Sediment samples were collected as part of the Clean Seas Environment Monitoring programme (CSEMP) ( Figure 1 and Table S1 ). Collected data were also presented according to UK regionals seas ( Figure S16 ) and OSPAR regions II & III ( Figure S11 ). Sediment samples were collected using a stainless steel 0.1 m² Day Grab, rinsed before deployment with Pentane to reduce sample contamination between sites. Surficial sediment (top 2 – 3 cm) was collected using a solvent rinsed metal spoon and placed in a pre-rinsed glass jar with foil and cap. Samples were stored at -18°C until ready for analysis. Plastic items stability was assumed following the freezing process from direct observation and no impact on microplastics stability was reported under simulated freeze-thaw cycles by Koutnik et al. (2022).

The chemicals used in this study are listed in supplementary information ( Table S2 ).

Standard Operating Procedures (SOPs) for the collection of sediment samples were developed to ensure reproducibility of sampling. Dedicated field technicians were trained using SOPs to minimise field contamination during sampling for the time-period 2018-2021. Empty pre-rinsed reverse osmosis water (RO) glass jars were also used as field blanks and were open for the time required to transfer a sediment sample from the Day Grab to the clean, pre-rinsed RO glass collecting jars. For the time-period 2013-2017, archive sediment samples collected for the investigation of organic contaminants were used. The same sampling protocol was used for the collection of sediments for organic contaminants; however, glass jars were Pentane rinsed only and some particle contamination could have occurred during collecting jar preparation and sample handling.

Prior to use, all glassware was cleaned using a laboratory detergent and rinsed using RO water. Care was taken during the field sampling to minimise sample contamination by only opening jars for the minimum amount of time during sample transfer. All chemical solutions used in this study were previously filtered using a 47 mm diameter, 0.2 μm regenerated cellulose membrane filter. Sample handling was carried out inside a biological safety cabinet (BSC). Contamination monitoring within the laboratory was carried out by using blank filters processed in the same way as environmental samples for each batch of samples processed. For time efficiency, a procedural control was prepared with every batch of samples and subjected to the same preparation steps.

Alongside samples for microplastic analysis, additional sediment was collected for particle sized analysis (PSA). This was subsampled from the Day grab using the same method as described in section 2.1. Samples were placed in a plastic bag within a plastic tub and stored at -18°C until analysis. All samples were analysed at the Cefas laboratory, Lowestoft, following the NMBAQC method (Mason, 2011). This involves taking a subsample for laser diffraction and then wet-splitting the remaining sediment over a 1mm sieve. The sediment >1mm is oven dried at 70°C and then dried sieved at half phi intervals. The <1mm sediment is left to settle for 24 hours, before the overlying water is syphoned off and the sediment is oven dried at 70°C to calculate total weight of the >1mm sediment. The subsample taken for laser diffraction is wet screened at 1mm and subsequently the <1mm is analysed using the Beckman Coulter LS 13 320 laser sizer to produce detailed size distributions of the <1mm at half phi intervals, down to 0.004 μm. The dry sieve and laser data are then combined to produce the final results for the full PSA distribution of the sediment. Accompanying statistical summaries were calculated using GRADISTAT (Blott and Pye, 2001).

Prior to processing for PSA, a subsample was taken from the sediment sample at each station for total particulate organic carbon and nitrogen (OCN) analysis. The subsample is taken from the whole sample and frozen, then freeze dried for a minimum of 7 days or until completely dry. The freeze-dried sediment is then screened over a 2mm sieve, the >2 mm sediment is discarded and the <2 mm sediment is ground to a fine powder using the Pulverisette 5 model IV planetary ball mill. All grinding equipment is cleaned with denaturated alcohol (IMS) before and between each sample to ensure there is no cross-contamination. Samples are then sent to the University of East Anglia for Organic Carbon and Nitrogen analysis. Inorganic carbon is removed using a sulphurous acid digest, samples are then freeze dried and ground. Ground samples are weighed into tin capsules for analysis. Total carbon samples are dried and ground. Samples are then weighed into tin capsules for analysis. Samples are analysed using an Exeter Analytical CE440 CHN combustion analyser. Quality assurance procedures include analysis of 3 repeats completed for 1 in 10 samples with RSD% <10%, certified reference materials (CRM) and in-house reference materials (IHRM) are run every 4 determinations. Limits of detection are for organic carbon (<0.01%) and nitrogen (<0.001%), and measurement limits are calculated as 10 times the detection limit.

Quantification of microplastics was achieved using an automatic counting tool of the fluorescent items onto filters described below. The final count of microplastics was also compared to a manual counting for validation from the acquired digital imaging.

A Python-based model was created for the fast recognition and counting of the fluorescent images onto filters from the digital images. The tool utilises the package scikit-image within the Python scripting language (van der Walt et al., 2014). The image is first divided into separate wave bands for analysis (red, green, and blue). Segmentation of the image into background and foreground (plastics) is undertaken using a threshold value derived by the tool or by the user. The tool calculates a threshold value based on the Otsu method histogram of the red and green fluorescence of the whole image. The Otsu method looks to separate the image into two classes by identifying a threshold which reduces intra-class variance by maximising inter-class variability (Otsu, 1979). This type of threshold is particularly suited for where the histogram is bimodal, such as with the microscope images.

The tool uses a decision tree to decide whether to use the derived threshold from either the red band or the green band of the image or whether to use one of the pre-determined user defined thresholds to segment the image ( Figure S9 ). Once the image is separated into two classes (bright spots and background) the areas are merged to form objects from each class. Cleaning of the objects then takes place removing objects smaller than the user defined minimum area and filling in holes in objects below 10 pixels in size. For each of the plastic objects a series of statistics are calculated including: i) mean red value, ii) mean green value, iii) mean red value, iv) area (pixels) and v) eccentricity (scale between 0 and 1 on how round an object is). Outputs from the tool include an image of the plastics identified and a tale of statistics for each object identified as a plastic. The tool can run iteratively on a folder of images or multiple folders of images providing separate results tables for each folder. Units were expressed in number of particles kg^-1 dry weight (d.w.) sediment. Additional parameters were reported such as type of particle (e.g., spheres, fragments, film, pellets, or foam), polymer type and sizes of the items analysed using μ-FTIR.

The mean number of fluorescent items onto the laboratory procedural controls (i.e. negative controls) were subtracted to the final count of fluorescent items onto filters for samples.

Suspected microlitter were identified for microplastic confirmation and polymer identification. A LUMOS II (Bruker, UK) using micro-ATR and transmission FT-IR with a liquid nitrogen-cooled MCT detector. For micro-ATR FT-IR, spectra (32 scans) were collected in reflectance mode in the range 4000-500 cm^-1 at a resolution of 4 cm^-1. For particles that couldn’t be analysed using micro-ATR FT-IR, transmission mode was used. For transmission mode, particles of interest were manually transferred to 25 mm Anodiscs filters (0.2 μm porosity, Whatman, VWR, UK). Spectra (32 scans) were collected in transmission mode in the range 4000-1250 cm-1 at a resolution of 4 cm-1. For all cases, polymer identification was verified based on the % match against provided polymer libraries (ATR-FTIR-Library complete, Vol.1-4; Bruker Optics ATR-Polymer Library; IR-Spectra of Polymers, Diamond -ATR, IR-Spectra of Polymers, Geranium-AT & IR-Spectra of Additives, Diamond-ATR). Only matches above 60% were selected for a positive microplastic validation and polymer identification (Leistenschneider et al., 2021).

Due to the discrete nature of the variable, a Poisson or Negative binomial distribution was initially considered. However, because of numerical convergence difficulties for the statistical models for Poisson and Negative Binomial, a different approach was adopted: the raw data were square root transformed and Normal errors were assumed. A Mann-Kendall (MK) analyses and linear regressions (LR) were applied to investigate any trends. The MK analysis was conducted using the R function mannkendall in the R-library emon (Barry et al., 2017). For the statistical analysis and to make plots clearer, square root transformations were applied to the microplastic counts kg^-1 d.w. sediment. Because replications from the same station are not independent, means are taken of the transformed counts for each station. These mean values are used in the trend analyses.

To study variation among and between groups (i.e. sampling sites and sampling years), an ANOVA test was used to analyse the differences among group means, followed by Tukey HSD post hoc test as the multiple comparison procedure using IBM SPSS^® Statistics (version 28.0.1.1) with α=0.05. Both ANOVA and Tukey’s test assume independence of samples, homogeneity of variance and normality of residuals (Zar, 1999). Analysis was conducted as ANOVA is considered robust to such departure from normality where large data sets are employed (Underwood et al., 1997).

A spectroscopic based method was also developed in parallel to the NR method using a LUMOS II FTIR microscope with focal plane array (FPA) detector (Bruker, UK) and is summarised in Appendix 6. As a summary, 10 g of dried sediment were weighed into three 50 mL polypropylene centrifuge tubes in a BSC with ventilation. Dried sediments were digested using the dropwise addition of filtered (0.2 μm) H2O2 (30%) until bubbling stops. Following the digestion step, the tubes were covered using a RO rinse filter membrane secured into place using a fine metal wire and subsequently placed in a drying cabinet < 50°C for 48 hours prior to density separation. Density separation was carried out by using a 1.5 g mL^-1 solution of zinc chloride (ZnCl2). Approximately 35 mL of zinc chloride was added to each of the centrifuge tube and 37 mL was added to an empty tube as a control. The tubes were then well shaken to homogenise the samples. Each tube was centrifugated at 3900xg for 5 minutes. Each supernatant was transferred to a previously cleaned filtration unit and filtered using a 25 mm diameter 0.2 μm porosity Whatman Anodisc (VWR, UK). The whole process was repeated one more time and the supernatants combined on the same filter. Residues of zinc chloride were rinsed with 100 mL of RO water and particles stuck onto the funnels were rinsed using RO water. Each filter was then carefully transferred to previously cleaned glass petri dish and transported to a drying cabinet for 24 to 48 hours under 50°C prior to analysis using a LUMOS II μ-FTIR with FPA detector (Bruker, UK). The whole area (100%) of the 25 mm Anodiscs filters (0.2 μm porosity, Whatman, VWR, UK) was scanned. FTIR spectra were collected using FPA detector in transmission mode using a single scan in the range 4000-1300 cm-1 at a resolution of 8 cm-1 using a 4x4 binning (LOD of ~ 20 μm). Spectra were converted using a macro in Bruker OPUS (version 8.5) and particle identification was carried out using the siMPle software developed by Aalborg University (Denmark) and Alfred Wegener Institute (Germany) (https://simple-plastics.eu/). Results were expressed as number of microplastics, corresponding sizes (in μm). The software also allowed for the identification and counting of non-microplastic items including natural items (e.g., chitin) and fibres (e.g., cotton) decreasing the reporting of non-plastic items and related false positives as compared to the NR method.

Visual inspection of the filters confirmed the presence of chemical resistant materials onto filters. 10% KOH appeared to interfere with the fluorescence tagging method using NR and a 30% KOH : NaClO (v:v) was investigated instead. KOH : NaClO has been applied in several studies as a chemical digestion solution for sediment and biota with no impairment of the identification of the tested polymers by Raman micro-spectrometry (Strand and Tairova, 2016; Enders et al., 2017). Enders et al (2017) developed and tested an alkaline digestion protocol to preserve small plastic particles while removing organic tissue materials. They concluded that using a 30% KOH : NaClO (v:v) solution was effective for the digestion of fish stomachs. The combination of KOH : NaClO was also found to be more effective than KOH and NaClO alone. As a result, a 30% KOH : NaClO (v:v) was prepared and tested in the laboratory for sediment samples. KOH and NaClO combined was effective at suppressing most of the fluorescence of biological residues while leaving intact the fluorescent behaviour of polymers ( Figure S2 ). As a result, a 30% KOH : NaClO (v:v) was selected as the alkaline digestion method for sediment samples.

Recoveries of microplastics from spiked clean sediments ranged from 71 ± 10% (mean ± SD) for PP to 100 ± 0.01% for PA and PES ( Figure S3 ).

Contamination detected on laboratory negative controls averaged 3.31 items per filter (n=24), which was subsequently subtracted from the total particle count for the environmental samples. Fibres were mainly found in the negative controls and were identified as predominantly natural (i.e. cotton and Rayon fibres) indicating an effective contamination control procedure for other types of microplastics such as fragments. Particles identified in the field blanks were mainly cellulose-based fibres and natural polyamides. ( Figure S10 )

A positive relationship was observed between the abundance of microplastics and the silt/clay content of sediments (MK: p=0.001) ( Figure 7 ). Maes et al. (2017a) also reported a higher abundance of microplastics from seafloor sediment from the Southern North Sea for sediments with a lower grain size Maes et al, 2017b

It has been suggested that the transport of microplastics in rivers may behave similar to the transport of sediments with fine sediments and microplastics being carried in suspension ( Galgani et al., 2000 ; Browne et al., 2010; Costa et al., 2011; Ballent et al., 2012; Bakir et al., 2014). Particle position in the water column and sinking and deposition velocities onto sediments is mainly influenced by the environment’s turbulence level. With a decreasing level of turbulence along a depth-gradient from coastal environments, seafloor environments are likely sinks for the very fine sediments (< 63 μm) and thus microplastics correlating with that particle fraction size such as fibres, low density polymers and microplastics < 500 μm (Enders et al., 2019). Enders et al., 2019 also suggested that higher density polymers exhibited a sedimentation closer to source, limiting their long-distance transport in the marine environment. However, such a hypothesis is difficult to validate as additional factors such as particle shape, density, conglomeration with suspended particulate matter, interactions with biota (e.g., ingestion and excretion), biofouling and physical resuspension will affect deposition mechanisms and affect short to long-range transport of microplastics.

Results from this study demonstrated the integration of microplastics in sediment aggregates along with other pieces of organic matter, especially for sediments with high silt/clay content (section 6.1). Microplastics are also known to influence soil organic carbon fraction distribution as well as contributing to sediment carbon storage (Rillig, 2018; Yu et al., 2021). As a result, microplastics in sediment can be detected as carbon using methods used to quantify sediment organic carbon (Kim et al., 2021) and may also affect sediment parameters such as bulk density which is important to estimate carbon concentration to stocks (e.g. in kg C ha^-1) (Rillig, 2018).

Positive relationships were observed between the abundance of microplastics and the organic carbon (OC) (LR: p=0.003 and MK: p=0.002) and nitrogen (< 2 mm) (LR: p=0.003 and MK: p=0.002) in the present study ( Figure 8 ). Maes et al. (2017a) also reported a positive relationship between the abundance of microplastics in seafloor sediment collected from the Southern North Sea and total organic carbon (TOC) in sediment (Maes et al, 2017b). It is however unclear whether the observed positive relationship with organic carbon is due to the presence of microplastics or due to the high silt/clay % in sediment. Further work is recommended to understand the relative contribution of microplastic-carbon in sediment.