Introduction

Soils are a major sink for microplastics (MPs, <1000 µm) during their global migration through the environment1,2 leading to high MP concentration in soils worldwide3,4. Once introduced to soil systems, MPs can influence physicochemical soil properties5, biogeochemical processes6, soil microorganisms7 and, as a consequence, plant performance and key chemical transformation processes (e.g., C and N cycling)8.

Previous research has focused mainly on agricultural or urban soils, where high MP emissions from direct sources, such as plasticulture, fertilization, and littering, are expected9. Forests, as a key type of terrestrial ecosystem and covering 31% of the global land surface10, remain as a blind spot in our understanding of terrestrial MP occurrences11. Previous studies have demonstrated the presence of MPs in forest soils in the Republic of Korea with mean concentrations of 160 particles per kg soil dry weight (p kg–1)12, primary and secondary rainforests in China with mean concentrations of 630 p kg–113, neotropical rainforest and pine plantations in Mexico with mean concentrations of 1500 p kg–114, and alluvial forests in Serbia with mass concentrations ranging of 0–6 g kg–1 15. However, these studies are limited because they did not measure small sizes of MPs (<40 µm), have no comprehensive polymer identification, and mostly focus on mineral forest soil and negating organic soil horizons formed from leaf litter. In contrast to, for example, agricultural soils, forest mineral soils are covered by one or more organic soil horizons formed by biogeochemical turnover processes of leaf litter. Up to three horizons form, consisting of minimally decomposed (Oi), intermediate (Oe), and strongly (Oh) decomposed material on top of mineral soil horizons (e.g., A and B horizons)11.

In the absence of direct MP sources, atmospheric deposition is thought to act as the major source of MPs to forest ecosystems. Forest filter effects and the trapping of atmospheric particulate matter are already known for trace metals16 and organic compounds (e.g., DDT, PFAS, PAHs)17,18. For MPs, mean MP throughfall depositions of 365 p m–2 d–1 and a majority of MP particle sizes of <50 µm were found in remote forests in the Pyrenees (France)19, whereas other studies found mean MP throughfall depositions of 331–512 p m–2 d–1 during short-term20 and 147 p m–2 d–1 (during long-term sampling in near-urban forests (Germany)21, both with the majority of MPs occurred in sizes <64 µm20 or <30 µm21. Similarly, MP deposition on leaves was already demonstrated under laboratory conditions, with rates of capture reaching 0.87 p cm–2 during a two-week exposure to airborne MPs22. However, all previous research on atmospheric deposition of MP in forest systems did not differ between throughfall deposition (including resuspension of previously deposited MPs) and the direct atmospheric deposition (accounting only for the entry of MPs to the forest systems)19,20,21.

Combining previous studies on MPs in temperate or tropical forests soils with the outcome of atmospheric deposition studies, it is clear that soil-related studies show higher MP size detection limits; consequently, the majority of atmospheric deposited MPs cannot be detected12,13,14. Additionally, the removal of forest litter during sampling12,14,15 leads to a missing link between atmospheric deposition and MP occurrence in forest soils. The potential importance of forest soil litter was highlighted by the verification that atmospheric MPs can become trapped on leaf surfaces in tree canopies23,24. Concentrations of 0.14–25 p cm–2 for MPs >10 µm in size were found after washing leaves with water25, whereas 57 p m–2 d–1 was found on near-urban oak leaves (Japan) after extraction using a 10% potassium hydroxide solution26.

Following the suggestion of a forest trapping function for MPs, the question arises to which extent canopy-trapped MPs and atmospheric MPs reach forest soils via throughfall deposition, as well as whether forest soils act as a sink for atmospheric MPs? Therefore, this study aimed to apply an aligned extraction method on a representative study site in central Germany to answer the following questions:

  1. i.

    What stocks of MPs (>20 µm) occur in the organic and mineral soil horizons of forest soils?

  2. ii.

    What are the MP fluxes (>20 µm) in throughfall deposition, and can MP stocks in forest soils originate solely from throughfall deposition over time?

  3. iii.

    Does the MP identified in forest soils originate from atmospheric deposition?

Our results indicate that direct atmospheric MP deposition and canopy-trapped MPs reaching the soil by throughfall and litter fall are the major sources of MPs to forest soil. Ground-based MP sources, such as littering or outdoor activities, are a possible additional minor source. MPs enter the soil from the surface and are finally accumulated in lower mineral soil by litter turnover processes. The total MP stocks and concentrations in the soils are high, indicating diffuse MP pollution.

Results

MP concentrations within the investigated temperate forest soils ranged between 120–13,300 p kg–1 with a mean of 4440 ± 3690 p kg–1 (mean ± SD) considering all samples (Supplementary Table S3). The highest mean concentrations were found in more decomposed organic forest soil horizons (Oe, Oh) with 5940 ± 4800 p kg–1, followed by mineral soil horizons (A, B) with 3670 ± 2800 p kg–1 and the uppermost litter horizon (Oi) with 3020 ± 1600 p kg–1. No statistically significant (p ≥ 0.05) differences were observed between MP concentrations in different forest soil horizons. The mean MP concentrations at the different forest sites were not significantly different (p ≥ 0.05) and were 3300 ± 2800 p kg–1 (FS1), 5200 ± 4400 p kg–1 (FS2), 2500 ± 850 p kg–1 (FS3) and 6800 ± 3500 p kg–1 (FS4).

MP stocks calculated from MP concentration using Eq. 3 at different sites and in different horizons ranged between 310–870,000 p m–2 with a stepwise increase from the uppermost litter horizon (Oi) with 1900 ± 900 p m–2, to the organic horizons (Oe, Oh) with 14,300 ± 12,600 p m–2, then to the mineral horizons (A, B) with 230,000 ± 290,000 p m–2 (Fig. 1a, Supplementary Table S3). MP stocks in the lowest forest soil horizon (B) were significantly higher (p ≤ 0.05) than all other horizons. Within the different forest sites, MP stocks occur without significant differences (p ≥ 0.05). The total MP storage in forest soils occurred in sums of 550,000 p m–2 (FS1), 200,000 p m–2 (FS2), 240,000 p m–2 (FS3), and 990,000 p m–2 (FS4).

Fig. 1: Microplastic stocks in forest soil and microplastic flux with throughfall deposition.
figure 1

a Microplastic stock (particles per square metre soil, p m-2) in different forest soil organic (Oi, Oe, Oh) and mineral (A, B) horizons (n = 4). P-values from the ANOVA test are provided and letters indicate significant differences among means from Tukey’s test (p ≤ 0.05). b Microplastic flux with throughfall deposition (particles per m2 per day, p m2 day-1) on sampling dates at each of the four forest sampling sites.

Full size image

Throughfall deposition fluxes of MPs occurred with an overall mean deposition of 9.1 ± 9.4 p m2 day–1 (range 0.6–35 p m2 day–1). MP fluxes in throughfall deposition occurred with a high spatial and temporal variance (Fig. 1b), but no significant (p ≥ 0.05) mean difference was observed between sampling dates or sites. Comparing MP fluxes over time along the four forest sites, the highest mean concentrations were found at site FS1 with 15 ± 15 p m2 day–1 followed by site FS2 (8.0 ± 7.8 p m2 day–1), FS4 (7.4 ± 7.8 p m2 day–1) and FS3 (5.7 ± 2.9 p m2 day–1) (Fig. 1b) with overall high variances in MP deposition fluxes.

The composition of the MP polymers showed differences in forest soil samples and throughfall deposition. Within atmospheric deposition samples, the majority of MPs were polypropylene (PP, 61%), followed by polyethylene (PE, 20%), polyamide (PA, 9%), and polystyrene (PS, 3%) and polyurethan (PU, 3%), together with others (4%). In forest soil samples, MPs occur as PP (48%), followed by PA (23%), PE (15%), ethylene-vinyl acetate (EVAc, 5%), and poly(methyl methacrylate) (PMMA, 3%), together with others (6%) (Fig. 2a). In contrast, MP particle properties showed no significant differences (p ≥ 0.05) between MPs in throughfall deposition and forest soil samples. The average particle size of MPs in throughfall deposition was 62 ± 65 µm, which is comparable to that in forest soil samples, i.e., 65 ± 76 µm (Fig. 2b). The average aspect ratio as a numerical ratio of each MP longest to shortest dimension, was the same, i.e., 1.8 ± 0.7 in both throughfall deposition and forest soil samples (Fig. 3c). Cumulative particle size distributions showed that 85% of the MP had a size <250 µm in both sample sets but MP particles 500–1000 µm only occurred in soil samples. Average MP sizes in forest soils were significantly (p ≤ 0.01) smaller at sites FS3 (54 ± 44 µm) and FS4 (50 ± 47 µm) relative to FS1 (78 ± 67 µm) and FS2 (80 ± 112 µm). Between forest soil horizons, no significant MP size differences were detectable. Within forest throughfall deposition, site FS4 showed a significantly (p ≤ 0.01) smaller average particle size (44 ± 38 µm) compared to all other sampling sites ranging between 52 and 74 µm. No significant variations in MP particle size were observed over time during the atmospheric deposition sampling period.

Fig. 2: Microplastic particle properties in forest throughfall deposition and forest soil.
figure 2

a Cumulative MP polymer composition (percentage%) at all sampling sites in throughfall deposition and forest soil. b Microplastic particle size (µm) in both sample types with Wilcoxon test p-value. c Microplastic particle aspect ratio (longest to shortest dimension of each particle) in both sample types with Wilcoxon test p-value. Sample size in throughfall deposition samples (n) = 152 particles and in forest soil samples (n) = 513 particles.

Full size image
Fig. 3: Estimated cumulative throughfall deposition of microplastics (p m-2) under different scenarios ranging from minimal to extreme combined with European Union (EU) plastic production (Mt-1) according Plastic Europe (2014, 2023)66,67 and fit to total microplastic stocks (p m-2) in forest soils.
figure 3

Estimated throughfall deposition of microplastics per year (p m-2) coupled to the increase factor of European resin (plastic) production with minimal scenario (minimal value of microplastic loads), conservative scenario (1st quartile of microplastic loads); median scenario (median of microplastic loads) and progressive scenario (3rd quartile of microplastic loads) and extreme scenario (maximal value of microplastic loads). Sum of microplastic stocks (p m-2) combining organic (Oi, Oe, Oh) and mineral (A, B) horizons of the studied forest soils.

Full size image

Assuming throughfall deposition as the major MP source for the investigated forest soils, a rough estimate of different scenarios reveals that the current total MP storage in forest soils with a median of 394,000 p m–2 (organic and mineral soil) would result from the highest rate of MP deposition measured in this study after 70 years (Fig. 3). Because this estimation considers cumulative throughfall deposition since 1950, coupled to the annual increase factor of European plastic production, the different scenarios reveal that the lower range of total MP storage (229,000 p m–2) can be reached under a progressive (3rd quartile of MP deposition loads) scenario, whereas the median or conservative scenarios reach up to 50,000–125,000 p m–2, i.e., remaining clearly below the estimated total MP storage (Fig. 3).

Discussion

Microplastics in forest soils

MP concentrations in the forest soils studied here fall within the upper range of previously reported MP concentrations in soils corresponding to different land uses (Fig. 4)4. Compared to the global database4 and considering interstudy methodological differences, mean and median MP concentrations of the studied forest soils exceed the interquartile range of the other datasets reported for agricultural soils, wetland soils, and coastal soils, and are comparable to the ranges found in urban soils (Fig. 4). When directly compared to previous studies that report mean microplastic concentrations in mineral forest soils of 200 ± 260 µg kg–1 12,13,14, the mean values observed in the investigated forest soils are approximately 20 times higher. The most important factor explaining this difference is that previous studies applied mainly optical microscopy methods, which are known to have limitations in MP analysis27 and also had higher size detection limits ranging between 40–200 µm13,14. Our aligned method approach using a LODsize of approximately 20 µm captures a higher frequency of smaller MPs, which constitute the majority of airborne MPs19,22,26. However, our method may underestimated the presence of MP fibres, because of the lower recovery rates for fibres compared to spherical MP particles, we find in our recovery tests. Another reason for this difference could be the sampling procedure, as previous studies did not consider organic soil horizons12,13,14. Our method shows a low recovery for fibres and thus the true amount of deposition is expected to be even higher than that reported here, even if the majority of MPs detected on tree leaves occur as fragments or films19,21,26.

Fig. 4: Comparison of microplastic concentrations (p kg-1) from forest soils with global microplastic concentration data in soils of different land uses.
figure 4

Data not derived from this study obtained from other studies for forest soil13 as well as global database4 for all other land uses. ANOVA analysis between forest soil (this study) and each other group (p-values: **** = p ≤ 0.001; ns = not significant).

Full size image

No differences in the MP stocks between the studied forest sites were observed; this is probably due to a limited difference in forest stand density and stand age (Supplementary Fig. S1, Supplementary Table S1) and distance from an emission source21,25. A top-down increase of MP stocks occurs as a function of soil horizon thickness and bulk density. Atmospheric deposition and leaf capturing are thought to be the major entry pathways of MPs to forest systems16,25,26. MPs show a strong retention on leaf surfaces as they are primarily bound within the epicuticular wax layer26. However, the decomposition of litter and subsequent delivery of new litter on top causes constant relocation of MP into deeper horizons23. Furthermore, the vertical migration of MPs released from decomposed litter has already been proven to occur for mineral soils with different textures and porosities by gravity15 and bioturbation9,23.

Microplastics in forest throughfall deposition

MP fluxes in throughfall deposition within the studied forest systems (9.1 ± 9.4 p m2 day–1) fall within the lower range of previously measured atmospheric MP fluxes in near-urban or remote areas. In general, atmospheric MP deposition rates are highly variable. For example, average deposition rates of 12 p m–2 day–1 (Mount Derak, Iran)28, 41 p m-2 day–1 (Ganges River catchment, India)29 and up to 365 p m–2 day–1 (French Pyrenees)19 have been reported. In German beech-oak and Douglas fir temperate forests average deposition rates were 53–512 p m2 day–1 20,21. Although studies on MP atmospheric deposition with a comparable MP size detection limit (e.g., 30 µm) report a comparable MP deposition rate to our study, significantly higher rates are reached with lower size detection limits (e.g., 1.0 µm via Raman spectroscopy)19,29. It seems that the <10 µm (PM10) fraction of total suspended particulates is of particular importance for MP deposition in the forest30,31. Thus, MP concentrations in atmospheric deposition may depend strongly on size detection limits.

Regarding spatial and temporal variances within atmospheric MP deposition, previous studies with a larger spatial scale found correlations with population density29 or meteorological conditions, including precipitation and wind direction28. Studies with a narrower spatial scale found no correlation with precipitation or wind speed, nor meteorological seasons20,21 but a correlation with vegetation periods as a result of the canopy comb-out effect11. While coniferous forests showed a constant MP deposition rate, deciduous beech-oak forests exhibited lower rates during the leafless period. This finding is consistent with the comparable low deposition rates found in our studied forest areas; however, we observed no significant differences between our mixed-forest site and deciduous beech or beech-oak forest sites. Compared to short-term direct MP sources for soils, such as biosolid fertilizers (e.g., sewage sludge) resulting in MP concentrations between 546 and 10,400 p kg–1 (1 to 5 applications)32 and strongly dependent upon wastewater treatment and application intensity33, the mean MP flux during one year was found to reach deposition values of 3321 p m–2 assuming constant deposition rates.

Sources of microplastics in forest soils

The overall cumulative polymeric composition, with a dominance of PE, PP, and PA in both sample matrices, follows known compositions found in other soils3,4 or on leaf surfaces25,26 and is similar to the compositions of polymers involved in global polymer production3. Additionally, MP numbers in soils and atmospheric deposition decrease with particle size, a typical pattern found in all environmental compartments. The mean particle sizes found in our soils and atmospheric deposition match previous studies of atmospheric deposition in forests, with the majority of MPs occurring in size fractions <63 µm20 or <40 µm21, while MPs trapped on leaf surface showed smaller sizes of 10–30 µm26. MP particle morphology was dominated by fragments and films, consistent with previous findings for forest soils and leaves MPs25,26, but different from the reported dominance of fibres in airborne MPs22,34. However, this difference might be partly due to the extraction methods used, which have a comparatively low recovery of fibres and are missing quantification of the <20 µm fraction30,31. We did not find any significant differences in average particle sizes and aspect ratios in forest throughfall deposition and forest soils. This finding supports the assumption that MPs enter forest soils directly via atmospheric deposition19,20 or via leaf litter fall, including canopy-trapped MPs21,26.

A rough estimate of annual MP deposition focusing on atmospheric MP inputs and neglecting MP in throughfall deposition resulting from forest internal cycling showed that current MP storage within the studied forest soils could be realistically reached after 70 years, considering the higher range of atmospheric deposition found in this study (i.e., under an extreme or progressive scenario). However, considering general variations in atmospheric MP fluxes19,21, the complexity of forest structures35, and the forest filter effect17, MP deposition might change depending on the vegetation period and the age of forests. The performed sampling of throughfall deposition and missing control of sole atmospheric deposition could lead to overestimation of MP fluxes21. However, our comparison of the estimated cumulative deposition with the MP stocks in the forest soils (Fig. 3) rather indicates an under- than an overestimation. The uncertainty can be due to the comparatively short sampling period of two months outside the growing period of trees, which could cause an uncertainty in MP fluxes by not covering the full yearly cycle of foliage development, seasonal weather conditions, and possible variations in deposition17,30. For further studies, long-term datasets of throughfall and atmospheric deposition covering all vegetation periods would be needed for enhanced estimations for MP accumulation in forest soils. Furthermore, internal cycling of MP in forest systems should be estimated in the future by investigating, e.g., direct MP deposition from the atmosphere at crown height and MP transport via stem flow.

Despite the MP forest filter effect, additional MP sources should not be ruled out. Because the studied forests are managed, forest management could release MPs36; for example, tree shelters commonly used in forestry to protect seedlings are typically not collected after their useful period (e.g., 5 years)37 and could act as MP sources. In addition to forest management sources, littering, waste dumping, and outdoor activities are further possible MP sources15,38,39. In summary, the properties of MPs in soils and from atmospheric deposition are mostly consistent, and the estimated cumulative deposition could explain MP stocks found in forest soils. However, data limitations and uncertain model assumptions prevent an exact estimation, and contributions from forestry, littering, and outdoor activities might also exist. We assume atmospheric deposition to be the major source of MPs, while other sources have minor or spatially limited contributions to MP in forest soils. This study also highlights that diffuse atmospheric deposition alone can cause high MP concentrations in soils. Thus, deposition should be considered as a significant background when specific sources of MPs are discussed, and their cumulative effects might even be higher than those of well-known high MP sources (e.g., sewage sludge applications) because MPs from deposition accumulate constantly over long timescales.

Methods

Forest soil sampling

Forest soil sampling was conducted at four forest sites (FS1-FS4) located in a managed forest area east of Darmstadt (Hesse, Germany) (Fig. S1). The four sampling sites cover beech (Fagus sylvatica L.), oak (Quercus robur L.) and pine (Pinus sylvestris L.) as dominant tree species with different forest structures developed on Cambisols and Dystric Cambisols (Supplementary Table S1). Site FS1 consists of beech trees with a stand age of 70 years, site FS2 of oak and subdominant beech trees with a stand age of 50 years, while site FS3 and FS4 consists of dominant pine trees with beech or ash trees and a stand age of 60 years (Supplementary Table S1). Composite soil samples were collected from a stainless-steel frame (20 × 20 cm) for litter (Oi) and decomposed litter (Oe, Oh) horizons and by core drilling (Ø 20 mm) for mineral (A, B) horizons, at 8 locations around a central point (spacing 5 m) at each sampling site. Horizon thickness ranged between 2.5–3.0 cm (Oi), 0.5–1.0 cm (Oe), 0.5–1.0 cm (Oh), 2.0–2.5 cm (A), and 7.5–8.0 cm (B) (Supplementary Table S3). Composite samples (~300–500 ml volume) were mixed within stainless-steel bowls and stored in aluminium coated paper bags (METRO Starpak)40. During soil sampling, only stainless-steel tools were used and cleaned between each sample with MilliQ water (Direct-Q® Water Purification system) transported in glass vessels, and pure cotton overalls were worn. Field contamination was controlled with field blank samples, consisting of a 5 L–1 wide neck jar filled with 1000 ml MilliQ and opened during the full sampling process at each side41,42. In-situ soil parameters (pH, carbonate content, texture) were measured and soil classification according WRB43 conducted after the sampling. Bulk densities of the organic (O) and upper mineral (A) soil horizons were measured according the excavation method using a 20 × 20 cm (400 cm–2) metal frame, while the bulk densities of lower mineral (B) soil horizons were measured by soil cores44. Coarse soil fragments (>2 mm) were excluded from bulk density calculation in mineral soil horizons44.

Throughfall deposition sampling

Sampling of throughfall deposition was performed from 19.10.2023 until 19.12.2023, at each sampling site (FS1-FS4) within the forest area. To allow a sampling of throughfall deposition and throughfall including litterfall, we used a plastic-free passive funnel sampler (DIN 19739-2, 2003). At each sampling site (5 × 5 m), one passive sampling system was installed in the centre of the site, with a minimal distance of 2 m to the next tree trunk. The passive bulk deposition sampling system consists of a borosilicate glass funnel (Ø 250 mm at inlet and Ø 10 mm at outlet) connected to an empty borosilicate glass cartridge as a continuous transition to a 5 L–1 sample collection vessel with glass lid (borosilicate glass, wide neck jar), housed in an aluminium box (Supplementary Fig. S2)31. Samplers were installed with the upper funnel opening 1 m-1 above ground surface. During the two-month sampling period, we replaced the sample collection vessels four times (Supplementary Table S2). After each sampling period, the funnel and cartridge were rinsed with MilliQ to collect deposited MPs from the glass surface. The litterfall inside the funnel was collected with tweezers and rinsed with MilliQ into the sample collection vessel. To ensure a plastic-free sampling environment, we pre-cleaned every component of the sampling system with ethanol and MilliQ and covered those with aluminium foil until installation42. During system installation and sample collection, field blanks were collected (see chapter 2.2) and pure cotton overalls were worn41.

Microplastic extraction method

The development of an aligned method for MP extraction from organic forest soil horizons was optimized for the extraction of MPs from litterfall of different decomposition states to be compatible with already established extraction protocols for mineral soils45,46. We tested ultra-pure water25,47,48, ethanol49 and tetrasodium pyrophosphate (Na4P2O7)50 in different molar ratios to enable a separation of MP particles attached to leaf litter and decomposed litter. Furthermore, we performed experiments with different detachment protocols including various soaking, shaking as well as sonification conditions with different time intervals and different amounts of sample aliquots (0.5 g, 1.0 g, 2.5 g and 5.0 g dry litter or decomposed litter). From those experiments, it became evident, that a 0.1 M tetrasodium pyrophosphate solution, a common dispersing agent in grain size analysis51 or organic matter extraction from mineral surfaces52, showed the best performance.

The new MP extraction and purification procedure aligned for forest litter, decomposed litter and mineral horizon material as well as throughfall deposition samples, consists of five steps (Supplementary Fig. S3). For mineral soil samples with an organic matter content <15%, we applied the following protocol45: Soil samples were dried (45 °C, 48 h), sieved (<2 mm), homogenized and representatively divided using the quartering method53 to an aliquot of 5 g–1 dry soil material stored in corning tubes (Sigma Aldrich, Corning® 50 ml centrifuge tubes). Mineral soil phase was separated using a pre-filtered (Whatman, Nuclepore Membrane, 0.8 μm) sodium bromide solution (NaBr, ρ = 1.5 g cm–3). Each sample undergoes sonification (10 min), shaking (100 rpm for 30 min) and centrifugation (2500 x g, 30 min) before transfer using a rubber disc decanting aid and supernatant vacuum-filtration under washing with MilliQ on 10 µm stainless-steel filters (Rolf Körner GmbH, material 1.4401, Germany) before further sample purification.

For leaf litter and decomposed litter in organic soil horizons (O-horizons), we developed the following protocol: Organic horizon samples were dried (45 °C, 48 h), homogenized and divided using the quartering method53. An aliquot of 5 g-1 dry organic material mixed with 30 ml 0.1 M tetrasodium pyrophosphate solution (Na4P2O7) undergoes shaking (200 rpm, 5 min), ultra-sonification (10 min) and a second shaking (200 rpm, 5 min). Afterwards sample material was washed on stainless-steel sieves (5 mm, 2 mm, 10 µm) and rinsed with MilliQ to remove remaining Na4P2O7 solution. The washing residues were captured within a 10 µm sieve and transferred on a 10 µm stainless-steel filter via vacuum-filtration (Supplementary Fig. S2a). Throughfall deposition samples, stored in solution in the sample collection vessel, undergo the same washing procedure like organic soil horizon samples including sieving (5 mm, 2 mm, 10 µm), rinsing of larger leaf parts with MilliQ and capturing of washing residues within a 10 µm sieve as well as the later transfer on a 10 µm stainless-steel filter via vacuum-filtration (Supplementary Fig. S2a).

MP extraction of pre-treated samples from organic soil horizons and throughfall deposition was performed via the “binary-solvent extraction” method54 enabling a separation of most organic matter (OM) and MPs in two separate liquid layers. A solution with the ratio of 80% EtOH and 20% MilliQ was added to the pre-treated samples and left soak for 2 h. During this soaking time, the solvent-water mixture acts an anti-hydrophobic agent and allows water to penetrate in the OM matrix to overcome its hydrophobicity and allows a separation of MP from OM54. After the soaking time, the upper solution within each Corning tube was carefully removed by pipetting (2- and 5-ml steps). After EtOH solution removal, 20 ml MilliQ was added to the remaining liquid and the sample shaken by hand. The supernatant was vacuum-filtrated on a 10 µm stainless-steel filter and the filter was washed with MilliQ.

The combined sample purification protocol for all sample matrices (Supplementary Fig. S3c) includes the transfer of remaining particles after MP extraction with 25 ml NaUT-solution (sodium-urea-thiourea: 8% NaOH, 8% urea, 6% thiourea) to a fresh corning tube. Samples were afterwards frozen at –16 °C for 1 h to enable mini-crystals the degradation of particulate organic matter (e.g., cellulose) and the surface structure modification of chitin55. Thawing of samples was conducted under constant shaking (50 rpm, 30 min) before vacuum-filtration with MilliQ washing and transfer to a fresh corning tube with 15 ml MilliQ. Finally, we applied the digestive removal of remaining particulate organic matter with a temperature-controlled (ice-bath) Fenton reaction, using 15 ml hydrogen peroxide (H2O2, 30%, Roth, ROTIPURAN®) and 15 ml iron(II)sulphate solution (FeSO4, pH 3, >99%, Roth, ACS reagent) added in small and alternately steps over the reaction (1 ml to 5 ml). After the Fenton reaction, remaining iron oxides were flushed away with MilliQ under vacuum-filtration on stainless-steel filter and a final NaBr (ρ = 1.5 g cm–3) density separation. Final filtration of the supernatant was done on Anodisc filters (Whatman, Anodisc 25, 0.2 μm, diameter 25 mm) with MilliQ water. Residual iron oxides were removed by adding 2 M H2SO4 in 150 µl steps directly on the filter and diluting it with 15 ml MilliQ (repeated up to 5 times). Anodsic filters were stored in glass petri dishes with aluminium foil until MP analysis.

Microplastic analysis

Anodics filters were analysed using chemical imaging via focal plane array (FPA, 32 × 32, Bruker Cooperation, Billerica, MA) detector at an µFTIR (Lumos II, Bruker Cooperation, Billerica, MA). Measurements were performed in transmission mode with 1 scan per pixel, spectral resolution of 4 cm-1 within the wavenumber range of 4000–1250 cm–156,57. Binning of 2 × 2 pixels was used to reduce file size, resulting in a single pixel resolution of 10.84 × 10.84 µm. Chemical images were processed within Purency software (version 4.17, Purency GmbH, Austria) and analysed by a random forest machine learning model58. To ensure proper particle identification, thresholds for relevance (relevance of classification) with 0.3 and similarity (similarity of sample spectra and reference spectra) with 0.15 were applied59. To avoid overestimation of MP identification, identified particles were manually expert verified for plausibility and fragmented particles merged60. Finally, single pixel identifications were excluded, leading to a limit of detection in size (LODsize) of 21.68 µm for particle length and 10.84 µm for particle width.

Method performance and quality control

We spiked 1 g triplicates of each Oi, Oe and Oa-horizon with cryo-milled (Retsch, Mixer Mill) and red coloured PE films (<500 µm) as well as red PP fibres (100–500 μm length). To enable attachment of spike MPs to litter and decomposed litter, we incubated the samples for 4 weeks within aluminium dishes and covered with transparent film to enable sunlight access under room-temperature. We applied wet-dry cycles by adding MilliQ water with a spray bottle every three days. Afterwards, the incubated samples undergo the aligned pre-treatment and extraction protocol. The mean recovery for PE film spikes was 95 ± 11.1% (Oi), 60 ± 5.3% (Oe) and 90 ± 50% (Oh) and 30–40% for PE fibre spikes in all three organic horizons (Supplementary Fig. S4b). The comparatively low recovery rate of PE fibres likely results from the use of 10 µm stainless-steel filters during the MP extraction protocol, as the fibres with a diameter of approx. 5 µm can partly pass through these filters, independent of their fibre length.

Contamination prevention measures include the use of previously cleaned steel- and glass-tools during field sampling, wearing cotton overalls during field work and cotton lab coats during laboratory work. During laboratory work, all MP extraction steps were conducted within a laminar flow box (SPETEC®, Laminar Flow Box Serie SuSi), all solutions previously filtered using 0.8 μm membrane filters and all lab tools consequently pre-cleaned or rinsed in-between with MilliQ. Contamination control was performed via field blanks (one per sampling date, n = 4) and laboratory blank samples (one per processing batch, n = 6)42,61. We found a mean contamination of 1.5 particles per sample (p sample–1) for all blanks, with 1.6 p sample–1 in field blanks and 1.3 p sample–1 in lab blanks. Contaminant particles consist of PP (55%), PE (27%), and PA (18%) with average particle sizes of 50 ± 26 µm (Supplementary Fig. S4a). Overall, blank concentrations are significantly lower (p ≤ 0.01) compared to particle counts in forest soil or throughfall deposition samples.

Statistical analysis

Data analysis was performed in R (Version 4.2.2) via RStudio Version (2022.02.3)62 using the standard R-packages, “dplyr”63 for data manipulation and “ggplot2”64 for visualization. Limit of detection (LOD) was calculated based on blank contamination65 following Eq. 1:

$$LOD={\underline{M}}_{blank}+(3\times ST{D}_{blank})$$
(1)

with Mblank mean MP concentration in blank samples and \({{STD}}_{{blank}}\) standard deviation of MP concentration in blank samples. The overall LOD was 3.4 particles per sample. Furthermore, reported MP concentrations were corrected (MPcorr) without calibration to the recovery rates but by subtracting the calculated LOD61,65 following Eq. 2:

$${{MP}}_{{corr}\,}\left(p\,{{kg}}^{-1}\right)=\left({{MP}}_{{count}}-{LOD}\right)$$
(2)

with \({{MP}}_{{count}}\) count of detected MP per sample and limit of detection (LOD). Corrected MP concentration were reported in particles per kg (p kg–1) soil dry weight for organic and mineral forest soil horizons and in particles per m–2 per day (p m–2 day–1) for throughfall deposition fluxes.

\({{MP}}_{{stocks}}\) were calculated for each forest soil horizon using the corrected MP concentrations (MPcorr), soil horizon thickness (correspond to sampling depth) and horizon bulk densities (Supplementary Table S3) following Eq. 3:

$${{MP}}_{{corr}}\left(p\,{g}^{-1}\right) \times {horizon\; thickness}\left({cm}\right)\times {bulk\; density}\left(g\,{{cm}}^{-3}\right) \\ = {MP}{{cm}}^{-2}\times 10,000={{MP}}_{{stock}}\left(p\,{m}^{-2}\right)$$
(3)

Sums of \({{MP}}_{{stocks}}\) per sampling site were calculated as the sum of all horizon stocks and considered as the total MP storage at each site. Bulk densities occurred with means of 0.02 g cm–3 for the litter Of horizon, 0.22 g cm–3 for Oe and 0.42 g cm–3 for Oh horizons, while mineral horizons show means of 0.78 g cm–3 for Ah and 1.47 g cm–3 for B horizons.

The throughfall deposition since 1950 was estimated coupled to the increase factor of European Union (EU) resin (plastic) production according the following equation Eq. 4:

$${D}_{\left(t\right)}={D}_{0}\times {\left(\,\frac{1}{f}\right)}^{\left({T}_{0}-t\right)}$$
(4)

With \({D}_{\left(t\right)}\) estimated throughfall deposition of MP (p m–2) in year \(t\) calculated from quantiles of daily MP fluxes (p m–2 day–1) and extrapolated to one year; \({D}_{0}\) the measured throughfall deposition of MP (p m–2) in the reference year \({T}_{0}\) = 2023; \(f\) the annual increase factor of EU plastic production (e.g., 1.05 for 5% annual growth) according data from PlasticsEurope (2013)66 and PlasticsEurope (2024)67; \({T}_{0}\) the reference year 2023 and \(t\) the year for which the back-calculation is performed (e.g., 1951).

Statistical operations include descriptive statistics, non-parametric Kruskal–Wallis test and follow-up Wilcoxon test, analysis of variance (ANOVA) applied to MP concentrations and repeated measures ANOVA (rmANOVA) applied to MP particle properties. Overall, MP concentrations and particle data (e.g., size distributions) occur not normally distributed and with unequal variances. We interpreted statistical analysis results as significant with a p-value < 0.05.