Plastic pollution risks in bioretention systems: a case study

ABSTRACT We investigated plastic pollution in soil-based stormwater bioretention systems (BRS), which are potentially important pollutant receptors and pathways. Our integrated study is the first of its kind, focusing on plastic abundance, size fractionation, composition, and interactions with urban metrics (including housing density and auxiliary stormwater treatment infrastructure) in BRS filter media. Our results revealed that mesoplastic (MEP) and microplastic (MP) concentrations in BRS are comparable with those reported in other stormwater systems (e.g. wetlands) as well as soils in other land use areas (e.g. agriculture). Distributional sampling within the BRS revealed MP abundances do not change with horizontal distance from the inlet to the outlet. However, MEP abundances drastically decreased towards the outlet, indicating plastic accumulation within BRS. This is important because MEPs can breakdown into MPs which can affect BRS function as well as mobilise downstream. Yet our data uncovered more complex mechanisms involved in BRS plastic fate, with composition data revealing that MPs are not simply breakdown products of MEPs but are instead derived from different sources. Composition, morphology and colour analysis confirmed that BRS polymer liners are a key source of MPs and MEPs in soil filter media. Multivariate analysis of the data with urban design metrics showed gross pollutant traps are effective at decreasing MP concentrations in BRS but not as effective at controlling MEPs. Our results point to complex plastic transmission and accumulation pathways in BRS. Interception measures can partially alleviate plastic risk, but more work is needed to elucidate plastic long-term fate in BRS. GRAPHICAL ABSTRACT


Introduction
Plastic products are ubiquitous. Population growth and urbanisation have led to increases in the quantity of plastic waste generated and contamination of environmental ecosystems. As a major solid waste component, plastic pollution is of considerable environmental concern due to its prevalence, persistence in the environment, and overall environmental health impacts.
Plastic particles termed microplastics (plastic particles ≤ 5 mm) [1] and mesoplastics (plastic particles between 5 and 25 mm) [2] are ubiquitous forms of terrestrial plastic litter. These particles have been reported in many environments including the atmosphere [3], and soils from agricultural lands [4] to flood plain soils [5]. The covid pandemic is also possibly triggering a new plastic pollution crisis via the disposal of facemasks [6].
Due to their microscopic size, toxicity, and bioaccumulative properties, microplastics can cause alteration to a variety of soil properties [7] and biota (plants, microbes, and animals), including reduced metabolism [8] and long-term toxicity to soil-dwelling biota [9].
In terrestrial environments, microplastics can be transported large distances by stormwater runoff [10], and eventually, be deposited in green stormwater infrastructure such as bioretention systems (BRS). BRS provide valuable ecological services such as; reducing flooding, remediating stormwater pollutants, providing habitats for wildlife, conserving water, modulating micro-climate, and capturing and storing atmospheric carbon dioxide [11]. BRS pollution is problematic as they serve as an intermediary between urban and aquatic environments which help prevent further discharge of pollutants into the aquatic environment. Recently, several studies have reported microplastics in treated effluent from bioretention ponds [10,12]. However, limited information is available on the accumulation and distribution of microplastics in the filter media of bioretention systems (BRS). Two relevant studies focused on microplastic pollution in sediment samples of Denmark's stormwater retention ponds, with reported concentrations upto 127,986 items.kg −1 sediment [13,14].
South East Queensland (SEQ) is one of Australia's most densely populated regions (after Sydney and Melbourne) and its area of 22,420 km 2 accommodates 12 local government areas (LGA) including Gold Coast, Ipswich, and Logan ( Figure 1). With population growth and economic development in recent years, the region has been under environmental stress due to increasing urbanisation activities including residential developments. Additionally, the SEQ region is known for its integrated approach of using Water Sensitive Urban Design elements in urban landscapes to curb the negative impacts of pollution in its waterways and other aquatic environments [15].
To date, studies on microplastics in green stormwater infrastructures have been carried out on constructed wetlands [16] in South East Queensland. However, no studies have been carried out on solid media from BRS basins. These soil-based systems comprise four layers: a mulch layer (typically 50 mm depth), a sandy loam filter media layer (typically 400 mm depth), a coarse sand transition layer (typically 150 mm depth), and a gravel drainage layer (typically 100 mm depth).
Given the important ecological roles of BRS, this study aimed to quantify micro(meso)plastic accumulation in BRS in suburban residential areas. In addition to contributing novel information on the presence of micro (meso)plastic accumulation in BRS, we evaluate potential relationships between micro-and mesoplastic abundance and urban design factors including BRS age, presence of gross pollutant traps (GPTs), and population/housing density.

Sampling sites
The filter media samples were collected from 20 selected BRS with access permitted from the three councils. The specific BRS sites were selected to encompass a broad range of values for key factors including age and size in order to minimise any potential bias in observations across the regions. Site characteristics are described in Table 1. The studied systems have been constructed and implemented according to Water by Design [15], and for systems constructed pre-2014, the Facility for Advancing Water Biofiltration [20]. These two guides are similar with Water by Design [15] recommending loamy sand as filter media while FAWB [20] suggesting sandy loam. Five BRS in Ipswich and Logan respectively, as well as 10 BRS in Gold Coast, were chosen for analysis. Three sampling locations were selected along the longitudinal transect of each bioretention basin following the stormwater runoff pathway ( Figure 2); (1) inlet to the bioretention cell herein referred to as 'inlet', 4 m from the inlet pipe around the sediment forebay; (2) bioretention cell herein referred to as 'middle' around the dense vegetated areas of the BRS cell and (3) outlet of the bioretention cell herein referred to as 'outlet', approximately 5 m from the outlet pipe.

Sample collection and preparation
Sampling took place from November 2020 to January 2021, with a total of 60 filter media samples collected. At each system, soil samples were taken as a whole from a depth of 0-400 mm within 250 mm × 250 mm plots using a metallic hand auger (80 mm in diameter) and transferred into glass jars. The sampling depth corresponds to the minimum filter media depth required for a BRS [15,20]. Soil samples were kept cool until returned to the laboratory where they were stored at ∼4°C until analysis. In the laboratory, the samples were air-dried to a constant weight at 40°C. The samples were then homogenised and sieved through 4 mm to 300 µm stainless sieves. Only fractions ≤2 mm were used for analyses given that plastic particles > 2 mm can be visually sorted. When present, microplastics > 2 mm were culled, counted, and stored for further analysis.

Sample digestion and plastic extraction
Once at the laboratory, purification and extraction of the collected samples was performed using an enzymatic digestion method. For enzymatic digestion, soil samples were purified using a sequential enzymatic digestion procedure [21]. Briefly, soil samples (100 g) were first pre-treated with 30% w/w H 2 O 2 (Chemsupply, Australia) at 60°C for 24 h to increase the susceptibility of organic materials to enzyme degradation [22] and then treated with 10% SDS solution (Sigma Aldrich, Australia) to soften biogenic matter [23]. Then a three-step enzyme purification began with the introduction of cellulase-hemicellulase, lipase, and protease (Novozymes, Australia). Hydrogen peroxide treatment (30% w/w) was carried out at 60°C for 24 h to further solubilise digestion-resistant materials. Post-digestion samples were then subjected to a density separation process using ZnCl 2 (1.7 g cm −3 )(Chem-supply) saturated solution. Soil was stirred vigorously and left to settle for a day. The resulting supernatant from each sample was filtered through 11 µm stainless steel filters (Merck Millipore, 47 mm Ø) in order to recover plastic particles.

Analytical methods
Filter papers were air-dried at room temperature and inspected visually under a trinocular optical microscope (Nikon SMZ745 T and Eclipse E-200) at up to 100× magnification, using the method described by Hidalgo-Ruz et al. [24]. Furthermore, the extracted plastic particles were examined using attenuated total reflectance-Fourier transform infrared spectroscopy (ATR-FTIR) (PerkinElmer Spotlight 400N FT-NIR imaging system with Spectrum TM 3 FTIR spectrometer) to investigate their chemical composition. All extracted plastic particles were counted and categorised by morphology (fibre, spherical, fragments, film, cylindrical, and foam), size (≤ 5 mm as microplastics; ≥5 mm as mesoplastics), and colour as described by Helm [25]. Cylindrical particles identified herein are elongated plastic particles with three dimensions.

Quality control
Laboratory blank samples and a control sample (i.e. BRS filter soil supplied by local BRS construction contractor) were prepared to check for any microplastic interference, especially atmospheric, that might compromise analysis. A section of BRS liner supplied by Logan City Council was also analysed compositionally by ATR-FTIR. To mitigate contamination during field sampling, preventive measures were implemented such as covering soil samples with aluminium foil when not in use. Laboratory surfaces were wiped down with isopropyl alcohol. All materials and equipment were washed and rinsed with deionised water before each use. Cotton fibre clothing and nitrile gloves were worn during field sampling and laboratory analysis.

Data analysis
Statistical data analysis was performed using Microsoft excel ® and SPSS software. Results on the abundance of microplastics are expressed as the number of microplastic particles per kilogram (N.kg −1 ) d.w. soil. The relationship between the abundance of mesoplastics and microplastics was investigated using linear regression analysis. Multivariate statistical analysis using principal component analysis (PCA) was then applied to investigate the relationship among population density, system age, presence of gross pollutant traps and micro(meso)plastics abundance. Population density expressed as housing density (dwellings per hectare) was mapped out on a contributory catchment area basis to gain an insight into their contribution to microplastics that enter BRSs. Esri ® geographic information software ArcGIS was used to calculate and map out these variables. Site area and number of dwellings were obtained from Queensland Globe and were used to determine the average housing density for each catchment area. One-way Analysis of Variance (ANOVA) was used to test the significance in the difference between micro(meso)plastics abundance in study sites, study systems and sampling locations. A significance level of 0.05 was chosen.

Mesoplastic contamination: abundance, distribution, and characterisation
The number of mesoplastics (MEPs) extracted from the 20 different BRS in the SEQ region is presented in Table 2. In total, 820 MEPs with a total count varying from 0 to 110 particles.kg −1 soil (average 41 ± 30.5 particles.kg −1 d.w. soil) were extracted from 90% of the BRS examined. We found the highest amount of MEPs in Ipswich with a mean (±SD) of 78 ± 28.6 particles.kg −1 d.w. soil, followed by Logan (32 ± 21.6 particles.kg-1 d.w. soil) and Gold Coast (27 ± 19.4 particles.kg −1 d.w. soil) ( Table 3). One-way ANOVA did not indicate a significant difference between study sites for MEP abundance (p > 0.05).
MEP size ranged from 5.03 mm to approximately 23 mm. MEPs 5-15 mm comprised 76% of the total MEP by number across all sites. The highest abundance was found in the 5.03-8 mm (37.8%) and 11-15 mm (23.2%) ranges (Table 3).
Size fractionation revealed that MPs ranged between 0.072 and 4.934 mm, with the size range from 0.07 to 1 mm accounting for 63% of all MPs detected. It was observed that the occurrence of MPs decreased with an increase in size, except in the size class of 2-3 mm where a 50% increase in abundance was observed from the 1-2 mm size range (Table 4). There was no significant difference between the average MP size found across all study sites (ANOVA, F = 2.21, p = 0.11).
MPs in the inlet were significantly larger in size than those from all the middle and outlet sampling locations (ANOVA, p < 0.05, F = 9.06, P < 0.05). However, MPs in the 0.07-0.5 mm size range were mostly abundant in the outlet (42.9%) while the middle (34.7%) and inlet (22.4%) contained lower abundances (Figure 4). Generally, particles < 0.5 mm dominated across the three study regions.
In terms of composition, LDPE was the most dominant MP polymer detected contributing 50% of total MP abundance, followed by PP (34%) (Figure 5(a)). Examples of the FTIR spectra and representative images of MPs extracted from BRS filter media are depicted in Figure 6. Three different shapes of MPs were identified including fragments, sheets, and cylinders. Fragments were the most dominant shape (84%), followed by sheets (13%) and cylinders (3%) ( Figure 5 (b)). Blue and green dominated the MP colours detected, accounting for 54% and 43% ( Figure 5(c)).

Relationship between meso-and microplastic abundance
A total of 1920 plastic particles (820 mesoplastics;1100 microplastics) were observed in this study from the bioretention filter media. A weak but statistically significant positive correlation exists between mesoplastic and microplastics (p < 0.05).

Relationship between plastic particles and urban design factors
The presence of gross pollutant traps, population density (housing density), and system age was evaluated to ascertain how these factors affect micro/meso plastic abundance in BRSs. Study systems assigned  two categories for gross pollutant traps: YES (systems with gross pollutant traps, n = 6) and NO (systems without gross pollutant traps, n = 14). As BRS are designed per catchment area, the surrounding housing density (dwellings ha −1 ) was based on the assumed drainage pipe network for population  density. Finally, on the basis of system age, BRS were classified into four classes according to their ages, I: 1-3 years (n = 4), II: 4 -6 years (n = 3), III: 7-9 years (n = 6), and IV ≥ 10 years (n = 7). The resulting PCA biplot relating each of these factors on plastic abundance is shown in Figure 7. Microplastic abundance was observed to have no correlations with system age and a weak correlation with housing density. However, microplastic abundance shows a strong negative correlation with GPT. The statistics for each factor of each BRS can be found in supplementary information. Furthermore, according to Figure 7, mesoplastic abundance was also observed to have no correlation with system age and a weak correlation with housing density. Interestingly, the presence of a GPT was not strongly correlated with mesoplastic abundance in the BRS filter media.

Overview of findings
We present the first comprehensive studyevaluating not only plastic abundance but also composition, distribution, and correlation with urban design metrics into micro(meso)plastic pollution in BRSs. Indeed, only five studies have evaluated plastic abundances in the sediments of green stormwater systems [13,14,16,26,27]. Our study revealed the presence of plastic particles in BRS soil filter media, with MEP and MP abundances ranging between 0-110 particles.kg −1 d.w. soil, and 0-180 particles.kg −1 d.w. soil, respectively (Table 5). This indicates plastic pollution is occurring in green stormwater infrastructure. Similar abundancesranging between 2 and 147 particles.kg −1 d.w. sedimentwere found in sediment samples from wetlands in Melbourne, Australia [26]. In contrast, other studies in Denmark, Australia and Iran which reported MP abundance in retention ponds (similar to BRS), documented generally higher MP abundance, ranging from 320-10 6 particles.kg −1 d.w. sediment [13,14,16,27] The high occurrence of MPs in these other studies can be attributed to the surrounding land use conditions and the ages of the retention ponds. Intensive land uses, especially those associated with commercial and industrial activities increases the probability of plastic particle transportation into nearby stormwater infrastructures [28]. Additionally, older stormwater infrastructure (>20 years) are likely to retain more plastic particles, resulting in higher concentrations of plastic particles in the soil media [14].   [14] Denmark Stormwater pond Sediment 10 6 Liu et al. [13] Denmark Retention ponds Sediment 17,490 Ziajahromi et al. [16] Australia Floating treatment wetland Sediment 595 ± 120 (inlet) b 320 ± 42 (outlet) b Townsend et al. [26].
Australia Wetland Sediment 2-147 Rasta et al. [27] Iran Wetland Sediment 113-3690 a Reported as mean concentrations. Abundance is presented for microplastic, except where otherwise stated.

New data on MEP abundances in BRS
By comparison with other MEP sediment studies, the abundance of MEP in this study (in soils) is similar to that found by Sarkar et al. [29] ranging between 4 and 215 items.kg −1 d.w. in the sediments of river Ganga, India, but higher than reported in other land use areas, e.g. 0.62-5.37 items.kg −1 d.w. in floodplain soils, Germany [30]. Importantly, the presence of MEP in the filter media of BRS has not been reported in previous studies.

Trends by location and position in BRS
Based on sampling locations, we found that the overall plastic particle abundance decreased with increasing distance from the inlet; inlet (990 particles.kg −1 d.w. soil) > middle (530 particles.kg −1 d.w. soil) > outlet (400 particles.kg −1 d.w. soil). This is in agreement with a recent study that reported higher plastic particle concentrations in sediments at the inlet than at the outlet of a stormwater floating treatment wetland [16]. The higher accumulation of plastic particles at the inlet of the BRS could be attributed to the presence of a coarse sediment forebay, which allows the settling of particles [15]. A significant problem of plastic accumulation in BRS relates to filter soil health. Although different factors such as polymer type, size, concentration, and morphology influence the effects of plastic contamination on soil ecosystems, various effects have been associated with micro(meso)plastic presence [31]. In soil organisms, increased plastic abundance can lead to ingestion, resulting in reduced survival rate, growth, reproduction, feeding behaviour, as well as alter metabolic activity rates and gut bacterial community [32]. Plastic particles have also been reported to be absorbed by plants, affecting their development and shoot length [33]. Bioretention systems rely on their physiochemical (filter media) and biological (flora and fauna) components for stormwater pollutant remediation. Thus, the presence of microplastic particles can potentially impact stormwater remediation effectiveness. In addition to effects on soil health, mesoplastic contamination can affect the aesthetic appeal of the catchments in which these bioretention systems are implemented [34].

Influence of plastic particle size on distribution
To our best knowledge, there are few studies on mesoplastic size range in terrestrial ecosystems and none for BRS. This lack of data limits comparison of our results with other studies for MEPs.
The MP size range encountered in our study is similar to that analysed in the sediments of retention ponds in Denmark (0.01-2 mm) [13]. Liu et al. [13] showed that 97% of the MP sampled were between 0.01 and 0.5 mm. In our work, MPs displayed a reduction in average size at the outlet compared to the middle and inlet of the BRS across all locations. This suggests that BRS cells could trigger settling of larger MP particles.

Implications of plastic composition
Polymer composition varied between meso-and microplastics in the studied BRS. Extracted MEP particles were composed mainly of PS, PP and Nylon whereas PE dominated MP composition. These differences signify differences in plastic sources and the resistance of certain polymers to fragmentation. Although PE accounts for 36% of plastic production globally [35], they are not abundant as MEPs (only approx. 10%). The reason for PE dominating MPs in our study is unclear but could  . Fourier transforms infrared spectroscopy of some isolated microplastics and mesoplastics of different shapes. Fourier transform infrared spectroscopy of (a) a PE sheet from system B2 (b) PP fragment extracted from system B5, (c) PS foam obtained from system B11, (d) Nylon cylindrical recovered from system B5. be due to its rapid fragmentation in water. A study has shown that weathering of PE macroplastics in water can result in elevated PE MPs [36]. Nonetheless, the overall ranking of plastic particle composition in the filter media of bioretention systems was PE (33%) > PP (30%) > PS (16%) > nylon (10%). These are similar to the composition of polymers found by Treilles et al. [37] where PE was the most abundant followed by PP and PS in the analysed urban runoff samples. PE and PP, therefore, seem to represent the dominant plastic particles conveyed to soils via stormwater runoff [30].
Plastic shapes identified in this study were predominantly fragments (75%) which is consistent with several studies [13,14,26]. This finding indicates that the disintegration of discarded plastic materials in the open environment is a primary source of plastic pollution in BRS. Furthermore, the second dominant shape observed in this studyi.e. sheetsuggests that these particles originated from the use of geotextile polymer liners. Synthetic impermeable sheets are utilised in bioretention basins to limit the infiltration of stormwater into the underlying and adjacent soil [15]. ATR-FTIR analysis of sheet MP particles extracted from the filter media as well as a section plastic liner obtained from Logan City Council confirmed matching polymer composition (i.e. polypropylene). This implies that after a period, these liners disintegrate, increasing plastic particle abundance in BRS.
Interestingly, in our study, we did not find any plastic fibre, commonly present in stormwater runoff [37], stormwater treatment infrastructures [12][13][14]16,26], and urban residential areas [38]. There are several possible reasons for our non-detection of plastic fibres. It may be an artefact of our sampling technique, although no mechanistic reason for their escape is obvious. One is that most studies either analysed runoff and sediment samples in environments with high population and anthropogenic activities, such as agricultural and commercial environments. Townsend et al. [26] evaluated sediment samples from urban wetlands in commercial, industrial, highway, and residential catchments and reported a total of 120 microfibres. Ziajahromi et al. [16] also reported the predominance of fibrous particles in the sediment of a stormwater floating treatment wetland constructed from a matrix of PET fibresalthough tyres rather than the wetland polymer liner were considered the main source of fibres in that study. Another possible reason for the lack of fibrous plastics relates to the context of BRS function within the landscape. Lightweight fibrous MPs may simply be washed from BRS surfaces into downstream receptors. This could explain why they were not detected in the soil filter media studied here, but are commonly encountered in downstream sinks such as wastewater treatment plants and wetlands.
Microplastic abundance is influenced by the presence of plastic litter [39]. Exposure to environmental factors such as sunlight, rain, abrasion and wind make plastic items brittle thus increasing the abundance of even smaller particle of plastics [40]. Our study indicates that MEP moderately influences the abundance of MP. These results agree with the fragmentation phenomena which states that after a period, large plastic debris is disintegrated into microplastic particles under environmental conditions [40].

Plastic colour and potential significance
Mesoplastic particles contained a greater variety of colours than MPs (see Figure 3(c) and Figure 5(c)). The colour variation indicates that not all MEPs were disintegrated. This finding is in contrast with other sediment studies that reported white as the most dominant colour followed by blue and green [41,42]. The reason for the dominance of blue coloured plastic particles is unclear as white plastic materials are the most manufactured and used in common plastic products such as plastics bags and films [43]. On the other hand, the presence of green coloured particles, the second most abundant colour observed in our work (Total = 33%; 20% mesoplastic and 43% microplastic) is attributed to the usage of green synthetic impermeable liners, supporting the matching polymer composition observed for the liner material and the sheet forms of MPs detected in the filter media. Green plastic particles share colour similarities with plants which serve as a food source for soil animals. Van Cauwenberghe et al. [44] suggest the similitude of plastic particles with food sources contributes to the likelihood of ingestion. Thus the high abundance of green coloured particles observed in the present study could lead to its ingestion, impacting soil fauna health. There is no data on the influence of plastic colour on ingestion by soil organisms.

Framing findings in context of urban design metrics
In our study, the presence of GPTs was strongly associated with the reduction of plastic pollution in BRS. Gross pollutant traps are effective structures of stormwater runoff treatment designed or retrofitted for the capture of litter ≥ 5 mm, as well as coarse sediments [45]. This is observed in the relative proportion of MEP (43%) and MP (57%) extracted. Although GPTs had no significant effect on MEP, they influenced MP abundance. This can be seen in Gold Coast region where the MP abundance was the least (290 particles.kg −1 ) as a result of retrofitted GPTs most BRS in that region. The results indicate that the importance of GPTs as a management strategy for the reduction of plastic pollution in bioretention basins.
Although previous studies found a positive correlation between MP concentration and population density [5,46], a different relationship was found in our study, which suggests that population density does not influence plastic particle abundance in BRS. Finally, it was revealed that the effect of BRS age was insignificant on both MEP and MP abundance. The result signifies that the age timeframes assessed in this study (<20 years) cannot be used to predict plastic particle build-up in BRS. However, over longer terms, age may well prove to be a strong indicator of plastic accumulation risk in BRS. Given the fact that technologies for breaking down MPs in soil environments are in their infancy [47], our findings warrant follow-up research into the impacts and fate of plastics in BRS media.

Conclusions
. MEP and MP were found to be abundant in bioretention systems (BRS) in an urban region of Australia. . Plastic accumulation was observed in the studied BRS highlighting potential for in-system problems as well as downstream mobilisation. . Composition, morphology and colour analysis confirmed that BRS polymer liners are a key source of MPs and MEPs in soil filter media. . Composition analysis showed that MPs in these systems are not simply breakdown products of coarser MEPs indicating a variety of sources contribute to plastic ingression into BRS filter media. The presence of GPTs was a significant factor affecting MP abundance in BRS filter media while the influence of other factors such as system age and population density was not clear. As a result, GPTs are highly recommended as a management strategy to curb plastic pollution into stormwater infrastructures. . Further research is needed to gain insight into the vertical distribution of plastic particles within BRS soil media. . Overall our results show that BRS designed to capture, store, and remediate urban stormwater runoff also act as a sinks and possible transmitters for plastic particles.