, and the processing and interpretation analysis of aerial imagery remains time-consuming even with recent advances in machine-learning-based approaches [5,13–15].
Considering these constrains, there is a growing need for alternative, cost-effective, and minimally invasive tools for monitoring waterfowl populations. In this context, high-throughput sequencing of macro-organismal DNA from environmental samples, known as eDNA metabarcoding, has proven to be a robust and efficient technique to characterize biological communities across a diverse array of aquatic taxa (e.g., [16–20]. Several lines of evidence suggest that extending eDNA monitoring to waterfowl populations should be feasible.
First, eDNA metabarcoding surveys using fish and mammal primers typically “bycatch” bird species [21–28]. Second, analyses of water samples from zoo bird enclosures and natural environments have demonstrated that avian metabarcoding primers can be used to describe bird communities [29,30]. Third, eDNA assays have detected geese and swan species in both controlled and natural aquatic systems [31,32]. In this study we take a first step towards the future implementation of robust eDNA-based waterfowl biosurveillance programs by addressing two questions. Because waterfowl communities are highly dynamic, with rapid turnover during migration, effective monitoring must resolve temporal changes in community composition rather than relying solely on static presence or abundance. Accordingly, we test whether (1) a diverse waterfowl community can be reliably detected from water samples, and (2) whether eDNA-derived patterns of waterfowl abundance and turnover are consistent with those obtained from intensive ground-based visual surveys.
To answer these questions we first tested a series of published bird metabarcoding primers [30,33–39]. Although most of these primers successfully amplified waterfowl eDNA, our initial results showed PCR primer bias and cross-amplification with other abundant taxa, including fish, severely affected the sensitivity of the assay. To minimize these limitations we designed a metabarcoding assay [sensu [40]] specifically to detect North American waterfowl species. We then examined the applicability of this method by tracking the temporal variation in relative abundance of waterfowl species during fall migration at Montezuma National Wildlife Refuge, New York, a major resting area for waterfowl on their migration to and from North American nesting areas.
Materials and methods
Study areas
Our primary study area was Montezuma National Wildlife Refuge, in Seneca County, New York, United States (Fig 1). Given the water flow within the aquatic system, our water samples primarily represented the Main Pool (Fig 1), but would also have contained water deriving from other parts of the system.
The map shows the Seneca Spillway where water samples were taken in 10 consecutive weeks during the peak of the 2020 waterfowl fall migration. Black arrows represent the general direction of the water flow during this period.
Montezuma NWR is a 4,000-hectare wetland complex located on a major migratory corridor in the Atlantic Flyway, sustaining up to a million waterbirds during non-breeding seasons [41]. The Main Pool (also referred to as Black Lake on some maps) is a shallow 309-hectare impoundment whose water levels are managed to provide emergent marsh and other habitat types for migratory birds and resident wildlife. The water level was 1 m deep during our field sampling, from September to November 2020. Other than precipitation, the primary intake for the Main Pool is water from Cayuga Lake (to the south), which is provided through a ditch with a gravity-fed connector [42]. Water leaves the Main Pool via the Seneca Spillway outlet, which empties into the Cayuga-Seneca Canal. The overall flow of water is from Cayuga Lake northward through the Main Pool to the Cayuga-Seneca Canal (Fig 1). To conduct the initial assessment of both general and targeted bird metabarcoding primers, we collected water samples from both Sapsucker Woods Pond in Tompkins County, New York (42°28’48.1"N, 76°27’05.7"W) and two water bodies at the Huyck Preserve (Lincoln Pond – 42°31"35"N, 74°9’31"W and Lake Myosotis – 42°31’ 38.3"N, 74°09’30.4"W) in Albany County, New York (S1, S2 and S3 Figs). Then, to test the validity of targeted metabarcoding as a monitoring tool, we took weekly water samples (n = 3 + 1 field blank) at Montezuma NWR during the peak of the 2020 fall waterfowl migration (September 24 -November 29). All weekly water samples were taken near the Seneca Spillway outlet (Fig 1). Sampling was carried out in compliance with the Montezuma National Wildlife Refuge Research and Monitoring Special Use Permit #20−05 issued by the U.S. Fish and Wildlife Service. Sampling at the Huyck Preserve was conducted in compliance with the funding received from the station.
Water sampling
At each sampling location, we sampled eDNA by collecting and filtering 250 ml water samples and one distilled water negative control. This sample volume is commonly used in aquatic eDNA protocols and metabarcoding studies, providing a practical balance between capturing sufficient DNA and limiting filter clogging and field processing constraints (e.g., [43]).
Water was collected by submerging a sterilized wide-mouth Nalgene container (Thermo Fisher Scientific) < 0.1 m beneath the water’s surface following the recommendations of [44]. Each of the samples was then filtered through a 25 mm diameter cellulose-nitrate filter, 1.0 µm pore size, (Whatman, GE Healthcare) using a luer-lok sterile syringe system. Filtration blanks of distilled water were processed in the same way. Filters from each sample were immersed in 700 μl Longmire’s buffer in the field, transferred to the lab within 24 h, and stored at −80°C until DNA extraction. Water samples described above were taken just after our visual surveys.
Visual surveys
From the shore, we conducted exhaustive visual surveys of all waterfowl on the Montezuma NWR Main Pool during weekly visits from September 24 to November 29, 2020. These ground-based surveys were performed along a 1.3 km stretch of the Montezuma NWR Wildlife Drive, immediately adjacent to the Main Pool, using Zeiss 10 x 40 binoculars and a Swarovski spotting scope with a 20–60x eyepiece. The average survey start time was 9:28 am (range = 7:47 am to 10:33 am) and the average survey duration was 131 minutes (range = 92–149 minutes). To supplement our ground visual surveys, we compiled waterfowl observational data from eBird, the world’s largest curated database of citizen science bird sightings [45], for each water sampling date plus a five-day look-back period. We followed best practices for using eBird data [46], utilizing only complete checklists (i.e., all the birds observed are reported) that followed eBird’s “Stationary” or “Traveling” protocols. For dates with more than one complete eBird checklist, we calculated species abundance as the mean number of individuals from all complete checklists.
Avian metabarcoding primer tests using natural eDNA
Based on previous bird eDNA and phylogenetics studies we tested 11 unique primer pairs targeting different fragments of 5 mitochondrial loci: cytochrome c oxidase I [35,37], 12 S rRNA [30,34], NADH dehydrogenase subunit 2 [38,39], and Cytochrome b [33,36] (S1 Table). Insert length and amplicon size for each genetic marker were presented in S2 Table. The performance of each primer pair was tested on the eDNA samples from Sapsucker Woods Pond and Huyck Preserve (see above).
Anatidae primer design
We chose the mitochondrial gene NADH Dehydrogenase, ND2, as the molecular marker for waterfowl species because it is a well-established locus in avian genetics that provides high taxonomic resolution among closely related bird species and has been widely used in phylogenetic and species-level identification studies, often outperforming more conserved mitochondrial loci at shallow evolutionary scales (e.g., [47]). In addition, ND2 is represented by an extensive and well-curated set of avian reference sequences in the GenBank database, facilitating reliable taxonomic assignment in metabarcoding applications when appropriate quality controls are applied.
All ND2 gene sequences from NCBI corresponding to all North American Anatidae tribes as of May 27, 2020 were downloaded, and up to two DNA sequences for each species from distant geographic locations were selected. A total of 85 sequences representing 54 species and 19 genera in the family Anatidae were then aligned using the ClustalW algorithm. The manual inspection of the alignment did not reveal suitable ND2 regions for a single “universal” set of primers for Anatidae. Thus, we followed a strategy similar to that of [48] to generate a Tamura-Nei NJ tree to identify groups of similar sequences sharing relatively conserved regions (S5 Fig). For each of the three groups we found, we designed a single set of PCR primers using the software Geneious Prime version 2020.1.1 and default parameters (Table 1).
In-silico assessment of Anatidae ND2 primers
The performance of the primers was tested as follows. First, we used CRABS [49] to generate a custom reference database including all ND2 vertebrate sequences found in the NCBI nr repository (n = 157,730). Using the same pipeline, we performed in-silico PCR to produce a list of unique metabarcodes and assessed the specificity of the primers by calculating the percentage of Anatidae species amplified over non-target organisms using a maximum mismatch of 2 nucleotides in each primer.
DNA extraction and library construction
All DNA extractions were carried out in a dedicated eDNA laboratory at Cornell University (Ithaca, NY), with UV light treatment overnight, HEPA filtered air under positive pressure, personnel full body suits, breathing masks and face cover shields. We used the DNeasy Blood and Tissue extraction kits (Qiagen Inc.) and the protocol described in [50]. All lab reusable materials, such as forceps, containers, and the lab counters, were regularly sterilized with 50% bleach solution and treated under UV light before and after each round of DNA extractions. Extraction negative controls were included in each of the eDNA extractions rounds. Extracted samples were placed in a −20ºC freezer until library preparation. First-stage eDNA amplification was conducted in triplicate for each sample and negative control (i.e., field, filtration, extraction, and PCR blanks) in PCR volumes of 20 μl containing 8.475 μl molecular biology grade (MBG) water, 5 μl of 5x OneTaq Standard Reaction buffer, 0.2 μl of 10 μM primer (forward and reverse), 0.5 μl of 10mM dNTPs, 0.125 μl of One Taq Hot Start polymerase, 2.5 μl of bovine serum albumin (0.4 μg/μl), and 3 μl of eDNA extract as template. PCR cycling for each waterfowl ND2 primer pair was described in the supplementary material.
PCR products were then pooled by sample and tagged in a second-stage PCR using Illumina Nextera N7 and N5 indexes. The second PCR was carried out with a 20-mL reaction volume per eDNA sample containing: 3 μl DNA, diluted with MGB water 1:1, 10.9 μl MGB water, 4 μl 5x OneTaq buffer, 0.4 μl 10mM dNTPs, 0.8 μl each of 10 mM Nextera N7 and Nextera N5, and 0.1 μl OneTaq Hot Start DNA polymerase (New England – BioLabs Inc.).
PCR was run using the following thermal cycling parameters: six cycles of 30 seconds denaturation at 95ºC, 1 minute annealing at 62ºC, and 1 minute extension at 68ºC, with a final extension for 10 minutes at 68ºC. The resulting amplicons were purified and size-selected using Agencourt AMPure XP beads (reaction ratio AMPure beads 0.9: PCR product 1; Beckman Coulter Genomics), and the concentration of each library was estimated using the Qubit dsDNA Broad Range Kit and Qubit 2.0 fluorometer (Life Technologies). We pooled the barcoded PCR products for Anatini and Aythyini primers and all the libraries were diluted to 2 nM for paired-end sequenced V2 (2 x250bp) on a Miseq platform (Illumina, Inc.).
Bioinformatic analyses
Initial quality control of raw sequence data was performed using FastQC/MultiQC [51]. Degenerate bases and adaptors were trimmed from demultiplexed paired reads and sequences 35 bp in length were removed using Trimmomatic 0.33 [52]. Processed sequences were then analyzed using DADA2 1.16 [53], which involved removing forward and reverse primers, trimming reads to 200 bp for Anatini/Aythyini primers, and 390 bp for Anserini/Cygnini, and discarding sequences with expected number of errors >2 (EE, calculated based on Phred scores). The sequences were then denoised using the DADA2 error filtering algorithm. Forward and reverse reads for Anserini/Cygnini primers were merged with a minimum overlap of 20 bp and a maximum of one accepted error in the overlap region, and putative chimeras were removed. For Anatini/Aythyini primers, because of the small size of the amplicon (200 bp and 179 bp respectively), only forward reads were used for subsequent analysis. The resulting dataset contains amplicon sequence variants (ASVs) and the associated read counts for each sample. Taxonomic assignments for the resulting amplicon sequence variants (ASVs) were obtained using the BLASTn algorithm [54] and the nucleotide and taxonomy (nt/taxdb) databases from NCBI [55,56]. We retained the top five target sequence matches for each ASV and assigned species-level taxonomy to ASVs that matched a single species with a sequence identity of ≥98%. If multiple species matched the query sequence equally well, ASVs were assigned the lowest common taxonomic rank (genus or family) among the target sequences with equal percent identity. ASVs that could not be classified to the family level were excluded from further analysis. ASVs with taxonomic assignments were filtered by removing ASVs with read counts fewer than the average number of non-zero reads summed across the DNA extraction and PCR negative controls. To account for variation in sequencing depth among samples, read counts were standardized by proportionally scaling each sample to the minimum sequencing depth prior to downstream metabarcoding analyses.
Statistics
To account for abundances that differed across species by five orders of magnitude for both visual surveys and ASVs, we used log-log correlations to reflect the relationship between number of ASVs and the visual estimates of bird number, as previous quantitative eDNA metabarcoding studies have done [57,58].
Because waterfowl daily density is likely to exhibit short-term temporal autocorrelation [59] and because the rate of eDNA degradation can vary from hours to weeks depending on the ecosystem, target species, and eDNA capture method [60], we calculated the correlations between the number of ASVs and individuals, as well as between the number of birds/species (from eBird lists) on different dates, up to 5 days prior to each eDNA sampling event. We selected this period as a pragmatic approximation of “recent presence”, consistent with the timescales over which migrating waterfowl commonly use major stopover sites such as Montezuma NWR [61], and with the upper limit interval over which macro-organismal eDNA is often detectable in aquatic systems [62–64].
To visualize the temporal covariation of waterfowl and their associated eDNA reads during migration, we constructed a correlation-based network (CNA) including eDNA and visual count datasets. In the CNA, the nodes represent dates during the fall migration and the edges between them the significant, positive, log-log correlation coefficients (≥ 0.6) between visual surveys or between visual surveys and the standardized number of DNA reads obtain by combining the Anserini/Cygnini and Anatini/Aythyini datasets. We applied a correlation threshold of r ≥ 0.6 to retain only strong, positive associations, reducing network density and limiting the influence of weak or potentially spurious correlations that can arise from shared trends, compositional effects, or sampling noise. This threshold is commonly used in correlation-based ecological and microbial network analyses to improve interpretability and focus on biologically meaningful structure rather than marginal relationships (e.g., [65,66]).
To compare waterbird relative abundance estimates based on visual surveys with those based on eDNA read counts we focused on sampling dates with at least three independent complete eBird checklists (September 24th, October 15th, November 7th, and November 29th). For each of these dates, we compared the log-log correlation coefficient between our visual observations and eDNA counts to the log-log correlation coefficient between our visual observations and eBird data. All statistical analyses were performed with R.4.0.4 [67] and igraph.
Results
Evaluation of avian metabarcoding primers on eDNA samples
The pilot experiments at Sapsucker Woods Pond and the two Huyck Preserve waterbodies revealed a high level of cross-amplification of non-targeted taxa for all previously published avian metabarcoding primers. COI primers showed very little specificity, amplifying mostly bacteria (72% of reads) and fishes (20% of reads) (S4 Fig). Similarly, CytB primers amplified mostly other taxonomic groups (fish 54% of reads; mammals 44% (S4 Fig). 12S and ND2 primers amplified mostly avian species (86% and 78% of reads respectively), but the majority of ASVs were represented by only one waterfowl species (Canada goose) that was the predominant species in the two locations during the field sampling (S4 Fig). 12S and ND2 also cross-amplified teleosts (12S: 8% of reads; ND2: 20% of reads), potentially compromising the ability to detect waterfowl species in water bodies where fish species are abundant.
In-silico and experimental evaluation of waterfowl-targeted primers
To avoid fish cross-amplification and increase waterfowl detection, we designed three pairs of primers, each of which targeted groups that corresponded to Anatidae tribes (Table 1).
In silico-PCR simulated against all vertebrate sequences available on the nt NCBI repository revealed that the targeted primers could amplify all ND2 waterfowl sequences available, encompassing 76% (132/174) of the species recognized by the IOC World Bird List International Ornithological Committee [68]. None of the targeted primers are waterfowl-specific when considering a maximum of two mismatches between each primer and the template. However, within these parameters the newly designed primers exclusively amplify bird species (S6 Fig). The experimental evaluation in two water bodies with contrasting waterfowl species diversity indicated high sensitivity and accuracy. At the Huyck Preserve, where visual surveys indicated low abundance and diversity of waterfowl (Canada goose, n = 406; Wood duck, n = 56; hooded merganser, n = 5; mallard, n = 11; American black duck, n = 1), we generated approximately 4 million high-quality reads, identifying 46 ASVs corresponding to four of the five observed species, with the Hooded Merganser being the only one not detected. Additionally, we identified DNA from three species that were not visually observed: American Wigeon, Northern Pintail, and Redhead. At Montezuma NWR, where 25 waterfowl species were reported during our sampling period (Table 2), the eDNA dataset comprised over 11 M high-quality sequence reads (~650,000 reads/sample) corresponding to 322 waterfowl ASVs. eDNA successfully detected all waterfowl species present in Montezuma NWR, whose abundance spanned 5 orders of magnitude (from less than 10 to over 5,000 individuals). While most of the ASVs could be undoubtedly assigned to a single species, the targeted ND2 fragments could not be distinguished three species pairs: mallard/American black duck, tundra swan/trumpeter swan, and greater/lesser scaup. Accordingly, both visual and ASV counts for these species pairs were pooled for subsequent analyses.
Temporal variation in the abundance of waterfowl species
Overall, there was a weak correlation between the standardized number of ASVs and the total abundance of birds, as determined by the number of individuals observed during the visual survey on the day of sampling (Fig 2). Only 8 out of 25 species (including blue-winged teal, bufflehead, canvasback, green-winged teal, and wood duck) showed a statistically significant correlation (Pone-tail> 0.05). For example, Canada goose ASVs were consistently overrepresented, while ring-necked duck reads were underrepresented. Although both species exhibited strong seasonal variation in the number of individuals, there was little seasonal variation in their eDNA read counts. Thus, the observed variation in the standardized number of ASVs did not accurately reflect the temporal fluctuations in the absolute abundance of waterfowl species observed during fall migration.
In contrast, the standardized ASV counts showed significant correlations with relative abundance (Fig 3). A correlational network analysis of significant positive log-log relationships between ground visual surveys and standardized ASVs revealed a highly dynamic and variable system. Importantly, eDNA data captured relative waterfowl abundance not only on the day of sampling, but also correlated well with visual surveys conducted up to five days prior (Fig 3). The exception was October 31st, when the relative abundance of species in eDNA did not show any significant correlation with the relative abundance based on individual ground visual surveys (Fig 3).
The nodes represent the dates during the waterfowl fall migration and the edges between them represent the significant, positive, and log-log correlation coefficients (≥ 0.6) between ground visual surveys (orange nodes) or between visual surveys and the standardized number of ASVs (pink nodes) obtain by combining the Anserini/Cygnini and Anatini/Aythyini and datasets.
At any given sampling date, the relative abundance of ASVs showed a positive correlation (average τ ≥ 0.62) with the relative abundance of species present in the area. On 5 out of the 6 sampling dates for which at least three independent ground visual surveys (eBird) were available, the strength of the eDNA-visual correlation exceeded that of the correlation between ground visual surveys (Fig 4).
Gray: Mean and 95% confidence intervals of Kendall correlation coefficients between the number of birds present as determined by eBird checklists from different observers including our own visual survey. Orange: Kendall correlation coefficients between the standardized number of ASVs and number of birds present. We did not include the dates Oct1, Oct8, Oct31 and Nov 14 in this analysis because there were fewer than three checklists available for correlation.
Discussion
Our study adds to the growing body of research demonstrating the usefulness of eDNA approaches for monitoring bird species, and extends this work by showing how eDNA can be applied to track species turnover and community dynamics during migration. While most previous avian eDNA studies have focused on the detection of specific target species, such as the Gouldian Finch [69], Black Rail [70], Ridgway’s Rail [71], and various wading birds [72,73], our study employs targeted metabarcoding, a technique inherently designed to characterize entire communities. By leveraging repeated sampling through a migration period, we demonstrate that eDNA metabarcoding can resolve temporal changes in community composition, rather than static presence alone. This work complements recent applications of metabarcoding to assess bird community composition across structurally complex habitats, including bush ecosystems [74], rice-based agricultural landscapes [75], and managed wetland systems supporting waterbird assemblages such as geese and swans [31,32,76], and highlights the potential of eDNA for dynamic biodiversity monitoring in systems where community composition is inherently transient.
By employing a suite of primers specifically designed to target Anatidae tribes, with reduced cross-amplification to non-target vertebrates relative to existing avian primers, we detected both common and rare waterfowl species during fall migration in an important bird conservation area, highlighting the value of eDNA metabarcoding for community-level presence–absence detection in birds. Notably, eDNA detections included all waterfowl species recorded during visual surveys, as well as Ross’s goose, an uncommon and difficult-to-identify species that often blends into mixed flocks with the more abundant snow goose.
Presence data are an important component of ecological inference, and can be used to determine migration time, stopover use, turnover, and habitat suitability [77,78]. Our results show that eDNA accurately determines the species composition of waterfowl communities and can, therefore, be used to capture broad-scale community dynamics, turnover, and temporal patterns. In some circumstances, eDNA signals vary substantially over short distances due to localized shedding, retention, and environmental heterogeneity. However our hydrologically integrative approach successfully aggregated eDNA signals over space (e.g., [79–81]: the eDNA samples collected at the outflow of the Main Pool at Montezuma NWR obtained a pooled, system-level assessment of species presence.
Our recent work on eDNA transport in Cayuga Lake, has shown that flow pathways, residence time, and vertical mixing strongly regulate eDNA signal strength, and that eDNA concentrations reflect the coupling of production with physical transport processes rather than abundance alone [82]. In impounded wetland systems such as Montezuma NWR, this complexity is likely to be further amplified by emergent and submerged vegetation, which can trap, retain, or locally concentrate eDNA [44,83,84]. Therefore, our quantitative inferences should not be generalized to study designs that employ spatially replicated sampling. For most waterfowl species in our study, the number of ASVs was a poor proxy of the total number birds present. Such a mismatch between ASV counts and the number of individuals present is often reported in metabarcoding studies and has been often highlighted as an intrinsic limitation of metabarcoding studies. In contrast, as in our study, the proportion of ASVs attributed to different waterfowl species reliably reflected their relative abundance, and the eDNA signal captured the seasonal shifts in the relative abundance of functional groups, with dabbling ducks more prevalent during early to mid-fall migration and diving ducks becoming more abundant later in the season. The ability of eDNA metabarcoding results to detect such shifts is important because, outside a migratory context, similar changes in community composition are widely used as early warning indicators of habitat degradation, climate-driven range shifts, and biodiversity loss [85–87]. Even in relatively resilient guilds such as waterfowl [9], changes in the balance between dabbling and diving ducks can signal alterations in vegetation structure or water depth, providing actionable information for adaptive wetland management [88].
The correlation coefficients between standardized ASV counts and visual surveys were similar in magnitude to the correlation coefficients obtained between different visual surveys. Visual surveys, including eBird, are subject to observer effort and skill, species detectability, and reporting biases, and therefore uncertainty. Consequently, the agreement between eDNA and eBird likely reflects shared sensitivity to broad temporal dynamics in migration phenology rather than a proof of methodological superiority or a direct validation of abundance estimates.
Altogether, our results demonstrate that eDNA-targeted metabarcoding captures ecologically meaningful signals of relative waterfowl abundance and temporal community dynamics. While this approach is not yet a replacement for traditional monitoring methods, it provides a powerful complementary tool for informing adaptive management, particularly in dynamic wetland ecosystems where environmental pressures, rapid change, and logistical constraints challenge conventional survey approaches.
Supporting information
S1 Fig. eDNA sampling site at the Cornell Lab of Ornithology, Ithaca, NY, sampling site SWP1.
https://doi.org/10.1371/journal.pone.0337508.s001
(TIF)
S2 Fig. eDNA sampling sites at the Huyck Preserve, Rensselaerville, NY. Lake Lincoln Pond (P1 and P2) and Lake Myosotis (L1 – L3).Ten Mile Creek (C1), which links the two waterbodies.
https://doi.org/10.1371/journal.pone.0337508.s002
(TIF)
S3 Fig. Workflow of the pilot experiments to conduct the initial assessment of both general and targeted bird metabarcoding primers (the Cornell Lab of Ornithology, Phase I, and the Huyck Preserve, Phase II).
https://doi.org/10.1371/journal.pone.0337508.s003
(TIF)
S4 Fig. Percentage of sequence reads per organism group for each genetic marker tested to amplify bird DNA in environmental samples collected at Sapsucker Woods Pond and Huyck Preserve.
https://doi.org/10.1371/journal.pone.0337508.s004
(TIF)
S5 Fig. Phylogenetic tree of waterfowl species selected in this study to design waterfowl ND2 primers.
We used the Neighbor-Joining tree-built method and Tamura-Nei genetic distance model in Geneious Prime 2020.2.1. Colors represent different Tribes: red (Anserini and Cygnini), green (Anatini, Cairinini, and Oxyura), and orange (Aythyini and Mergini).
https://doi.org/10.1371/journal.pone.0337508.s005
(TIF)
S6 Fig. Graphical summary of the taxonomic families amplified by waterfowl ND2 primers.
(A) Anatini and Cairinini ND2 Primers, (B) Aythyini and Mergini ND2 Primers, and (C) Anserini and Cygnini ND2 Primers.
https://doi.org/10.1371/journal.pone.0337508.s006
(PDF)
S1 Table. Mitochondrial genes and multiple primers tested in the pilot experiments to amplify bird DNA from eDNA water samples.
Different colors in the table represent the primer pair. Mibird-U-F1, F2, F3, and F4 were used in pair with Mibird-U-R.
https://doi.org/10.1371/journal.pone.0337508.s007
(PDF)
S2 Table. Insert length and amplicon size of each genetic marker tested in this study.
https://doi.org/10.1371/journal.pone.0337508.s008
(PDF)
Acknowledgments
We thank Matthew Medler (Cornell Lab of Ornithology, Cornell University) for support to L. Andrade during the visual bird surveys and field sampling and for reviewing the manuscript. We also thank Linda Ziemba (Wildlife Biologist, Montezuma NWR) for all her support in the field sampling at Montezuma NWR, and Anne Rhoads for all her support during the field work at the Huyck Preserve and for sample collections at the station. We thank Sara Cristina Pereira de Souza and Larissa S. M. Sugai for the support with map constructions. We thank Irby J. Lovette, Bronwyn Butcher, and the Lovette Lab for all their support to L. Andrade with the pilot molecular analyses. We also would like to thank Kara Andres, Paul Simonin, and Timothy Lambert for feedback on the field sampling and molecular analyses.
References
- 1. Kumar AV, Rice MB. Optimized survey design for monitoring protocols: a case study of waterfowl abundance. J Fish Wildlife Manag. 2021;12(2):572–84.
- 2.
Baldassarre GA, Bolen EG. Waterfowl ecology and management. 2nd ed. Malabar, Florida: Krieger; 2006.
- 3.
Baldassarre GA. Ducks, geese, and swans of North America. Baltimore: Wildlife Management Institute, Johns Hopkins University Press; 2014.
- 4. Casazza ML, McDuie F, Jones S, Lorenz AA, Overton CT, Yee J. Waterfowl use of wetland habitats informs wetland restoration designs for multi-species benefits. J Appl Ecol. 2021;58:1910–20.
- 5. Dundas SJ, Vardanega M, O’Brien P, McLeod SR. Quantifying waterfowl numbers: comparison of drone and ground-based survey methods for surveying waterfowl on artificial waterbodies. Drones. 2021;5(1):5.
- 6. Hennig JD, Benson TJ, Stafford JD, Yetter AP, Hagy HM, Stodola KW. An aerial grid-based sampling method to estimate nonbreeding duck abundance. Wildlife Society Bulletin. 2017;41(4):678–84.
- 7. McEvoy JF, Hall GP, McDonald PG. Evaluation of unmanned aerial vehicle shape, flight path and camera type for waterfowl surveys: disturbance effects and species recognition. PeerJ. 2016;4:e1831. pmid:27020132
- 8. Soulliere GJ, Loges BW, Dunton EM, Luukkonen DR, Eichholz MW, Koch ME. Monitoring waterfowl in the midwest during the non-breeding period: challenges, priorities, and recommendations. J Fish Wildlife Manag. 2013;4(2):395–405.
- 9.
U.S. Fish and Wildlife Service. Waterfowl population status, 2023. Washington, D.C. USA: U.S. Department of the Interior; 2023. https://www.fws.gov/media/waterfowl-population-status-2023
- 10. Cordts SD, Zenner GG, Koford RR. Comparison of helicopter and ground counts for waterfowl in Iowa. Wildlife Soc Bulletin. 2002;30(2):317–26.
- 11. Wen D, Su L, Hu Y, Xiong Z, Liu M, Long Y. Surveys of large waterfowl and their habitats using an unmanned aerial vehicle: a case study on the siberian crane. Drones. 2021;5(4):102.
- 12. Fleming PJS, Tracey JP. Some human, aircraft and animal factors affecting aerial surveys: how to enumerate animals from the air. Wildlife Research. 2008;35(4):258–67.
- 13. Francis RJ, Lyons MB, Kingsford RT, Brandis KJ. Counting mixed breeding aggregations of animal species using drones: lessons from waterbirds on semi-automation. Remote Sensing. 2020;12(7):1185.
- 14. Hong S-J, Han Y, Kim S-Y, Lee A-Y, Kim G. Application of deep-learning methods to bird detection using unmanned aerial vehicle imagery. Sensors (Basel). 2019;19(7):1651. pmid:30959913
- 15. Pöysä H, Kotilainen J, Väänänen VM, Kunnasranta M. Estimating production in ducks: a comparison between ground surveys and unmanned aircraft surveys. Eur J Wildl Res. 2018;64(6):74.
- 16. Bálint M, Nowak C, Márton O, Pauls SU, Wittwer C, Aramayo JL, et al. Accuracy, limitations and cost efficiency of eDNA-based community survey in tropical frogs. Mol Ecol Resour. 2018;18(6):1415–26. pmid:30155977
- 17. Andrés J, Czechowski P, Grey E, Saebi M, Andres K, Brown C, et al. Environment and shipping drive environmental DNA beta-diversity among commercial ports. Mol Ecol. 2023;32(23):6696–709. pmid:36799015
- 18. Coghlan SA, Shafer ABA, Freeland JR. Development of an environmental DNA metabarcoding assay for aquatic vascular plant communities. Environmental DNA. 2020;3(2):372–87.
- 19. Fraija-Fernández N, Bouquieaux M-C, Rey A, Mendibil I, Cotano U, Irigoien X, et al. Marine water environmental DNA metabarcoding provides a comprehensive fish diversity assessment and reveals spatial patterns in a large oceanic area. Ecol Evol. 2020;10(14):7560–84. pmid:32760549
- 20. Gleason JE, Elbrecht V, Braukmann TWA, Hanner RH, Cottenie K. Assessment of stream macroinvertebrate communities with eDNA is not congruent with tissue-based metabarcoding. Mol Ecol. 2021;30(13):3239–51. pmid:32799390
- 21. Miya M. Environmental DNA metabarcoding: a novel method for biodiversity monitoring of marine fish communities. Annu Rev Mar Sci. 2022;14:161–85.
- 22. Harper LR, Lawson Handley L, Hahn C, Boonham N, Rees HC, Lewis E, et al. Generating and testing ecological hypotheses at the pondscape with environmental DNA metabarcoding: a case study on a threatened amphibian. Environmental DNA. 2019;2(2):184–99.
- 23. Macher T-H, Schütz R, Arle J, Beermann AJ, Koschorreck J, Leese F. Beyond fish eDNA metabarcoding: field replicates disproportionately improve the detection of stream associated vertebrate species. MBMG. 2021;5.
- 24. Mariani S, Harper LR, Collins RA, Baillie C, Wangensteen OS, McDevitt AD, et al. Estuarine molecular bycatch as a landscape-wide biomonitoring tool. Biol Conserv. 2021;261:109287.
- 25. Port JA, O’Donnell JL, Romero-Maraccini OC, Leary PR, Litvin SY, Nickols KJ, et al. Assessing vertebrate biodiversity in a kelp forest ecosystem using environmental DNA. Mol Ecol. 2016;25(2):527–41. pmid:26586544
- 26. Ritter CD, Dal Pont G, Stica PV, Horodesky A, Cozer N, Netto OSM, et al. Wanted not, wasted not: searching for non-target taxa in environmental DNA metabarcoding by-catch. Environ Adv. 2022;7:100169.
- 27. Thomsen PF, Kielgast J, Iversen LL, Møller PR, Rasmussen M, Willerslev E. Detection of a diverse marine fish fauna using environmental DNA from seawater samples. PLoS One. 2012;7(8):e41732. pmid:22952584
- 28. Zhang S, Zhao J, Yao M. Urban landscape-level biodiversity assessments of aquatic and terrestrial vertebrates by environmental DNA metabarcoding. J Environ Manage. 2023;340:117971. pmid:37119629
- 29. Coleman HT, Matthews TG, Sherman CDH, Holland OJ, Clark ZSR, Farrington L, et al. Contrasting patterns of biodiversity across wetland habitats using single‐time‐point environmental DNA surveys. Aquat Conserv. 2023;33(12):1401–12.
- 30. Ushio M, Murata K, Sado T, Nishiumi I, Takeshita M, Iwasaki W, et al. Demonstration of the potential of environmental DNA as a tool for the detection of avian species. Sci Rep. 2018;8(1):4493. pmid:29540790
- 31. Honka J, Kvist L, Olli S, Laaksonen T, Aspi J. Optimised PCR assays for detecting elusive waterfowl from environmental DNA. Ecol Evol. 2024;14(4):e11224. pmid:38571801
- 32. Shuai H, Liu X, Zhang Y, Sun Y, Li H, Li Z. eDNA enhances detection efficiency but reveals lower waterbird diversity: a comparison with point counting method. Avian Research. 2025;16(2):100236.
- 33. Awad A, Khalil SR, Abd-Elhakim YM. Molecular phylogeny of some avian species using Cytochrome b gene sequence analysis. Iran J Vet Res. 2015;16(2):218–22. pmid:27175180
- 34. Epp LS, Boessenkool S, Bellemain EP, Haile J, Esposito A, Riaz T, et al. New environmental metabarcodes for analysing soil DNA: potential for studying past and present ecosystems. Mol Ecol. 2012;21(8):1821–33. pmid:22486821
- 35. Kerr KCR, Stoeckle MY, Dove CJ, Weigt LA, Francis CM, Hebert PDN. Comprehensive DNA barcode coverage of North American birds. Mol Ecol Notes. 2007;7(4):535–43. pmid:18784793
- 36. Kocher TD, Thomas WK, Meyer A, Edwards SV, Pääbo S, Villablanca FX, et al. Dynamics of mitochondrial DNA evolution in animals: amplification and sequencing with conserved primers. Proc Natl Acad Sci U S A. 1989;86(16):6196–200. pmid:2762322
- 37. Patel S, Waugh J, Millar CD, Lambert DM. Conserved primers for DNA barcoding historical and modern samples from New Zealand and Antarctic birds. Mol Ecol Resour. 2010;10(3):431–8. pmid:21565042
- 38. Sorenson MD, Ast JC, Dimcheff DE, Yuri T, Mindell DP. Primers for a PCR-based approach to mitochondrial genome sequencing in birds and other vertebrates. Mol Phylogenet Evol. 1999;12(2):105–14. pmid:10381314
- 39.
Sorenson MD. Avian mtDNA primers. 2014. https://people.bu.edu/msoren/primers.html
- 40. Wu Y, Colborne SF, Charron MR, Heath DD. Development and validation of targeted environmental DNA (eDNA) metabarcoding for early detection of 69 invasive fishes and aquatic invertebrates. Environ DNA. 2022;5(1):73–84.
- 41.
BirdLife International. Important bird areas factsheet: montezuma wetlands complex. 2023. http://www.birdlife.org
- 42.
Wurster FC, Hunt KJ, Burns RE. Water resources inventory and assessment. Seneca Falls, New York: Montezuma National Wildlife Refuge; 2013.
- 43. Sevellec M, Lacoursière-Roussel A, Normandeau E, Bernatchez L, Howland KL. Effect of eDNA metabarcoding temporal sampling strategies on detection of coastal biodiversity. Front Mar Sci. 2025;12.
- 44. Goldberg CS, Turner CR, Deiner K, Klymus KE, Thomsen PF, Murphy MA, et al. Critical considerations for the application of environmental DNA methods to detect aquatic species. Methods Ecol Evol. 2016;7(11):1299–307.
- 45. Sullivan BL, Aycrigg JL, Barry JH, Bonney RE, Bruns N, Cooper CB, et al. The eBird enterprise: an integrated approach to development and application of citizen science. Biol Conserv. 2014;169:31–40.
- 46. Strimas-Mackey M, Hochachka WM, Ruiz-Gutierrez V, Robinson OJ, Miller ET, Auer T. Best practices for using eBird data. Ithaca, New York: Cornell Lab of Ornithology; 2020.
- 47. Barker FK, Benesh MK, Vandergon AJ, Lanyon SM. Contrasting evolutionary dynamics and information content of the avian mitochondrial control region and ND2 gene. PLoS One. 2012;7(10):e46403. pmid:23071562
- 48. Cannon MV, Hester J, Shalkhauser A, Chan ER, Logue K, Small ST, et al. In silico assessment of primers for eDNA studies using PrimerTree and application to characterize the biodiversity surrounding the Cuyahoga River. Sci Rep. 2016;6:22908. pmid:26965911
- 49. Jeunen G-J, Dowle E, Edgecombe J, von Ammon U, Gemmell NJ, Cross H. crabs-A software program to generate curated reference databases for metabarcoding sequencing data. Mol Ecol Resour. 2023;23(3):725–38. pmid:36437603
- 50. Spens J, Evans AR, Halfmaerten D, Knudsen SW, Sengupta ME, Mak SST, et al. Comparison of capture and storage methods for aqueous macrobial eDNA using an optimized extraction protocol: advantage of enclosed filter. Methods Ecol Evol. 2016;8(5):635–45.
- 51.
Andrews S. FastQC: a quality control tool for high throughput sequence data. 2010. http://www.bioinformatics.babraham.ac.uk/projects/fastqc/
- 52. Bolger AM, Lohse M, Usadel B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics. 2014;30(15):2114–20. pmid:24695404
- 53. Callahan BJ, McMurdie PJ, Rosen MJ, Han AW, Johnson AJA, Holmes SP. DADA2: High-resolution sample inference from Illumina amplicon data. Nat Methods. 2016;13(7):581–3. pmid:27214047
- 54. Camacho C, Coulouris G, Avagyan V, Ma N, Papadopoulos J, Bealer K, et al. BLAST+: architecture and applications. BMC Bioinformatics. 2009;10:421. pmid:20003500
- 55. Benson DA, Karsch-Mizrachi I, Lipman DJ, Ostell J, Wheeler DL. GenBank. Nucleic Acids Research. 2013;41:D36–42.
- 56. Federhen S. The NCBI taxonomy database. Nucleic Acids Res. 2012;40(D1):D136–43.
- 57. Martin JL, Santi I, Pitta P, John U, Gypens N. Towards quantitative metabarcoding of eukaryotic plankton: an approach to improve 18S rRNA gene copy number bias. MBMG. 2022;6.
- 58. Stoeckle BC, Kuehn R, Geist J. Environmental DNA as a monitoring tool for the endangered freshwater pearl mussel (Margaritifera margaritifera L.): a substitute for classical monitoring approaches?. Aquatic Conserv: Mar Freshw Ecosyst. 2016;26:1120–9.
- 59. Acosta S, Kelman T, Feirer S, Matchett E, Smolinsky J, Pitesky M, et al. Using the california waterfowl tracker to assess proximity of waterfowl to commercial poultry in the central valley of California. Avian Dis. 2021;65(3):483–92. pmid:34699147
- 60. Li J, Lawson Handley LJ, Harper LR, Brys R, Watson HV, Di Muri C, et al. Limited dispersion and quick degradation of environmental DNA in fish ponds inferred by metabarcoding. Environ DNA. 2019;1(3):238–50.
- 61. O’Neal BJ, Stafford JD, Larkin RP. Stopover duration of fall‐migrating dabbling ducks. J Wildl Manag. 2012;76(2):285–93.
- 62. Skinner EM, Hamann M, Strugnell J, Rath C. Investigating the limits of detection of environmental DNA from loggerhead and hawksbill turtles. Aquatic Conserv: Marine and Freshwater Ecosyst. 2025;35(8):e70199.
- 63. Minamoto T. Environmental DNA analysis for macro-organisms: species distribution and more. DNA Research. 2022;29(3).
- 64. Pont D. Predicting downstream transport distance of fish eDNA in lotic environments. Mol Ecol Res. 2024;24: e13934.
- 65. Barberán A, Bates ST, Casamayor EO, Fierer N. Using network analysis to explore co-occurrence patterns in soil microbial communities. ISME J. 2012;6(2):343–51. pmid:21900968
- 66. Guo B, Zhang L, Sun H, Gao M, Yu N, Zhang Q, et al. Microbial co-occurrence network topological properties link with reactor parameters and reveal importance of low-abundance genera. NPJ Biofilms Microbiomes. 2022;8(1):3. pmid:35039527
- 67.
R Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2021.
- 68.
Gill F, Donsker D, Rasmussen P. IOC World Bird List (v14.1). 2024. https://doi.org/10.14344/IOC.ML.14.1
- 69. Day K, Campbell H, Fisher A, Gibb K, Hill B, Rose A, et al. Development and validation of an environmental DNA test for the endangered Gouldian finch. Endang Species Res. 2019;40:171–82.
- 70. Neice AA, McRae SB. An eDNA diagnostic test to detect a rare, secretive marsh bird. Global Ecol Conserv. 2021;27:e01529.
- 71. Guan X, Lance RF, Feist SM. Environmental DNA (eDNA) assays for the detection of Ridgway’s rail (Rallus obsoletus) in the United States. Conserv Genet Resour. 2023;15(4):179–85.
- 72. Feist SM, Guan X, Malmfeldt MP, Lance RF. Two novel qPCR assays to enhance black rail (Laterallus jamaicensis) eDNA surveys in the United States. Conserv Genet Resour. 2022;14(3):321–9.
- 73. Schütz R, Tollrian R, Schweinsberg M. A novel environmental DNA detection approach for the wading birds Platalea leucorodia, Recurvirostra avosetta and Tringa totanus. Conserv Genet Resour. 2020;12(4):529–31.
- 74. Newton JP, Bateman PW, Heydenrych MJ, Kestel JH, Dixon KW, Prendergast KS, et al. Monitoring the birds and the bees: environmental DNA metabarcoding of flowers detects plant–animal interactions. Environ DNA. 2023;5(3):488–502.
- 75. Katayama N, Yamamoto S, Baba YG, Ito K, Yamasako J. Complementary role of environmental DNA for line-transect bird surveys: a field test in a Japanese rice landscape. Ecolog Indicat. 2024;166:112442.
- 76. Ke Y, Liu T, Han C, Yu X, Wang J, Ding L, et al. A review of eDNA technology in avian monitoring: current status, challenges and future perspectives. Avian Res. 2025;16(2):100235.
- 77.
Newton I. The migration ecology of birds. Academic Press; 2007.
- 78. Robinson WD, Bowlin MS, Bisson I, Shamoun-Baranes J, Thorup K, Diehl RH, et al. Integrating concepts and technologies to advance the study of bird migration. Frontiers in Ecol Environ. 2009;8(7):354–61.
- 79. Deiner K, Fronhofer EA, Mächler E, Walser J-C, Altermatt F. Environmental DNA reveals that rivers are conveyer belts of biodiversity information. Nat Commun. 2016;7:12544. pmid:27572523
- 80. Civade R, Dejean T, Valentini A, Roset N, Raymond J-C, Bonin A, et al. Spatial representativeness of environmental DNA metabarcoding signal for fish biodiversity assessment in a natural freshwater system. PLoS One. 2016;11(6):e0157366. pmid:27359116
- 81. Shogren AJ, Tank JL, Andruszkiewicz E, Olds B, Mahon AR, Jerde CL, et al. Controls on eDNA movement in streams: transport, retention, and resuspension. Sci Rep. 2017;7(1):5065. pmid:28698557
- 82. Li Z, Ramón CL, Bogdanowicz S, Koeberle AL, Wang D, Rueda FJ, et al. Tracing environmental DNA transport in a large lake with synthetic DNA microparticles and hydrodynamic modeling. Environ Sci Technol. 2026;60(4):3519–31. pmid:41556542
- 83. Mayne SR, Manning JA, Henderson SM, Parsley MB, Strickler KM, Nielson JR, et al. Hydrodynamics and aquatic vegetation drive spatial patterns of environmental DNA in ponds. Environmental DNA. 2024;6(6).
- 84. Harrison JB, Sunday JM, Rogers SM. Predicting the fate of eDNA in the environment and implications for studying biodiversity. Proc Biol Sci. 2019;286(1915):20191409. pmid:31744434
- 85. Both C, Bouwhuis S, Lessells CM, Visser ME. Climate change and population declines in a long-distance migratory bird. Nature. 2006;441(7089):81–3. pmid:16672969
- 86. Hitch AT, Leberg PL. Breeding distributions of north American bird species moving north as a result of climate change. Conserv Biol. 2007;21(2):534–9. pmid:17391203
- 87. Rosenberg KV, Dokter AM, Blancher PJ, Sauer JR, Smith AC, Smith PA, et al. Decline of the North American avifauna. Science. 2019;366(6461):120–4. pmid:31604313
- 88. Beatty WS, Webb EB, Kesler DC, Raedeke AH, Naylor LW, Humburg DD. Landscape effects on mallard habitat selection at multiple spatial scales during the non-breeding period. Landscape Ecol. 2014;29(6):989–1000.