Atmospheric MP found at PDM
MP samples were collected between June 23, 2017 and October 23, 2017 (15 × 7 day durations). MP fragments or fibres were found in all samples analysed during the summer–autumn monitoring period. Active aerosol sampling presented MP counts of 0.09–0.66 MP/m3 at the PDM sampling platform, with an average of 0.23 MP/m3 (standard deviation ±0.15) (Fig. 1). In total, 13 of the 15 samples presented MP counts >0.1 MP/m3, with 4 samples presenting >0.33 MP/m3 (the upper 25th percentile).
Samples show a tentative association between the proportion of MP <10 µm and the fragment content of the sample. Fibres were found to generally occur in greater quantities in the larger MP size fraction, while the smaller size fraction contained a greater proportion of fragment shaped particles (r = 0.65, P < 0.05). On average, 51% of the MP particles fell within the smaller particle fraction (MP ≤ 10 µm) (21–74% range, standard deviation ±17%), with almost all MP particles characterised as ≤20 µm (aerodynamic diameter) (96%, standard deviation ±0.1). Due to their small size (primarily ≤20 µm), particles were primarily identified as fragments or fibres (70%, 30%, respectively), with fibres defined as presenting a 3:1 length to width ratio22,23. The larger MP particles (MP <30 µm) identified were fibres, presenting a maximum aerodynamic diameter of 53 µm. The smallest spectrally characterised particle within this study was 3.5 µm (limit of quantification adopted for this study was 3.5 µm).
MP particles were comprised of polyethylene (LD/HDPE), polystyrene (PS), polyvinylchloride (PVC), polyethylene terephthalate (PET) and polypropylene (PP) (in order of abundance 44%,18%,15%,14%,10%). The polymer type did not correlate significantly with fibre or fragment quantities in the samples or their relative particle-size distribution, with all samples presenting a mix of the five polymer types analysed in the study. This suggests that polymer type or density may not be a key variable influencing MP occurrence at PDM within this study period (however source availability is potentially important).
The wind velocity (monitored at the site for the sampling duration) ranged from 1.4–22.6 m/s (mean 7.9 m/s ± 3.6 m/s), with the stronger winds (≥10 m/s) occurring predominantly from the W to SW (71% of the monitoring period). There is notable intra-sample variation in wind velocity resulting in no clear local wind velocity trend relative to total MP counts. All sample periods present both peak wind velocities for short periods above 10 m/s and calmer wind periods (<5 m/s). Larger MP particles and fibres (MP > 10 µm) show a trend with the maximum recorded wind velocity from the northerly projection (rs = 0.79, P < 0.05). Stronger wind velocities in general (from any direction) appear to tentatively occur with higher MP >10 µm counts (larger MP) (rs = 0.61, P < 0.05). This may suggest regional wind velocity could be a transport vector for larger MP at PDM but further long-term monitoring and shorter sampling time step may be necessary to elucidate this potential association.
There appears to be limited correlation or comparable trends between the local meteorological atmospheric conditions monitored (air temperature, relative humidity, precipitation or wind direction) and the total MP particle counts found for the corresponding monitoring periods (P ≥ 0.05 for all datasets). MP found in the PDM samples, collected at high altitude, cannot easily be attributed to any clear or specific local influence. The occurrence of MP, especially smaller MP (<10 µm), at PDM may therefore be a result of more complex atmospheric transport, mixing and distal source influence.
Air mass and particle history and long-distance transport of MP arriving at PDM
Atmospheric transport and particle dispersion models can be used to consider the possible trajectory and dispersion of atmospheric air masses and particulates. The models use reanalysis of meteorological data which allows high confidence in estimating mixing (ground collision/influence) and local weather effects enroute. However, assumptions for particle transport characteristics (density, dry deposition rate, wet removal and scavenging etc.) have to be made as microplastics have limited field validation data. Air-mass back trajectory and particle dispersion analyses were run for 7 days (168 h), long enough to consider possible distal sources and trajectory elevation decline (a commonly used duration adopted in backward modelling for long-distance analysis)17,24,25. Back trajectories were created for each hour during the monitoring periods, resulting in 1713 model runs with hourly latitude, longitude and elevation data points. These hourly releases (back trajectories) over the full 24 h provide for the inclusion of any diurnal (day/night) influences on the trajectory elevation (including the change in PBL/FT elevation and associated convection activity). This was complemented by dispersion modelling for the same period (particle dispersed from PDM field location for the active daily sampling period, and backward tracked for the 168 h period).
The back-trajectory and dispersion modelling was overlaid onto the corresponding surface levels to illustrate historic air mass/particle altitude above surface level (ASL) (Fig. 2 and Supplementary Fig. 1). On average, the air mass/particle dispersion backward trajectories maintained an elevation of >2000 m ASL and over the 168 h monitoring period travelled a minimum of 275 km from the PDM sample site.
In general, sample periods with elevated MP quantities (MP > 0.33 MP/m3, upper 75th percentile) illustrated atmospheric transport at lower elevations than samples with lower MP particle counts (MP > 0.33 MP/m3 average back-trajectory elevation = 2747 ± 373 m ASL; MP < 0.33 MP/m3 average back-trajectory elevation = 3276 ± 425 m ASL, MP < 0.13 MP/m3 (lower 25th percentile) average back-trajectory elevation = 3123 ± 158 m ASL) (Fig. 2a, b). Correspondingly, the percentage of backward modelled trajectories and dispersion illustrating FT/PBL mixing (the point where the trajectory or particle fell within the PBL elevation above surface level) was notably greater for samples presenting higher MP counts (MP > 0.33 MP/m3 average percentage of trajectories falling within PBL = 9% ± 6%; MP < 0.33 MP/m3 average percentage of trajectories falling within PBL = 2% ± 1%). The minimum FT atmospheric transport duration and distance (without any PBL influence) prior to reaching PDM was also notably greater for the lower 25th percentile samples (MP < 0.13 MP/m3, 887 km (Supplementary Fig. 4) compared to the upper 75th percentile (MP > 0.33 MP/m3, 343 km). This tentatively suggests that long-distance FT transported MP forms a potentially greater proportion of MP particles for MP < 0.13 MP/m3 samples and more proximal MP (PBL/FT mixing closer to PDM) form a comparatively greater proportion of the MP > 0.33 MP/m3 samples. It is noted that while the MP > 0.33 MP/m3 samples backward air/particle modelling illustrates a greater frequency of PBL/TF mixing in general, the maximum PBL/FT mixing frequency at any one point in time was <30% and on average 33% of all trajectories (range = 15–57%) included PBL/FT mixing at some point prior to arriving at PDM. This is notably higher (and expresses a greater difference than the general PBL/FT mixing frequency) than the equivalent calculated for the MP < 0.33 MP/m3 samples air mass/particle transport histories, where on average <8% (range = 0–18%) of trajectories showed PBL/FT mixing at any one point in time and <12% of all trajectories illustrated any PBL/FT mixing. This reinforces the theory of frequency of PBL/FT mixing being a potentially influential factor of elevated PDM atmospheric MP concentration (MP/m3).
When individual samples are considered, only three of the sample periods illustrate modelled backward trajectories meeting surface levels (0 m A.S.L) (A2, A8 and A9) within the modelled 168 h period (a minimum of 114, 122 and 96 h prior to arriving at PDM). Of these, sample A2 and A8 present the highest two atmospheric MP concentrations of the total dataset, while A9 is one of the lower MP concentrations. Five sample periods (A2, A8, A9, A12 and A15; a mixture of elevated and low MP concentration samples) show air/particle histories to be elevated above 50 m ASL for the entire 168 h backward modelling period. This suggests that while proximity to surface level is an important consideration in identifying atmospheric MP sources, the PBL entrainment and PBL/FT mixing is complex and an important atmospheric MP consideration for elevated (FT) observation and sampling sites.
There is a positive correlation between the frequency or percentage of modelled back trajectories falling within the PBL (PBL/FT mixing) and the quantity of MP in the resultant PDM samples. Higher MP concentrations at PDM appear to occur when a greater the number of back trajectories undergo PBL/FT mixing prior to reaching the PDM sampling location (MP counts r = 0.69, P < 0.05; MP counts >10 µm r = 0.78, P < 0.05) (Supplementary Information S5). All samples with the exception of A6 (the lowest MP > 10 µm count sample, 0.066 MP/m3) illustrated atmospheric transport that included PBL/FT mixing within the modelled backward 168 h period (free tropospheric transport for greater than 168 h). However, while the frequency of back trajectories with PBL/FT mixing is a potential influence on PDM atmospheric MP concentration, the duration of time the overall trajectories (air mass/particle) spend within the PBL relative to each sample is not significantly correlated to the PDM atmospheric MP concentration. Despite this lack of statistical correlation, there is a visual trend suggesting a possible link between greater average duration of back-trajectory occurrence within the PBL and an elevated MP particle count in the PDM samples (Supplementary Information S5).
Back-trajectory modelling illustrates an average air-mass movement of 4550 km from PDM over the 168 h modelled period (2047–6631 km average trajectory distances for samples A1–A15). The shortest distance travelled (in a straight line from PDM) is 275 km (A1, average trajectory elevation 2775 m ASL) while the longest distance is 10,212 km (A13). Backward modelling of samples with MP < 0.33 MP/m3 travelled an average of 4992 km (±1097 km), 1660 km further than samples with MP < 0.33 MP/m3. While the overall trend appears to suggest lower atmospheric MP concentration samples at PDM occur in concurrence with greater modelled atmospheric transport pathways, a statistically significant correlation is not evident in this dataset. This is suggested to be due to individual influence of PBL mixing and trajectory elevation above surface-level occurrence along the trajectory rather than the overall atmospheric transport distance.
The projection of these air mass/particle histories is generally westerly or southerly, across the Atlantic Ocean towards North America or across the Mediterranean Sea towards Northern Africa (Fig. 3a–h, Supplementary Fig. 2 and Supplementary Fig. 3). It is noted that all atmospheric backward modelling from PDM suggest long-distance transport (>100 km)6.
All sample periods illustrate an Atlantic Ocean (westerly) trajectory influence. Samples with a higher MP quantity (MP > 0.33 MP/m3) show a greater proportion of trajectories over the Mediterranean Sea and Northern Africa (52%) compared to lower MP samples (MP < 0.33 MP/m3) (21%) (Fig. 3a, b). Comparatively, backward air/particle history modelling presented a greater proportion of trajectories across the Atlantic Ocean and over North America (53%) for samples with MP > 0.33 MP/m3 compared to samples with higher MP particle counts (34%). When lower trajectory elevations are considered (Fig. 3c, d), elevated MP samples were found to have a greater number of trajectory points below 500 m ASL over the Mediterranean Sea and North Africa compared to lower MP samples (MP > 0.33 MP/m3 = 73%, MP > 0.33 MP/m3 = 55%). Correspondingly, there is a positive trend between the larger particle-size MP findings (MP > 10 µm) and the number of Northern African trajectories in each sample period (Log10(MP > 10 µm) r = 0.8, P < 0.05). This North African influence alongside the Mediterranean influence is of interest given these areas are identified as presenting high plastic concentrations26,27. This mapping suggests that that Northern Africa and the Mediterranean may be areas of potential MP entrainment and influence to the air mass and MP particles at PDM (for the monitored duration).
The elevated MP content in the PDM samples appears to be influenced by both the number of back trajectories that fall within the PBL (frequency) and the location over which the PBL mixing occurs in conjunction with the proportion of the sample that is >10 µm in particle size (Supplementary Data and Supplementary Fig. 5). Sample A2, A8, A14 and A15 (samples with MP > 0.33 MP/m3) have an elevated number of modelled back trajectories that fall within the PBL (>10 trajectories points), PBL mixing points over Northern Africa, central and western Europe, the Mediterranean Sea (Fig. 3e, f). The modelling suggests that samples A2 and A8 MP may be transported from Northern Africa, the Mediterranean Sea, Europe and the Atlantic Ocean. Sample A14 appears to primarily have PBL mixing occurring over Northern Africa while A15 back-trajectory modelling suggests Northern African and Atlantic Ocean PBL-mixing points.
Samples with <0.33 MP/m3 (A1, 3–7, 9–13) generally have limited the Mediterranean Sea or European PBL-mixing points and a higher proportion of trajectories falling within the PBL over the Atlantic Ocean. The trend between the percentage of trajectories falling within the PBL correlates with both the MP count and MP > 10 µm (r = 0.7, P < 0.05, see Supplementary Information). While there is notable variance within the PBL-mixing locations, there is also positive trend between the number of PBL-mixing points over land and the number over ocean/sea (MP/m3 and land PBL-mixing frequency: r = 0.74, P < 0.05, R2 = 0.7; MP/m3 and sea PBL-mixing frequency: r = 0.81, P < 0.05, R2 = 0.69). This does not suggest land to be a greater MP source influence, with regard to PBL/FT mixing occurrence, and there is a possibility the MP transported into the FT through mixing occurring over the marine environment may be both terrestrial or marine MP8,28. This is confirmed by the potential emission sensitivity (PES, Fig. 3g, h) outputs, illustrating elevation particle occurrence over Europe complemented by extensive PES over the Atlantic, Northern Africa and extending to North American continent. Rather than suggesting that terrestrial or marine PBL-mixing locations of are of greater or lesser importance, this tentatively suggests that it is the overall frequency of PBL mixing may be an important influence on atmospheric MP quantities at PDM during this sample period.