Comparison of two procedures for microplastics analysis in sediments based on an interlaboratory exercise

Abstract

Microplastics (MP) are distributed throughout ecosystems and settle into sediments where they may threaten benthic communities; however, methods for quantifying MP in sediments have not been standardized. This study compares two methods for analyzing MP in sediments, including extraction and identification, and provides recommendations for improvement. Two laboratories processed sediment samples using two methods, referred to as “core” and “augmentation”, and identified particles with visual microscopy and spectroscopy. Using visual microscopy, the augmentation method yielded mean recoveries (78%) significantly greater than the core (47%) (p = 0.03), likely due to the use of separatory funnels in the former. Spectroscopic recovery of particles was lower at 42 and 54% for the core and augmentation methods, respectively. We suspect the visual identification recoveries are overestimations from erroneous identification of non-plastic materials persisting post-extraction, indicating visual identification alone is not an accurate method to identify MP, particularly in complex matrices like sediment. However, both Raman and FTIR proved highly accurate at identifying recovered MP, with 96.7% and 99.8% accuracy, respectively. Low spectroscopic recovery of spiked particles indicates that MP recovery from sediments is lower than previously assumed, and MP may be more abundant in sediments than current analyses suggest. To our knowledge, likely due to the excessive time/labor-intensity associated with MP analyses, this is the first interlaboratory study to quantify complete method performance (extraction, identification) for sediments, with regards to capabilities and limitations. This is essential as regulatory bodies move toward long-term environmental MP monitoring.

Graphical Abstract

1. INTRODUCTION

Microplastics (MP; plastic particles 1 nm – 5 mm) are globally distributed throughout ecosystems and pose increasing ecotoxicological concerns, though the extent of their impacts is still unknown (de Sá et al., 2018; Ajith et al., 2020; Huang et al., 2021; Koelmans et al., 2022; Perumal and Muthuramalingam, 2022). Once plastics enter the environment, physical and chemical processes such as aggregation and coating by biofilms consisting of algae, bacteria, environmental particles (e.g., organic matter), and small benthic organisms cause many to settle into sediments (Barnes et al., 2009). An estimated 70% of marine litter is expected to sink into sediments (Frias et al., 2016) and MP concentrations in benthic zones can be up to six orders of magnitude greater than in surface waters (Hoellein et al., 2017). Though the term “microplastics” is widely used to describe small plastic fragments, it comprises a diverse suite of plastic particles with varying chemical compositions, sizes, shapes, color, densities, and chemical additives (Rochman et al., 2019). Denser plastics, such as polyvinyl chloride (1.38 g mL⁻¹), are expected to sink rapidly into sediments, while less dense particles, like polystyrene (0.96 g mL⁻¹), will likely sink more slowly and be dependent on other processes, such as aggregation with other particles or growth of biofilms. Physical changes from weathering can influence the settling velocity (Kowalski et al., 2016), as can varying shapes, given that flakes and films are more likely to float compared to particles with three-dimensional shapes.

The pervasiveness of MP in sediments presents potential ecotoxicological concerns for the many benthic organisms that are filter or detrital feeders and are therefore at higher risk of ingesting MP (Bellasi et al., 2020). Once ingested, MP can have detrimental physiological impacts on aquatic organisms including molecular, cellular, and systemic effects, while damaging apical endpoints and causing hepatic stress (Cole et al., 2011; Wright et al., 2013; Auta et al., 2017; Franzellitti et al., 2019; Ajith et al., 2020; Huang et al., 2021). Because benthic invertebrates contribute up to 90% of fish prey biomass, exposure to MP in sediment may lead to bioaccumulation and trophic transfer of MP (Bellasi et al., 2020; Huang et al., 2021). At the community scale, exposure to MP and nanoplastics may decrease the abundance of macroinvertebrates and other benthic fauna altogether (Green, 2016; Redondo-Hasselerharm et al., 2020). Therefore, it is essential to better understand MP contamination in sediments.

Many methods for isolating, extracting, and identifying MP in sediments exist, but have not been standardized. A density separation step with salt solution is present in most methods, though this step varies in the composition and densities of salt solutions employed, oxidation techniques applied, and the equipment and instrumentation used for processing and identification (Hidalgo-Ruz et al., 2012; Cashman et al., 2020, Maisto et al., 2022). Method variations make it difficult to compare results and understand spatial and temporal trends in reported MP abundances; therefore, standardized methodologies or a system for evaluating methodologies is needed to provide clear, reproducible guidelines for processing and analyzing MP (Underwood et al., 2017; Prata et al., 2019). Method evaluations and comparisons often occur within a single laboratory or research group, but it is important to demonstrate that methods can be reproduced among laboratories and achieve similar results (Hanvey et al., 2017). Further, while visual microscopy has been widely used in MP studies, differentiation between plastic and non-plastic particles can be difficult, especially for particles smaller than 1 mm (Song et al., 2015), and visual identification can often lead to false positives and negatives, resulting in over- or under-estimation of MP (Lenz et al., 2015). Raman and Fourier-transform infrared (FTIR) spectroscopy have become integral to identifying the chemical composition of MP (Lu et al., 2021), and are more effective than visual identification alone (Käppler et al., 2016). Some studies have compared various MP extraction methods for sediments (Pagter et al., 2018; Han et al., 2019; Cadiou et al., 2020; Cashman et al., 2020, Radford et al., 2021) and others have compared identification instrumentation (e.g., Raman spectroscopy, FTIR, pyrolysis GC-MS) (Song et al., 2015; Käppler et al., 2016; Käppler et al., 2018; Müller et al., 2020). To our knowledge, no study compares two procedures side-by-side in an interlaboratory study from start to finish, i.e., from extraction to polymer identification, and doing so is essential for real-world applications (e.g., source determination and control). This is, in part, because performing complete MP analyses on sediment samples is very time consuming and labor intensive; for example, a complete MP analysis of a sediment requires approximately 240 hours. Consequently, only a limited number of investigations have performed any sort of interlaboratory comparison.

The surge in research reporting the abundance, distribution, and potential impacts of MP over the last decade has resulted in new policies and legislation regarding MP production and monitoring (Halfar et al., 2021). The Save Our Seas 2.0 Act directs U.S. federal agencies to foster research that broadens scientific understanding of how MP affect ecosystems and human health (https://www.congress.gov/bill/116th-congress/senate-bill/1982/text). As noted, sediments play a key role in this understanding as a major sink and, ultimately, as a potential source of MP. California recently adopted both a legislatively mandated statewide microplastics strategy for its coastal waters (https://www.opc.ca.gov/webmaster/ftp/pdf/agenda_items/20220223/Item_6_Exhibit_A_Statewide_Microplastics_Strategy.pdf), and standardization of methods for analyzing MPs in drinking water (https://www.waterboards.ca.gov/drinking_water/certlic/drinkingwater/microplastics.html). In response, the Southern California Coastal Water Research Project (SCCWRP) organized a method evaluation study to provide a scientific foundation for selecting and standardizing methods for extracting and identifying MP. The study examined the performance of candidate methods for measuring MP across four matrices: clean water (De Frond et al., 2022a, this issue), “dirty” water (to simulate contaminated surface water or wastewater), sediment, and fish tissue, to understand method accuracy, precision, reproducibility, and required resources (i.e., supplies, equipment, labor, time). In addition to the core SCCWRP study, additional augmentation studies were conducted to assess different aspects of method performance. The augmentation presented here tested a sediment extraction method with three primary differences from the core method adopted by SCCWRP: i) using sodium bromide as the salt versus calcium chloride, ii) performing a second density separation with a denser solution versus a single density solution, and iii) using separatory funnels rather than beakers.

The objective of this study was to compare the performance of two methods (i.e., core versus augmentation) for extracting MP from sediments using laboratory-spiked sediment samples, followed by visual and spectral identifications. A particular focus of this interlaboratory comparison was assessing MP recovery when using only simple visual microscopy versus more complex spectroscopy. We hypothesize the augmentation method will achieve higher recovery, due to the use of salt solutions with greater densities, and the application of separatory funnels which allows rinsing to ensure complete removal of particles. This study provides recommendations for improving MP quantification in sediments and other porous media for effective monitoring of MP contamination in the environment.

2. MATERIALS AND METHODS

2.1. Participating Laboratories and Study Design

Two laboratories participated in this study: the U.S. Environmental Protection Agency Atlantic Coastal Environmental Sciences Division in Narragansett, Rhode Island, USA (USEPA) and SCCWRP (Costa Mesa, CA). Both laboratories extracted MP from sediment using both the core and augmentation methods. Six identically prepared spiked sediment samples were provided to each laboratory for the method comparison – three each for the core and augmentation methods (SI Figure 1). The number of replicates per method (n=6) was chosen with consideration of balancing adequate performance data with time/labor-intensive analytical processes. Further, this sample size was used in the main study of the SCCWRP intercalibration study and is consistent with other MP methodological studies (Cadiou et al., 2020; De Frond et al., 2022a).

2.2. Sample Preparation

SCCWRP prepared the sediment samples using dated pre-industrial oceanic sediment cores from Woods Hole Oceanographic Institute (Falmouth, MA, USA). In brief, the sediment was sieved, freeze-dried, and homogenized, then aliquoted into 25 g (dry weight) samples. The samples were spiked with gelatin pills containing known amounts of plastic polymers (i.e., recovery standards) in various shapes, sizes, and colors including polystyrene (PS), polyethylene (PE), polyethylene terephthalate (PET), and polyvinyl chloride (PVC) (SI Table 1). Samples were also manually spiked with natural particles as false positives: cotton fibers, cellulose fibers, marine shell fragments, and domestic rabbit fur (SI Table 1). Blanks were prepared in an analogous manner. Further details are in Supplementary Information (SI).

2.3. MP Extraction and Size Fractioning

We used the SCCWRP core sediment method and a novel augmentation method (Cashman et al., 2022) (SI Figure 2). In the core method, the sediment sample was combined with CaCl₂ solution (1.4 g mL⁻¹) in glass beakers three times (SI Figure 3). The particles were then sieved into three size fractions. In the augmentation method, separatory funnels and NaBr solution at two different densities (1.3 and 1.5 g mL⁻¹) were used instead of the beakers and single-density solution (SI Figure 4). The particles were sieved into two size fractions and underwent oxidation with H₂O₂. Both laboratories counted and distinguished particles with visual microscopy and then identified the polymers with Raman or FTIR. The smallest size fraction (1–20 μm) was excluded in the present study due to interference from the gelatin pills. Brief method summaries follow, with additional details in SI.

2.4. Core Method

The sediment sample was transferred to a 1 L glass beaker (Beaker 1) and submerged in a CaCl₂ solution. The sample slurry was stirred manually for 2 minutes and left to settle for 2 hours. The floating debris was then scooped into a new beaker (Beaker 2) using a metal spoon, and the supernatant was decanted into Beaker 2. A second 2-hour density separation using CaCl₂ solution was performed with sediments remaining in Beaker 1. The floating debris was again scooped into Beaker 2 and the supernatant decanted into Beaker 2, for which a third density separation was conducted using the CaCl₂ solution, stirred for 2 minutes, then left to settle overnight (12–24 hours). The floating debris was then scooped onto stacked sieves (500 μm, 212 μm, and a sieve pan) and the supernatant was also decanted through the sieves. The debris on the sieves was rinsed with MP-analysis-grade (MAG, see SI) water into a clean beaker. The solution of each size fraction was filtered through a polycarbonate track-etched (PCTE) filter (20 μm pore size). By this method, the MP particles were separated into three size fractions: 20–212 μm, 213–500 μm, and >500 μm.

2.5. Augmentation Method

The sediment sample was first sieved into two size classes: 251 – 1000 μm (size class A) and 45 – 250 μm (size class B). Each fraction was transferred into a separatory funnel (1 L). An aliquot of low density NaBr solution (1.3 g mL⁻¹) was added to submerge the sediments, and the mixture was manually shaken for 3 minutes. The sediment was left to settle for 2 hours. The settled material was transferred gently through the stopcock into a new separatory funnel. The supernatant with floating debris was drained through the stopcock directly into a filtration cup and filtered through a 20 μm pore PCTE filter. Two filters of particles, i.e., 251–1000 μm and 1.3 g mL⁻¹ NaBr (A-1), and 45–250 μm and 1.3 g mL⁻¹ NaBr (B-1), were produced by this step. Another aliquot of NaBr solution with higher density (1.5 g mL⁻¹) was added into the new separatory funnel which was shaken for 3 minutes and left to settle for 2 hours. The settled particulates were discarded by gently opening the stopcock. The remaining liquid and floating debris was filtered through a 20 μm PCTE filter. Another two filters of particles, i.e., 251–1000 μm and 1.5 g mL⁻¹ NaBr (A-2) and 45–250 μm and 1.5 g mL⁻¹ NaBr (B-2) were produced. To remove biofilms, all filters were moved to 30 mL crystallization dishes and submerged in ~ 10–20 mL 30% H₂O₂. The dishes were covered with aluminum foil and moved to a 60 °C oven to oxidize for 2 hours, after which the samples were re-filtered through new 20 μm pore PCTE filters, respectively.

2.6. Particle Identification

Suspected plastic particles were hand-picked using fine-tipped forceps under a stereoscope (USEPA: SMZ745-T stereoscope (Nikon, Minato, Tokyo, Japan), SCCWRP: LAXCO (Washington, USA)) and placed on a 5.1 cm × 7.6 cm glass microscope slide with double-sided tape. The USEPA picked suspected plastic particles from the large size classes only (251 – 1000 μm in the augmentation method, and 212–500 μm and >500 μm fractions in the core method). SCCWRP picked particles from all size fractions in both extraction procedures. Magnification, background lighting, and illumination settings were recorded for each filter. Color, shape, and tactile sensation tests (e.g., checking for hardness with forceps) were used to choose suspected MP. These characteristics were also recorded. Each picked particle was then photographed and measured along the longest perpendicular axes (length and width). SCCWRP executed this picking process for up to 30 particles of each distinct color and morphology (e.g., blue fragment, white spheres) (De Frond et al., 2022b), while the USEPA picked all suspected particles. For both methods, the USEPA used Raman spectroscopy to identify MP particles, while SCCWRP used FTIR. Detailed descriptions of the spectroscopy methods can be found in the SI.

2.7. Quality Assurance/Quality Control (QA/QC)

MP extraction and particle identification were conducted in laboratories with HEPA air filtration. The SCCWRP laboratory has positive pressure to prevent contamination from outside particles and the USEPA conducted all vacuum filtration under a laminar flow hood. All staff wore 100% cotton laboratory coats to minimize contamination by synthetic polymers. Laboratory plastic use was minimized. Glassware and samples were covered when not in use, and glassware and sieves were pre-cleaned with soap and water using a natural sponge. Blank samples, including a laboratory air blank and a procedural blank, were taken for every three samples to establish background particle levels.

2.8. Data Analysis

Raw data were analyzed to build three metrics: visual recovery, spectroscopic recovery, and spectroscopic accuracy. Visual recovery was calculated as follows:

$$\mathit{\text{Visual}}\mathit{\text{Recovery}}\left(\%\right)=\left(\frac{\#\mathit{\text{recovered}}\mathit{\text{particles}}\mathit{\text{after}}\mathit{\text{visual}}\mathit{\text{identification}}}{\#\mathit{\text{spiked}}\mathit{\text{particles}}}\right)\times 100$$ (1)

where ‘# recovered particles after visual identification’ was the particle count visually identified by comparing morphology, size, and color of particles on the filter to known spiked particles, and ‘# spiked particles’ was the total number of spiked particles. This metric is identical to that used elsewhere in the SCCWRP intercalibration study (De Frond et al., 2022a, this issue), and is possible here given the characteristics of the spiked microplastic particles is known.

Spectroscopic recoveries were calculated for each spiked polymer type (Eq. 2 with PE as an example) and for all spiked polymers (Eq. 3), respectively.

$$\mathit{\text{Spectroscopic}}\mathit{\text{Recovery}}\mathit{\text{of}}\mathit{\text{Each}}\mathit{\text{Polymer}}\left(\%\right)=\left(\frac{\#\mathit{\text{recovered}}\mathit{\text{particles}}\mathit{\text{identified}}\mathit{\text{as}}PE}{\#\mathit{\text{spiked}}PE}\right)\times 100$$ (2)

$$\mathit{\text{Total}}\mathit{\text{Spectroscopic}}\mathit{\text{Recovery}}\left(\%\right)=\left(\frac{\#\mathit{\text{recovered}}\mathit{\text{particles}}\mathit{\text{identified}}\mathit{\text{as}}PE+PS+PET+PVC}{\#\mathit{\text{spiked}}PE+PS+PET+PVC}\right)\times 100$$ (3)

Spectroscopic accuracy was determined using a two-step verification process: SCCWRP first examined images of the recovered particles submitted by both laboratories to verify whether they were those spiked; then SCCWRP confirmed those recovered spiked particles were identified correctly by comparing the reported chemical ID with the known chemical composition of the spiked MP (Eq. 4).

$$\mathit{\text{Spectroscopic}}\mathit{\text{Accuracy}}\left(\%\right)=\left(\frac{\#\mathit{\text{spiked}}\mathit{\text{particles}}\mathit{\text{identified}}\mathit{\text{correctly}}}{\#\mathit{\text{spiked}}\mathit{\text{particles}}\mathit{\text{recovered}}}\right)\times 100$$ (4)

These results excluded fibers, as it is very difficult to distinguish spiked fibers from contamination. In addition, for the smallest size class for both methods (45–250 μm for augmentation and 20–212 μm for core), the USEPA performed analyses using point maps on the Raman spectrometer, which does not allow the user to examine particles after the analysis to verify whether they were spiked.

These three metrics (mean ± standard deviation) allowed examination of the data from different perspectives. Visual recovery readily quantifies particles based on simple discrimination of color and morphology but cannot identify polymer type and therefore is only a rough evaluation of MP recovery. Spectroscopic recovery is an absolute recovery value, based on the known number of spiked particles, which provides a more exact evaluation of method performance. Spectroscopic accuracy is useful for examining performance of Raman or FTIR to identify known polymers correctly.

To determine statistically significant differences between treatments (i.e., core versus augmentation methods, or Raman versus FTIR), one sided t-tests were performed. Analysis of variance (ANOVA) was performed (one-factor and two-factor, when appropriate), with post-hoc Tukey tests as appropriate, to determine significance among recovery (visual and spectroscopic) with multiple treatments (method and morphology/color, etc.). Statistical significance was set at α = 0.05. All statistical analyses were conducted in R Studio (RStudio, PBC, Version 1.4.1717). In all figures, error bars represent a single standard deviation.

The size classes for the two methods differed slightly; for the small size class the core used 20–212 μm while the augmentation used 45–250 μm, for the large size class the core used 212–500 μm and > 500 μm while the augmentation used 251–1000 μm. The two large size classes from the core method were combined (i.e., >212 μm) before performing statistical analyses to make the core and augmentation datasets as comparable as possible. No spiked particles were between 20–45 or 212–250 μm, therefore the size classes were analogous (SI Table 1).

3. RESULTS

3.1. Blanks

There was an average of 21±13 particles per sediment blank across both methods. Fibers and fragments were the most abundant morphologies present, while black and clear were the most common colors of particles detected (SI Table 2). We did not blank correct, as the blank contamination from all sources (sediment collection, spiking, shipping, extraction, and analysis) was low compared to the number of spiked particles.

3.2. Visual Recovery

Using visual microscopy, the augmentation method (78±11%) achieved a statistically significant greater mean recovery than the core method (47±29%) for total particle counts (Figure 1, SI Table 3). The augmentation method also had greater precision than the core method, with the overall standard deviation decreasing by 61%. Although mean percent recovery was greater in the augmentation method for all morphologies and colors except blue particles, no statistically significant differences were observed between these groups. In addition, no significant differences in recovery were found by size class (i.e., small vs. small; large vs. large) (Figure 2a) or between laboratories using either method (Figure 2b). The accuracy of the visual recoveries will be discussed in detail below.

Figure 1.

Total recovery (%) summary statistics for the core and augmentation methods for visual microscopy and spectroscopy. Statistics include mean (x), the range (error bars), and 25, 50 (median), and 75 percentiles (lower, middle, and upper horizontal lines, respectively). Each method had a sample size of six (n = 6). A statistically significant difference (*) was detected between the two methods for visual microscopy (p = 0.03).

Figure 2.

Visual recovery results: a) Mean recovery (%) between methods by size class using visual microscopy. No statistically significant differences were detected in comparisons between size classes and methods. b) Mean recovery (%) between USEPA and SCCWRP laboratories using visual microscopy. No statistically significant differences were detected in comparisons between laboratories using either method. c and d) Mean recovery (%) by color and morphology for the (c) core method and (d) augmentation method. This data does not include the smaller size class (< 212 or < 250 μm) for either methods due to inconsistencies in interpreting color or shape on the Raman spectrometer. For the core method, statistically significant differences in recoveries between the particle morphology and color with the blue, green and sphere particles are shown with a, b and c, respectively (p ranging < 0.001 to 0.046). No significant difference was found between morphologies or colors within the augmentation method.

Morphology and color seemed to play an important role in recovery for the core method; mean recovery for blue, green, and spherical particles using the core method were all greater than 60%, with recovery for the clear, white, and fragment particles significantly lower than the blue particles (Figure 2c). Similarly, recoveries of white and fragment particles were significantly lower than green particles and spherical particle recovery was significantly greater than fragments. In contrast, augmentation recovery for all types of particles exceeded 60% without any significant differences among particles (Figure 2d). To summarize, in the core method fragments had lower recovery than other shapes, and blue and green particles had higher recovery than other colors.

3.3. Spectroscopic Recovery

Although the augmentation method achieved a higher mean spectroscopic recovery, results did not indicate any statistically significant differences in overall MP recovery between the core (24±6%) and the augmentation (31±8%) methods (Figure 1, SI Table 4). Similarly, there was no significant difference between methods by size class; however, significantly greater recovery was observed for large size classes compared to small size classes for both methods (Figure 3a). There was no significant difference in recovery between laboratories for either method, though the USEPA achieved greater recovery values (Figure 3b). Despite overall higher recovery, none of the individual polymer recoveries were significantly greater in the augmentation method than in the core method (Figure 3c, d). However, within each method, significant differences were detected: recoveries of both PE and PS were statistically significantly greater than those of PVC and PET in both methods (Figure 3c, d).

Figure 3.

Spectroscopic recovery results: a) Mean recovery (%) between methods by size class using spectroscopy. No statistically significant differences were detected in comparisons between size class and method, though both large size classes were significantly greater than both small size classes (p ≤ 0.0005). b) Mean recovery (%) between USEPA and SCCWRP laboratories using spectroscopy. No statistically significant differences were detected in comparisons between laboratories in either method. c and d) Mean recovery (%) by polymer for the (c) core method and d) augmentation method. For the core method, the significant difference between PE and PS was p = 0.011. The significant differences between PE and PS and the other two polymers (PET and PVC) were p ≤ 0.006. For the augmentation method, the significant difference between PE and PS was p = 0.04. The significant differences between PE and PS and the other two polymers (PET and PVC) were p ≤ 0.004. There was no significant difference between PVC and PET recovery in either method (p > 0.98).

3.4. Spectroscopic Accuracy

Raman and FTIR both achieved high accuracy and precision at 96.7±2.3% and 99.8±0.4%, respectively for both methods combined (Figure 4, SI Table 5). There was no significant difference in spectroscopic accuracy between the augmentation and core methods; however, FTIR did have a significantly greater accuracy than Raman overall. This difference lies in the core Raman accuracy, which had the lowest mean percent accuracy (94.9%) (Figure 4). There was a significant difference between the core Raman and core FTIR accuracy measurements, along with a significant difference between core and augmentation Raman, which indicates Raman performance improved in the augmentation method, likely due to gained experience by the analysts.

Figure 4.

Mean spectroscopic accuracy (%) of Raman and FTIR for both methods. Each dataset (i.e., Raman augmentation, FTIR augmentation, Raman core, FTIR core) has a sample size of three (n = 3). There was a significant difference (*) between the core Raman and core FTIR accuracy (p = 0.003) and between core Raman and augmentation Raman (p = 0.02).

4. DISCUSSION

4.1. Recovery using Visual Identification & Differences between Methods

Previous visual microscopy method comparison and validation studies for MP in sediment have achieved recoveries of 40 – 100% (Quinn et al., 2017), 78 – 100% (Han et al., 2019), 2 – 87% (Cashman et al., 2020), 87% (Scherer et al., 2020), 88 – 100% (Liu et al., 2021), and 37.9 – 100% (Cashman et al., 2022), depending on polymer type and method. Both methods in this study achieved recoveries within this broad range, although the augmentation recovery is more similar to the higher recovery studies. Regardless, there is no accepted recovery threshold for MP sediment investigations; Cashman et al. (2020) used 70% as a threshold for method performance and Brander et al. (2020) suggested 80% though they acknowledged that recovery might be lower for complex matrices.

A key difference between the methods potentially influencing recovery of different particle types was the use of beakers in the core method versus separatory funnels in the augmentation method. The former required scooping particles out of beakers, which may have led to particles adhering to the glassware walls, while the latter rinsing of the inside funnel walls resulted in more plastic particles being recovered, and more sediment debris, falling on the filters (Fries et al., 2013) (SI Figure 3). This may explain the lower core recovery of fragments than of other morphologies and colors. In addition, brightly colored particles (e.g., blue, green) were more easily recovered as they were easier to spot adhered to the side of the beaker and on the filters than clear and white particles. Further, the density separations at two different solution densities in the augmentation method (1.3 and 1.5 g mL⁻¹) versus the single density solution used in the core method (1.4 g mL⁻¹) might have resulted in additional particles recovered. The higher density solution allowed for extraction of MP with a greater density and using four filters reduced the likelihood that any would be too crowded with MPs for spectral analysis. Additional modifications could be introduced to the augmentation method to alleviate sediment and organic matter buildup on the filters, such as the vacuum method (https://www.seed.world/build).

4.2. Spectroscopic Accuracy

Spectroscopic accuracy was quite high for both Raman and FTIR (96 and 99%, respectively). These results were very similar to previous measures of accuracy for Raman at 91 to 99% (De Frond et al., 2022a; Jin et al., 2022) and FTIR with values of 95 to 98% (Renner et al., 2019; De Frond et al., 2022a). Both analytical techniques are effective tools to identify MP particles extracted from porous media and low recovery is not due to spectroscopic error but likely poor extraction efficiency.

4.3. Recovery using Spectroscopic Identification

Because this is the first study to our knowledge to measure spectroscopic recovery, we are unable to compare our results with previous studies. Comparing to the recovery thresholds previously discussed for visual microscopy, these recoveries are quite low at 24% and 31% for the core and augmentation methods, respectively. However, if we exclude the low PVC and PET recoveries, overall recovery increases to 42±11% (core) and 54±16% (augmentation) (Figure 3c, d). Reasons to exclude PVC and PET are: 1) the PET particles are primarily fibers, which tend to have low recovery in MP studies due to their small diameter and probable loss during the sieving steps of the sediment extraction (Cashman et al., 2020) and 2) PVC is relatively dense (1.38–1.41 g mL⁻¹ (Lusher et al., 2020)) and very close to the density of the salt solutions used in both methods in this investigation.

While exclusion of PVC and PET increases the overall spectroscopic recovery, we acknowledge that it is still relatively low (42–54%). Reasons for these low recoveries may include losses during the extraction, filtration, and oxidation steps, which are generally more common in complex matrices such as sediments. Specifically, there are several opportunities for MP loss during the extraction process: sieving, transfer to separation equipment (i.e., beaker, separatory funnel), vacuum filtration, and transport of the filter from both the filtration apparatus and to the analytical instrument. Additional losses may occur if there is an oxidation step that requires moving the filter to a new container and repeating vacuum filtration, as in the augmentation method. These losses may be mitigated or quantified using appropriate standards for measuring extraction efficiency (i.e., recovery standards - see discussion below).

Our low recoveries of spiked particles using spectroscopic recovery relative to visual identification (SI Table 5) suggests that visual microscopic identification of spiked MP standards for QA/QC purposes (e.g., recovery) is not accurate unless the spiked particles are very distinct or can otherwise be identified using methods such as unique fluorescent signatures. Our results also suggest that PE and PS show promise as recovery standards, and polymers and particles that are not as dense as PVC nor fiber-shaped should be further investigated as recovery standards. However, if PVC is a polymer of interest, a separation solution with greater density may capture denser particles. While this will capture more sediment particles, it may lead to higher PVC recoveries. The inability to estimate fiber recovery is a significant research gap as MP fibers are thought to be highly abundant in aquatic systems and have detrimental toxicological effects on organisms due to their elongated shape (Gago et al., 2018; Qiao et al., 2019; Pirsaheb et al., 2020; Rebelein et al., 2021; Ugwu et al., 2021). In addition, white and clear particles may be less suitable as recovery standards than brightly colored particles, although white and clear particles may give a more realistic measure of particles found in the environment. That said, researchers should recognize that recovery of fibers and denser plastics such as PVC are limited, and particles chosen as recovery standards should be based on data quality objectives of the study (Radford et al., 2021).

4.4. Visual Microscopy vs. Spectroscopy

The difference between visual and spectroscopic recovery results is most likely due to incorrect visual identification. In porous sedimentary media (e.g., sediment and soil), non-plastic particles, such as silicates, carbonates and other minerals, will persist in the sample through extraction, making it extremely difficult to select only plastic particles during visual microscopy. Several studies have noted how matrix properties, such as clay content, charge, bulk density, and organic matter, may affect the extent of MP transport and recovery, aggregation with other particles, and persistence of non-plastic particles post-extraction (Li et al., 2020, Wang et al., 2020, Li et al., 2021, Radford et al., 2021, Maisto et al., 2022). In this study, visual microscopy overestimated the recovery of plastic particles, confirming that visual identification alone is not an accurate method to identify MP particles when sediment or other confounding particles are co-extracted (Lenz et al., 2015; Song et al., 2015; Kroon et al., 2018). The standard deviation dropped from 27% for visual recovery overall to 8% for spectroscopy overall, indicating that spectroscopic recovery is more precise than visual recovery. Low recoveries of spiked particles using spectroscopic identification suggest that MP recovery from sediments and other porous media may be lower than previously assumed. Thus, current MP analyses may underestimate levels of MP contamination in the environment.

5. CONCLUSIONS

To our knowledge, this is the first interlaboratory study to quantify method performance with regards to both capabilities and limitations, from start to finish, including both extraction and chemical identification of MP from sediment. Overall, the augmentation method yielded greater recovery for total MP from the sediment than the core method. No significant differences were detected between the two laboratories regarding performance of the augmentation, thus confirming reproducibility of this method. The high accuracy of both spectroscopic techniques proves the successful performance of both Raman and FTIR for identifying MP. However, methods for extracting and isolating MP from sediments, and by extension other porous sedimentary media, underestimate MP quantities.

We recommend the use of separatory funnels for MP extraction as well as sufficiently high-density solutions for the study objectives. In addition, development of appropriate recovery standards should be pursued to achieve experimental objectives. Optimizing polymers for recovery standards to assess environmental particle recovery will improve method performance. For example, a general monitoring program may use a diverse suite of recovery standards, while targeted studies (e.g., tire wear particles) would use more relevant materials. Studies targeting fibers or dense polymers, such as PVC, may need to develop specific recovery standards and methods. Researchers are also encouraged to include spectroscopic identification in future method validation studies to better understand its role in environmental MP quantification. Finally, despite the excessive time and labor intensity of interlaboratory comparisons, they are critical to the continued optimization and standardization of MP methodologies for wide-ranging regulatory and monitoring applications.

As regulatory bodies move toward regular monitoring of environmental MP levels, accurate methods demonstrating acceptable and reproducible extraction efficiencies are crucial for long-term applicability and comparisons. For example, the US EPA’s National Coastal Condition Assessment program has included quantification of sediment MP in a 2020 pilot study, and California’s MP management strategy for coastal waters (https://www.opc.ca.gov/webmaster/ftp/pdf/agenda_items/20220223/Item_6_Exhibit_A_Statewide_Microplastics_Strategy.pdf) will include sediments in a statewide monitoring program. Different environmental matrices and data quality objectives of studies will likely dictate the method chosen; however, development and use of recovery standards to demonstrate extraction efficiencies are necessary to allow meaningful comparison among data obtained from different approaches. Regardless of the methods used, it is crucial to recognize that MP levels in the environment are likely much higher than current investigations indicate.