Wet and dry deposition of MPs >11 µm
In the studied year, only a small amount of MP particles >11 µm in the shape of fragments were detected in all monthly collected wet deposition samples and in 5 out of 12 dry deposition samples (Table 1). In total, 8 particles found in dry deposition samples and 13 particles in wet deposition samples were assigned to common synthetic polymers, therefore, all statements based on particle count should be taken with some caution. The average particle size in wet deposition samples (50±37 µm, mean±SD) was slightly larger than the average particle size in dry deposition samples (45±21 µm, mean±SD) but not significantly different (t-test, p>0.05). MP particles <100 µm were observed both in dry and wet deposition samples, whereas particles larger than 100 µm were identified in wet deposition samples only. Calculated dry deposition velocities ranged from 0.05 to 0.2 m s-1.
Table 1 MP fragments detected in wet and dry atmospheric deposition samples using µFTIR down to 11 µm
Sample
|
MPs detected in wet deposition
|
MPs detected in dry deposition
|
Plastic type
|
Max. [µm]
|
Min. [µm]
|
Colour
|
Plastic type
|
Max. [µm]
|
Min. [µm]
|
Colour
|
Dry deposition velocity [m s-1]
|
May-19
|
PP
|
44
|
22
|
white
|
n. d.
|
n. d.
|
n. d.
|
n. d.
|
|
Jun-19
|
PBT
|
11
|
11
|
grey
|
n. d.
|
n. d.
|
n. d.
|
n. d.
|
|
|
PP
|
60
|
27
|
grey
|
|
|
|
|
|
Jul-19
|
PP
|
34
|
27
|
grey
|
n. d.
|
n. d.
|
n. d.
|
n. d.
|
|
Aug-19
|
PP
|
22
|
11
|
transp.
|
PP
|
33
|
22
|
grey
|
0.061
|
|
PE
|
130
|
50
|
grey
|
PP
|
33
|
33
|
transp.
|
0.065
|
|
PP
|
111
|
45
|
transp.
|
|
|
|
|
|
Sep-19
|
PE
|
22
|
11
|
transp.
|
PVC
|
11
|
11
|
grey
|
0.053
|
|
PE
|
56
|
35
|
transp.
|
PP
|
66
|
45
|
transp.
|
0.222
|
|
PP
|
27
|
19
|
white
|
PP
|
44
|
11
|
transp.
|
0.056
|
Oct-19
|
PP
|
70
|
43
|
grey
|
PP
|
45
|
25
|
transp.
|
0.063
|
Nov-19
|
PE
|
11
|
11
|
grey
|
n. d.
|
n. d.
|
n. d.
|
n. d.
|
|
Dec-19
|
PE
|
45
|
45
|
grey
|
SI
|
46
|
32
|
grey
|
0.066
|
Jan-20
|
n. a.
|
n. a.
|
n. a.
|
n. a.
|
n. d.
|
n. d.
|
n. d.
|
n. d.
|
|
Feb-20
|
n. a.
|
n. a.
|
n. a.
|
n. a.
|
PC
|
81
|
62
|
black
|
0.228
|
Mar-20
|
n. a.
|
n. a.
|
n. a.
|
n. a.
|
n. d.
|
n. d.
|
n. d.
|
n. d.
|
|
Apr-20
|
n. a.
|
n. a.
|
n. a.
|
n. a.
|
n. d.
|
n. d.
|
n. d.
|
n. d.
|
|
Max. and Min. – maximum and minimum dimension of the MP particle; transp. – transparent; n. d. – not detected, n. a. – not analyzed; PP – polypropylene; PE – polyethylene; PBT – poly(butylene terephthalate); PVC – poly(vinyl chloride), SI – silicone; PC – polycarbonate.
Close to MP sources, efficient gravitational settling of MP particles >100 µm should be expected under dry conditions, while transport distances of these large particles are generally small. Brahney et al. (2020) noted similar observations for deposition to protected areas in the US. The plastics deposited under dry conditions were smaller and potentially subjected to long-range transport [21]. However, it is accepted that wet deposition favours the scavenging of smaller particles of the same chemical composition [16]. No fibres, but plastic fragments only were detected in wet and dry deposition samples, which is consistent with findings in the Hamburg metropolitan area, where 95 % of all detected MPs in the total deposition were in the form of fragments [28].
In the period from May 2019 to April 2020, calculated number wet DFs varied between 5 MPs m-2 day-1 and 19 MPs m-2 day-1 (10±5 MPs m-2 day-1, mean±SD), whereas dry DFs ranged between 0 and 23 MPs m-2 day-1 (5±7 MPs m-2 day-1, mean±SD) (Fig. 1a). Note that wet deposition samples from January to April 2020 are missing due to technical problems, while all 12 monthly dry deposition samples are considered. In 5 out of 8 monthly measurements, the wet DFs of MP exceeded the corresponding dry DFs. However, in all four months when MP particles were identified both in dry and wet deposition samples (Aug, Sep, Oct, Dec), wet and dry DFs were similar. Overall, the wet deposition from May to December was 62 % of the total (wet + dry) MP number atmospheric deposition. This value is in the range of the typical estimate of one-half of the total input of pollutant surface flux by wet deposition (18) and the estimated loss of submicron particles by wet deposition of greater than 80 % used in atmospheric models [29].
The seasonal variation of dry deposition was similar to that of wet deposition. Among the wet/dry deposition measurements from May to December, the highest MP total DFs were observed in August and September.
MP concentrations in precipitation varied from 2 to 18 MPs L-1 (Fig. 1b), reflecting higher concentrations in August and September. The lowest concentration of 2 MPs L-1 was found in May, and the precipitation weighted mean concentration was 7±5 MPs L-1 (mean±SD). Thus, based on our samples, the estimated annual plastic wet deposition at the study site comprises 4284 MPs m-2 a-1. Daily precipitation sums can be found in Supplementary information Fig. S1.
Mass DFs showed similar trends (SM, Fig. S3). For the detected MP particles that exist as supermicron (coarse) particles (>11 µm), wet deposition dominates the total mass flux at the study site, comprising 70 % of the total plastic mass deposition (from May to December 2019).
Correlation of MP DFs and concentrations in precipitation with meteorological conditions and population density
Dry and wet DFs are affected by a multitude of factors. Dry DFs depend on particle deposition velocities, small-scale meteorological effects near the surface, surface characteristics, and atmospheric stability and friction velocity, which are influenced by variables such as particle size, wind speed, and temperature [30]. Coarse atmospheric particles in the diameter range larger than 10 µm mainly deposit by gravitational settling and impaction or interception [31]. In contrast, the scavenging efficiency by wet deposition was found to be related to the duration and intensity of the precipitation event, as well as to the hydrometeor size distribution, the settling speed, the average collection efficiency, the shape of the hydrometeors and aerosol hydrophilicity [32–34]. Moreover, Wu et. al (2018) found that removal of particles smaller than 2.5 µm by wet deposition to be more efficient than by dry deposition [35].
We investigated the relationship between dry and wet DFs, concentrations in precipitation, and meteorological variables by means of Spearman rank correlation analysis. MP dry DFs did not correlate significantly with temperature, wind speed, RH, and with none of the other studied meteorological parameters (Table 2, expanded table in Supplementary information, Table S2). Furthermore, wet DFs did not show significant correlations with precipitation amount, intensity, or duration but showed a strong negative correlation with the maximal dry period, i. e. the longest continuous period without precipitation during sampling intervals (rs = -0.96, p = 0.0009). This points to low accumulation capability for MP particle >11 µm in the atmosphere. Coarse MP particles tend to sediment faster and since no positive correlation between dry DFs and maximal dry period was observed, MP sources might have local and temporally irregular emission patterns.
Table 2 Correlation between meteorological variables and microplastic dry, wet, and total DFs, and concentrations in precipitation*
Parameter
|
MP dry DF
[MPs m-2 day-1]
|
MP wet DF
[MPs m-2 day-1]
|
MP total DF
[MPs m-2 day-1]
|
MP concentration in precipitation [MPs L-1]
|
Average dry period [days]
|
-0.56
|
-0.21
|
-0.42
|
-0.12
|
Maximal dry period [days]
|
-0.40
|
-0.96
|
-0.61
|
-0.68
|
Average wet period [h]
|
-0.09
|
-0.58
|
-0.40
|
-0.87
|
Maximal wet period [h]
|
0.16
|
-0.28
|
-0.05
|
-0.58
|
Precipitation [mm]
|
0.29
|
-0.41
|
-0.20
|
-0.78
|
Wind speed [m s-1]
|
-0.04
|
-0.19
|
0.02
|
-0.55
|
Temperature [°C]
|
0.17
|
0.09
|
0.03
|
0.34
|
Relative humidity [%]
|
0.17
|
0.07
|
0.17
|
-0.09
|
*Spearman rank correlation coefficients between meteorological variables and MP DFs and concentrations in precipitation: p < 0.05 (in red), p> 0.05 (in black)
MP concentrations in precipitation showed negative correlation with the precipitation amount and average wet period (Table 2). This indicates dilution of precipitation samples and efficient MP scavenging by precipitation.
The dominant wind direction for the study period in relation to the MP total DFs is shown in Fig. 2. The wind rose plot for the study site and timeframe indicates that the prevailing wind direction was NE and SSW (see also Supplementary information, Fig. S4). In May and July 2019, and in March and April 2020 the dominant wind was blowing from the NE, and in the other months from the SSW. During the dominance of SSW wind, we collected 92 % of all detected MP particles in the total (wet + dry) deposition, 86 % of all MP particles in wet deposition, and 100% of MP particles in dry deposition samples. The city centre of Kassel is SW of the sampling site; thus, the dominant SSW winds may provide the most probable vector for local MP pollution at the sampling site. Calculated deposition velocities of detected plastic particles in dry deposition samples ranged from 0.05 to 0.2 m s-1 (see Table 1), indicating relatively short-range transport phenomena. This supports the idea that MP pollution (>11 µm) to the study site was brought from the nearest urban centre.
Since single-point local wind fields can be misleading and can differ from the regional wind conditions, air mass origin was also evaluated by calculating 24-hour back trajectories for the study site and sampling periods. Anthropogenic influence on MP DFs may be estimated by averaging population densities along air mass back trajectories during each sampling period. Fig. 3 shows the monthly evolution of MP DFs (dry DF, wet DF, total DF), of MP concentrations in precipitation, and of the trajectory-averaged population density during the sampling periods. MP concentrations in precipitation were higher when air masses arrived from more densely populated areas and correlated significantly with the trajectory-averaged population density (rs = 0.91, p = 0.002, Supplementary information, Table S3). Thus, anthropogenic activities appear to contribute to airborne MP abundances.
Polymer type
In total, six different polymer types were found in wet and dry deposition samples by µFTIR, and polypropylene (PP) particles predominated in both deposition modes (Fig. 4).
Minor fractions of the polymer-type distribution differ among the deposition modes. In wet deposition samples, polyethylene (PE) and poly(butylene terephthalate) (PBT) particles were detected, whereas particles made of poly(vinyl chloride) (PVC), polycarbonate (PC), and silicone-based compounds (SI) were detected in dry deposition samples only. Since MP particle sizes were similar in wet and dry deposition samples, we cannot explain the observed variation due to particle size. If MP particles would be scavenged by cloud droplets in the free troposphere, the particles observed in precipitation might differ from those removed from the boundary layer by dry deposition [36]. However, we propose that variable MP emission patterns and the low number of detected MP particles are the main factors leading to compositional differences among the minor components in wet and dry deposition samples. A study by Brahney et al. in 2020 [21] suggests that wet-deposited MPs originate from different source regions than those that are dry deposited (studied particle size ≥ 4 µm). Wet plastic DFs were significantly correlated to population metrics, as determined by the intersection of the air mass with population centres, whereas dry deposition correlated negatively with regional dust DFs and suggests that dry-deposited plastics are subject to large-scale, global transport. Regarding wet deposition, our findings are consistent with the work by Brahney et al.
PE particles scavenged by precipitation comprised 38 %, but no PE particles were detected in dry deposition samples, while PP particles were observed in similar amounts in both wet and dry deposition. We expect that dry deposition mechanisms for PP and PE particles do not differ strongly and that particles of all studied sizes are removed by dry deposition independent of chemical composition. PE and PP polymers have similar densities, and they were detected as fragments in both removal modes, so significant differences in settlings velocities due to shape are not expected. However, we expect that all atmospheric MP particles will quickly age and form surface coatings, e. g. by condensation of low-volatile compounds. The surface properties of ambient plastic particles can thus change significantly compared to the native polymer, making particle deposition difficult to assess. Moreover, airborne MP originated from marine or freshwater systems for example via bubble-bursting or raindrop impacts (e.g. [37]) possess most likely reduced hydrophobicity due to an eco-corona increasing the density and surface charge of particles
[38, 39], or a biofilm [40], which typically features a rather sticky matrix of extracellular polymeric substances
[41] and may support adhesion to other airborne particulates. The ability to form larger particle aggregates, which may support dry deposition due to the increased settling velocities, is likely similar to both plastic types. Thus, we believe that temporal variation in plastic emission is the crucial factor determining the composition of plastics in wet and dry deposition samples. However, further research is required to investigate the potential role of chemical composition of plastic particles, aging degree as well as size in wet and dry removal mechanisms.
Exemplary Raman analysis of dry deposited MP
It may be expected that MP particle numbers are underestimated by the sole application of µFTIR having a detection limit around 10 µm since several studies confirmed an increase in airborne MP particle numbers with decreasing particle size [42, 43]. Unfortunately, most of the current analytical techniques used for MP identification are limited with respect to particle size.
[43]. Detecting plastic particles in the size range most relevant for inhalation (< 10 µm) is time-consuming and expensive [2]. Therefore, the vast majority of studies focusing on airborne MP use techniques with an analytical threshold of particle diameters >10 µm. Only a few methods are suitable for the detection of polymer fragments in the lower µm-range (between 10 µm and 1 µm), submicron range (between 1 µm and 100 nm), and ultrafine or nano range (≤ 100 nm), and up to date lack the routine methodology [44]. Currently, by using Raman imaging the smallest MP particles measured in a real sample were down to 100 nm in size [45]. FTIR and Raman spectroscopy are the most frequently applied methods in MP studies, and it has been suggested to use them in tandem for complete and reliable chemical characterization of microplastics [46]. Even though both methods are vibrational spectroscopic methods, they complement each other and may provide different numbers and polymer types of detectable MPs. In this study, Raman spectroscopy was applied in order to gain insight into the abundance of dry deposited MP particles in the size range <11 µm, i. e. particles which may enter the respiratory system via inhalation and may therefore affect human health [47].
In total, 1361 particles were found over the selected area (Fig 5, area 1), employing the WITec ParticleScout particle-analysis tool of which 3 were assigned to plastics, i. e. polydimethylsiloxane (PDMS), poly(vinyl chloride) (PVC), and polypropylene (PP) (Fig. 5b). Thus, MP particles composed 0.2 % of all analysed particles in this 3 x 3 mm2 area (see also Supplementary information Fig. S5, S6). In comparison, scanning the entire particle loaded filter area by using µFTIR, only one PP particle (Fig. 5a, position 4) was detected (see also Supplementary information, Fig. S7).
Two identified particles made of PDMS (14 µm × 4 µm) and PP (9 µm × 9 µm) (Fig. 6a,c) fell in the size range where µFTIR is no longer applicable for reliable identification, but a slightly larger PVC particle (21 µm × 15 µm) (Fig. 6b) was likely overlooked by µFTIR. As suggested by Käppler et al. (2016) PVC particles seem better detectable by Raman spectroscopy compared to FTIR. This is potentially due to the characteristic relatively broad C–Cl stretching vibration at 690 cm−1, which was not detectable due to the limited spectral range of the FPA detector (4000–900 cm−1) in µFTIR in their study, or categorised as another chemical especially when containing high amounts of plasticisers [46]. Most of the other particles showed Raman spectral patterns typical for inorganics i. e. black carbon (soot), quartz, titanium dioxide, and dolomite. Identified organic non-plastic particles were assigned to cellulose and bacteria (see Supplementary information, Fig. S6).
Extrapolation of these additional results to the total effective filter surface (7.85 × 10-5 m2), sampling surface (8.01 × 10-3 m2) and collection time (35 days) gave a total dry DF of 207 particles per m2 per day instead of 17 MPs m-2 d-1 as calculated from µFTIR analysis only. Thus, FTIR analysis compared to Raman analysis may lead to underestimation of MP DFs by an order of magnitude; however, uncertainties due to the subsample analyses and extrapolation should be considered. Scanning a larger number of particles would give a more robust estimation of the dry DF. It was shown by other studies that FTIR imaging may lead to underestimation of MP number by about 35 %, especially in the size range <20 μm compared to Raman imaging
[46], and automated single-particle exploration coupled to μ-Raman (ASPEx-μ-Raman) quantified two-times higher MP numbers in the size range <500 µm compared to FTIR imaging [48]. This suggests that Raman analyses might typically yield higher deposition fluxes compared to FTIR imaging.
The large area Raman scan with a step size of 2 µm pixel-1 (Fig. 5a, area 2 (500 x 500 µm2), enclosing PDMS MP particle) is presented in Fig. 7. As shown in Fig. 7b (also Supplementary information, Fig. S8), a large number of single particles in the lower micron range were detected. However, comparison of the particles’ Raman spectra to the Raman spectra of the most common synthetic polymers did not result in a reliable particle assignment to plastics.
Moreover, large area Raman scans with a step size of 0.5 µm pixel-1 of three randomly selected 100 x 100 µm2 quadrants did not result in plastic particle assignment in the submicron range (component distribution from Raman imaging and respective Raman spectra for one quadrant are given exemplarily in Supplementary information, Fig. S9). Submicron-range particles made of quartz, calcite, soot, and most probably amylose or another polysaccharide were identified. If particles of the most common synthetic polymers would be present, the spectral assignment would be achieved (also in the submicron range). It is worth mentioning that only 0.04 % of the sample area were scanned with a step size of 0.5 µm pixel-1, and therefore, the analysed subsample is not representative. However, scanning only 1 % of the filter area would require the analysis of 78 quadrants and >500 h of instrumental time. With the applied spatially resolved detection of submicron particles it would take several days to scan a representative area of the sample. However, our filtration technique distributes the particles on the filter equally, and if a vast amount of submicron plastic particles would be deposited on the sampling area, we would be able to see at least some of them.
Our results suggest that atmospheric dry deposition samples do not contain a large number of fine plastic particles >500 nm. This is consistent with the lowest deposition velocities of atmospheric particles in the so-called accumulation range from 100 nm to 1 µm [49]. Even if there is a large number of submicron MP particles in the atmosphere, reduced dry DFs must be expected. Further Raman mapping analyses of aerosol samples collected by active pump sampling would give new insights into the abundance of airborne submicron plastic particles.