Environmental Science and Pollution Research (2020) 27:5559–5571

https://doi.org/10.1007/s11356-019-07274-5

RESEARCH ARTICLE

Comparison between manta trawl and in situ pump filtration

methods, and guidance for visual identification of microplastics
in surface waters
Therese M. Karlsson 1 & Anna Kärrman 2 & Anna Rotander 2 & Martin Hassellöv 1

Received: 23 July 2019 / Accepted: 3 December 2019 / Published online: 18 December 2019
# The Author(s) 2019

Abstract
Owing to the development and adoption of a variety of methods for sampling and identifying microplastics, there is now data
showing the presence of microplastics in surface waters from all over the world. The difference between the methods, however,
hampers comparisons, and to date, most studies are qualitative rather than quantitative. In order to allow for a quantitative
comparison of microplastics abundance, it is crucial to understand the differences between sampling methods. Therefore, a
manta trawl and an in situ filtering pump were compared during realistic, but controlled, field tests. Identical microplastic
analyses of all replicates allowed the differences between the methods with respect to (1) precision, (2) concentrations, and (3)
composition to be assessed. The results show that the pump gave higher accuracy with respect to volume than the trawl. The
trawl, however, sampled higher concentrations, which appeared to be due to a more efficient sampling of particles on the sea
surface microlayer, such as expanded polystyrene and air-filled microspheres. The trawl also sampled a higher volume, which
decreased statistical counting uncertainties. A key finding in this study was that, regardless of sampling method, it is critical that a
sufficiently high volume is sampled to provide enough particles for statistical evaluation. Due to the patchiness of this type of
contaminant, our data indicate that a minimum of 26 particles per sample should be recorded to allow for concentration
comparisons and to avoid false null values. The necessary amount of replicates to detect temporal or spatial differences is also
discussed. For compositional differences and size distributions, even higher particle counts would be necessary. Quantitative
measurements and comparisons would also require an unbiased approach towards both visual and spectroscopic identification.
To facilitate the development of such methods, a visual protocol that can be further developed to fit different needs is introduced
and discussed. Some of the challenges encountered while using FTIR microspectroscopic particle identification are also critically
discussed in relation to specific compositions found.

Keywords Microplastics . Surface water sampling . Monitoring . Particle quantification . Method development . Plastic
pollution . Microlitter

Introduction

Early plastic pollution research built on lessons from

Responsible editor: Christian Gagnon planktologists. For example, one of the metrics that have been
Electronic supplementary material The online version of this article used to report microplastic concentrations is the ratio of
(https://doi.org/10.1007/s11356-019-07274-5) contains supplementary microplastics/plankton (Moore et al. 2005). Sampling was of-
material, which is available to authorized users. ten done with surface water trawls, and identification was
done visually with the aid of stereomicroscopy—methods that
* Martin Hassellöv
are still commonly used. Consequentially, a common cutoff
[email protected]
size limit in many microplastic studies is 0.3 mm since that is
1
Department of Marine Sciences, University of Gothenburg, also a common mesh size for studying zooplankton, fish eggs,
Kristineberg 566, 45178 Fiskebäckskil, Sweden larvae, and other organisms in the nekton.
2
MTM Research Centre, School of Science and Technology, Örebro It was, however, recognized early on that one method
University, Orebro, Sweden would not be adequate to assess microplastic pollution,
5560 Environ Sci Pollut Res (2020) 27:5559–5571

particularly since microplastics come in many different organic structures, and the presence of clear and homogenous
shapes, sizes, and densities (Moore et al. 2005). This was colors, while fibers should be equally thick throughout their
further emphasized in a recent paper by Rochman et al. length (Hidalgo-Ruz et al. 2012). It has also been suggested
(2019). Additionally, a majority of plastic pollutants are ex- that uneven crooked edges and/or distinctive colors can be
pected to sediment (Woodall et al. 2014; Koelmans et al. used as characteristics to distinguish microplastics from natu-
2017) or beach for periods of time. As the research field ma- rally occurring particles (Viršek et al. 2016). The categories
tured, other methods, specifically adapted to different situa- used for classification also vary but often refer to source, type,
tions and research questions, were developed (reviewed in shape, color, and/or degradation stage (Hidalgo-Ruz et al.
Hidalgo-Ruz et al. 2012, Renner et al. 2017a, Prata et al. 2012). Of those studies that then move on to further charac-
2018). The different methods have allowed research to specif- terize and analyze the particles, 28% use Fourier transform
ically identify microplastics in surface waters, biota, and sed- infrared (FTIR) spectroscopy (Renner et al. 2017a), where
iments from all over the world. Several scientists have, how- there are also discrepancies regarding how to obtain and inter-
ever, noted the need to harmonize sampling methods pret the spectra (Renner et al. 2017a, b, 2018). Guidelines for
(European Commission Joint Reseach Centre 2013; Lusher environmental monitoring protocols recommend spectroscop-
2015; Van Cauwenberghe et al. 2015; Setälä et al. 2016; ic analysis to complement visual identification, but there are
Rochman et al. 2017). no established recommendations on how the subset of parti-
Today, surface waters remain an important area of study cles should be selected (European Commission Joint Reseach
and particles above 0.3 mm are commonly sampled with Centre 2013; GESAMP 2019).
trawls, such as the manta trawl, or filtering pumps. The degree To be able to provide advice for future microplastics and
of comparability between trawl and pump samples is however microlitter sampling and analysis, in both research and mon-
poorly understood. Previous studies using both trawl and itoring, the aims were to (1) use field tests to examine differ-
pump methods with varying mesh sizes reported inconclusive ences between the most common methods for sea surface
results. A larger number of particles are quantified when using sampling (manta trawl and filtering pump), (2) use collected
smaller mesh sizes, as shown for sea water samples taken by field data as the basis on which to discuss appropriate sam-
both a bongo net (mesh size 0.3 mm) and a submerged pump pling volumes, and (3), in line with recent calls for harmoni-
(mesh size 0.044 mm) resulting in 0.045 ± 0.093 and 2569 ± zation (Hartmann et al. 2019), propose a comparable and less
1770 particles/m3, respectively (Cai et al. 2018). A sampling ambiguous protocol for identification and coupled visual and
campaign in the Baltic, using both a submersible pump and a spectroscopic classification. This study has been focused on
manta trawl (both with a mesh size 0.3 mm) on 10 different sea surface sampling, but many of the prerequisites and ob-
stations, resulted in a similar microlitter (including combus- servations are equally valid for sampling microplastics in sed-
tion particles) concentration range between 0.9 and 1.9 parti- iment and biota.
cles/m3 (Setälä et al. 2016). However, since there was only
one replicate from each station, with low particle counts, a full
statistical comparison was not possible, and there was no clear Materials and methods
correlation with any category of microlitter for the two
methods. Given the highly heterogeneous distribution of Sampling
microplastics, a method comparison puts high demands on
the study design. So far, information on the variation between Sampling was performed in the Gullmar fjord on the Swedish
replicates for different methods is lacking, due to the absence west coast on October 10, 2017 (sampling center:
of replication in most studies. Hence, comparison between 58°15,756 N 011°26,770 E), during a day with calm winds
different methods is currently impossible. and little current. A nearby weather buoy was used to retrieve
Another challenge encountered when trying to compare data on winds and currents before and during the sampling
results from microplastic surveys is the lack of comparable (station number 33025 ID Kristineberg, latitude: 58°15.41′N,
protocols for identification of microplastic, and other litter, longitude: 11°27.06′E).
in the samples. Identification is primarily done by visual iden-
tification, often with the aid of stereomicroscopy (Renner et al. Trawl
2017a). Of those that use visual identification for
microplastics in environmental samples, half use it as a way The trawl consists of an aluminum frame with an opening
to narrow down potential microplastics that are then further width of 67.5 cm and height of 20 cm. A nylon net with a
analyzed. The other half uses it as a stand-alone method mesh size of 0.3 mm was attached to the frame. At the
(Renner et al. 2017a). Researchers use different protocols for end of the net, detachable “cod-ends” were attached using
distinguishing plastics from natural particles. For larger parti- hose clamps. The cod-ends were made from 3-mm-thick
cles (> 1 mm), indicators include the absence of cellular or gray polypropylene tubes with a length of 23 cm and a
Environ Sci Pollut Res (2020) 27:5559–5571 5561

diameter of 11 cm. Nets with the mesh size of 0.3 mm In situ filtration pump
were attached to the tubes, also using hose clamps
(Fig. 1). One tube was used per sample, and to avoid Six replicate pump samples were taken simultaneously with
contamination, each tube was cleaned and prepared with the trawl samples from a boat anchored in the center of the
Milli-Q (< 0.22 μm) in the lab, back-flushed, and sealed trawl trajectory. The filtration pump used is commercially
with aluminum foil before and after the sampling. Three available, and is the second generation developed by KC
different heights were marked on the sides of the trawl Denmark in the EU CleanSea project 2012–2014 (Fig. 2).
frame opening, to show a water height of 12.5 cm, The main differences from the first CleanSea pump relate to
10.0 cm, and 7.5 cm, from the bottom of the trawl net the motor and flowmeter. The current pump is fitted with a
level. To test the variation in water height, a miniature higher-capacity motor than its predecessor, and the original
waterproof camera (GoPro) was attached. The water digital flowmeter has been replaced with an electromagnetic
height in relation to the three marks was then recorded one. The pump is a stainless steel multi-stage Grundfos sub-
by converting the film to stacked images. The stacked mersible pump. The 1-phase motor (0.55 kW/230 V, Grundfos
images were analyzed in Fiji 64 for Windows, using a MS402) is located at the top, followed by the water intake and
macro detailed in Online Resource 1a. The data were then a filter stack that fits three different filters. The filters are
analyzed based on times above, between and below the 14 cm in diameter and made in stainless steel using a laser
three different marks. Net size was analyzed by taking cutter. Although this study focused on a 0.3-mm mesh size, it
images with a Leica M205C camera stereomicroscope, is possible to use sequential filtering with three different mesh
and measuring differences in mesh width, length, and di- sizes. The mesh size of the 0.3-mm filters was analyzed by
agonal length. taking images with an Axiocam ERc 5 s mounted on a light
Trawl samples were taken onboard RV Oscar von Sydow. microscope (Stemi 508, Zeiss) and measuring differences in
The trawl was pulled alongside the ship from a boom for 0.5 mesh width, length, and diagonal length. At the bottom of the
nautical miles. The trawl was lifted as the ship turned, and was pump, the electromagnetic flowmeter (PD340) measures the
then pulled along the respective counter course to compensate
for any differences due to currents. In total, ten samples were
collected, and for each sample the course was altered by 72°.
This created a cartwheel-shaped sampling pattern, in relation
to which the pump was positioned at the center. The trawl
repeatedly sampled across and back, Fig. 4). Each new trawl
replicate started in five different directions (shifted by 36°)
and then repeated again for a replicate size to ten. Three blanks
were collected by preparing tubes and bringing them into the
field, attaching them to the trawl and then analyzing in the
same way as the samples in the lab. In addition, 2 carryover
samples were taken after the first and last samples, by
attaching clean code-ends and rinsing the trawl from the out-
side with seawater, to assess the carryover between samples.

Fig. 2 The pump consists of a motor above a water inlet grid (4 ×

20 mm). The water is passed through a filter stack, and the filtered
Fig. 1 Detachable cod-ends, made from a polypropylene tube with a volume measured by an electromagnetic flowmeter. A synthetic rope is
mesh attached to it, were prepared and cleaned in the lab to avoid pictured here, but a stainless steel wire could also be used to further avoid
contamination or loss of materials risk of contamination
5562 Environ Sci Pollut Res (2020) 27:5559–5571

water volume passing the filters with an accuracy of (± From the probability density function, the probability for
0.006% or 1.2 L at 20,000 L/h). different random results can be calculated based on a true
In this study, the pump was suspended horizontally about mean and a set of observations.
0.5 m out from the boat. The 360° water intake, which mea- If the distribution approximates a Gaussian distribution, the
sured approximately 20 cm in diameter, was situated 5–10 cm number of replicates necessary to detect a difference between
below the water surface. Prior to sampling, the filters were samples can be calculated. For a t test between two samples,
ultrasonicated, rinsed with deionized water, and stored in jars with a power of 80 (meaning that the difference would be
lined with aluminum foil. Two blank filters were processed detected 80% of the time), Eq. 4 can be used:
and stored in the same way, except that they were not sub-  σ 2
mersed underwater. Assuming a stainless steel wire is used for N ≈16 ð4Þ
deploying the pump, the only plastic materials associated with D
the pump are the external electrical cables, a silicone rubber where N is the number of replicates and D denotes the differ-
seal, and two internal black nitrile o-rings. ence that should be detectable between samples.

Statistical evaluation Visual analysis

Within a given sampling area, there is a true concentration that All samples were kept in the cooling room at 5 °C between
can only be measured by sampling the entire area (the true analyses. Visual analyses were performed using a stereomi-
population). As that is not practically possible, subsamples are croscope, following the MyVISPEC protocol outlined in
taken to estimate the true concentration. A number of assump- Online Resource 2 and 3. One analyst analyzed all the samples
tions are then made in regard to the subsamples. These as- to maximize comparability. Samples with a lot of seagrass and
sumptions can for example be based on the assumption of other macro debris were analyzed by carefully investigating
Gaussian distributions (commonly referred to as a normal dis- both sides of the macro debris under the stereomicroscope and
tribution). As microplastics typically arrive to the sampler as moving any adhered microlitter. All particles identified as an-
whole particles, i.e., discrete numbers, it is more suitable as a thropogenic were sorted into categories related to shape, so-
first approximation to adopt the assumption of a Poisson dis- lidity, color, longest length and perpendicular width, and de-
tribution, at least until the number of particles per analysis scriptive categories (see MyVISPEC protocol in
reaches high enough levels that the abundance (expressed as Online Resource 2 and 3). Each particle was assigned a parti-
integers per sample) approaches a continuous Gaussian distri- cle ID and photographed using a Leica M205C camera
bution, such that the Poisson statistical uncertainty is less than microscope.
other measurement uncertainties. Based on the measured con-
centrations in the current study, these relationships are inves-
tigated as described below. FTIR analysis
In order to calculate the probability value for a sample with
the volume V (m3) of filtered water, one can use Eq. 1 FTIR analyses were performed using a Nicolet iN10, reflec-
tance mode with an MCT detector cooled with liquid nitrogen.
μV ¼ λ ð1Þ
On each day of analysis, the inferometer was calibrated and a
where λ corresponds to the arithmetic mean of particles per reference clear PE film was analyzed at the beginning and end
sample. In Poisson distributions, the variance σ2 has been of the day, as a quality control sample.
shown to be equal to the mean. Therefore, the standard devi- Although it is common to use a spectral resolution of
ation for that probability value can be calculated through Eq. 2 8 cm−1 (Renner et al. 2017a), a higher resolution of 4 cm−1
was employed for this work to provide (1) better discrimina-
σV ¼ √μV ð2Þ tion between low-density polyethylene (LDPE) and high-
density polyethylene (HDPE) (Gulmine et al. 2002), and (2)
The probability value only approximates the true value at a to distinguish characteristic peaks related to degradation, and
high number of observations (which, in this case, can be ob- increase the accuracy with which the automated library search
tained by a large sample volume). For a lower number of functions. In this study, the spectra were analyzed with the
observations, Poisson statistics (e.g., Haight 1967) provide a software Omnic Picta (v9.1), using correlation with the soft-
probability density function P(x = k), as in Eq. 3: ware libraries Thermo Wizard library, Hummel polymer sam-
λk e−λ ple library, and Polymer laminate films. In addition, compar-
P ðx ¼ k Þ ¼ ð3Þ isons were made with two internal libraries, one with clean
k!
reference samples and one with household plastics and
Environ Sci Pollut Res (2020) 27:5559–5571 5563

degraded plastics. All matches were confirmed with visual evaluation. For the pump (Fig. 3), the majority (32/40) of the
inspection of the key peaks for the respective polymer types. anthropogenic particles were visually identified as microplastics,
while 5 were identified as semisolids, 1 as fat and 1 as a com-
bustion particle. The microplastics were mainly fragments (20/
Results and discussion 32) but also had some expanded cellular plastics (3/32), some air-
filled spheres 3/32), and a couple of fibers (2/32).
Pump
Trawl
Precision
Precision
The pump filtered 20 m3 (± 1.2 L) per sample, as measured by
the magnetic flow transmitter L, corresponding to a variation For the trawl, the video analysis resulted in a data set showing
of 0.006%. Random measurements of the pump mesh size on the times that the water height was between marks, or above or
three 0.3 mm filters (ten replicates for each filter) showed that below the marks, in terms of images and corresponding time
the average width was 0.299 mm, with a standard deviation of as detailed in Table 1.
0.014 mm and values ranging from 0.279 to 0.328 mm. The For the actual trawled distance, the ship nautical data ren-
diagonal length was on average 0.444 mm with a standard dering was used, which translated to an estimated average
deviation of 0.016 mm. volume of 62 m3 per sample at the middle depth mark
(Table 2). This estimate, however, depends if there is a devi-
Concentration ation from the estimated trawl depth. A given trawling dis-
tance of 0.5 nautical miles and estimated water height position
In the pump samples, between 0 and 13 particles identified as in line with the middle part of the trawl (10 cm) give a volume
microplastics were found, a concentration that corresponds to of 63 m3. On the other hand, if the trawl would be steady at the
0–0.4 microplastics/m3. The highest concentration was found lower mark, the sampling volume would only be 47 m3, and
for the first replicate, and the lowest concentration for the last conversely, with an assumed water level on the higher mark,
replicate. Additionally 0–2 other types of microlitter particles the volume would be 78 m3, which would be a significant
were found per sample as detailed under composition (Fig. 3). difference. Based on the video analysis, the water level was
No microplastics were found on the blank filters. observed to be within the low and the top mark for the most
part. If the results are extrapolated to a full trawl sample of 0.5
Composition nautical miles, the volume would be 64 m3, 1 m3 more than
the visually estimated volume. These are still quite crude mea-
In total, only 35 particles were identified as microlitter particles in surements, but they indicate an error of at least 1% in volume
the pump samples. Compositional differences between the repli- estimations for the trawl.
cates are therefore not reliable. In order to compare the compo- Random measurements on 10 spots in the mesh used for
sitions between the sampling devices, all particles found in the the trawl code-ends showed that although they were marketed
trawl and in the pump samples, respectively, were pooled for the with a specified mesh size of 0.300 mm, the average width

Fig. 3 Composition of all

particles found in the six pump
samples
5564 Environ Sci Pollut Res (2020) 27:5559–5571

Table 1 Height of the water in trawl in relation to the different marks Composition
Middle mark Top mark Low mark
For the trawl, the number of particles per sample ranged be-
Above Below Above Below Above Below tween 23 and 83 microlitter particles. This is again insufficient
for detailed compositional analysis between samples, but it
Images 199 118 12 305 292 25
can be used to illustrate the variation between replicates, as
Time (s) 19.4 11.5 1.2 29.7 28.4 2.4
seen in Fig. 5. Among the first samples, the portion of frag-
ments was higher than among the last samples, which were
was 0.258 mm, with a standard deviation of 0.01 mm and dominated by expanded cellular plastics.
values ranging between 0.258 and 0.271 mm. The diagonal For the trawl 256/376, anthropogenic particles were visu-
length was on average 0.379 mm with a standard deviation of ally identified as microplastics (Fig. 6). Other types of
0.008 mm. microlitter included occasional combustion particles (10/
376), fat particles (5/376), and also semisolid synthetic parti-
cles (103/376), which made up a total of 20% of the total
Concentration anthropogenic particles. Among the microplastics, the major-
ity of the particles were classified as fragments (139/256),
The trawl had between 11 and 57 microplastic particles per even when the fragments of expanded cellular plastics were
sample, giving concentrations ranging between 0.18 and 0.92 excluded (62/256).
microplastics/m3. Additionally, 5–26 other types of microlitter The particles from the trawl were grouped according to
particles were found per sample. approximate lengths and widths. The length was measured
Trawl blanks showed one paraffin particle. Therefore, con- as the longest length, and the width was measured perpendic-
tamination due to air exposure can be assumed negligible in ular to the length. The size grouping (Fig. 7) showed that the
these samples. Inter-sample contamination (extra trawl rinse) amount of particles increased with decreasing sizes down to
showed 1–3 particles per replicate, suggesting a potential car- 0.3 mm (the lower size limit). A few particles with a smaller
ryover effect, although relatively small. width were recorded, but these were typically long fibers.
Concentrations of microlitter in the trawl samples were Among all particles in the trawl, 82% were found to be trans-
observed to change throughout the day. When sorted accord- parent or white (Fig. 8).
ing to trawl direction, it appeared that the difference between
samples was more closely linked to trawling direction than to
time of the day (Fig. 4). The samples that were sampled in the Statistical evaluation of the sampling results
same direction had very similar concentrations, with the ex-
ception of sample 1 (denoted *), which was noted to pass The modeled values and the observations (expressed as parti-
through a more polluted streak (Online Resource 1b). cles per sample and particles per volume) follow each other,
The small scale changes in current and wind direction, as although the trawl samples had more values towards the lower
measured by a nearby weather buoy, did not show any clear and the higher end of the modeled distribution (Fig. 9). The
relationship to the sample volumes (Online Resource 1c). The modeled values for the pump also illustrate the increased risk
currents during the day were between 3.6 and 9.6 cm/s, which of false zero values with lower particle counts per sample
could give a 5–16% variation in effective trawling speed de- since the curve cuts the y-axis. For the trawl, the curve is closer
pendent on the direction, but since all samples included a to a Gaussian distribution.
counter course haul, it should in principle be compensated In order to calculate size distribution, morphology, or com-
for in the experimental design. position, an even higher number of particles would be desir-
This also emphasizes the importance of including environ- able, for an acceptable level of Poisson relative to standard
mental parameters during sampling conditions. deviations. For large size fractions (> 300 μm) of

Table 2 Comparison between

pump and trawl Pump Trawl

Volume per hour (m3) 25 > 140

Volume precision (%) 0.01 >1
Microplastic concentration in comparative study (particles/m3) 0.23 0.51
Possibility to operate at varying depths Yes No
Possibility to operate with varying mesh sizes Yes No
Potential of plastic contamination from sampling device Small Medium
Environ Sci Pollut Res (2020) 27:5559–5571 5565

one with twice the amount, assuming a similar standard devi-

ation to that observed here, at least 8 replicates would be
necessary. If the minimum difference is instead raised to 50
particles/sample, only two replicates would be necessary.
Conversely, if the same assumptions on distribution and stan-
dard deviation are used to examine a difference of 5 particles
per sample, more than 200 replicates would be necessary to
detect the difference 80% of the time.
The acceptable variation and standard deviation will ulti-
mately be a balance between the research questions and what
is economically and practically feasible. It is however vital to
recognize the limitations imposed by low particle counts and
replicate numbers for under-sampled fractions (especially
larger sizes and uncommon microlitter types) in order to avoid
drawing false conclusions not supported by the obtained data.

Comparison between sampling methods

The repeatability of the volume measurements was notably

Fig. 4 Concentrations of microlitter in trawl samples arranged according higher for the pump. The precision of the trawl was found to
to trawl direction primarily depend on the ability to keep it leveled in the water
and visually estimate the water height during sampling.
microplastics and microlitter, this may not be practically fea- Although our tests using a waterproof camera indicated that
sible. Therefore, it is crucial that studies in this field do not the estimates used in this study might be 1 m3 higher than
draw conclusions beyond the inherent limitations. actual trawled volumes, it should be noted that the height
Analogously, here, from Eqs. 1 and 2 the inverse square root has been observed to vary more in rougher weather condi-
of the sample mean (26 particles) is taken, which would give tions, which would increase the uncertainty in the volume
0.2 or a percentage of 20% relative standard deviation. Since estimation. The actual trawled volume will also depend on
that is less than the observed sample variability (here approxi- the current and wind conditions during sampling, which might
mated to 55% by taking the sample mean through the Gaussian also to some extent explain the differences in concentration
standard deviation), 26 particles could be considered enough associated with trawling direction. The uncertainty in volume
for these types of samples. However, due to the high variations is probably the biggest uncertainty in sampling with a trawl,
noted here, and the general patchiness that is associated with and there is room for innovative approaches to measuring it
plastic pollutants, 26 particles per sample are towards the lower more precisely.
extreme, and should be considered a minimum. The accuracy of the mesh size is likely to depend on the
In order to test how many replicates would be needed to supplier, but should be recorded prior to sampling. Moreover,
allow for spatial or temporal comparisons, Power calculations due to the variations and accuracy level in mesh size, it is more
according to Eq. 4 were performed. The results indicate that in appropriate to report the sizes as > 0.3 mm than the conven-
order to distinguish between a sample with 26 particles and tional 333 μm.
Sampled concentrations of microplastics and other
microlitter differed between the two methods. The trawl sam-
pled higher concentrations than the pump. The pump also
sampled a lower volume and, consequently, had lower number
of particles per sample. Statistical evaluations showed that the
low number of particles gave rise to increased variability when
recalculated to concentration. Low particle counts per sample
also resulted in higher probability of obtaining a false zero
value. However, variations in statistical uncertainty between
the methods are mainly due to differences in sampling vol-
ume, and can therefore be overcome by increasing the sam-
pling volume, or sampling in areas with higher levels of con-
tamination. Our results indicate that a minimum of 26 particles
Fig. 5 Compositional variation between replicates in the trawl samples per sample is necessary, but determining appropriate sample
5566 Environ Sci Pollut Res (2020) 27:5559–5571

Fig. 6 Sample composition in all

ten trawl samples pooled

sizes is not straightforward and will depend on the research necessary criteria (Geelhoed and Glass 2004). The relation-
question. Geelhoed and Glass examined theoretical ways of ships are therefore complex, but it is nonetheless clear from
determining the minimum sample size based on the relation- the current study and relevant literature that future
ship between the sample size and the sample variation. In their microplastics studies need to consider (1) the sample volume;
review, they found that none of the existing theories met all the to get a representative amount of particles per sample and (2)
replication; to accommodate for the plastic pollution patchi-
ness. The latter becomes particularly important for stationary
sampling devices such as filtering pumps compared with
trawls that cover a larger sampling area and therefore are less
affected by the patchiness of the pollutants. Concentrations in
surface water studies are known to differ over several orders of
magnitude. If the results here are recalculated to microplastics/
km2, the average levels (41,000 microplastics/km2) are similar
to concentrations found in accumulation zones according to a
global inventory of microplastics carried out over several
years (Van Sebille et al. 2015). Higher concentrations have
been reported in surface waters closer to urban areas (Mani
et al. 2015; Zhang et al. 2017; Hassellöv et al. 2018). Aside
from urban waters, our results are therefore comparable with
commonly sampled areas in terms of particles per volume.
The volumes used for the pump sampling may therefore be
more suitable for samples taken closer to cities or more pol-
luted areas. It should also be noted that an important merit of
the filtration pump is its capability to sample lower sizes in
cascade filtration setup, which manta trawls cannot.

Fig. 7 Length (a) and width (b) size distribution among microlitter
particles in the trawl samples. The compositional difference between
trawl replicates illustrates a potential temporal variation and the
importance of replicates. The size distributions showed increasing
numbers of particles in the lower sizes. The lower recorded numbers of
larger particles also show that this type of sampling might not be suitable
for particles above 2 or even 1 mm as only 48/376 particles were longer
than 1 mm and 26/376 were wider. These results support recent recom-
mendations regarding the definition of microplastics, where it has been
suggested to use 1 mm as a cutoff rather than the commonly but arbitrarily
adopted 5 mm cutoff limit (Hartmann et al. 2019). Consequently, a dif-
ferent approach (larger mesh and much larger sample volumes) would be
required for 1–5-mm fractions. Fig. 8 Color composition of microlitter particles in the trawl samples
Environ Sci Pollut Res (2020) 27:5559–5571 5567

Fig. 9 Modeled probability 0.2 a 2

Probability density function P(x=k)

density function based on the
measured data compared with

# observations of k
actual observations of 0.15
microplastics per sample for the
pump and the trawl based on a 0.1 1
number of particles per sample
and b concentration of
microplastics/m3 0.05

0 0
0 10 20 30 40 50 60
k (number of MP in sample)

0.2 b

# observations per concentration

Probability density function P(x=k)
1
0.15

0.1

0.05

0 0
0 0.2 0.4 0.6 0.8 1
Concentration (# MPs>300µm/m3)
Average pump (simulation) Average trawl (simulation)
Measurements pump Measurements trawl

The differences in concentrations between the trawl and the not visually weathered and seemed to be intact primary plas-
pump in the current study were mainly attributed to the com- tics. The composition results are based on insufficient sample
positional differences between the methods, although the low sizes for extrapolation, but give a snapshot image of the con-
particle count in the pump sample did not allow for any in- tamination at the time of the sampling in this area.
depth analysis of the compositional difference. The two
methods sample at slightly different depths. The trawl samples Visual analysis
the full surface microlayer, wherefore the trawl might be more
efficient at sampling microplastics that float on top of the Most studies on microplastics in the sea surface have used
surface, such as expanded cellular plastics. The trawl also visual identification and sorting of the particles (Hidalgo-
had a higher amount of microplastic spheres (62/256) with Ruz et al. 2012). Intra- and inter-lab comparisons showed that
entrapped gas, sometimes called micro-balloons. The pump different analysts get different results. Therefore, one analyst
has its sample intake at 5–10 cm below the surface wherefore did all the analyses in this study. As tactile parameters were
it might not be as efficient in sampling the sea surface included to distinguish between solid and semisolid particles,
microlayer, where Song et al. (2014) showed that lighter no false positives were identified with the FTIR analysis,
microplastics tend to accumulate (Song et al. 2014). The dif- which also indicates that careful visual analysis performed
ference in sampling depth might also explain the higher by experienced analysts may be suitable for samples down
amount of spheres found in the trawl, as the spheres were to 0.3 mm. The risk of underestimation should however not
often filled with air. Since the majority of the microplastics be overlooked, as previously illustrated (Karlsson et al. 2016).
were classified as fragments (84% when fibers and expanded The authors therefore encourage a combination of visual, tac-
cellular plastics were included), it is likely that the greatest tile, and spectroscopic methods to rule out false positives, and
source of microplastics in this area is fragmentation of larger to help avoid false negatives, such as the spheres which could
items, which has also been noted in previous studies be mistaken for air bubbles, glass spheres (Karlsson et al.
(Hassellöv et al. 2018; Karlsson et al. 2019). The spheres were 2019), or possibly eggs. The risk of false negatives and
5568 Environ Sci Pollut Res (2020) 27:5559–5571

positives increases with smaller sizes (Primpke et al. 2017), Although the use of the more objective protocol turned out
and it is therefore appropriate that JRC Guidelines (European to be slightly more time consuming, it provided a less ambig-
Commission Joint Reseach Centre 2013) recommend that vi- uous way of categorizing the particles (moving away from
sual identification should not be done without complementary qualified guessing). It also provided more detailed data on
identification techniques for samples below 0.1 mm. The the particle composition and proved to be a practical way of
same guidelines state that most of the microlitter consists separating the particles. It is therefore recommended that fu-
of microplastics, an assertion corroborated by the present ture studies on microplastics or microlitter include a detailed
study. However, the high percentage of other types of description of identification parameters. The suggested proto-
microlitter (> 20%) warrants their inclusion in future col, MyVISPEC, can be found in Online Resource 3 and is
surveys/monitoring programs. Through including the iden- free to use and alter for related purposes.
tification of particles such as paraffin, false positives are
also avoided.
Categories used for sorting particles typically vary be- Fourier transform infrared spectroscopy analysis
tween analysts (Hidalgo-Ruz et al. 2012) but often require
the analyst to make assumptions about the type or source of Similar to visual analysis, there are few protocols for analyz-
the particles, based on visual identification alone. These ing environmental plastics with FTIR spectroscopy and it has
subjective assumptions can increase the risk of drawing been emphasized that the parameters for acquisition and espe-
false conclusions; for example, expanded particles can be cially spectral identification are rarely specified in scientific
several polymer compositions and a weathered piece of col- articles (Renner et al. 2017a). Environmental transformations
ored polyethylene may visually be similar to a paint flake. of plastics (such as oxidative weathering, hydrolysis, biofilm
Additionally, this type of protocol requires a highly experi- formation) give rise to new functional groups, chain scissor-
enced analyst to visually recognize different particles. The ing, and crystal state changes. Consequently, spectral charac-
importance of experience and training cannot be understat- teristics change, in that peaks decrease and broaden, and new
ed in these types of analysis, but in order to increase objec- distinct peaks appear, while broad regions characteristic of –
tivity, a protocol based on a set of subcategories related to OH and –NH groups appear as a result of biofilm formation
particle morphology (such as shape, color, solidity, and (Gulmine et al. 2003; Jakubowicz et al. 2006; Karlsson et al.
size) is suggested (Online Resource 2 and 3). These subcat- 2018). Manual scrutiny of automatic spectral identification,
egories can then be coupled with spectroscopic composition even though it is more time-consuming, gives a more accurate
identification. In accordance with the recent article on def- identification of weathered microplastics (Renner et al.
initions of microplastics (Hartmann et al. 2019), it is sug- 2017a). Additives can also change spectra to some degree.
gested that particle composition and potential sources are Identification of unknown particles is therefore not always
only stated when they can be conclusively determined. (or rarely) possible with pristine polymer library matching,
Therefore, the term “filament” to describe fibers that appear but requires both environmental plastic reference libraries
to be fragments of long filaments is avoided, and instead the and some expert judgment, to scrutinize the computer
all-encompassing term “fiber” is used. matching. Sometimes, the authors state that they have used a
A section on color is also included in the protocol, matching or hit quality index cutoff limit of 60% or 80%, but
since it can facilitate visual separation, and a potential without specifying which software was used (e.g., Yang et al.
bias towards more bright colors cannot be ruled out 2015) or which library (e.g., (Woodall et al. 2014; Avio et al.
(Hidalgo-Ruz and Thiel 2013). 2015; Yu et al. 2016), or which criteria/settings (e.g., (Woodall
Initially, the protocol differed between foils and other et al. 2014; Yang et al. 2015; Avio et al. 2015; Castillo et al.
types of fragments, but as these were sometimes ambigu- 2016). In such cases, a matching rate cutoff is in fact rather
ously identified, it was decided to group these together and nonspecific. Microplastics analysis with FTIR requires
use a more neutral differentiation between them, based on knowledge of polymer spectroscopy (Song et al. 2015;
the tactile section as either rigid or pliable (Online Resource Mecozzi et al. 2016) and can be aided by better-adapted pre-
2). This section also assisted in distinguishing between par- processing algorithms and analytical algorithms, tailored to
ticles that can and cannot be quantified. Pliable particles environmentally weathered materials. As developed, for ex-
and fibers with a width smaller than the mesh size cannot ample, by Renner et al., they managed to increase the accuracy
be assumed to have been reliably quantified by these sam- of identification from 76 to 96% compared with conventional
pling methods, since they can slip through the mesh and library searches, by limiting the comparison to regions with
their presence is therefore dependent on several factors vibrational bands and applying new search algorithms
such as clogging and entanglement. Representative exam- (Renner et al. 2017b). Even with the improved identification
ples of the different types of microlitter are shown in rate, Renner and colleagues still also recommend visually
Online Resource 2. double-checking the spectra after the automatic recognition.
Environ Sci Pollut Res (2020) 27:5559–5571 5569

As there were not enough particles among the pump sam-

ples to compare chemical compositions between the two sam-
pling methods, the total amount of particles was pooled for
chemical analysis. All particles that were visually identified as
microplastics in the pump samples, and in four random trawl
samples, were analyzed. In total, 144 suspected microplastics
were analyzed corresponding to 50% of the total 286
microplastic particles (Fig. 10). The analyses identified all of
the analyzed expanded cellular plastics (10) as polystyrene
(PS) and all spheres (22) as polymethylmethacrylate
(PMMA). Fibers were either polypropylene (PP) polyamide
(PA) or high-density polyethylene (HDPE). Fragments were
identified mainly as low-density polyethylene (LDPE), poly-
propylene (PP), or HDPE. Two fragments were also identified
as PS. Of the fragments, 15 spectra were of insufficient quality
for reliable identification the composition.
Additionally, 64 of the semisolid synthetic particles were an-
Fig. 11 To distinguish between LDPE and HDPE, the peak at 1377 cm−1
alyzed, corresponding to 59% of the total. Seventy-seven percent is often used. Additionally, the peaks at 1471 and 1464 cm−1 can give
were identified as paraffin due to their similarities with reference some indication as these indicate the CH3 and CH2 groups, respectively
spectra of paraffin, PE, and PP. Twenty-three of the particles did
not produce spectra of sufficient quality for identification. was wrong, even though it could in fact be due to a false negative
Another benefit of visual inspection of the spectra is that it FTIR identification. Here, we have instead chosen to write “incon-
allows for further interpretation. One example is the distinc- clusive” if we cannot make a positive identification with FTIR. No
tion between LDPE and HDPE which is typically done false positives were identified with the FTIR analysis.
through looking at the presence or absence of a peak at Microspheres have proved challenging to visually differentiate
1377 cm−1 (Gulmine et al. 2002). This difference would not from glass microspheres, and FTIR identification facilitated recog-
be noted by a typical library search. HDPE and LDPE also nizing visual characteristics (Karlsson et al. 2019). If only visual
differ from each other in the relative height of the peaks at studies are available, the microspheres can be tested with acetone,
1471 cm−1 and 1464 cm−1 (Fig. 11). In this study, 83% of the which would have a strong effect on PMMA, but not on glass.
PE was identified as LDPE and 17% as HDPE. Paraffin also proved challenging to identify using only vi-
Microplastic studies sometimes conclude that some particles sual or spectroscopic tests. The paraffin polymer can generate
could not be identified as plastics by FTIR, and that they therefore false positives as it shares several similarities with PE and PP,
are not plastics—without stating the identity of those particles. both visually and spectroscopically. As paraffin has shorter
This risks giving the false impression that the visual identification chains, some differences in the spectra can be noted in the
ratio between 1464 and 1471 cm−1, but this difference was
not always clear. There are also some indications that the CH2
band around 2902 cm−1 can be helpful in separating paraffin
from PE and PP, although this relationship needs to be verified
with several types of paraffin.

Conclusions

The results presented here illustrate some of the challenges

associated with sampling microplastics. A visual identification
protocol is presented to encourage more objective sample
analysis. The high prevalence of paraffin and its visual and
spectroscopic similarity with polyethylene and polypropylene
show the importance of including tactile probing during iden-
tification. It also exemplifies why smaller fractions would not
be possible to quantify without supplementary chemical tech-
Fig. 10 FTIR analysis of 144 particles from the pump and the trawl niques, such as vibrational spectroscopy, and adds an
5570 Environ Sci Pollut Res (2020) 27:5559–5571

important argument as to why visual identification and FTIR need to obtain permission directly from the copyright holder. To view a
copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
can be used as complementary techniques for microplastics
identification. For subsequent analysis with FTIR, the results
obtained here further illustrate the many uncertainties and pit-
falls potentially encountered when using an automated library
search. Therefore, a high level of caution is recommended.
References
Due to the patchy nature and complexity of these contami- Avio CG, Gorbi S, Regoli F (2015) Experimental development of a new
nants, high sample volumes and replicates are necessary to protocol for extraction and characterization of microplastics in fish
provide quantitative and compositional data. In comparing tissues: first observations in commercial species from Adriatic Sea.
the two sampling devices, it was found that the pump was Mar Environ Res 111:18–26
Cai M, He H, Liu M, Li S, Tang G, Wang W, Huang P, Wei G, Lin Y,
more accurate in volume measurement and versatile for point
Chen B, Hu J, Cen Z (2018) Lost but can’t be neglected: huge
sampling and filter size choice. Due to the lower sampling quantities of small microplastics hide in the South China Sea. Sci
volume, it might be most suitable for sampling in areas with Total Environ 633:1206–1216
higher level of contamination. The trawl on the other hand Castillo AB, Al-Maslamani I, Obbard JP (2016) Prevalence of
integrates samples over a larger area and therefore helps cir- microplastics in the marine waters of Qatar. Mar Pollut Bull
111(1–2):260–267
cumvent some of the problems related to patchiness. It was European Commission Joint Reseach Centre (2013) Guidance on moni-
also more efficient in sampling the sea surface microlayer and toring of marine litter in European seas. MSFD Technical Subgroup
therefore resulted in higher concentrations and more micro- on Marine Litter
spheres and expanded cellular plastics than the pump. Geelhoed B, Glass HJ (2004) Comparison of theories for the variance
caused by the sampling of random mixtures of non-identical parti-
Regardless of sampling method, the tests showed that it is cles. Geostand Geoanal Res 28(2):263–276
important to ensure that a high enough volume is sampled to GESAMP (2019). Guidelines or the monitoring and assessment of plastic
obtain a sufficient particle count. It was concluded here that a litter and microplastics in the ocean (Kershaw P.J., Turra A. and
minimum of 26 particles per sample was needed to allow for a Galgani F. editors), (IMO/FAO/UNESCO-IOC/UNIDO/WMO/
IAEA/UN/UNEP/UNDP/ISA Joint Group of Experts on the
quantitative comparison, although several replicates would be
Scientific Aspects of Marine Environmental Protection). Rep.
necessary to detect differences between the location sampled Stud. GESAMP No. 99, 130p
here and a hypothetical location with double the microlitter Gulmine J, Janissek P, Heise H, Akcelrud L (2002) Polyethylene charac-
concentration. For compositional comparisons, even higher terization by FTIR. Polym Test 21(5):557–563
particle numbers would be necessary. Gulmine, J. V., Janissek, P. R., Heise, H. M., & Akcelrud, L. (2003).
Degradation profile of polyethylene after artificial accelerated
weathering. Polymer degradation and stability, 79(3), 385–397.
Acknowledgments Open access funding provided by University of
Haight, F. A. (1967) Handbook of the Poisson distribution
Gothenburg. The authors are grateful for all the help from the personnel
Hartmann N, Hüffer T, Thompson RC, Hassellöv M, Verschoor A,
the Sven Loven Centre, Kristineberg, especially to Hans Olsson, Ursula
Daugaard AE, Rist S, Karlsson TM, Brennholt N, Cole M,
Schwarz, and Carl Kristensson for assistance during sampling. We also
Herrling MP, Hess M, Ivleva NP, Lusher AL, Wagner M (2019)
want to thank Sinja Rist, Micah Landon-Lane, Felix Hollert, and Max
Are we speaking the same language? Recommendations for a defi-
Threapleton for assistance during parts of the sampling and analysis. We
nition and categorization framework for plastic debris. Environ Sci
also thank Julian Gallego for assistance with the paraffin measurements
Technol
and Peter Tiselius for good advice on trawls and statistics, and Tim
Hassellöv M, Karlsson T, Mattsson K, Magnusson K, Strand J, Lenz R,
Wilkinson for feedback on the manuscript.
van Bavel B, Eidsvoll DP (2018) Progress towards monitoring of
microlitter in Scandinavian marine environments: state of knowl-
Funding information The following funding bodies financially support- edge and challenges: Nordisk Ministerråd
ed this study: the Swedish Agency for Marine and Water management Hidalgo-Ruz V, Thiel M (2013) Distribution and abundance of small
(agreement number: 2175-17), the Swedish Environmental Protection plastic debris on beaches in the SE Pacific (Chile): a study supported
Agency (agreement number: 2219-17-015), Nordic Council of by a citizen science project. Mar Environ Res 87:12–18
Ministers project HARMIC, and JPI Oceans project BASEMAN. The
Hidalgo-Ruz V, Gutow L, Thompson RC, Thiel M (2012) Microplastics
development of the pump was funded under the CleanSea project
in the marine environment: a review of the methods used for iden-
European Union Seventh Framework Programme (FP7/2007–2013) un-
tification and quantification. Environ Sci Technol 46(6):3060–3075
der grant agreement no. 308370.
Jakubowicz I, Yarahmadi N, Petersen H (2006) Evaluation of the rate of
Open Access This article is licensed under a Creative Commons abiotic degradation of biodegradable polyethylene in various envi-
Attribution 4.0 International License, which permits use, sharing, ronments. Polym Degrad Stab 91(7):1556–1562
adaptation, distribution and reproduction in any medium or format, as Karlsson TM, Grahn H, van Bavel B, Geladi P (2016) Hyperspectral
long as you give appropriate credit to the original author(s) and the imaging and data analysis for detecting and determining plastic con-
source, provide a link to the Creative Commons licence, and indicate if tamination in seawater filtrates. J Near Infrared Spectrosc 24(2):
changes were made. The images or other third party material in this article 141–149
are included in the article's Creative Commons licence, unless indicated Karlsson TM, Hassellöv M, Jakubowicz I (2018) Influence of
otherwise in a credit line to the material. If material is not included in the thermooxidative degradation on the in situ fate of polyethylene in
article's Creative Commons licence and your intended use is not temperate coastal waters. Mar Pollut Bull 135:187–194
permitted by statutory regulation or exceeds the permitted use, you will Karlsson TM, Ekstrand E, Threapleton M, Mattsson K, Nordberg K,
Hassellöv M (2019) Undersökning av mikroskräp längs
Environ Sci Pollut Res (2020) 27:5559–5571 5571

bohuslänska stränder och i sediment. Swedish environmental pro- Rochman CM, Regan F, Thompson RC (2017) On the harmonization of
tection agency methods for measuring the occurrence, fate and effects of
Koelmans AA, Kooi M, Law KL, van Sebille E (2017) All is not lost: microplastics. Anal Methods 9(9):1324–1325
deriving a top-down mass budget of plastic at sea. Environ Res Lett Setälä O, Magnusson K, Lehtiniemi M, Norén F (2016) Distribution and
12(11):114028 abundance of surface water microlitter in the Baltic Sea: a compar-
Lusher A (2015) Microplastics in the marine environment: distribution, ison of two sampling methods. Mar Pollut Bull 110(1):177–183
interactions and effects marine anthropogenic litter (pp. 245-307): Song YK, Hong SH, Jang M, Kang J-H, Kwon OY, Han GM, Shim WJ
Springer (2014) Large accumulation of micro-sized synthetic polymer parti-
Mani T, Hauk A, Walter U, Burkhardt-Holm P (2015) Microplastics cles in the sea surface microlayer. Environ Sci Technol 48(16):
profile along the Rhine River. Sci Rep 5 9014–9021
Mecozzi M, Pietroletti M, Monakhova YB (2016) FTIR spectroscopy Song YK, Hong SH, Jang M, Han GM, Rani M, Lee J, Shim WJ (2015)
supported by statistical techniques for the structural characterization A comparison of microscopic and spectroscopic identification
of plastic debris in the marine environment: application to monitor- methods for analysis of microplastics in environmental samples.
ing studies. Mar Pollut Bull 106(1–2):155–161 Mar Pollut Bull 93(1):202–209
Moore C, Lattin G, Zellers A (2005) Density of plastic particles found in Van Cauwenberghe L, Devriese L, Galgani F, Robbens J, Janssen CR
zooplankton trawls from coastal waters of California to the North (2015) Microplastics in sediments: a review of techniques, occur-
Pacific Central Gyre. Paper presented at the The Plastic Debris rence and effects. Mar Environ Res 111:5–17
Rivers to Sea Conference, Redondo Beach, California, USA
Van Sebille E, Wilcox C, Lebreton L, Maximenko N, Hardesty BD, Van
Prata JC, da Costa JP, Duarte AC, Rocha-Santos T (2018) Methods for
Franeker JA, Eriksen M, Siegel D, Galgani F, Lavender Law KL
sampling and detection of microplastics in water and sediment: a
(2015) A global inventory of small floating plastic debris. Environ
critical review. TrAC Trends Anal Chem
Res Lett 10(12):124006
Primpke S, Lorenz C, Rascher-Friesenhausen R, Gerdts G (2017) An
automated approach for microplastics analysis using focal plane Viršek MK, Palatinus A, Koren Š, Peterlin M, Horvat P, Kržan A (2016)
array (FPA) FTIR microscopy and image analysis. Anal Methods Protocol for microplastics sampling on the sea surface and sample
9(9):1499–1511 analysis. J Vis Exp 118
Renner G, Schmidt TC, Schram J (2017a) Analytical methodologies for Woodall LC, Sanchez-Vidal A, Canals M, Paterson GL, Coppock R,
monitoring micro (nano) plastics: which are fit for purpose? Curr Sleight V, Calafat A, Rogers AD, Narayanaswamy BE, Thompson
Opin Environ Sci Health RC (2014) The deep sea is a major sink for microplastic debris. R
Renner G, Schmidt TC, Schram JR (2017b) A new chemometric ap- Soc Open Sci 1(4):140317
proach for automatic identification of microplastics from environ- Yang D, Shi H, Li L, Li J, Jabeen K, Kolandhasamy P (2015)
mental compartments based on FT-IR spectroscopy. Anal Chem Microplastic pollution in table salts from China. Environ Sci
89(22):12045–12053 Technol 49(22):13622–13627
Renner G, Nellessen A, Schwiers A, Wenzel M, Schmidt TC, Schram J Yu X, Peng J, Wang J, Wang K, Bao S (2016) Occurrence of
(2018) Data preprocessing & evaluation used in the microplastics microplastics in the beach sand of the Chinese inner sea: the Bohai
identification process: a critical review & practical guide. TrAC Sea. Environ Pollut 214:722–730
Trends Anal Chem Zhang W, Zhang S, Wang J, Wang Y, Mu J, Wang P, Lin X, Ma D (2017)
Rochman CM, Brookson C, Bikker J, Djuric N, Earn A, Bucci K, Athey Microplastic pollution in the surface waters of the Bohai Sea, China.
S, Huntington A, McIlwraith H, Munno K, de Frond H, Kolomijeca Environ Pollut 231:541–548
A, Erdle L, Grbic J, Bayoumi M, Borrelle SB, Wu T, Santoro S,
Werbowski LM, Zhu X, Giles RK, Hamilton BM, Thaysen C, Kaura Publisher’s note Springer Nature remains neutral with regard to jurisdic-
A, Klasios N, Ead L, Kim J, Sherlock C, Ho A & Hung C (2019) tional claims in published maps and institutional affiliations.
Rethinking microplastics as a diverse contaminant suite.
Environmental toxicology and chemistry, 38(4):703–711.