Two Canadian Arctic ports were used to investigate the effect of tidal dynamics on eDNA detection in contrasting coastal environments. Churchill was historically Canada’s principal port on the Arctic Ocean (Chan et al., 2012), located in a substantial estuary at the mouth of the Churchill River where it enters Hudson Bay (Manitoba, Canada, 58°46′N, 94°11′O). Milne Inlet is a primarily marine port, located at the northern end of Baffin Island (Nunavut, Canada, 72°15′N, 80°30′W) and currently subject to the greatest shipping activity in the Canadian Arctic (Babin et al., 2019; Radtke et al., 2023). Churchill shows characteristics consistent with an estuarine ecosystem whereas Milne Inlet corresponds to a marine habitat as evidenced by representative water quality profiles collected during the various stages of the tide cycle in these two respective ports ( Supplementary Figure 1 ). The tides are semidiurnal in both ports with amplitude ranges of 1.25 m and approximately 1 m in Churchill and Milne Inlet, respectively (Fissel, 1982; Prinsenberg and Freeman, 1986; Kleptsova and Pietrzak, 2018).

Water samples (250 ml each) from both locations were collected in the nearshore subtidal zone (10–20 m depth at chart datum) using a Niskin bottle deployed from a boat. Samples were collected at the surface and ~1 m from the bottom at each of three sites per port during the low, middle and high tides, over three consecutive days (Churchill: August 10 to 12, 2016; Milne Inlet: August 18 to 20, 2017) for a total of 54 water samples for each port ( Supplementary Table 1 ).

Each water sample was filtered within <2 h of collection using a syringe (BD 60 mL, Kranklin Lakes, NJ, USA) equipped with a filter head containing a 0.7 μm glass microfiber filter (Whatman GF/F, 25 mm). Filters were preserved at 4°C in 700 µl of Longmire’s lysis preservation buffer within a 2 mL tube (Wegleitner et al., 2015) and then frozen at -20°C until DNA extraction. To test for background contamination in the field, a negative control (250 mL sterile distilled water) was filtered for every 10 water samples collected (total of N=8). Moreover, to reduce the risk of cross-contamination, individual and sterile sampling kits were used for each sample. These sampling kits, composed of bottles and filter housing sterilized with a 10% bleach solution and new sterilized gloves as well as syringes and tweezers, were bagged and sealed then exposed to UV for 30 minutes following assembly.

Meticulous care was taken to avoid contamination risk during the DNA extraction and library preparation. In brief, (i) eDNA extraction, PCR preparation, and post-PCR steps were performed in three separate rooms; (ii) All PCR manipulations were performed in a UV hood; (iii) samples within a specific port were all treated together but processed in a randomized order; (iv) all laboratory bench space and tools were bleached with a 10% solution and exposed to UV for 30 minutes before performing any manipulations.

DNA from filters was extracted using a QIAshredder and phenol/chloroform protocol as described in Lacoursière-Roussel et al. (2018). Three negative control extractions were performed with 950 µl sterile distilled water [one for every 23 samples (i.e., one batch)] and were treated as normal samples for the remaining manipulations until sequencing.

To maximize detection of biodiversity and reduce experimental biases associated with the PCR primers, four different universal primer pairs were used to amplify the eDNA. Two pairs of mitochondrial cytochrome c oxidase subunit I (COI) primers were used (mlCOIintF/jgHCO2198 and LCO1490/ill_C_R) amplifying 313 bp and 325 bp respectively (Folmer et al., 1994; Geller et al., 2013; Leray et al., 2013; Shokralla et al., 2015). Additionally, two other pairs in the ribosomal 18S gene were also used to amplify eDNA in the V4 region (the primers F-574/R-952 and TAReuk454FWD1/TAReukREV3) which amplify regions of 378 bp and 399 bp respectively (Stoeck et al., 2010; Hadziavdic et al., 2014). The primer genes 18S and COI have been shown to be complementary with respect to types of detected taxa (Zhan et al., 2014; Leduc et al., 2019). They are commonly used together with the goal of detecting large numbers of metazoan taxa present in the sampled environment and helping to reduce the effects of primer bias (Kelly et al., 2017). Details of the eDNA amplification using a one-step dual-indexed PCR approach with Illumina barcoded adapters are provided in Lacoursière-Roussel et al. (2018). Three PCR replicates were done for each sample and primer pair combination. The 12 resulting aliquots were pooled in equal molar concentrations after purification using Ultra AMPure beads and quantification by PicoGreen. Because barcodes were different for each sample, a negative PCR control was performed for each sample and each primer pair. Although PCR negative controls were checked but not sequenced, the 16 negative controls from the field and extraction steps were treated as regular samples and sequenced. Sequencing was carried out for each port separately using two separate Illumina MiSeq runs (Illumina, San Diego, CA, USA) at the Plateforme d’Analyses Génomiques (IBIS, Université Laval, Québec, Canada; www.ibis.ulaval.ca).

Preparation and analysis of the raw sequencing was done using the metabarcoding pipeline Barque version v1.5.2 (www.github.com/enormandeau/barque). Adaptor and primer sequences were first removed, then raw sequencing reads were demultiplexed into individual samples files using the MiSeq Control software v2.3. Raw forward and reverse reads were then quality trimmed and merged using several criteria. Contigs were split following their primer pairs (COI or 18S) then sequences with lengths outside the expected range and chimeric sequences were removed. The SILVA and BOLD databases were used to classify the sequences with a 0.03 cutoff level (http://v3.boldsystems.org/ and https://www.arb-silva.de/). Further details on data cleaning are described in Sevellec et al. (2021).

Sequences from non-marine species (insects, birds and most mammals) as well as those that could not be taxonomically assigned were removed from our dataset ( Supplementary Table 2 ). Moreover, sequences with multiple hit assignment were annotated to the genus level and the species was left as “sp.”. More details about the removal of species and the management of multiple hits are described in Lacoursière-Roussel et al. (2018). Results were highly similar between primers (i.e. COI1 and 2 vs 18S1 and 2; see Supplementary Tables 3 , 4 ). Thus, the sequences were summed between COI1 and COI2 as well as between 18S1 and 18S2 for each taxon within each sample. Hereafter, the terms COI and 18S correspond to the combined gene primers (COI1+COI2 and 18S1 + 18S2, respectively).

Finally, to limit potential bias in the dataset, taxa were removed if the total number of sequences detected in negative controls (field and PCR; N=16) was greater than 2% of the total number of sequences detected across all samples for a given genus or species (Leduc et al., 2019; Sevellec et al., 2021). Thus, we removed ten (18S) and five (COI) genera as well as eight (18S) and six (COI) species from the remaining analyses for both ports ( Supplementary Table 5 ). Furthermore, for both gene primers, one sample from Milne Inlet and one sample from Churchill with no final annotated marine sequences were removed from the data set.

All analyses were performed with the combined sequences for the four gene primers (18S and COI) to document the change in eDNA metabarcoding based on: (i) tides, (ii) depth in the water column, (iii) daily temporal variation (iv) and interactions between these factors in two different ports (Churchill and Milne Inlet) representing contrasting coastal ecosystems (estuary and marine). The analyses were done with the vegan R package in R studio (Dixon, 2003) using the genus or species level matrix with two different statistical transformations: the Hellinger transformation (abundance score) and presence-absence transformation (presence/absence score) (Legendre and Gallagher, 2001). Although the eDNA abundance may be ascertained with a metabarcoding approach, the comparison between abundance and presence/absence scores can provide some reliable relative abundance score results (Deiner et al., 2017). PERMANOVAs (number of permutations = 10,000) were performed on the eDNA communities at the genus level. To document the variation in dominant detection rates among the low, middle and high tides, pie charts based on relative abundance were generated for both ports. The diversity based on inverse Simpson index and richness (number of species) was generated from the genus matrix with Hellinger transformation. These biodiversity indices were then sorted and plotted according to the tide, water column depth and day for comparison between ports. We also used a generalized linear model (GLM) with a Gaussian distribution followed by analysis of variance (ANOVA; number of permutations = 999) to test for differences in inverse Simpson index among the above-described groups. Similarly, a simple ANOVA (number of permutations = 999) was performed to compare richness among these groups.

To investigate the ecological status of species observed in both port ecosystems, taxa were classified as holoplanktonic or meroplanktonic, freshwater or marine mammals. Histograms were then constructed from abundance scores with a Hellinger transformation for each tide, sampling depth and days. Canonical correspondence analysis (CCA) was used to illustrate the correlation between environmental factors (temperature, salinity and pH) and detected eDNA communities for each port. CCA graphics were generated using presence-absence and Hellinger scores at the genus level which were plotted according to tide and depths for each port and gene primers. ANOVAs were performed on the environmental variables (number of permutations = 999) in R to test for correlations between environmental factors and eDNA communities. Finally, balanced accumulation curves, based on randomized subsampling, were used to estimate the change in rate of detection at the genus level depending on the number of sampling days, sampling depth and different tide levels using the BiodiversityR package and the function “balanced.specaccum” (Kindt and Coe, 2005).

Canonical correspondence analysis (CCA) highlighted three distinct clusters representing the different tides in the estuarine ecosystem (i.e., Churchill) using both gene primers ( Figure 4 ; Supplementary Figure 4 ). These clusters followed a trend of continuity from low tide to high tide in both surface and bottom waters. In surface waters, low and middle tide community structure, were generally associated with higher water temperatures (and lower salinities), whereas high tide communities were associated with greater salinities (and somewhat lower temperatures) for both genes (COI: Figure 4 , Table 3 ; 18S: Supplementary Figure 4 ; Supplementary Table 6 ). In contrast, in bottom waters, higher temperatures (and lower salinities) were only associated with low tide community structure (with COI only; Table 3 ), while greater salinity and bottom depth (and lower water temperatures) were strongly associated with middle and high tide community structure, respectively. Unlike temperature and salinity, the pH factor did not have a significant effect on community structure. In the marine ecosystem (i.e., Milne Inlet), there were no distinct clusters according to the tide and no correlation with local environmental factors was observed in surface water (COI: Figure 4 ; Table 3 ; 18S:  Supplementary Figure 4 ; Supplementary Table 6 ). In bottom waters, there was a slight significant correlation of salinity and the temperature with community structure (with COI only; Table 3 ), but these correlations were not linked to any distinct clusters by tide (COI: Figure 4 ).

As shown by the accumulation curves of Figure 5 , sampling at multiple days, depths and tides can further increase detection of taxa (genus) in both estuarine and marine ecosystems. However, this improvement in detection was found to be more pronounced in the estuarine system. Interestingly, in the estuarian ecosystem, the greatest improvements in taxa detection by sampling at different days and depths were achieved for mid and low tides (see slopes in Figure 5 for the upper four panels, Figure 5 ).

A strong effect of tide was observed in the estuarine community structure detected with eDNA, whereas there was no change in composition within the marine ecosystem. Both vertical and horizontal eDNA transport may explain the observed differences in the influence of tides on communities in estuarine versus marine environments. Although this study is the first to analyze the effect of tides on metazoan taxa detections with eDNA metabarcoding in an estuary ecosystem, Ahn et al. (2020) used eDNA metabarcoding to investigate the effect of incoming and outcoming tides on detection of fishes in five different estuaries of Japan. They also found that tidal movements had a strong influence on community composition. Indeed, in estuarine ecosystems, horizontal eDNA displacement through tides and currents may be especially strong, which would increase eDNA dispersion (Hansen et al., 2018; Rourke et al., 2022), particularly in the presence of larger, faster flowing rivers such as the Churchill River (Deiner and Altermatt, 2014; Pont et al., 2018; Cole et al., 2022). In marine ecosystem, previous studies have reported similar results with weak or no influence of tides on taxa detection using eDNA metabarcoding (Kelly et al., 2018; Jeunen et al., 2019; van Bleijswijk et al., 2020; Lafferty et al., 2021). The physical hydrodynamics associated with coastal morphology in marine areas, e.g., in inlets or fjords, can lead to slower and multi-directional flows which might be significant drivers limiting horizontal eDNA dispersion (Hansen et al., 2018; Rourke et al., 2022). Except for certain free-floating cells such as eggs, eDNA would likely sink rapidly through the water column with gravity in the absence of significant mixing or horizontal forces, explaining the higher diversity with depth in marine ecosystems (Hansen et al., 2018; Ratcliffe et al., 2021). Therefore, particles released in estuarine, marine or freshwater communities might vary depending on different sources of eDNA (e.g. nekton/benthos, phenology such as reproduction, feeding, nesting, resting) related with seasonality or time of the day in addition to different fates and dynamics associated with the local mixing and horizontal forces.

Variation in eDNA production rates may also contribute to differences in tidal effects on detection of estuarine and marine communities. In this study, we observed the strongest detection of the intertidal gastropod, Littorina saxatilis, at high tide in the Churchill River estuary. During the rising tide, the intertidal zone is submerged with tidal flood currents, allowing a potential accumulation of eDNA released by intertidal species. In contrast, the eDNA signal from intertidal species was generally found to be weaker at low and middle tide, likely due to the tidal ebb current (Jeunen et al., 2019). Thus, while in marine ecosystems the eDNA may be transported over shorter distances through the low-level influence of tides and currents and rapidly settle in the sediment, resuspension of eDNA from sediments may be possible in estuarine ecosystems (Hansen et al., 2018; Jeunen et al., 2019; van Bleijswijk et al., 2020).

In estuarine ecosystem, the eDNA from freshwater taxa may flow from the river over long distances and likely contributes to increasing the detected diversity at the river mouth (Deiner and Altermatt, 2014; Wood et al., 2021). Given that the strength of riverine inputs varies with tides, this could lead to complex spatio-temporal patterns of diversity. Our results reveal a shift in community composition from low tide to high tide that is characterized by a reduction of freshwater species, and an intermediate community structure at mid-tide.

A strong effect of depth on detected community structure was observed in both the estuarine and marine ecosystems, regardless of tides. While several studies have reported differences in eDNA communities between surface and bottom waters (Moyer et al., 2014; Yamamoto et al., 2016; Andruszkiewicz et al., 2017; Yamamoto et al., 2017; Lacoursière-Roussel et al., 2018; Uthicke et al., 2018; Jeunen et al., 2020; Littlefair et al., 2021), none have considered the potential dynamic interaction with tides. The vertical distribution of eDNA is well-described and known to have a consistent pattern with distinct communities being found over depth ranges of meters (Chappuis et al., 2014; Deiner et al., 2017). Indeed, water column stratification is a main physical process restricting eDNA vertical dispersion, while other factors, such as wave action (Jeunen et al., 2020) and biological transport from natural species behavior (e.g. daily vertical migration and predator–prey interactions), may increase eDNA dispersion through the water column (Hänfling et al., 2016; Easson et al., 2020; Rourke et al., 2022).

In the marine ecosystem of Milne Inlet, the diversity and the richness were higher in bottom waters, whereas in the estuarine ecosystem (i.e. Churchill), the surface samples tended to possess a greater biodiversity, an observation that is consistent with previously documented patterns in the Churchill estuary (Lacoursière-Roussel et al., 2018). In estuaries, lower density freshwater coming from the river accumulates in the surface layer of the estuarine water column (Prandle, 2009). As a result, eDNA transported by upstream freshwater into estuaries can explain the greater species diversity measured in these ecosystems through eDNA metabarcoding (Yamamoto et al., 2017). In marine ecosystem such as Milne Inlet, where there is an absence of stratification in the water column ( Supplementary Figure 1 ), the non-floating eDNA would be expected to lead to rapid downward vertical advection, leading to greater detected biodiversity at the bottom. In addition, abiotic factors such as temperature, salinity, hydrodynamics may also influence food availability, and thus influence the species distribution across the water column. Although there can be seasonal variations in vertical distribution of organisms, zooplankton and benthic species are generally distributed deeper in the water column rather than near surface waters (Pados and Spielhagen, 2014; Bandara et al., 2016). These observations may also explain the greater richness detected with eDNA in the bottom waters of Milne Inlet.

We found that the eDNA community structure was distinct among the three sampling days in both the marine and estuarine ecosystems. Although the importance of temporal variation to understanding eDNA community dynamics is widely agreed upon, only a few studies have analyzed daily variation (Kelly et al., 2018; Ely et al., 2021; Jensen et al., 2022). This eDNA pattern is consistent with natural daily distribution of species. Ely et al. (2021) noted high variability in species observations on a day-to-day basis with many taxa being observed in all samples whereas the vast majority of species were only intermittently detected. This dynamic aspect to communities detected by eDNA may, in part, be explained by biological variation such as daily behaviors and variation of highly mobile or migratory species (Jensen et al., 2022). For example, Holm-Hansen et al. (2019) found that abundance of different types of fish (demersal, benthic and pelagic) varied with time of day and could explain observed diurnal differences in fish community structure with eDNA metabarcoding (e.g., as observed by Suzuki et al., 2022). Furthermore, for some taxa, the persistence times of eDNA in dynamic marine environments have been shown to be as short as a few hours rather than days, as previously noted in temperate freshwater samples (Jo et al., 2019). Such rapid degradation in marine environments may strongly influence taxa detection, leading to an incomplete picture when communities are characterized using eDNA samples from a single sampling site or time point. Indeed, Kelly et al. (2018) observed that temperature and salinity changes caused by the movement of water masses influenced the daily variation in eDNA-based community composition. To overcome this inherently heterogeneous distribution of eDNA in the water (Jensen et al., 2022), sampling over multiple days or short-term replication within a day would maximize taxa detection rates, thus leading to a better overall representation of actual biodiversity.

Our findings clearly demonstrate the importance of sampling eDNA at multiple time points within a day or over several days to help reduce the impact of stochasticity associated with sub-day and day-to-day variation. Moreover, we found substantial improvement in detected taxa rates when sampling over multiple tides in marine ecosystem but especially in estuarine ecosystem, suggesting that sampling more than one tide should maximize the chances of detecting taxa. Our results also show the need for considering tides as well as depth in water eDNA metabarcoding sampling designs for coastal biodiversity monitoring and how the relative importance and interaction between these factors can vary among ecosystems. Based on our findings, the first considerations for maximizing eDNA detection of biodiversity in marine environments should be choosing the appropriate depth and tide ( Supplementary Table 7 ). Indeed, sampling closer to the bottom of the water column is likely to provide more robust estimates of coastal biodiversity. In estuarine conditions, the effect of tides on detected biodiversity was found to be stronger and is thus of primary importance in sampling design, and collecting samples at low and mid tides have a greater likelihood of generating higher biodiversity estimates. Our findings also show that, in an estuarine setting, surface samples contain a greater diversity of detected upstream (freshwater) and local marine taxa, whereas a lower diversity of local marine and brackish taxa were observed at the bottom of the water column. For studies that aim to achieve broad spatial coverage, we suggest that the inclusion of samples from both portions of the water column in estuaries can substantially increase the numbers of detected taxa that are distributed over a greater spatial area, including upstream portions of the estuary. Thus, sampling a broader range of physical conditions across multiple tides and depths may be particularly important in order to ensure accurate representation of biodiversity in estuarine environments which tend to be more dynamic (Nagarajan et al., 2022).

We demonstrate that sampling over consecutive days is particularly important for obtaining more comprehensive estimates of biodiversity in an estuarine setting when using eDNA metabarcoding. Provided resources allow for it, eDNA sampling and integration of associated biodiversity information across broader temporal scales, such as over different seasons, should provide even more accurate representations of the full breadth of diversity in a given location (Sevellec et al., 2021). We thus recommend that future studies explore how the dynamics of eDNA detection related to tides and depth may also vary between estuarine and marine systems in different seasons and regions.