A first assessment of microplastics and other anthropogenic particles in Hudson Bay and the surrounding eastern Canadian Arctic waters of Nunavut

Surface water, sediment, and zooplankton samples were collected from onboard the CCGS Amundsen in July and August 2017 from Hudson Bay and the Canadian Arctic Archipelago of Nunavut (Fig. 1; Table 1). Surface water samples were collected from 22 stations, and zooplankton and sediment samples were collected from 20 stations. Sampling was done opportunistically aboard the vessel at sites where sampling was permitted. At some sites it was only possible to take one type of sample due to scheduling or feasibility. In addition, snow samples were collected from one site near the research station in Alert, Nunavut, in the spring of 2018.

In this study, the ship covered a range of sites across a large geographical area. Ice levels varied throughout sampling, with areas towards the pole tending to be more ice heavy and southern areas less so. The samples taken in Baffin Bay are influenced by Atlantic origin waters and glacier melting (Oksman et al. 2017). In the Hudson Bay, there are several rivers that drain into the area, none of which flow through large populated areas.

At each station a surface water sample was collected with a triple-rinsed metal bucket lowered from the port side of the ship with a metal wire and transferred to a clean 40-L stainless-steel soda keg prior to filtration onboard the vessel. Immediately after collection, samples were filtered onto a 142-mm diameter polycarbonate filter (10-μm pore size, MilliporeSigma, Burlington, Massachusetts, USA). If the samples were turbid and the filter began to clog, <40 L of water was used for the sample. Each filter was transferred to a clean glass petri dish (142-mm diameter), sealed with tape and stored at room temperature until analysis (roughly one year). To avoid contamination of microplastics during sampling, the soda keg and bucket were rinsed with surface water three times before collection. The filter holder was rinsed between each sample with 1 L of reverse osmosis (RO) water that had been filtered through a 10-μm polycarbonate filter. To account for any procedural contamination, three field blanks were randomly sampled at different times during the cruise. Field blanks were sampled identically to field samples by lowering RO water down the side of the boat in the bucket, transferring it to the soda keg, running it through the filter and sealing it in a clean petri dish for future analyses. Samples were shipped to the University of Toronto, Toronto, Ontario, to be analysed.

At the University of Toronto, the filters were sonicated for 1 h to remove particles in precleaned jars with RO water. After sonication, the sample was placed in the oven at approximately 50 °C for 24 h to dissolve any salts that may have been present. The sample was refiltered through a 47-mm diameter, 10-μm pore size polycarbonate filter. Anthropogenic particles (down to ∼100 μm) were extracted visually under a dissecting microscope. The particles were picked off the filter with tweezers and placed on double-sided tape. Particles were categorized according to shape and colour. They were then photographed and measured using ImageJ (Schneider et al. 2012) software. To determine how successful we were at identifying anthropogenic particles, and to get an idea of the types of materials in our samples, Raman spectroscopy (Xplora Plus; Horiba Scientific with LabSpec 6 software; New Jersey, USA) was conducted on a random number generated selection of 14% of the total particles aiming for 10% of each colour morphology combination (e.g., blue fiber, white fragment) per sample. Following Raman spectroscopy, we used the data from the analysis to normalize the final particle counts per sample to report an anthropogenic particle count (which includes microplastics) and a microplastic particle count. We include the category “anthropogenic particle” to include dyed cellulosic textiles (e.g., cotton), dyed textiles of unknown material, and other materials such as paint where the Raman signal gave us a dye spectrum only. We also report a microplastic count, which includes particles we can confirm as plastic via Raman spectroscopy.

Zooplankton were collected using a Tucker trawl with two individual sampling nets. One of the nets had a mesh size of 750 μm, with openings of 1 m². The Tucker is an oblique tow pulled alongside the boat that collects zooplankton as it descends to approximately 10 m above the sea floor. The net was held for 1 min before being brought slowly back to the surface. Zooplankton were picked from the samples with metal forceps and placed into filtered RO water with 10% isopropyl alcohol. The samples consisted of multiple species that varied across samples. Examples of organisms include chaetognaths and krill. To account for any procedural contamination, five field blanks were collected. Field blanks consisted of RO water in clean glass jars that were left open during sampling. Fixed zooplankton samples were stored at room temperature until analysis (roughly one year). Samples were shipped to the University of Toronto to be analysed.

Zooplankton samples were poured from the jars they were stored in through a 500-μm stainless-steel mesh sieve, where they were rinsed using RO water. This was done to remove any noningested microplastics from the sample. Zooplankton were analysed in bulk, and as samples varied greatly in the number (some having <10 and others having >50), and species, their wet weight, in grams, was recorded. Because it was not possible to count zooplankton in samples with more than 50 very small organisms, we do not report this value. Samples were then photographed and transferred to a clean jar that had been tripled rinsed with RO water. They were then covered in a 20% KOH solution in RO water, capped, and held at room temperature for 7–14 d, or until the zooplankton were fully digested. Each sample was then sieved using a 45-μm stainless-steel sieve to remove the KOH solution and retain the microplastics (Lusher et al. 2017). Anthropogenic particles (down to ∼100 μm) were extracted visually under a dissecting microscope. The particles were picked off the filter with tweezers and placed on double-sided tape. Particles were categorized according to shape and colour. They were then photographed and measured using ImageJ (Schneider et al. 2012) software. Using the same method as the surface water samples, we analysed 13% of the particles, based on our subsampling strategy described above. Following Raman spectroscopy, we used the data from the analysis to normalize the final particle counts for each sample to report an anthropogenic particle count (which includes microplastics) and a microplastic particle count.

Twenty sediment samples were collected using a box corer. A clean metal trowel was used to scrape the top 5 cm (measured with a ruler) of the sediment. Surface sediment was placed in glass jars and capped with aluminium foil under Teflon-lined lids. A field blank consisted of RO water in a clean glass jar that was left open during collection of one sediment sample. The sediment samples were processed in a sample separation facility at the University of Western Ontario.

Each sample was wet sieved using a 20-μm stainless-steel sieve to remove the size fractions finer than medium silt. This step was required to avoid flocculation of very fine particles during drying. Samples were emptied onto aluminum pie trays, covered with aluminum foil, and dried at 70 °C in a drying oven. Each sample was weighed and then emptied into a glass beaker containing 250 mL of sodium polytungstate (SPT) solution with a specific gravity of 1.5 g/cm³. The samples were magnetically stirred for 5 min and were then transferred to glass separatory funnels. Additional SPT solution was used to rinse the glass beaker to ensure a quantitative transfer. Once the sediment settled completely, the nonbuoyant material could pass through the stopcock into a 750-mL glass beaker. The buoyant material was then drained into a separate 750-mL glass beaker containing a glass conical funnel lined with filter paper. The separatory funnel was rinsed with SPT and drained onto the filter paper. Using RO water in a squirt bottle, the material on the filter paper was transferred onto a 53-μm, 7.5-cm diameter polycarbonate–polyester sieve. The material was rinsed, transferred using the squirt bottle from the sieve to a glass petri dish, covered with aluminum foil, and redried at 70 °C. One sample (A19) was composed entirely of shells and was processed simply by pouring the material into an aluminum pie plate, rinsing the jar with RO water, pouring the water into the same pie plate, and allowing the water to evaporate in the oven. The sample was then weighed and placed into a petri dish for microscopic examination.

The material in each petri dish was examined using a dissecting microscope to separate microplastics from any remaining sediment and organic material. The particles that were visually identified as microplastics were categorized according to their colour, morphology (fibre, film, fragment), and size. Each particle was removed from the petri dish using tweezers and placed in a glass vial labelled according to sample number. A total of 29 (13%) particles identified as potential microplastics were randomly selected and analyzed using micro-Fourier Transform Infrared Spectroscopy (micro-FTIR) at Surface Science Western, University of Western Ontario, London, Ontario. Following FTIR spectroscopy, we used the data from the analysis to normalize the final particle counts for each sample to report an anthropogenic particle count (which includes microplastics) and a microplastic particle count. Because sediment samples were analysed at the University of Western Ontario and surface water and zooplankton were analysed at the University of Toronto, chemical identification methods differed due to access to different equipment.

Seven snow samples (∼450 mL melted volume) were collected from Alert, Nunavut, near the Dr. Neil Trivett Global Atmosphere Watch Observatory. Untouched snow, upwind from the station and sampler, was collected to limit procedural contamination. Two field blanks were also sampled to account for procedural contamination during sampling. Snow was collected into a clean glass (rinsed 3× with RO water) jar using a clean stainless-steel spoon. The field blank was collected by leaving a clean jar open during sampling. Samples were shipped to the University of Toronto for analysis.

All melted snow samples were processed in a clean cabinet to prevent contamination from dust. The volume of each sample of melted snow was measured using a clean graduated cylinder and then filtered onto a 47-mm, 20-μm polycarbonate filter (Merck Millipore Ltd.). After filtration, the filter was transferred into a clean petri dish. Anthropogenic particles (down to ∼100 μm) were extracted visually under a dissecting microscope. Using the same method as both surface water and zooplankton samples, particles were photographed and measured using ImageJ software. All particles underwent Raman spectroscopy (Xplora Plus; Horiba Scientific) with LabSpec 6 software (Schneider et al. 2012). We did not subsample due to the small number of particles in total.

Throughout the study, consistent efforts were made to minimize sample contamination. Bright orange flotation suits were worn while collecting samples on board the CCGS Amundsen, and cotton laboratory coats were worn in both laboratories. In Rochman’s laboratory (Univeristy of Toronto), all materials used were made of stainless steel or glass and were rinsed a minimum of three times with RO water prior to use. When collecting all samples, clothing type and colour were noted. In the laboratory, samples were filtered under a clean cabinet to prevent contamination. Prior to using the microscopes, the area was wiped down to reduce contamination. For surface water samples, three field blanks were collected on the CCGS Amundsen. Lastly, one laboratory blank for surface water, and two for zooplankton, were included during laboratory extraction and processing. All blanks were extracted by following the sample protocol as outlined above, using RO water instead of real samples. These samples were then analysed under a dissecting microscope and particles were picked and transferred to double sided tape. Particles found in blanks from the field and the laboratory were subtracted from samples according to morphology and colour. This was done by averaging the field blanks and then the laboratory blanks separately for each matrix type and then adding them together to remove this total value from each sample within that matrix (e.g., 10 blue fibres were removed from each surface water sample). See Table S1 for data related to blank subtraction in each sample and sample type.

In Corcoran’s laboratory (University of Western Ontario), all materials were made of stainless steel, aluminium, or glass, except for the small polycarbonate–polyester sieve used to transfer material from the filter paper to the glass petri dish. The dissecting microscope had a metal enclosure along the sides and top of it, protecting the petri dish from airborne particles. Corcoran’s laboratory has two portable air filtration systems with HEPA filters, one for the processing laboratory and one in the microscope lab. The filter papers were composed of white translucent cellulose and a minor amount of clear resin. Squirt bottles were made of clear polyethylene. Cellulose fibres derived from the paper were easily identified when found in the samples and were not included in the counts. No clear fragments that might represent resin were identified in any of the samples. The samples were covered with aluminium foil while stored in aluminium pie plates, beakers, and separatory funnels. When wet sieving, cross contamination between samples was avoided as the sieve was cleaned in an ultrasonic bath filled with RO water for 1 h between sieving each sample. The sediment sample was then immediately poured into the sieve then placed under RO running water and gently stirred with the flat part of a metal tablespoon until the clay and fine silt fraction had seeped through the sieve. The amount of time that each sample was exposed to the air (during wet sieving and transfer from one vessel to another) is estimated to be no longer than 30 min. The average time required for visual analysis was ∼1 h. Two laboratory blanks were sampled; one from the processing laboratory for a period of 30 min, and one from the microscope laboratory for a period of 1 h. The laboratory blanks were sampled using a petri dish containing a small mixture of pre-examined sedimentary material. The blanks were examined under a dissecting microscope and any particles resembling microplastics were counted. The number of plastic particles in the field blank and the laboratory blanks were subtracted from the samples according to morphology and colour. Table S2 contains initial totals, types, and colours of microplastic particles as well as final totals following blank subtraction and normalizing to FTIR results.

Statistical analyses were completed in Excel 2016 (version 16.13.1) and R Studio (version 1.0.143). We used simple linear regression to measure the relationship between the concentration of microplastics sampled from each station and the population size nearby and upstream of the station location. Population data came from the most recent or available census data (Statistics Canada. 2020a, 2020b). If the sample site was within 100 km of a community and the water current flowed towards the sample site, then the population from that community was used and included (Canadian Coast Guard, 2019). If a site was not within 100 km of a community or was upstream of a current, then a value of 0 was used; 100 km was selected as it was deemed local enough to influence plastic quantities, whereas anything beyond this could be seen as a long-distance source and therefore not local.

To assess the relationship between microparticles among media, a nonmetric multidimensional scaling (nMDS) plot was created. We used nMDS to examine patterns of assemblage structure of microplastics (i.e., by particle morphology—fibre, fragment, film, and foam) among matrices. For nMDS plots, we made two-dimensional ordinations using Euclidean distances. We used the function “metaMDS” in the vegan community ecology package (Oksanen et al. 2009) in R.

Following blank subtraction, microplastics and other anthropogenic particles were found in 19 of 21 (90%) samples. After normalization following Raman spectroscopy, the anthropogenic particle concentration in the samples ranged from 0 to 0.61 particles per litre (Fig. 2A). All particles were categorized as either fibres or fragments. Fibres were more abundant in samples, comprising 98% of all particles found in surface water (Fig. 2A) where the remaining 2% were fragments. The predominant colour of particles was blue, making up 72% of the particles found. Roughly 50% of particles were <1 mm in size (Fig. S1). Overall, 86% of the particles analysed with Raman were anthropogenic. Anthropogenic particles are defined here as having been created or processed by humans, including dyed cellulosic fibres. Of these, 29% were definitively plastic—which included polyester and polypropylene (PP). The mean number of microplastics per sample was thus 0.07 ± 0.08 (per litre). Because of band overlay from the dyes used in textiles or due to fluorescence, anthropogenic fibres of an unknown base were the most dominant category found (Fig. 4). This is a common issue for coloured microfibres (Lenz et al. 2015). Nonanthropogenic, natural fibres comprised ∼14% of particles analysed. The densities of microplastics found in surface water samples were not consistently buoyant and ranged from lower density plastics such as polyurethane to higher density plastics such as polyvinyl chloride. Across all samples, we found that there was not a significant relationship between population density and the quantity of anthropogenic particles found in surface water (p = 0.694) (Fig. 5).

Following blank subtraction, microplastics and other anthropogenic particles were found in 18 of 20 (90%) samples. The anthropogenic particle concentrations in the samples ranged from 0 to 16 particles per gram of zooplankton (wet weight; Fig. 2B). All particles were categorized as either fibres, fragments, film, or foam. Fibres were the most abundant, consisting of 92% of particles, followed by fragments (6%), films (1%), and foam (one particle was found in all samples). Similar to surface water, the most predominant colour of particle found in zooplankton samples was blue, making up 68% of particles found. Additionally, ∼42% of particles were <1 mm in size (Fig. S1). Overall, 96% of the particles subsampled for Raman were anthropogenic. Again, anthropogenic fibres with unknown base were the most dominant category (Fig. 4). The remainder of the particles were composed of natural material (4%). Plastic particles consisted of 20% of all analysed particles. Following Raman spectroscopy, data were normalized similarly to surface water (Fig. 2). The mean number of microplastics per sample was 0.7 ± 0.9 (per gram). The densities of microplastics found in zooplankton samples were also not consistent, but they tended towards higher density plastics such as polytetrafluoroethylene. Across all samples, we found no significant relationship between population density and the quantity of anthropogenic particles found in zooplankton (p = 0.712). (Fig. 5).

Following blank subtraction, microplastics and other anthropogenic particles were found in 18 of the 20 sites (90%). After blank subtraction, we found at total of 106 fibres, 38 fragments, and 42 films. The mean number of particles per gram of dry weight sediment (g⁻¹ dw) was 2 g⁻¹ dw (Table S2). All particles were characterized as fibres, fragments, or film (Fig. 2C). Fibres were more abundant in samples, comprising 57% of all particles found in sediment, with fragments comprising 20% and film the remaining 23%. The fragments ranged in size from 0.28 to 300 μm, and a variety of colours were identified. From most to least abundant, fragment colours were blue, grey, red, orange, purple, clear, black, green, and white. Fibre lengths ranged from 205 to 380 μm, and from most to least abundant the colours identified included blue, black, grey, purple, pink, white, yellow, and green. The 42 films were identified only in the sample from station Belcher Glacier and they were all the same colour (cobalt blue) with sizes ranging from 0.25 to 125 μm (Fig. 3D). The films were very soft and easily broken into smaller size particles.

The results of FTIR analysis showed that of the 23 fibres analysed, 11 were composed of polyethylene terephthalate (PET) and the remaining 12 of cellulose (all dyed either blue, black, purple, or green) (Fig. 4C). Three of the five fragments analysed are synthetic polymers (PP, polyurethane, polystyrene-PS). The composition of the remaining two fragments is unknown, but both are dyed blue. The cobalt blue film analysed from sample 101 is a prepolymer, and as it appears to be the same colour and texture as the remaining 41 blue films in the same sample, we infer these all to be prepolymers. These results indicate that 48% and 66% of the fibres and fragments, respectively, are composed of plastic. The mean number of microplastics per sample was 1.66 ± 3.67 (per gram) and of the total particles, 52% were plastic. The densities of microplastics found in sediment samples were not consistent, and predominantly had lower density plastics, including polystyrene and polyurethane. Across all samples, we found no significant relationship between population density and the quantity of anthropogenic particles found in sediment (p = 0.521).

After blank subtraction, microplastics and other anthropogenic particles were found in one out of seven (14%) sampling sites. In total, there were four particles that were picked because they appeared anthropogenic, three of which were determined to be natural. The remaining particle was polyethylene. All particles were categorized as fibres. Owing to the small sample volumes, and thus low plastic counts above the noise, it is not practical to discuss patterns of size.

To look at patterns across sites, we made heat maps showing the anthropogenic particles at each station for each media (Fig. 1). Figure 1 allows for a comparison of concentrations of anthropogenic particles across media. It is worth noting that samples were not taken for all matrices at all sites. For example, there were no sediment samples taken in Hudson Bay, one of the most populated areas in the study area. For the data we have, generally there is not a consistent pattern across all media. For example, at site 736 we see relatively high concentrations in surface water and relatively low concentrations in zooplankton. Still, we observe higher quantities of anthropogenic particles near Belcher Glacier across all media and Ungava Bay (sites 684, 682, 676, 670) for surface water and zooplankton.

For a closer look at patterns among locations and across media we constructed an nMDS plot (Fig. 6). The plot considers both the number of anthropogenic particles as well as the morphologies of particles within each sample. It groups together samples with similar “microparticle communities” based on both amount and morphology. Here, we see a pattern suggesting that zooplankton samples are different than sediment and surface water. Sediment and surface water do not separate from each other in space on the plot. Zooplankton had the most diverse morphologies of any sample type, having a mixture of fibres, fragments, foam, and film. This may be the reason for the pattern that we observe.