Swedish Sampling
Sampling took place on August 19th 2017 in the Gårdsjön research catchment and surrounding area in southwest Sweden. The catchment altitude is 100-170 m ASL, and the climate is maritime temperate with an annual precipitation for 2017 of 1145 mm (data from SMHI’s Komperöd site 49, 600 m from Lake Gårdsjön). The catchment has been extensively studied, particularly in relation to the effects of acid deposition on ecology and biogeochemistry 50,51, and both the terrestrial and aquatic components of the system have been exceptionally well-characterised 40,52. It is 12 km from the coast, and thus has been subject to moderately high levels of atmospheric acid deposition 41, the majority of which is the result of long-range pollutant transport from industry in the UK and central Europe 42. This deposition resulted in acidified lake waters 40 that have historically been managed by liming 53. All lakes within the catchment are oligotrophic and land use is forest with Picea abies (Norway spruce) and Pinus sylvestris (Scots pine) being dominant. Soils are thin and predominantly podzols, with outcrops of bedrock, and approximately 10% peat cover 40.
There are no settlements within the Gårdsjön catchment, and only two dwellings: one holiday cottage with no road access 70 m from Lake Gaffeln (a headwater lake) and the research cabin at the outflow of the catchment, with no permanent residents. Our project also considered two other lakes just outside the Gårdsjön catchment: Stora Bjurevatten and Västersjön. The surrounding land of Stora Bjurevatten is forested, but in the catchment, there are approximately 50 dwellings plus a negligible amount of farm land (~10 ha), whilst there are approximately 80 dwellings scattered immediately around Västersjön at a distance of 50-200 m from the lake, as well as a small road along one shoreline. The nearest town is 10 km to the west (Stenungsund, population = 10,000) and the city of Gothenburg (population = 600,000) is 40 km to the south.
Within the Gårdsjön catchment we collected ten surface water samples: four from streams and six from lakes. These sites comprised two locations within a headwater lake (Stora Hästevatten, which is divided into two separate waterbodies by a causeway); the other headwater lake (Gaffeln) and the stream draining it; the main inflow to Lake Gårdsjön (draining these lakes plus one additional lake, Lilla Hästevatten); the other three permanent streams flowing into Gårdsjön; and the lake outflow. We also sampled the next lake downstream of Gårdsjön, Stora Bjurevatten, and one lake in an adjacent catchment (Västersjön) with the intention of extending the trophic gradient of sampling sites to include mesotrophic systems. Lake samples were collected at outflows (including the culvert separating the divided northern and southern basins of Stora Hästevatten) where access permitted; exceptions were Lake Gaffeln, where water was sampled from an accessible shoreline location approximately 100 m from the outflow, and Stora Bjurevatten where we sampled from a point on the eastern shore. The Gårdsjön inflow sample was collected next to a bridge over the short section of channel between Gårdsjön and Lilla Hästevatten and can thus also be considered representative of the outflow from this lake. The three small inflow streams (F1, F2 and F3) were sampled at established v-notch weirs and the remaining stream sample (Gaffeln Stream) was taken from the lowest point with a defined channel before flow enters a wetland downstream of Stora Hästevatten.
All analyses on the Swedish samples were originally conducted for an experiment investigating dissolved organic matter (DOM) composition and reactivity 54. Water samples were filtered in-situ with 0.45 µm filters (pre-rinsed with lake/stream water) and stored in 100 ml glass bottles in the dark at 4°C until analysis
Siberian Sampling
Sampling took place on the 16th and 17th of July, 2017 in the Kytalyk Nature Reserve in the Indigirka River lowlands in north east Siberia. The site is 10-40 m ASL and the climate is Arctic, with annual precipitation of 232 mm and mean annual temperature of -13.8°C 55. The landscape is oligotrophic tundra with thaw-induced drained lake basins and a Yedoma ridge 35. The site has been well studied in relation to greenhouse gas emissions, vegetation communities, and soil organic matter 32,35,55.
There are no settlements in the area. There is a research station/camp that has 10-30 visitors per year. The camp consists of four cabins (two for local hunters and fishing, two for research visitors) and a small educational centre associated with the nature reserve. Rubbish and human waste are disposed of by burning with petrol. Samples were collected from the shorelines of three lakes (one large thermokarst lake and two small lakes), five ponds, runoff from thawing permafrost, and flooded tundra. At the largest lake (51 ha), three samples were collected, each from a different shoreline, but all other sites were sampled at only one location.
Siberian samples were filtered in-situ using 0.7 µm pre-combusted GF/F filters and stored in 15 ml Polypropylene tubes with PE caps. Note that no PP/PE was detected in any of these samples, thus no plastic contamination occurred from the tubes. The filter housing was made of Polysulfone (which our method does not detect), PP, and silicone, and therefore no contamination occurred during filtering. Water samples were acidified to pH < 2 with HNO3 and stored in the dark at 4°C until analysis. DOC concentrations were measured on an Aurora 1030 TOC Analyzer (OI Analytical, Texas, USA).
TD-PTR-MS analysis
For the nanoplastic analysis, aliquots of sample were filtered with 0.2 µm PTFE syringe filters (syringes were made of PE with rubber stoppers). The procedural blanks were exposed to the same potential impurities as the samples: i.e. the same vials and the same syringes and filters were used. Samples (0.5 ml) were then subjected to thermal desorption proton transfer-reaction mass spectrometry (TD-PTR-MS) using the method described in Materić et al. (2017 and 2020) 56,56. This method provides detailed information on the molecular composition of volatile and semi-volatile dissolved organic matter in the sample; it gives both the molecular weight and the concentration of molecules up to a size of ~500 m/z. Briefly, a low-pressure evaporation/sublimation process was used to remove the sample water, leaving behind residues of organic matter. Samples were then thermally desorbed by ramping the temperature up from 35°C to 350°C (at a rate of 40°C per minute) and measured on a PTR-TOF 8000 instrument (IONICON Analytik, Innsbruck, Austria). This assures complete thermal desorption of the sample as shown in thermograms of our method paper 19. Raw data of all the measurements in this work are available in the permanent reposition (see the SI). During the optimisation of the system, blanks were analysed to minimise contamination: system blanks (clean vials), dry blanks (clean vials exposed to the low-pressure evaporation/sublimation process) and ultrapure water blanks (vials containing 0.5 ml of HPLC water, filtered using the same filters used for the samples, then exposed to the low-pressure evaporation/sublimation process). Following optimisation all samples were analysed, interspersed with ultrapure water blanks (procedural blanks) every 3-4 samples. The mean signal generated from the blanks was subtracted for each ionic mass detected in the samples. The detection limit was calculated as 3-sigma of the ultrapure water blanks and ion signals below the detection limit were excluded. Thus only reliable ions were considered for further analysis. For the analysis of the Swedish samples, five samples were analysed in duplicate, whilst for the Siberian samples all samples were analysed in triplicate.
All the steps of the data analysis (including all the mass spectra, subtractions, detection limit calculation, and final mass spectra) are provided in the permanent repository (see the SI). Further details on the analysis are provided in Peacock et al (2018) 54, where the PTR-MS data were used to investigate the molecular composition of DOM.
Plastic fingerprinting and quantification
Nanoplastics detection and quantification was performed as described in Materić et al. (2020)19. TD-PTR-MS data were analysed using PTRwid 57, where the organic ion signals were integrated for 10 minutes starting when the TD oven reached a temperature of 50°C. Plastics polymers show a specific ion signal when heated up above the boiling point. PTR-ToF-MS is used to monitor these signals in real-time and at high resolution, and measures the quantity of each organic ion arising from the thermal desorption process. The ion list for each plastics type analysed in this work together with the concentrations for each ion produced are available in the permanent repository (see SI).
A mass spectrum is generated for each sample and the ions signals from the sample are compared with the signals coming from pure plastic standards, as described in previous work 19. The analysis software used 40 ions in the range m/z >100 from the library mass spectra (e.g. PET). The mass range of >100 Da is chosen to exclude lower volatility dissolved organic matter so as to increase the analytical performance in preselecting the volatility range for target polymers. The high number of ions in the fingerprint (40) assures a specific fingerprint, further improving the analytical performance of the technique.
A fingerprinting algorithm scores the similarity between the mass spectra of the sample and that of a plastic standard material. A positive fingerprint was assigned for samples with the fingerprint score significantly higher than the mean score of 1000 randomly generated spectra plus 2s (z-score > 2, p < 0.02275, one tail distribution). The fingerprint scripts used in this work, together with the outputs (scores) are available in the SI.
The concentration of the detected plastics was then calculated based on the known kinetics of the reaction chamber 58,59, as explained in our previous work 19,54,56,60.
In short, the concentrations of organic ions ate calculated according to the following 59,61,62:
where [Cppb] is the molar ratio of ions in gas phase in ppb, k is the reaction rate coefficient, t is the residence time of the primary ions in the drift tube (corrected for the ToF transmission efficiency 63), [M·H+] and [H3O+] are ion counts representing the protonated analyte and primary ions, respectively, (m/z)H3O+ and (m/z)M·H+ represent the mass-to-charge ratio of protonated water and the protonated analyte M, respectively. This step is carried out using PTRwid software 57, and the actual script is included in SI.
From deduced molar ratio values (in ppb), the concentration of each ion (in ng ml-1) was calculated as follows;
Where [Ci ng] is the concentration of the ion i in ng ml-1, mzi is the molecular mass of the particular ion in Da (given as mass-of-charge in our TOF-MS output), MV is molar volume, L is volume loaded in ml, and F is the flow of the TD unit in L min-1, and It is the integration time. After this, the concentrations of all the ions i released from the plastics were summed giving the final concentration. This step of data processing was included in the fingerprinting script, and the code is available in the SI.
Such calculated concentrations represent the amount of organic matter that is actually ionised in the system. The PTR ionization efficiency of organics is not 100% (e.g. there are losses in neutral molecular fragments and CO2 – not detected by the PTR-MS), so the values are considered as a minimum amount of the analyte 19.
Number concentrations of NPs particles are calculated from the mass concentrations, assuming 200 nm diameter (the mesh size of the filter used), spherical shape and density of 1 g/cm3, which is 238732.4 particles per ng of nanoplastics.
HYSPLIT model
Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) modelling was used to calculate footprint (trajectory frequencies) 64. We used a 48-hour backward trajectory of air movement with a frequency grid of 1.0 x 1.0 degree. We choose to use the air parametrisation rather than particle parametrisation as the nanoplastics measured by our method have an upper size of 200 nm, which we assume behave more like a gas than a PM10 particle. We used monthly footprints for the entire year prior to the sampling dates and the trajectories are available in the Supplement.
Quality control and data limitations
In this work, we took particular caution to provide conservative results of the plastic concentrations that we report. Thus several steps were taken to ensure that our concentrations are conservative: 1) We had several types of blanks to assess the potential contaminating (explained in details in the TD-PTR-MS analysis section, all mass spectra available in SI). Measured nanoplastics were in the range of 6 to 508 ng/mL, well above the detection limit, e.g. of <0.34 ng/mL (based on a PS nanoplastics standard) 19. 2) We used a strict fingerprinting method including 40 ions from the library of each plastic to assure the presence of the polymer in question; 3) the low-pressure evaporation/sublimation system was used to remove volatile organic matter from the organic-rich matrix; 4) slow thermal desorption was used for boiling point separation of the matrix, where only organic molecules of the volatility range similar to the plastics we measure in real-time and high resolution, and 5) the postprocessing quantification algorithm preserves the ion signal ratio of the plastics library, thereby preventing the overestimation caused by single or group of ions coming from the matrix. Further details are provided our methods paper 19.
The original sampling campaigns were planned for studies investigating dissolved organic matter; i.e. they were not designed with plastic analyses in mind. Thus, no field blanks (FB) were collected at the Swedish site. Two FBs were collected at the Siberian site, using the same filtering and storage procedures as for the field samples. However, the location is remote and off-grid, and there was no access to ‘fresh’ Milli-Q water. Deionized water was used instead and transported in available commercial 5L PET drinking water bottles. Each FB was analysed in duplicate. No plastics were detected in either duplicate of FB2 but one duplicate of FB1 gave a positive return for PS, with a concentration of 43 µg l-1. The PS algorithm match for this positive return was very low; just 27, compared to the matches for the field samples which were 30-65, with a mean and median of 53 and 55. We therefore cannot be entirely sure whether the signal in the FB1 duplicate is PS, or some other contaminant.
Where replicate samples were taken, agreement between them was consistently good for the Swedish samples, as shown by low standard deviations. Furthermore, if the presence or absence of a specific polymer was detected in any sample, the result (presence or absence) was the same for both replicates (Table 1). In contrast, standard deviations were larger for the Siberia samples, and there was a lower consistency between replicates; e.g. presence and absence of specific polymers was sometimes found for different replicates of the same sample.