In many agricultural landscapes where field drainage is required to enhance crop production, agricultural drainage ditches, and their associated banks and hedgerows can support riparian biodiversity, including bird communities. Against a global background of farmland bird and terrestrial insect decline due to agricultural intensification and extensification, emerging aquatic insects in these aquatic corridors can provide a pulse of energy-rich, nutritionally-important food for birds and other wildlife. In this paper, we quantify the value of drainage ditch habitats in terms of aquatic insect production as a potential food source for riparian foraging birds in a river basin in eastern Canada. Despite being highly managed, agricultural drainage ditches remained extremely productive in terms of emerging biomass of aquatic insects (high quantity), but large-bodied aquatic insects such as mayflies, stoneflies and caddisflies, which are rich in fatty acids, were more common in natural, forested streams and less common in agricultural streams and ditches. The proportion of riparian insectivorous birds was lowest along straight ditches running through agricultural fields and highest among meandering (sinuous) streams in more forested areas, suggesting that agricultural drainage systems may not be able to fully support resource use for foraging predators that rely on emerging aquatic insects. Agricultural producers can improve habitat provisioning for birds on their farms by supporting mosaicked farmscapes through careful conservation and management of ditches and ditch bank vegetation. Establishing larger forest blocks with natural or unmanaged streams between areas of more intense land use can ensure the provisioning of more high quality prey to riparian insectivorous birds, helping to find the balance between agricultural productivity and protection of declining bird populations.

The delivery of consistent and accurate fine‐resolution data on biodiversity using metabarcoding promises to improve environmental assessment and research. Whilst this approach is a substantial improvement upon traditional techniques, critics note that metabarcoding data are suitable for establishing taxon occurrence, but not abundance. We propose a novel hierarchical approach to recovering abundance information from metabarcoding, and demonstrate this technique using benthic macroinvertebrates. To sample a range of abundance structures without introducing additional changes in composition, we combined seasonal surveys with fish‐exclusion experiments at Catamaran Brook in northern New Brunswick, Canada. Five monthly surveys collected 31 benthic samples for DNA metabarcoding divided between caged and control treatments. A further six samples per survey were processed using traditional morphological identification for comparison. By estimating the probability of detecting a single individual, multispecies abundance models infer changes in abundance based on changes in detection frequency. Using replicate detections of 184 genera (and 318 species) from metabarcoding samples, our analysis identified changes in abundance arising from both seasonal dynamics and the exclusion of fish predators. Counts obtained from morphological samples were highly variable, a feature that limited the opportunity for more robust comparison, and emphasizing the difficulty standard methods also face to detect changes in abundance. Our approach is the first to demonstrate how quantitative estimates of abundance can be made using metabarcoding, both among species within sites as well as within species among sites. Many samples are required to capture true abundance patterns, particularly in streams where counts are highly variable, but few studies can afford to process entire samples. Our approach allows study of responses across whole communities, and at fine taxonomic resolution. We discuss how ecological studies can use additional sampling to capture changes in abundance at fine resolution, and how this can complement broad‐scale biomonitoring using DNA metabarcoding.

Abstract Background Identification of biomarkers, which are measurable characteristics of biological datasets, can be challenging. Although amplicon sequence variants (ASVs) can be considered potential biomarkers, identifying important ASVs in high-throughput sequencing datasets is challenging. Noise, algorithmic failures to account for specific distributional properties, and feature interactions can complicate the discovery of ASV biomarkers. In addition, these issues can impact the replicability of various models and elevate false-discovery rates. Contemporary machine learning approaches can be leveraged to address these issues. Ensembles of decision trees are particularly effective at classifying the types of data commonly generated in high-throughput sequencing (HTS) studies due to their robustness when the number of features in the training data is orders of magnitude larger than the number of samples. In addition, when combined with appropriate model introspection algorithms, machine learning algorithms can also be used to discover and select potential biomarkers. However, the construction of these models could introduce various biases which potentially obfuscate feature discovery. Results We developed a decision tree ensemble, LANDMark, which uses oblique and non-linear cuts at each node. In synthetic and toy tests LANDMark consistently ranked as the best classifier and often outperformed the Random Forest classifier. When trained on the full metabarcoding dataset obtained from Canada’s Wood Buffalo National Park, LANDMark was able to create highly predictive models and achieved an overall balanced accuracy score of 0.96 ± 0.06. The use of recursive feature elimination did not impact LANDMark’s generalization performance and, when trained on data from the BE amplicon, it was able to outperform the Linear Support Vector Machine, Logistic Regression models, and Stochastic Gradient Descent models (p ≤ 0.05). Finally, LANDMark distinguishes itself due to its ability to learn smoother non-linear decision boundaries. Conclusions Our work introduces LANDMark, a meta-classifier which blends the characteristics of several machine learning models into a decision tree and ensemble learning framework. To our knowledge, this is the first study to apply this type of ensemble approach to amplicon sequencing data and we have shown that analyzing these datasets using LANDMark can produce highly predictive and consistent models.

The soil fauna of the tropics remains one of the least known components of the biosphere. Long-term monitoring of this fauna is hampered by the lack of taxonomic expertise and funding. These obstacles may potentially be lifted with DNA metabarcoding. To validate this approach, we studied the ants, springtails and termites of 100 paired soil samples from Barro Colorado Island, Panama. The fauna was extracted with Berlese-Tullgren funnels and then either sorted with traditional taxonomy and known, individual DNA barcodes (“traditional samples”) or processed with metabarcoding (“metabarcoding samples”). We detected 49 ant, 37 springtail and 34 termite species with 3.46 million reads of the COI gene, at a mean sequence length of 233 bp. Traditional identification yielded 80, 111 and 15 species of ants, springtails and termites, respectively; 98%, 37% and 100% of these species had a Barcode Index Number (BIN) allowing for direct comparison with metabarcoding. Ants were best surveyed through traditional methods, termites were better detected by metabarcoding, and springtails were equally well detected by both techniques. Species richness was underestimated, and faunal composition was different in metabarcoding samples, mostly because 37% of ant species were not detected. The prevalence of species in metabarcoding samples increased with their abundance in traditional samples, and seasonal shifts in species prevalence and faunal composition were similar between traditional and metabarcoding samples. Probable false positive and negative species records were reasonably low (13–18% of common species). We conclude that metabarcoding of samples extracted with Berlese-Tullgren funnels appear suitable for the long-term monitoring of termites and springtails in tropical rainforests. For ants, metabarcoding schemes should be complemented by additional samples of alates from Malaise or light traps.

Freshwater systems are experiencing rapid biodiversity losses resulting from high rates of habitat degradation. Ecological condition is typically determined through identifying either macroinvertebrate or diatom bioindicator assemblages and comparing them to their known tolerance to stressors. These comparisons are typically conducted at family or genus levels depending on the availability of taxonomic keys and expertise for focal groups. The objective of this study was to test whether a more taxonomically comprehensive assessment of communities in benthic samples can provide a different perspective of ecological conditions. DNA metabarcoding was used to identify macroinvertebrates and diatoms from kick-net samples collected from sites with different habitat status. Sites with ‘good’ condition were associated with higher beta diversity as well as slightly higher directed connectance and modularity indicating higher resilience compared with ‘fair’ condition sites. Indicator value and correlation analyses used DNA metabarcoding data to detect 29 site condition indicator species consistent with known bioindicators and expected relative tolerances. DNA metabarcoding and trophic network analysis also recovered 11 keystone taxa. This study demonstrates the importance of taxonomic breadth across trophic levels for generating biotic data to study ecosystem status, with the potential to scale-up ecological assessments of freshwater condition, trophic stability, and resilience.

There is increasing need for biodiversity monitoring, especially in places where potential anthropogenic disturbance may significantly impact ecosystem health. We employed a combination of traditional morphological and bulk macroinvertebrate metabarcoding analyses to benthic samples collected from Toronto Harbour (Ontario, Canada) to compare taxonomic and functional diversity of macroinvertebrates and their responses to environmental gradients. At the species rank, sites assessed using COI metabarcoding showed more variation than sites assessed using morphological methods. Depending on the assessment method, we detected gradients in magnesium (morphological taxa), ammonia (morphological taxa, COI sequence variants), pH (18S sequence variants) as well as gradients in contaminants such as metals (COI & 18S sequence variants) and organochlorines (COI sequence variants). Observed responses to contaminants such as aromatic hydrocarbons and metals align with known patchy distributions in harbour sediments. We determined that the morphological approach may limit the detection of macroinvertebrate responses to lake environmental conditions due to the effort needed to obtain fine level taxonomic assignments necessary to investigate responses. DNA metabarcoding, however, need not be limited to macroinvertebrates, can be automated, and taxonomic assignments are associated with a certain level of accuracy from sequence variants to named taxonomic groups. The capacity to detect change using a scalable approach such as metabarcoding is critical for addressing challenges associated with biodiversity monitoring and ecological investigations.

Advances in high-throughput sequencing (HTS) are revolutionizing monitoring in marine environments by enabling rapid, accurate and holistic detection of species within complex biological samples. Research institutions worldwide increasingly employ HTS methods for biodiversity assessments. However, variance in laboratory procedures, analytical workflows and bioinformatic pipelines impede the transferability and comparability of results across research groups. An international experiment was conducted to assess the consistency of metabarcoding results derived from identical samples and primer sets using varying laboratory procedures. Homogenized biofouling samples collected from four coastal locations (Australia, Canada, New Zealand and the USA) were distributed to 12 independent laboratories. Participants were asked to follow one of two HTS library preparation workflows. While DNA extraction, primers and bioinformatic analyses were purposefully standardized to allow comparison, many other technical variables were allowed to vary among laboratories (amplification protocols, type of instrument used, etc.). Despite substantial variation observed in raw results, the primary signal in the data was consistent, with the samples grouping strongly by geographical origin for all data sets. Simple post hoc data clean-up by removing low-quality samples gave the best improvement in sample classification for nuclear 18S rRNA gene data, with an overall 92.81% correct group attribution. For mitochondrial COI gene data, the best classification result (95.58%) was achieved after correction for contamination errors. The identified critical methodological factors that introduced the greatest variability (preservation buffer, sample defrosting, template concentration, DNA polymerase, PCR enhancer) should be of great assistance in standardizing future biodiversity studies using metabarcoding.

The biodiversity-ecosystem function hypothesis postulates that higher biodiversity is correlated with faster ecosystem process rates and increased ecosystem stability in fluctuating environments. Exhibiting high spatiotemporal habitat diversity, floodplains are highly productive ecosystems, supporting communities that are naturally resilient and highly diverse. We examined linkages among floodplain wetland habitats, invertebrate communities and their associated traits, and ecosystem function across 60 sites within the floodplain wetlands of the lower Wolastoq | Saint John River, New Brunswick, using structural equation modelling and Threshold Indicator Taxa ANalysis. We identified key environmental filters structuring invertebrate communities, by linking increased niche differentiation through shoreline change, flood pulse dynamics, and macrophyte bed complexity with increased taxa and functional diversity. Examination of traits linked to ecosystem functions revealed that more resilient wetlands with balance between primary productivity and decomposition as carbon sources were associated with greater functional evenness and richness, while habitat patches with elevated decomposition rates had lower functional richness, reflecting a simplified, more disturbed habitat. While our more complex overarching SEM model was ultimately compromised by an overspecified number of pathways, our results nevertheless are indicative of a divergence between wetland and riverine ecosystems in their relationships linking biodiversity and ecosystem function, illustrating how to define ecosystem health in wetland habitats, and demonstrating how critical functions support healthy wetland habitats by providing increased resilience to disturbance. Read the free Plain Language Summary for this article on the Journal blog.

There is an urgent need for rapid, standardised, accurate and accessible monitoring techniques to better detect and quantify change given the increasing threat of degradation and biodiversity loss in freshwa-ter ecosystems. Community-based monitoring projects have been proven successful for the collection of meaningful biological data from a range of target species and ecosystems. The STREAM (Sequencing the Rivers for Environmental Assessment and Monitoring) project combines community-based monitoring with a DNA metabarcoding approach to assess aquatic ecosystem health by determining biodiversity of benthic macroinvertebrate species across Canadian watersheds. STREAM consists of outreach and recruit-ment, training and dissemination of results obtained from sequence data, allowing rapid generation of watershed biodiversity reports (e.g. in 2 months). We emphasise the benefits of partnering with commu-nity groups in these DNA biomonitoring efforts, highlighting the value of environmental stewardship and eliminating bottlenecks for scientific data collection. We believe the approach taken in STREAM is not only applicable to Canada, but functions as an ideal model for freshwater monitoring on a global scale. (C) 2021 Associacao Brasileira de Cieciencia Ecologica e Conservacao. Published by Elsevier B.V.

Floodplains are disturbance-driven ecosystems with high spatial and temporal habitat diversity, making them both highly productive and hosts to high biodiversity. The unpredictable timing of flood and drought years creates a mosaic of habitat patches at different stages of succession, while water level fluctuation directly influences macrophyte community dynamics, and thus habitat structure. This habitat complexity and diversity of disturbance regimes makes floodplains an ideal ecosystem in which to examine the links between biodiversity, traits and ecosystem function. With up to 90% of floodplains in North America and Europe altered to the point of functional extinction, it is particularly imperative to study and conserve those that remain intact, such as the Lower Saint John River and its associated floodplain, including the Grand Lake Meadows and Portobello Creek wetland complex. Despite the rise in trait-based science, taxonomic resolution has imposed limitations, especially in wetland and floodplain ecosystems where communities are vastly understudied compared to their riverine counterparts. Compared to traditional biomonitoring, DNA-based biomonitoring from high-throughput genomics sequencing methods is powerful in that it can reliably characterize community composition in unprecedented detail, allowing us to assess how disturbance and environmental filters interact with invertebrate traits and ecosystem function. Using structural equation analysis, we take a whole ecosystem approach to examine ecosystem health across a floodplain disturbance gradient. We focus chiefly on how anthropogenic alteration within watersheds affects downstream floodplain wetlands, how the resulting patch diversity shapes communities and, finally, how those communities influence ecosystem function through trait diversity metrics. We also examine and compare which traits are associated with crucial ecosystem gradients.

The complexity and natural variability of ecosystems present a challenge for reliable detection of change due to anthropogenic influences. This issue is exacerbated by necessary trade-offs that reduce the quality and resolution of survey data for assessments at large scales. The Peace-Athabasca Delta (PAD) is a large inland wetland complex in northern Alberta, Canada. Despite its geographic isolation, the PAD is threatened by encroachment of oil sands mining in the Athabasca watershed and hydroelectric dams in the Peace watershed. Methods capable of reliably detecting changes in ecosystem health are needed to evaluate and manage risks. Between 2011 and 2016, aquatic macroinvertebrates were sampled across a gradient of wetland flood frequency, applying both microscope-based morphological identification and DNA metabarcoding. By using multispecies occupancy models, we demonstrate that DNA metabarcoding detected a much broader range of taxa and more taxa per sample compared to traditional morphological identification and was essential to identifying significant responses to flood and thermal regimes. We show that family-level occupancy masks high variation among genera and quantify the bias of barcoding primers on the probability of detection in a natural community. Interestingly, patterns of community assembly were nearly random, suggesting a strong role of stochasticity in the dynamics of the metacommunity. This variability seriously compromises effective monitoring at local scales but also reflects resilience to hydrological and thermal variability. Nevertheless, simulations showed the greater efficiency of metabarcoding, particularly at a finer taxonomic resolution, provided the statistical power needed to detect change at the landscape scale.

Preservation of DNA in bulk environmental samples is conventionally achieved using ethanol; however, transportation restrictions on ethanol, particularly from remote locations, are problematic, and ethanol requires a lengthy evaporation period to avoid polymerase chain reaction inhibition. We examined the efficacy of an easily accessible, non-toxic, propylene glycol-based antifreeze as an alternative to molecular-grade ethanol for preserving macroinvertebrate DNA from bulk-benthos DNA samples. We used 2 processing methods (no evaporation of preservative vs full evaporation) to test the differences in both cytochrome oxidase I (COI) exact sequence variants (ESVs) and COI taxonomic orders detected in both ethanol- and antifreeze-preserved samples. Our results suggest that antifreeze is a suitable alternative to ethanol for preservation of DNA in freshly collected samples (e.g., up to 3 d) because of the comparable ESV richness detected in antifreeze-preserved samples. We have demonstrated that by using antifreeze, it is possible to achieve sufficient taxonomic coverage and assess macroinvertebrate assemblages within bulk-benthos DNA samples. The application of this non-regulated preservative is particularly important for remote sampling (i.e., only air accessible) and sampling for community-based biomonitoring projects within Indigenous territories where alcohol is prohibited or not available.

Robust data to refute or support claims of global insect decline are currently lacking, particularly for the soil fauna in the tropics. DNA metabarcoding represents a powerful approach for rigorous spatial and temporal monitoring of the taxonomically challenging soil fauna. Here, we provide a detailed field protocol, which was successfully applied in Barro Colorado Island (BCI) in Panama, to collect soil samples and arthropods in a tropical rainforest, to be later processed with metabarcoding. We also estimate the proportion of soil/litter ant, springtail and termite species from the local fauna that can be detected by metabarcoding samples obtained either from Berlese-Tullgren (soil samples), Malaise or light traps. Each collecting method detected a rather distinct fauna. Soil and Malaise trap samples detected 213 species (73%) of all target species. Malaise trap samples detected many ant species, whereas soil samples were more efficient at detecting springtail and termite species. With respect to long-term monitoring of soil-dwelling and common species (more amenable to statistical trends), the best combination of two methods were soil and light trap samples, detecting 94% of the total of common species. A protocol including 100 soil, 40 Malaise and 80 light trap samples annually processed by metabarcoding would allow the long-term monitoring of at least 11%, 18% and 16% of species of soil/litter ants, springtails and termites, respectively, present on BCI, and a high proportion of the total abundance (up to 80% of all individuals) represented by these taxa.

The Biodiversity-Ecosystem Function hypothesis postulates that higher biodiversity is correlated with ecosystem function by providing a high number of filled niches through species response types and resource use patterns. Through their high spatio-temporal habitat diversity, floodplains are highly productive ecosystems, supporting communities that are naturally resilient and highly diverse. We examined linkages among floodplain wetland habitats, invertebrate communities and their associated traits, and ecosystem function across 60 sites within the floodplain wetlands of the lower Wolastoq | Saint John River, New Brunswick, using structural equation modelling and Threshold Indicator Taxa ANalysis (TITAN2). We identified key environmental filters of invertebrate communities, namely linking increased niche differentiation through historical change, flood pulse dynamics, and macrophyte bed complexity with increased taxa and functional diversity. Examination of traits linked to ecosystem functions revealed that healthy wetlands with higher primary productivity were associated with greater functional evenness and richness, while habitat patches with increased decomposition rates had low functional richness, reflecting highly disturbed habitat. Our results highlight key differences between wetland and riverine ecosystems, relating to how critical functions support healthy wetland habitats by providing increased resilience to disturbance, here associated with differing levels of conservation protection.

Tropical forests are fundamental ecosystems, essential for providing terrestrial primary productivity, global nutrient cycling, and biodiversity. Despite their importance, tropical forests are currently threatened by deforestation and associated activities. Moreover, tropical regions are now mostly represented by secondary forest regrowth, with half of the remaining tropical forests as secondary forest. Soil invertebrates are an important component to the functioning and biodiversity of these soil ecosystems. However, it remains unclear how these past land-use activities and subsequent secondary forest developments have altered the soil invertebrate communities and any potential ecological consequences associated with this. DNA metabarcoding offers an effective approach to rapidly monitor soil invertebrate communities under different land-use practices and within secondary forests. In this study, we used DNA metabarcoding to detect community-based patterns of soil invertebrate composition across a primary forest, a 23-year-old secondary forest, and a 33-year-old secondary forest and the associated soil environmental drivers of the soil invertebrate community structure in the Maquenque National Wildlife Refuge of Costa Rica (MNWR). We also used a species contribution analysis (SIMPER) to determine which soil invertebrate groups may be an indication of these soils reaching a pre-disturbed state such as a primary forest. We found that the soil invertebrate community composition at class, order, family, and ESV level were mostly significantly different across that habitats. We also found that the primary forest had a greater richness of soil invertebrates compared to the 23-year-old and 33-year-old secondary forest. Moreover, a redundancy analysis indicated that soil moisture influenced soil invertebrate community structure and explained up to 22% of the total variation observed in the community composition across the habitats; whereas soil invertebrate richness was structured by soil microbial biomass carbon (C) (C mic ) and explained up to 52% of the invertebrate richness across the primary and secondary forests. Lastly, the SIMPER analysis revealed that Naididae, Entomobryidae, and Elateridae could be important indicators of soil and forest recuperation in the MNWR. This study adds to the increasing evidence that soil invertebrates are intimately linked with the soil microbial biomass carbon (C mic ) and that even after 33 years of natural regrowth of a forest, these land use activities can still have persisting effects on the overall composition and richness of the soil invertebrate communities.

The complexity and natural variability of ecosystems present a challenge for reliable detection of change due to anthropogenic influences. This issue is exacerbated by necessary trade-offs that reduce the quality and resolution of survey data for assessments at large-scales. The Peace-Athabasca Delta (PAD) is a large inland wetland complex in northern Alberta, Canada. Despite its geographic isolation, the PAD is threatened by encroachment of oil sands mining in the Athabasca watershed, and hydroelectric dams in the Peace watershed. Methods capable of reliably detecting changes in ecosystem health are needed to evaluate and manage risks. Between 2011 and 2016, aquatic macroinvertebrates were sampled across a gradient of wetland flood frequency, applying both microscope-based morphological identification, and DNA metabarcoding. Using multi-species occupancy models, we demonstrate that DNA metabarcoding detected a much broader range of taxa and more taxa per sample compared to traditional morphological identification, and was essential to identifying significant responses to flood and thermal regimes. We show that family-level occupancy masks high variation among genera, and for the first time, quantify the bias of barcoding primers on the probability of detection in a natural community. Interestingly, patterns of community assembly were near random, suggesting a strong role of stochasticity in the dynamics of the metacommunity. This variability seriously compromises effective monitoring at local scales, but also reflects resilience to hydrological and thermal variability. Nevertheless, simulations showed the greater efficiency of metabarcoding, particularly at a finer taxonomic resolution, provided the statistical power needed to detect change at the landscape scale.

Metabarcoding can rapidly determine the species composition of bulk samples and thus aids ecosystem assessment. However , it is essential to use primer sets that minimize amplification bias among taxa to maximize species recovery. Despite this fact, the performance of primer sets employed for metabarcoding terrestrial arthropods has not been sufficiently evaluated. Thus this study tests the performance of 36 primer sets on a mock community containing 374 species. Amplification success was assessed with gradient PCRs and the 21 most promising primer sets selected for metabarcoding. These 21 primer sets where also tested by metabarcoding a Malaise trap sample. We identified eight primer sets, mainly those including inosine and/or high degeneracy, that recovered more than 95% of the species in the mock community. Results from the Malaise trap sample were congruent with the mock community, but primer sets generating short amplicons produced potential false positives. Taxon recovery from the 21 amplicon pools of the mock community and Malaise trap sample were used to select four primer sets for metabarcoding evaluation at different annealing temperatures (40-60 Co) using the mock community. Temperature did only have a minor effect on taxa recovery that varied with the specific primer pair. This study reveals the weak performance of some primer sets employed in past studies. It also demonstrated that certain primer sets can recover most taxa in a diverse species assemblage. Thus there is no need to employ several primer sets targeting the same amplicon. While we identified several suited primer sets for arthropod metabarcoding, the primer selection depends on the targeted taxonomic groups, as well as DNA quality, desired taxonomic resolution, and sequencing platform employed for analysis.

Mixed community or environmental DNA marker gene sequencing has become a commonly used technique for biodiversity analyses in freshwater systems. Many cytochrome c oxidase subunit I (COI) primer sets are now available for such work. The purpose of this study is to test whether COI primer choice affects the recovery of arthropod richness, beta diversity, and recovery of target assemblages in the benthos kick-net samples typically used in freshwater biomonitoring. We examine six commonly used COI primer sets on samples collected from six freshwater sites. Biodiversity analyses show that richness is sensitive to primer choice and the combined use of multiple COI amplicons recovers higher richness. Thus, to recover maximum richness, multiple primer sets should be used with COI metabarcoding. In ordination analyses based on community dissimilarity, samples consistently cluster by site regardless of amplicon choice or PCR replicate. Thus, for broadscale community analyses, overall beta diversity patterns are robust to COI marker choice. Recovery of traditional freshwater bioindicator assemblages such as Ephemeroptera, Trichoptera, Plectoptera, and Chironomidae as well as Arthropoda site indicators were differentially detected by each amplicon tested. This work will help future biodiversity and biomonitoring studies develop not just standardized, but optimized workflows that either maximize taxon-detection or the selection of amplicons for water quality or Arthropoda site indicators.

Biomonitoring programs have evolved beyond the sole use of morphological identification to determine the composition of invertebrate species assemblages in an array of ecosystems. The application of DNA metabarcoding in freshwater systems for assessing benthic invertebrate communities is now being employed to generate biological information for environmental monitoring and assessment. A possible shift from the extraction of DNA from net-collected bulk benthic samples to its extraction directly from water samples for metabarcoding has generated considerable interest based on the assumption that taxon detectability is comparable when using either method. To test this, we studied paired water and benthos samples from a taxon-rich wetland complex, to investigate differences in the detection of arthropod taxa from each sample type. We demonstrate that metabarcoding of DNA extracted directly from water samples is a poor surrogate for DNA extracted from bulk benthic samples, focusing on key bioindicator groups. Our results continue to support the use of bulk benthic samples as a basis for metabarcoding-based biomonitoring, with nearly three times greater total richness in benthic samples compared to water samples. We also demonstrated that few arthropod taxa are shared between collection methods, with a notable lack of key bioindicator EPTO taxa in the water samples. Although species coverage in water could likely be improved through increased sample replication and/or increased sequencing depth, benthic samples remain the most representative, cost-effective method of generating aquatic compositional information via metabarcoding.